Require explicit via for linked-list for loops

Removes the header shorthand, extracts shared validation into
LinkedForLoopChecks.kt, fixes until-validation for pointer sentinels,
and updates docs/examples/tests.

.agents/skills/asm6502-coder/SKILL.md

@@ -1,291 +1,204 @@
 ---
 name: asm6502-coder
-description: Writing or understanding 6502/65C02 assembly code using 64tass syntax, for use as stand alone .asm or inline within Prog8 programs
+description: Writing or understanding 6502/65C02 assembly in 64tass syntax, standalone or embedded in Prog8
 ---
 
-# 65(C)02 Assembly Coder Skill
+# 6502 Assembly Coder
 
-You are an expert MOS 6502 assembly development assistant. Keep responses concise and practical — prefer short, correct code examples over lengthy prose.
-You are working with **6502/65C02 assembly** using **64tass syntax**, in separate `*.asm` files or embedded in a Prog8 program (inside `%asm {{ }}` blocks or `asmsub` routines). Follow all rules below.
+Use 64tass syntax, not ca65 or another assembler. Keep answers concise and
+practical. This applies to standalone `.asm` files and Prog8 `%asm {{ }}` or
+`asmsub` blocks.
 
-## Assembler: 64tass Syntax
-### Formatting Rules
-- Labels start in the first column of the line.
-- Instructions are indented at least 4 spaces.
-- Two spaces between the instruction opcode and its operand (e.g., `lda  #1`, `sta  $d020`).
-- End-of-line comments are preceded by two spaces before the `;` (e.g., `sta  $d020  ; border color`).
-- If a comment is the only thing on a line, it starts in the first column (no indentation).
-- Opcodes and operands are written in lowercase (e.g., `lda`, not `LDA`; `$d020`, not `$D020`).
+Reference documentation for the assembler is available locally in
+`~/Documents/prog8-dev-resources` (the 64tass manual, `64tass.md`). Search this
+folder first before looking online.
 
-- **NOT ca65/cc65** or other assemblers. Key differences:
-- `.proc` / `.pend` for procedures (scoping)
-- `_label` for local labels (prefixed with underscore, scoped to `.proc`)
-- **Symbol aliases**: inside a `.proc`, use `_name = P8ZP_SCRATCH_B1` to give a scratch variable a descriptive name.
-- **Anonymous labels**: defined as `+` (forward) or `-` (backward) at the start of a line.
-  - **CRITICAL DISTINCTION**: Label **definitions** are always just a single `+` or `-`. Label **references** use `+`, `++`, `+++` to refer to the 1st, 2nd, 3rd upcoming forward label, and `-`, `--`, `---` for the 1st, 2nd, 3rd preceding backward label.
-  - `+` refers to the NEXT upcoming `+` label, `++` refers to the ONE AFTER that, etc.
-  - `-` refers to the MOST RECENT `-` label, `--` to the one before that, etc.
-  - When a `+` label is passed, the forward-reference count resets (so `++` then refers to the next one after that new label)
-  - Example:
-    ```asm
-    -   dex         ; backward label '-'
-        bne -       ; branch to most recent '-' (the dex above)
-        ldx #5
-    +   dex         ; forward label '+'
-        bne +       ; branch to this same '+' (forward)
-        sta $400
-    +   lda #0      ; second '+' label definition (just '+')
-        bne ++      ; branch to the '++' below (third forward '+')
-        rts
-    +   inc $d020   ; third forward label definition (just '+', referred to as '++')
-    ```
-- Data directives: `.byte`, `.word`, `.dword`, `.fill` (allocate space)
-- Equates: `label = value` (not `label .equ value` or `#define`)
-- **Number Literals**: Hex `$1234`, Binary `%10101010`, Decimal `123`.
-- Zero-page variables defined with `=`
-- `.text` for inline string data
-- **Conditional assembly**: `.if`, `.elsif`, `.else`, `.endif`
-- **Memory Sections**: `.section <name>`, `.send <name>` (common sections: `CODE`, `DATA`, `BSS`, `BSS_NOCLEAR`)
+## Syntax
 
-## Instructions
-- **Instructions like `rol`, `ror`, `asl`, `lsr` (and `inc`, `dec` on 65C02) require an explicit operand for accumulator**: write `rol a`, `inc a`, etc., not just `rol` or `inc`.
-- Standard 6502 addressing modes: implied, immediate (`#`), zero-page (`zp`), zero-page,X (`zp,x`), absolute (`abs`), absolute,X (`abs,x`), absolute,Y (`abs,y`), indirect (`(abs)`), indirect,X (`(zp,x)`), indirect,Y (`(zp),y`), relative (branches), accumulator
-- Branches: `bne`, `beq`, `bmi`, `bpl`, `bcs`, `bcc`, `bvs`, `bvc` (relative, max +127/-128 bytes)
-- Jumps: `jmp` (absolute or indirect), `jsr`/`rts` (subroutine call/return)
-- No `push`/`pop` mnemonics — use `pha`/`pla` (byte) and `txa`/`phx`/`plx`/`tay`/`phy`/`ply` for registers
+- Labels begin in column 1. Instructions are indented at least four spaces.
+- Use lowercase opcodes and operands, two spaces between opcode and operand,
+  and two spaces before end-of-line comments.
+- Hex is `$1234`, binary is `%1010`, decimal is `123`.
+- Use `.byte`, `.word`, `.dword`, `.fill`, `.text`, and equates such as
+  `name = value`.
+- Sections use `.section name` and `.send name`.
+- Procedures use `.proc` and `.pend`; labels beginning with `_` are local to a
+  procedure.
+- Conditional assembly uses `.if`, `.elsif`, `.else`, and `.endif`.
 
-### Instruction Side Effects (Flags)
-- **`Z` (Zero)**: Set if the result of an operation is 0.
-- **`N` (Negative)**: Set if bit 7 of the result is 1.
-- **`C` (Carry)**: Used for unsigned overflow and shifts. `cmp` sets `C` if `Register >= Operand`.
-- **`V` (Overflow)**: Set if a signed arithmetic operation overflowed.
-- **Commonly affected by**: `lda`, `ldx`, `ldy`, `inx`, `dex`, `tax`, `tay`, `txa`, `tya`, `and`, `ora`, `eor`, `asl`, `lsr`, `rol`, `ror`, `adc`, `sbc`, `cmp`, `cpx`, `cpy`, `bit`.
-- **Note**: `lda`, `ldx`, `ldy` do NOT affect the Carry flag. Only `adc`, `sbc`, `cmp`, and shift/rotate instructions affect Carry.
+## Anonymous Labels
 
-## 6502 vs 65C02
-- **CX16 target only**: can use WDC 65C02 instructions — `stz`, `phx`, `plx`, `phy`, `ply`, `bra`, `trb`, `tsb`, `stp`, `wai`, `inc a`, `dec a`, `bit #imm`, `bit zp,x`, `bit abs,x`, `jmp (abs,x)`, and `(zp)` indirect addressing mode (e.g., `lda (zp)`).
-- **C64, C128, PET32 targets**: original 6502 only — no `stz`, no `phx`/`plx`/`phy`/`ply`, no `bra` etc.
-- **Note**: The Rockwell/bit-manipulation instructions (`rmb`, `smb`, `bbr`, `bbs`) are **NOT** available.
-- Check the target before using 65C02-specific instructions
-- **6502 / 65C02 instruction reference table**: https://www.pagetable.com/c64ref/6502/?cpu=65c02&tab=4 (provides exact instruction details for all opcodes: operation, addressing modes, byte length, and cycle count).
+Anonymous label definitions are always a single `+` or `-` in column 1.
+References use repeated signs:
 
-## Commander X16 Memory Map
+- `+` is the next upcoming `+` definition, `++` the second, `+++` the third.
+- `-` is the most recent `-` definition, `--` the preceding one, and so on.
+- Passing a `+` definition resets the forward-reference count.
+
+```asm
+-   dex
+    bne -       ; most recent '-' above
+    ldx #5
++   dex         ; first upcoming '+'
+    bne +       ; this '+'
+    sta $400
++   lda #0      ; second '+' definition
+    bne ++      ; next '+' after this one
+    rts
++   inc $d020   ; third '+' definition
 ```
-Bank    Offset      Content
-        $0000-$9EFF Fixed/System RAM
-        $9F00-$9FFF I/O Area (VIA, VERA, YM2151)
-$00-$FF:$A000-$BFFF Banked RAM (max 2 MiB, often 512 KiB) (256x8K banks)
-$00-$1F:$C000-$FFFF System ROM (512 KiB) (32x16K banks)
-```
-- Bank registers (in zeropage): `$00` selects the current RAM bank (0-255), `$01` selects the current ROM bank (0-31). For JSRFAR/banked calls from Prog8, use `extsub @bank ...` instead of manipulating these directly.
-- Full hardware specs at https://ayce.dev/emptyx16.html#emptyx16---x16-hardware-specs
 
-## Calling Convention / Register Conventions
-- **Accumulator (A)**: 8-bit, used for most arithmetic, data movement, return values
-- **X register**: 8-bit, often used for indexing, loop counters
-- **Y register**: 8-bit, often used for indirect addressing index
-- **Processor Status (P)**: flags — carry (C), zero (Z), negative/N (sign bit 7), overflow (V), decimal (D), interrupt (I), break (B)
-- **No caller-saved vs callee-saved convention** — list all modified registers in `clobbers (A, X, Y)` when writing `asmsub`
-- The CPU stack (SP, $0100-$01FF) is limited (usually ~128 bytes free). Do not overflow it
+## CPU Differences
 
-## Assembly within Prog8 Programs
+- `cx16` is the only Prog8 target that supports 65C02 instructions such as
+  `stz`, `phx`, `plx`, `phy`, `ply`, `bra`, `trb`, `tsb`, `wai`, `stp`, and
+  `bit #imm`. C64, C128, and PET32 use the original 6502.
+- Rockwell instructions (`rmb`, `smb`, `bbr`, `bbs`) are not available.
+- Use an explicit `a` for accumulator forms such as `rol a`, `lsr a`, `inc a`,
+  and `dec a`.
+- Branches are relative. Prog8's assembler invocation enables `--long-branch`
+  for out-of-range branches.
+- `lda`, `ldx`, and `ldy` do not affect carry. `cmp` sets carry for
+  `register >= operand`.
+- On NMOS 6502, `jmp ($xxff)` wraps the high-byte fetch within the page.
+- `BRK` skips the byte after its opcode. The byte is commonly a handler
+  signature.
+- `bit` memory forms set N and V from the operand and Z from `A AND operand`.
+  65C02 `bit #imm` sets only Z.
+- Decimal mode can persist into NMOS IRQ handlers. Use `cld` at handler entry
+  and explicitly control D before `adc` or `sbc`.
 
-### `%asm {{ }}` blocks
-- Embed arbitrary 64tass assembly directly in your Prog8 source
-- Access Prog8 symbols using their prefixed names (see below)
-- Can be placed inside subroutines or at block level
+## Prog8 Integration
 
-### `asmsub` (assembly subroutine)
-- For kernel (ROM) routines or low-level assembly
-- Parameters passed via registers: `@A`, `@X`, `@Y`, `@AX` (A low, X high), `@AY` (A low, Y high), `@R0`-`@R15`, `@FAC1`/`@FAC2` (float), `@Pc` (carry), `@Pz` (zero)
-- Return value: `-> type @register` — also via `@Pz`/`@Pc` for flags
-- Clobbers: `clobbers (A, X, Y)` — MUST list all modified registers
-- **CRITICAL**: Parameter names in `asmsub` are **documentation only**. You MUST use the actual registers in your assembly code, NOT the parameter names (unless you create aliases yourself).
-- Create symbolic aliases at assembly top for clarity: `x1 = cx16.r0`, `y1 = cx16.r0L`
-- Accessing Prog8 parameters if they were NOT mapped to registers: use `p8v_paramname`. (Mapping to registers is preferred for speed).
+`asmsub` parameter names are documentation only. Use the actual registers or
+define assembly aliases, and list every modified register in `clobbers`.
 
-Example:
 ```prog8
-asmsub line(uword x1 @R0, ubyte y1 @A, uword x2 @R1, ubyte y2 @Y) clobbers (A, X, Y) {
+asmsub increment(uword value @R0) clobbers (A, X) {
     %asm {{
-        x1 = cx16.r0
-        x2 = cx16.r1
-        lda  x1        ; use alias, not "_x1"
+        lda  cx16.r0L
+        clc
+        adc  #1
+        sta  cx16.r0L
     }}
 }
 ```
 
-### `asmsub` parameter annotation reference
-| Annotation | Register | Size |
-|------------|----------|------|
-| `@A` | Accumulator | 8-bit |
-| `@X` | X register | 8-bit |
-| `@Y` | Y register | 8-bit |
-| `@AX` | A (low) + X (high) | 16-bit |
-| `@AY` | A (low) + Y (high) | 16-bit |
-| `@R0`-`@R15` | cx16 virtual registers | 16-bit each |
-| `@FAC1`/`@FAC2` | Floating-point accumulators | 5-byte float |
-| `@Pc` | Carry flag | bool |
-| `@Pz` | Zero flag | bool |
+Annotations include `@A`, `@X`, `@Y`, `@AX`, `@AY`, `@R0`-`@R15`, `@FAC1`,
+`@FAC2`, `@Pc` (carry), and `@Pz` (zero). Unmapped parameters are accessed as
+`p8v_name`.
 
-## Accessing Prog8 Symbols from Assembly
-All Prog8 symbols are prefixed when accessed from assembly:
+Prog8 symbols use these prefixes: `p8v_` variables and parameters, `p8s_`
+subroutines, `p8b_` blocks, `p8c_` constants, `p8l_` labels, and `p8t_` struct
+types. Fully qualified names may look like
+`p8b_main.p8s_helper.p8v_local`. `%option no_symbol_prefixing` disables the
+prefixes. Split word arrays use `_lsb` and `_msb` symbols.
 
-| Prefix | Refers to | Example |
-|--------|-----------|---------|
-| `p8v_` | Variables, parameters | `p8v_myvar` |
-| `p8s_` | Subroutines | `p8s_mysub` |
-| `p8b_` | Blocks | `p8b_myblock` |
-| `p8c_` | Constants, enum members | `p8c_myconst`, `p8c_MyEnum_Member` |
-| `p8l_` | Labels | `p8l_mylabel` |
-| `p8t_` | Struct types | `p8t_MyStruct` |
-| `p8_` | Other symbols | |
+## Zeropage and Memory
 
-- **Fully qualified**: e.g., `p8b_myblock.p8v_myvar`, `p8b_myblock.p8s_mysub.p8v_localvar`
-- Within a `.proc` (subroutine) in your assembly, short names often work (assembler scoping)
-- **`%option no_symbol_prefixing`**: disables all prefixes. Stdlib modules (`cbm`, `cx16`, `txt`) use this — you can write `cbm.CHROUT` directly
-- **Split word arrays**: two separate byte arrays — append `_lsb` and `_msb` to the name: `p8v_myarray_lsb`, `p8v_myarray_msb`
+- Never hardcode zeropage addresses. Use `P8ZP_SCRATCH_B1`,
+  `P8ZP_SCRATCH_REG`, `P8ZP_SCRATCH_W1`, `P8ZP_SCRATCH_W2`, and
+  `P8ZP_SCRATCH_PTR`, or the appropriate `cx16.r0`-`cx16.r15` symbols.
+- Scratch variables are not necessarily consecutive. CX16 virtual registers
+  are consecutive and have `L` and `H` byte names.
+- Allocate additional temporary storage in BSS.
+- The hardware stack occupies `$0100-$01ff` and is limited. Do not overflow
+  it.
+- Prog8 words are little-endian: load the low byte first, then the high byte.
 
-## Zeropage Usage
-- Do NOT use arbitrary zeropage locations. Only use these predefined scratch variables:
-  - `P8ZP_SCRATCH_B1` (byte)
-  - `P8ZP_SCRATCH_REG` (byte)
-  - `P8ZP_SCRATCH_W1` (word)
-  - `P8ZP_SCRATCH_W2` (word)
-  - `P8ZP_SCRATCH_PTR` (word)
-- On CX16: virtual registers `cx16.r0`-`cx16.r15` are in zeropage (and their low/high bytes: `cx16.r0L`, `cx16.r0H`, etc.)
-- Virtual registers `cx16.r0`-`cx16.r15` are available on ALL targets, but only on CX16 in zeropage
-- Assume the scratch variables are not consecutive in zeropage. The CX16 virtual registers ARE consecutive in memory though.
-- For additional temporary storage, allocate regular variables in BSS
-- **CRITICAL: NEVER use hardcoded zeropage addresses** (like `$1e`, `$22`, `$7c`, etc.) to reference P8ZP_SCRATCH registers or cx16 virtual registers. Their addresses differ per compilation target. **ALWAYS** use the symbolic names (`P8ZP_SCRATCH_W1`, `cx16.r14`, `cx16.r14+1`, etc.) so the assembler resolves the correct address for the target.
-
-## Common 6502 Patterns
-
-### Common Branch Logic (Comparisons)
-| Logic | Unsigned | Signed |
-|-------|----------|--------|
-| `A == imm` | `cmp #imm`, `beq label` | (Same) |
-| `A != imm` | `cmp #imm`, `bne label` | (Same) |
-| `A < imm` | `cmp #imm`, `bcc label` | `sec`, `sbc #imm`, `bvc *+4`, `eor #$80`, `bmi label` |
-| `A >= imm` | `cmp #imm`, `bcs label` | `sec`, `sbc #imm`, `bvc *+4`, `eor #$80`, `bpl label` |
-| `A <= imm` | `beq label`, `bcc label` | (Use complex signed logic or reorder) |
-| `A > imm` | `beq +`, `bcs label`, `+` | (Use complex signed logic or reorder) |
-
-### Looping (downto with BNE)
-```asm
-        ldx #count
-loop    ; do work here
-        dex
-        bne loop        ; loop while X != 0 (runs 'count' times)
-```
-
-### Self-Modifying Code (SMC) — Avoid When Possible
-**Do NOT use self-modifying code unless absolutely necessary** — self-modifying code writes to the code segment at runtime, which means it **cannot run from ROM** and is generally fragile, hard to debug, and violates the expected behavior of modern tooling (emulators, debuggers, etc.). Prefer lookup tables, indirect jumps via a vector in RAM, or alternative algorithms that keep code in place.
-
-### Self-Modifying Code (SMC) Detection
-Look for `sta`, `stx`, or `sty` pointing into code labels:
-```asm
-        lda #$42
-        sta _target+1   ; Modifies the immediate operand of the LDA at _target
-        ...
-_target lda #$00        ; This #$00 will be replaced by #$42 at runtime
-```
-
-### Indirect indexed read (table of data)
-```asm
-        ldy #index
-        lda (ptr),y     ; read byte at address stored in zp ptr + Y
-```
-
-### 16-bit arithmetic (word add)
 ```asm
         clc
-        lda word1_lo
-        adc word2_lo
-        sta result_lo
-        lda word1_hi
-        adc word2_hi
-        sta result_hi
+        lda  word1
+        adc  word2
+        sta  result
+        lda  word1+1
+        adc  word2+1
+        sta  result+1
 ```
 
-### Calling a Prog8 subroutine from assembly
+## Common Patterns
+
 ```asm
-        jsr p8s_myblock.p8s_mysub
+        ldy  #index
+        lda  (ptr),y
+
+        jsr  p8s_main.p8s_helper
 ```
 
-### Reading a word variable
-```asm
-        lda p8v_myword          ; loads LSB
-        ldy p8v_myword+1        ; loads MSB (word variables are stored LSB-first)
-```
+Avoid self-modifying code unless it is explicitly required. It cannot run
+from ROM and complicates debugging and tooling. Prefer lookup tables, RAM
+vectors, or alternative algorithms.
 
-## CPU Quirks and Pitfalls
+## Reference
 
-### JMP ($xxFF) Page Wrap Bug (NMOS 6502)
-- **Problem**: On original 6502 CPUs, `jmp ($caff)` will fetch the LSB from `$caff` but the MSB from `$ca00` (instead of `$cb00`).
-- **Target**: Affects C64, C128, PET32.
-- **Solution**: Avoid placing indirect jump vectors on a page boundary, or use the CX16 (65C02) which fixed this bug.
+### Primary: 6502 Family CPU Reference
 
-### BRK Instruction and the "Signature Byte"
-- **Behavior**: After a `BRK` instruction, the return address on the stack is incremented by **2**. This means the CPU skips the byte immediately following the `BRK` opcode.
-- **Usage**: This skipped byte is often used as a "signature" or parameter byte for the BRK handler.
+<https://www.pagetable.com/c64ref/6502/?cpu=65c02&tab=2> is the reference to reach for.
+It covers the whole 6502 family and lets you **select the CPU variant**, which matters
+because the parts genuinely differ. The variants are `6502`, `6502rorbug`, `65dtv02`,
+`65c02`, `r65c02` (Rockwell), `65c02s`, `65ce02` and `65c816`. Change the `cpu=` query
+parameter to switch; for Prog8, use `cpu=65c02` when working on cx16 and `cpu=6502` for
+c64, c128 and pet32.
 
-### BIT Instruction Flags
-- **Absolute/Zero-page**: `bit $1234` copies bit 7 of the memory value to the **N** flag and bit 6 to the **V** flag. The **Z** flag is set based on `A AND memory`.
-- **Immediate (65C02 only)**: `bit #$01` only affects the **Z** flag; it does **NOT** modify N or V.
-- **NMOS 6502**: Does **not** support `bit #imm`.
+Per instruction it gives the operation, the opcodes, the byte length, the cycle count
+and which status flags are affected, with tabs for opcodes, instructions, addressing
+modes and a full opcode table (switchable between 4-4 and 3-3-2 layout). It can also
+show the NMOS undocumented opcodes, and each part lists what it is based on, so a
+variant's inherited instructions are visible.
 
-### The "B" (Break) Flag
-- **Quirk**: The B flag (bit 4 of the status register) doesn't actually exist in the hardware status register. It only exists on the stack after a `PHP` or `BRK` instruction (set to 1) or a hardware IRQ/NMI (set to 0).
-- **Detection**: To tell if an interrupt was caused by `BRK` or a hardware IRQ, your handler must `pla`, `and #$10`, and check the result.
+Cycle counts use a compact notation worth knowing: a trailing `t` means +1 if a branch
+is taken, `p` means +1 if a page is crossed, and `d` means +1 when the D (decimal) flag
+is set. So `d0  2+t+p` for `bne` is 2 cycles, 3 if taken, 4 if taken across a page.
 
-### Decimal Mode Flag (D) Persistence
-- **Pitfall**: On NMOS 6502, the `D` flag is **not** cleared on interrupt. Always use `cld` in IRQ handlers. On 65C02, it is cleared automatically, but `cld` is still good practice.
-- **ADC/SBC**: Be extremely careful with arithmetic if you haven't explicitly set or cleared the `D` flag, as its state might be unknown.
+Consult it whenever an exact figure is needed rather than recalled: the byte length or
+cycle cost of an instruction, which flags it clobbers, whether an addressing mode exists
+on the target CPU, or whether an instruction belongs to a different part at all. Typical
+reasons are costing a peephole or code-size tradeoff, checking a hand-written routine,
+and confirming a mnemonic is valid before using it.
 
-## Optimization Tips
-- **`stz` (65C02 only)**: Saves cycles and bytes compared to `lda #0`, `sta ...`.
-- **`bra` (65C02 only)**: Shorter and usually faster than `jmp`.
-- **Avoid `clc` before `bcc`**: `cmp` already sets the carry flag correctly for `bcc`/`bcs`.
-- **`inx` / `dex` vs `clc`+`adc #1`**: Incrementing/decrementing is faster and doesn't affect the carry flag.
-- **Zero-page usage**: Accessing variables in zeropage is 1 cycle faster and 1 byte shorter than absolute addressing.
-- **Cycle Counting**: Most instructions take 2-4 cycles. `jsr` takes 6, `rts` takes 6. Branches take 2 (no branch), 3 (branch taken), or 4 (branch taken across page boundary).
+Do not assume a timing from memory or from another part. Cycle counts are exactly the
+kind of detail that is misremembered, and the difference matters when it is the basis of
+an optimization claim. Read the number off the reference for the specific CPU being
+targeted.
 
-## 64tass Macros
-Define macros for common tasks:
-```asm
-pushax  .macro
-        pha
-        txa
-        pha
-        .endm
+### Secondary: WDC 65C02 reference
 
-popax   .macro
-        pla
-        tax
-        pla
-        .endm
-```
-Call them with `#pushax` and `#popax`.
+<https://cx16.dk/65c02/reference.html> is a 65C02-only reference. It is worth having
+because it states each instruction's effect in a single formula (for example
+`ADC:  A,Z,C,N = A+M+C`), which makes the flag behaviour quick to check, and it lists
+every status bit per instruction including the ones not affected. Use it for flag
+semantics; use the primary reference for timings and for anything CPU-specific, since
+it covers the 65C02 only and will not answer a question about the NMOS 6502.
 
-## IRQ Handler Best Practices
-- **Clear the Decimal Flag (`cld`)**: On the original 6502, the decimal flag (`D`) is **not** automatically cleared when an interrupt occurs. If the interrupted code was in decimal mode, your handler will also run in decimal mode, causing arithmetic errors. Always call `cld` at the beginning of your handler. The 65C02 clears it automatically, but `cld` is still recommended for portability.
+### Neither reference covers the assembler
 
-## Invoking the Assembler (64tass)
-If you need to manually invoke `64tass` to assemble a generated `.asm` file, you should be aware of the default arguments that `prog8c` supplies to ensure compatibility with the generated code:
+Both document the CPU, not 64tass. For whether 64tass accepts a mnemonic or a given
+syntax, assemble it: an instruction outside the selected CPU set fails with
+`error: general syntax` rather than being silently ignored. Note also that 64tass has
+no `w65c02s`; its CPU selection is `--m65xx` (default), `--m65c02` and `--m65ce02`, and
+Prog8 emits `.cpu 'w65c02'`.
 
-- `--ascii`: **CRITICAL.** Prog8 generates character and string data in ASCII. Without this flag, `64tass` defaults to PETSCII, which will garble your strings.
-- `--case-sensitive`: Prog8 is case-sensitive and expects the assembler to be as well.
-- `--long-branch`: Enables automatic conversion of relative branches (`beq`, `bne`, etc.) to absolute jumps if the target is out of range. Prog8 relies on this.
-- `-Wno-implied-reg`: Suppresses warnings when the accumulator `a` is omitted from instructions like `rol`, `lsr`, etc. (though the skill recommends always using `rol a` for clarity).
-- `-Wall`: Enables all warnings.
-- `--cbm-prg` (or `--atari-xex` / `--nostart`): Sets the output format and adds the appropriate load address header.
+## Tools
 
-### Optional but Recommended for Debugging:
-- `--vice-labels --labels=labels.txt`: Generates a label file that can be loaded into the VICE monitor (`load_labels "labels.txt"`) to see your Prog8 symbol names while debugging.
-- `--list=listing.txt`: Generates a full assembly listing file with addresses and opcodes.
+Prog8 invokes 64tass with `--ascii --case-sensitive --long-branch -Wall` plus
+the target output option such as `--cbm-prg`. For manual assembly:
 
-### Example manual invocation:
 ```bash
-64tass --ascii --case-sensitive --long-branch -Wall --cbm-prg -o myprogram.prg myprogram.asm
+64tass --ascii --case-sensitive --long-branch -Wall --cbm-prg \
+    -o myprogram.prg myprogram.asm
 ```
+
+For debugging, add `--vice-labels --labels=labels.txt` and/or
+`--list=listing.txt`.
+
+To check an instruction's encoding or size, assemble it in isolation and read the
+output bytes. The 2-byte load address comes first, so `stx $7b` / `ldy $7b` below
+shows as `86 7b a4 7b` - two instructions, four bytes:
+
+```bash
+printf '.cpu  %s\n* = $0801\nlbl\tstx\t$7b\n\tldy\t$7b\n' "'w65c02'" > sz.asm
+64tass --ascii --case-sensitive --cbm-prg -o sz.prg sz.asm && xxd sz.prg
+```
+
+Note that a 64tass label takes no trailing colon; `lbl:` is a syntax error reported as
+`error: label required`.

.agents/skills/m68k-coder/SKILL.md

@@ -0,0 +1,197 @@
+---
+name: m68k-coder
+description: Writing or understanding Motorola 68000/68020 assembly in vasm mot syntax, standalone or embedded in Prog8
+---
+
+# M68K Assembly Coder
+
+Use Motorola syntax for `vasm`, not Devpac or other assemblers. Keep answers
+concise and practical.
+
+Reference documentation for the assembler and linker is available locally in
+`~/Documents/prog8-dev-resources` (the vasm and vlink manuals). Search this
+folder first before looking online.
+
+## Syntax
+
+- Labels begin in column 1. Instructions are indented at least four spaces.
+- Use lowercase opcodes and operands, two spaces between opcode and operand,
+  and two spaces before end-of-line comments.
+- Use `.b`, `.w`, and `.l` for byte, word, and long operations. Do not rely on
+  default sizes.
+- Hex is `$1234`, binary is `%1010`, decimal is `123`, and characters use
+  `#'A'`.
+- Local labels start with `.`. Equates use `name = value`.
+- Data directives are `dc.b`, `dc.w`, `dc.l`; allocation directives are
+  `ds.b`, `ds.w`, `ds.l`.
+- Sections use forms such as `SECTION .text,code` and `SECTION .bss,bss`.
+
+```asm
+        moveq   #0,d0
+        move.w  value,d0
+        add.l   d1,d0
+.loop
+        dbra    d0,.loop
+```
+
+## FPU Size Suffixes
+
+FPU instructions that access memory **must** include a size suffix: `.s`
+(single, 4 bytes), `.d` (double, 8 bytes), or `.x` (extended, 12 bytes).
+Without a suffix, vasm defaults to `.x`, reading 12 bytes and corrupting
+results when the operand is smaller.
+
+**Use `.s` with memory operands:**
+```asm
+        fmove.s  value(pc),fp0      ; load float from memory
+        fcmp.s   other(pc),fp0      ; compare float in memory
+        ftst.s   value(pc)          ; test float in memory
+```
+
+**Omit the suffix for register-to-register operations:**
+```asm
+        fadd     fp1,fp0            ; fp0 = fp0 + fp1
+        fcmp     fp1,fp0            ; compare fp0 vs fp1
+        ftst     fp0                ; test fp0
+        fintrz   fp0,fp0            ; truncate fp0
+```
+
+The FPU registers are always 80-bit internally, so no size suffix is needed
+when both operands are registers.
+
+## Architecture
+
+- Registers `D0-D7` and `A0-A7` are 32-bit; `A7` is the stack pointer.
+- Pointers are 32-bit and must normally be accessed with `.l`.
+- Memory is big-endian: `$1234` is stored as `$12,$34` and a long's low byte
+  is at offset `+3`.
+- Word and long accesses must be even-aligned on 68000. 68020 permits
+  misalignment but it is slower.
+- **Byte `A7` auto-update steps by 2, not 1.** If A7 (== SP, the stack pointer) is the address
+  register and the operand length for post-increment or post-decrement is one byte, A7 is
+  decremented/incremented by 2 instead of 1, to maintain a word-aligned stack. This is the
+  specified behaviour of all the 680x0, not a 68000 quirk, so it is identical everywhere and
+  safe to rely on. Word and long operands step by 2 and 4.
+  ```asm
+        move.b  d0,-(sp)          ; A7 steps by 2, so SP stays word-aligned
+        ...
+        addq.l  #2, sp            ; and the matching pop is #2, not #1
+  ```
+  Do not "fix" this to an explicit `subq.l #1,sp`: that leaves SP odd and cause crash.
+- The 68000 (`amiga500`) has only word multiply/divide and no scaled indexing,
+  bit-field, or 32-bit multiply/divide instructions. The 68020 (`qemu68k`)
+  adds those features and `extb.l`.
+- On 68000, sign-extend a byte with `ext.w` followed by `ext.l`.
+- `move.b` and `move.w` to a data register do not clear its upper bits. Like
+  `move.w` only writes the lower 16 bits, `move.b` only writes the lower 8 bits;
+  both leave all other bits of the register untouched. So zero-extending a byte
+  to long requires `moveq #0,d0` first, then `move.b src,d0`, then `move.l d0,dst`
+  (or mask with `and.l #$ff` / `and.l #$ffff`). Relying on `move.b` to clear the
+  upper bits is a common bug that corrupts values.
+- `movea` does not set condition codes; `move` does. Keep the stack word
+  aligned and push words or longs, never bytes.
+
+## Registers and Calls
+
+- Return values use `D0` (primary), `D1` and `D2` (additional), or `A0` for a
+  pointer.
+- `D0-D1` and `A0-A1` are scratch registers. Preserve registers required by
+  the surrounding calling convention, especially `D2-D7` and `A2-A6`.
+- Never clobber `A5`: on every M68K target it is the Prog8 stack frame pointer
+  (locals at negative offsets, parameters at positive offsets). Overwriting it
+  corrupts the frame of the running subroutine and breaks unwinding.
+- Never clobber `A6` on the Amiga targets (`amiga500`, `amiga1200`): it holds
+  the AmigaOS library base pointer that library calls depend on. Prog8 itself
+  leaves `A6` free for that reason.
+- `move.l d0,-(sp)` pushes and `move.l (sp)+,d0` pops.
+
+### Prog8 M68K stack ABI
+
+- Normal framed subroutines use `link a5,#-N` and `unlk a5`; locals are at
+  negative `A5` offsets and incoming parameters are at positive offsets.
+- Arguments use word-packed slots of natural width: byte, bool, word and uword
+  arguments take a 2-byte slot; long, pointer and float arguments take a 4-byte
+  slot. The caller pushes arguments left-to-right (a byte sits at the base byte
+  of its 2-byte slot, the second byte is garbage; word values fill their slot)
+  and removes the complete argument area after the call. The area
+  size is always even, so the stack stays word-aligned; slot bases accumulate from
+  the rightmost argument at `8(a5)` according to each argument's width.
+- A normal call preserves the caller's frame. A `goto` or conditional branch
+  leaving the current subroutine must execute `unlk a5` on the taken path before
+  transferring control.
+- The built-in Prog8 `call(address)` is a zero-argument indirect call. It does
+  not marshal stack arguments. `asmsub` and `extsub` keep their explicit
+  register-annotated ABI, separate from normal framed calls.
+- Push narrow arguments into their own 2-byte slots, each with one instruction: a
+  byte uses `move.b ..., -(sp)` (A7 byte predecrement steps by 2, reserving the
+  slot; the byte lands at the slot's base byte), a word uses a plain
+  `move.w ..., -(sp)`. Never push a narrow value where the callee expects a
+  wide (4-byte) argument: the slot sizes must line up exactly.
+  Wide arguments (`move.l ..., -(sp)` / `fmove.s fp0,-(sp)`) push in one step.
+  Use `lea N(sp),sp` for the cleanup when the call returns a status flag
+  (`addq` clobbers CCR).
+
+## Prog8 Integration
+
+Assembly can appear in `%asm {{ }}` blocks or `asmsub` routines. `asmsub`
+parameter names are documentation only: use the actual annotated registers.
+Declare every modified register in `clobbers`.
+
+```prog8
+asmsub add100(uword value @D0) -> uword @D0 clobbers (D0) {
+    %asm {{
+        add.w  #100,d0
+    }}
+}
+```
+
+Do not use `%option no_sysinit` on `amiga500` or `qemu68k` targets when the goal
+is to create normal runnable programs. It skips critical startup logic including
+library initialization and CLI argument handling, causing programs to crash on
+start.
+
+Supported annotations include `@D0`-`@D7`, `@A0`-`@A6`, virtual `@R0`-`@R15`,
+floating-point `@FP0`-`@FP7`, and flags `@Pc` and `@Pz`.
+
+Prog8 assembly symbols are prefixed:
+
+| Prefix | Symbol |
+| --- | --- |
+| `p8v_` | variable or parameter |
+| `p8s_` | subroutine |
+| `p8b_` | block |
+| `p8c_` | constant or enum member |
+| `p8l_` | label |
+| `p8t_` | struct type |
+
+Use fully qualified names when needed, such as
+`p8b_main.p8v_value`. `%option no_symbol_prefixing` disables prefixes.
+
+```asm
+        move.l  p8v_pointer,a0
+        move.w  p8v_word,d0
+        jsr     p8s_main.p8s_helper
+```
+
+## Useful Patterns
+
+```asm
+        lea     source,a0
+        lea     dest,a1
+.copy
+        move.b  (a0)+,(a1)+
+        bne.s   .copy
+```
+
+Use `moveq` for small constants, `addq`/`subq` for constants 1 through 8, and
+`lea` for address arithmetic. Check divisors for zero and division overflow.
+
+## Tools
+
+For QEMU 68k, Prog8 invokes `vasmm68k_mot` with `-m68020 -m68881 -Felf
+-opt-speed -warnunaligned -ldots -spaces`, then links with `vlink`. For raw
+binary output use `-Fbin` and `-join=0x<addr>`. Add `-L listing.txt` when a
+listing is needed.
+
+Prog8 runs `amiga500` programs with `vamos --cpu 68000` and `qemu68k` programs
+with `qemu-system-m68k -M virt -cpu m68020 -m 1M -kernel program.elf -nographic`.

.agents/skills/prog8-coder/SKILL.md

@@ -1,248 +1,168 @@
 ---
 name: prog8-coder
-description: Writing or understanding Prog8 programs
+description: Writing or understanding Prog8 programs and Prog8 IR
 ---
 
-# Prog8 Coder Skill
+# Prog8 Coder
 
-You are an expert retro system development assistant, versed in the Prog8 programming language. Keep responses concise and practical — prefer short, correct code examples over lengthy prose.
-You are working with **Prog8** source code (`.p8` files) or its Intermediate Representation (`.p8ir` files). Prog8 targets 8-bit systems (C64, CX16, C128, PET32) with the 6502 CPU, plus a `virtual` target for testing.
-Follow ALL the rules below carefully.
+You are an expert Prog8 assistant. Keep answers concise and practical. This
+skill covers `.p8` source and `.p8ir` intermediate representation for the
+6502, M68000, and virtual targets.
 
-## General & Setup
-- A program = a `main` block containing a `start` subroutine entry point, plus optional other subroutines/blocks
-- Add `%zeropage basicsafe` at the top of your program to allow clean return on exit (instead of resetting the machine/emulator)
-- Module imports: `%import modulename` — no `as` aliasing. Use the module's defined prefix (e.g., `%import textio` → `txt.xxx`)
-- `.p8ir` files are the **Intermediate Representation** — target-independent, executable via `prog8c -vm file.p8ir`
-- **Library search**: `prog8c -libsearch <regex>` to find routines in the embedded stdlib (e.g., `prog8c -libsearch "txt\."`)
-- **Library dump**: `prog8c -libdump <dir>` to extract all embedded library source files
-- **`sys` module**: always available, no import needed
-- **CX16 programs**: add `%encoding iso`, call `txt.iso()` in `start()`, end with `sys.poweroff_system()`
+Various reference documentation is available locally in
+`~/Documents/prog8-dev-resources`: 
 
-## Datatypes & Variables
-- Primitives: `bool`, `byte`, `ubyte`, `word`, `uword`, `long`, `float`, `str`
-- `ubyte`/`uword` = unsigned; `long` = signed 4-byte; `float` = 5-byte MS format; `str` = 0-terminated ubytes (max 255 chars)
-- **float requires `%import floats`** at top of file, else compiler errors
-- Arrays: max 256 bytes (512 for split word arrays). For larger data, use `memory()` + pointers
-- `memory(name, size)` returns a `uword` address to a statically reserved block of memory
-- To point a typed pointer at a `memory()` block, assign the uword directly: `^^MyStruct ptr = memory("name", size)`. The `^^Type:expression` syntax (below) only works with array literals, not general uword expressions.
-- Struct initialization: `^^StructType ptr = ^^StructType:[val1,val2,...]` (the `^^StructType:` can be omitted if inferable). This `^^Type:[...]` syntax does NOT work with variables or `memory()` — only literal arrays.
-- Struct definitions must be inside a block, not at file level.
-- Struct fields: simple types, `str`, and fixed-size arrays. Array fields: `ubyte[108] name`. `str` in struct = `^^ubyte`
-- Word/pointer arrays split into LSB/MSB by default. Override with `@nosplit`
-- **No call stack**: all variables statically allocated. No recursion without manual stack management
-- Variables zero-initialized (globals at start, locals on subroutine entry)
-- `@shared`, `@zp`, `@requirezp`, `@dirty`, `@nosplit` are **tags** that go on variable declarations. **Place them after the datatype (and array specifiers), before the variable name(s):**
-  ```
-  private ubyte[8] @shared vera_storage     ; single var with tag
-  ubyte @zp @shared varname                 ; multiple tags allowed
-  ubyte @requirezp var1                     ; require zeropage
-  ubyte[] @shared names                     ; array with tag
-  ```
-  The grammar is: `[private] datatype [arraydims] [tags...] identifierlist`
-- `@shared` marks variables as "used by external code" (assembly), prevents the optimizer from removing them
-- `@zp`/`@requirezp`: use sparingly — only for pointers (limited zeropage space)
-- Pointer-like typed pointers (`^^type`) support C-style scaled arithmetic; `uword` pointers always treat element as 1 byte
-- `&` = untyped address (uword); `&&` = typed pointer
-- Available zeropage scratch: `P8ZP_SCRATCH_B1`, `P8ZP_SCRATCH_REG`, `P8ZP_SCRATCH_W1`, `P8ZP_SCRATCH_W2`, `P8ZP_SCRATCH_PTR` — and cx16 virtual registers R0-R15 on all targets
-- Virtual registers (`cx16.r0`–`cx16.r15`): global 16-bit, NOT preserved across calls.
-- **WARNING: Virtual registers in ISR/IRQ handlers**: The virtual registers R0-R15 are *not preserved* across the IRQ handler call. If your handler uses them, it will corrupt the interrupted program's state. Either avoid using them in the handler, or save/restore with `cx16.save_virtual_registers()` / `cx16.restore_virtual_registers()`. This applies to all targets, not just CX16.
-- **WARNING: Long operations clobber R12-R15**: Some operations on `long` values use R12-R15 as temporary storage and will silently overwrite them. Do not rely on R12-R15 values when working with longs, and avoid using R12-R15 explicitly if your code uses long arithmetic.
-- **WARNING: VERA registers in ISR handlers (CX16)**: If your IRQ handler reads or writes VERA control registers (e.g., `cx16.VERA_DATA0`, `cx16.VERA_ADDR_L`, etc.), you must save and restore the VERA context around the handler's work using `cx16.save_vera_context()` / `cx16.restore_vera_context()`. Without this, the handler will corrupt any VERA operations (tilemap updates, sprite positioning, etc.) happening in the interrupted main program.
-- **IRQ handler best practices**: Keep handlers extremely short and fast — they run with interrupts disabled and steal cycles from the main program. Do NOT do lengthy processing, I/O, or complex subroutine calls inside the handler. Instead, set a boolean flag or semaphore that the main loop checks periodically, and do the actual work there.
-- Math performance: integer trig (`math.sin8`, `math.cos8`) uses fast LUTs; float trig (`floats.sin`/`cos`) is much slower
+Prog8 reference documentation, emulator manuals such as VICE and the X16
+emulator, the 64tass/vasm/vlink assembler and linker docs, X16 hardware and register specs,
+the Amiga NDK (docs and includes), and other resources. 
+Search in this folder first before looking online!
 
-## Recursion & Stack Management
-No call stack for variable storage — recursion overwrites locals. To handle it:
-1. **CPU hardware stack**: `push()`/`pushw()`/`pushl()`/`pushf()` and `pop()`/`popw()`/`popl()`/`popf()`. Save/restore locals around recursive calls
-2. **Software stacks**: `buffers.stack` (uword) / `buffers.smallstack` (ubyte). Both provide `push_b()`/`push_w()` and `pop_b()`/`pop_w()`
-3. **Iterative rewrite (preferred)**: Many recursive algorithms work as `repeat` loops with explicit bounds — avoids all stack overhead
 
-## Strings, Arrays & Pointers
-- `str` / array: max 256 bytes. `long[]` limited to 64 entries (64x4=256). `str[]` for string arrays: `str[5] names = ["a","b","c","d","e"]`
-- 2D arrays: `type[rows][cols] name`, access `name[r][c]`. Flat init list only (no nested `[[...]]`). Total size still ≤ 256 bytes
-- str/array passed as pointer to subroutine (receiving subroutine gets `^^ubyte` or `^^element`)
-- **No const pointers** or pointer-to-pointer currently supported
-- **Parsing limitation**: `pointer[index].field` as assignment target needs `^^`: `pointer[index]^^.field = value`
-- **Pre-allocate buffers**: `str buffer = "." * 50` (empty `""` allocates nothing — `strings.append()` will fail)
-- **No reassignment**: can't `buffer = "new text"` after declaration. Use `strings.copy()`/`strings.ncopy()`
-- String concat is expensive on 6502. Prefer separate prints over concatenation
-- **Efficient buffer iteration**: prefer `ptr++` + `@(ptr)` over `@(buffer + offset)`. Exception: if offset is a `ubyte` (≤ 255), `buffer[offset]` works fine
-- `@(ptr)` = peek/poke byte. For words: `peekw`/`pokew`, longs: `peekl`/`pokel`, floats: `peekf`/`pokef`, bools: `peekbool`/`pokebool`
-- Assigning a uword address to a `str` field of a struct is done by direct assignment. Compute the address in a uword variable first, then assign: `cx16.r0 = &namebufs + offset; entry.name = cx16.r0`. The `^^ubyte:(expr)` cast syntax does not parse — use a temp uword instead.
-- Use `len(array)` instead of hardcoded sizes
-- Array indexing is 0-based: `arr[0]` is first element
-- Static memory only — real dynamic allocation impossible, but can emulate with a simple arena allocator over a `memory()` slab
+## Essentials
 
-## Logic & Control Flow
-- Logical operators (short-circuit, bool only): `and`, `or`, `xor`, `not`
-  - In `a and b`: if `a` is false, `b` is NOT evaluated
-  - In `a or b`: if `a` is true, `b` is NOT evaluated
-  - Important when `b` has side effects
-- Bitwise operators: `&`, `|`, `^`, `~`, `<<`, `>>`
-- Bit rotation: `rol()`/`ror()` (through carry), `rol2()`/`ror2()` (no carry)
-- CPU status flag branches: `if_cs`, `if_cc`, `if_z`, `if_nz` (compile to single 6502 branch instructions)
-- If-expressions for simple value assignments based on a choice
-- **Optional braces in if/else**: when the `if` or `else` body is a single statement, the `{ }` can be omitted. Place the statement on the next line, indented. Example:
-  ```
-  if x < 5
-      txt.print("small")
-  ```
-- `defer` defers statement execution until scope exit. Multiple defers fire in **reverse registration order** (LIFO / stack order — last deferred runs first). A defer is only registered if execution reaches that statement — conditional paths that skip the `defer` line will not register it.
-- `goto`, labels, jump lists allowed
-- **Common mistake**: `and`/`or` for bitmasking — use `&`/`|` instead!
+- A program normally has a `main` block with a parameterless `start` entry
+  subroutine. Other blocks and subroutines may be added.
+- `%import modulename` imports a module without aliases. Use its defined prefix,
+  such as `%import textio` followed by `txt.print(...)`.
+- `sys` is always available and needs no import.
+- `.p8ir` is target-independent IR and can be run with
+  `prog8c -vm program.p8ir`.
+- Use `%zeropage basicsafe` in test programs when a clean return is needed.
+- Find library routines with `prog8c -libsearch <regex>` or extract them with
+  `prog8c -libdump <dir>`.
+- Options accept single-dash or double-dash (`-target` == `--target`). Use
+  `-out <dir>` to place output in a directory, or `-o <file>` to name the exact
+  final program artifact (all other output files then land beside it).
 
-### The when Statement
-The `when` statement is a control flow construct that enables you to execute a specific action based on the value of an expression. It is generally more readable and often more efficient than a sequence of `if-else if` statements, as the compiler can optimize it into more efficient branching structures, such as a jump table.
-- **Expression**: Evaluates an expression and compares it against case values.
-- **Cases**: Defined by a value followed by the `->` operator.
-- **Blocks**: Use `{ }` to enclose multiple statements for a case.
-- **Else Clause**: Serves as a default handler; mandatory unless the expression type is fully covered.
-- **Efficiency**: Recommended for handling sets of fixed choices as it typically results in better assembly code.
+For CX16 emulator programs, use `%encoding iso`, call `txt.iso()` in `start`,
+and finish with `sys.poweroff_system()`. CBM targets use PETSCII by default;
+avoid uppercase in test output unless intentional graphics. The virtual target
+uses ISO when configured with `%encoding iso` and `txt.iso()`.
 
-## Loop Constructs
-Prog8 supports these loop types. All support `break` and `continue` (except `unroll`).
+On any CBM machine with PETSCII (this includes the Commander X16), files
+on disk must only consist of 'safe' characters (`a-z`, `0-9`, `.-+!%`).
+Do NOT mix casing in filenames; in Prog8 source code, filenames should
+usually be written in *lowercase* letters. Other characters (such as
+`_` or uppercase mixed with lowercase) can make kernal file operations
+fail, depending on the drive/host layer.
 
-### `for` loop — iterate over a range or array
-- Loop variable **must be declared separately** before the `for` statement
-- Works with `ubyte`, `byte`, `uword`, `word`, `long`, pointer types (NOT `float`)
-- Iterates over ranges (`start to end`), descending ranges (`start downto end`), or arrays/strings
-- Optional `step <constant>` for non-unit step sizes
-- Loop variable value **after the loop is undefined** — don't rely on it
-- Descending loops with `downto` usually produce more efficient 6502 code
-```
-ubyte i
-for i in 20 to 155 {
-    ; body
-    break       ; exit loop
-    continue    ; next iteration
-}
+Do not use `%option no_sysinit` on `amiga500`, `amiga1200`, or `qemu68k` targets when the goal
+is to create normal runnable programs. It skips critical startup logic including
+library initialization (DOS, graphics, intuition, timer) and CLI argument
+handling, causing programs to crash on start. It is appropriate for library
+modules, IRQ handlers, boot stubs, or code that deliberately avoids those
+libraries.
 
-; descending
-for i in 155 downto 20 {}
-for i in 155 to 20 step -1 {}
+## Types and Memory
 
-; iterate over array elements
-uword[] fib = [0, 1, 1, 2, 3, 5, 8, 13]
-uword num
-for num in fib {
-    ...
-}
+- Primitive types include `bool`, `byte`, `ubyte`, `word`, `uword`, `long`,
+  `float`, `str`, and `pointer`.
+- `pointer` is target-sized: 16-bit on 6502 targets, 32-bit on `amiga500`,
+  `qemu68k`, and `virtual`. Use it for portable addresses, not `uword` or
+  `long` chosen by assumption.
+- M68000 targets are big-endian and use 32-bit pointers. Load the `m68k-coder`
+  skill for assembly details.
+- `float` requires `%import floats`.
+- On 6502 and virtual targets, variables are statically allocated and
+  zero-initialized. There is no normal call stack for locals, so recursion
+  overwrites locals unless an explicit hardware or software stack is used.
+  Iterative rewrites are preferred. Arrays are cleared once at startup here,
+  not on every subroutine entry.
+- On the M68000 targets (`amiga500`, `amiga1200`, `qemu68k`) ordinary subroutine
+  locals and parameters live in an activation record on the CPU stack, so true
+  recursion and re-entrancy work. `link a5,#-N` / `unlk a5`, frame pointer `A5`,
+  locals at negative offsets, parameters at positive offsets. Block-level
+  variables, globals, `@shared`, memory slabs, and objects needing a permanent
+  external address stay static. Clean locals are zeroed on every call
+  including arrays, so the compiler warns for framed arrays over 8 bytes; use
+  `@dirty` when you assign first, otherwise you read stack garbage.
+- Normal M68K calls use word-packed argument slots: byte/bool/word/uword arguments
+  take a 2-byte slot, long/pointer/float take 4 bytes. The caller pushes arguments
+  left-to-right and removes the complete argument area after the call. Slot bases
+  accumulate right-to-left starting at positive `A5` offset 8, so a parameter's
+  offset depends on the widths of the parameters to its right; locals use negative
+  offsets. See `docs/source/technical.rst` for the full convention and static
+  fallback rules.
+- `memory(name, size)` reserves static memory and returns a `pointer`.
+- `str` and arrays have compile-time byte budgets. On 6502 targets the usual
+  limit is 256 bytes; M68000 targets allow up to 32768 bytes. Use `memory()`
+  for larger data.
+- Word arrays are split into LSB/MSB arrays by default. Use `@nosplit` when
+  contiguous storage is required.
+- Variable tags follow the type and array dimensions, before the name:
+
+```prog8
+ubyte[8] @shared vera_storage
+uword @requirezp address
 ```
 
-### `while` loop — repeat while condition is true
-```
-while condition {
-    ; body
-    break
-    continue
-}
-```
+Use `@shared` for values accessed externally by assembly. Use `@zp` and
+`@requirezp` sparingly because zeropage is limited. Never assume virtual
+registers survive calls. Long operations may clobber `R12-R15`.
 
-### `do`-`until` loop — always executes body at least once
-```
-do {
-    ; body
-    break
-    continue
-} until condition
-```
+## Syntax Pitfalls
 
-### `repeat` loop — repeat a fixed number of times (most efficient)
-- Most efficient code generation — prefer over `for` when loop variable not needed
-- Omit count for infinite loop (still supports `break`)
-```
-repeat 15 {
-    ; body
-    break
-    continue
-}
+- Hex uses `$FF`, binary uses `%1010`, and casts use `expression as type`.
+- There is no automatic type widening in expression results: `byte * byte`
+  remains a byte. Cast explicitly when a wider result is needed. Assignments,
+  call arguments, and mixed-type expressions do widen implicitly
+  (`ubyte`/`byte` to `word`/`uword` to `long`); only same-width signedness
+  reinterpretation (`uword` to `word` and back) needs an explicit `as` cast.
+- `&` is an untyped address and `&&` is a typed pointer. Complex pointer field
+  assignments may require `ptr^^.field`.
+- `and`, `or`, `xor`, and `not` are logical operators. Use `&`, `|`, `^`, and
+  `~` for bitwise operations.
+- `and` and `or` short-circuit, so the right operand may not be evaluated.
+- There is no `elif`, bare block, or semicolon statement separator. Semicolon
+  starts a comment. Prefer one statement per line and four-space indentation.
+- `defer` executes registered statements in reverse registration order.
+- Array indexing starts at zero. Use `len(array)` rather than hardcoded sizes.
+- There is no function overloading. Call type-specific library routines such as
+  `txt.print_ub` or `txt.print_w`.
 
-; infinite:
-repeat {
-    ; body
-    break if x==5
-}
-```
+## Control Flow
 
-### `unroll` loop — compile-time code duplication
-- Not a real loop — duplicates body N times at compile time
-- No `break`/`continue` allowed
-- Only simple statements (assignments, calls) in body
-- Constant iteration count required
-```
-unroll 80 {
-    cx16.VERA_DATA0 = 255
-}
-```
+Prog8 supports `if`/`else`, `when`, `for`, `while`, `do`/`until`, `repeat`,
+`unroll`, `break`, `continue`, `goto`, and labels. `unroll` duplicates a
+constant-count body at compile time and does not support `break` or `continue`.
+`repeat` is usually more efficient than `for` when no loop variable is needed.
+Replace a `for` loop with a `repeat` loop when the loop variable is not used in the loop body.
+Use `if_cs`, `if_cc`, `if_z`, and `if_nz` for direct CPU-flag branches.
 
-## Subroutines & Return Values
-- **Don't use `private` on subroutines and variables** (including nested ones) unless the user asks for it. Everything is public by default in Prog8 — follow that convention.
-- `inline` keyword for subroutines to suggest inlining
-- No **function overloading** (except builtins) and no **polymorphism** in general. This means you must call specific routines for different types (e.g., `txt.print_ub(val)` vs `txt.print_w(val)` instead of a generic `print(val)`). Cannot use builtin names (msw, lsw, msb, lsb, mkword, mklong, peek, peekw, peekl, etc.) as variable/sub names
-- Can return 0, 1, or multiple values: `a, b, c = routine()`. Use `void` to skip: `void routine()`, `a, void, c = routine()`
-- Nested subroutines access parent scope variables directly
+## Subroutines and Assembly
 
-## Assembly Subroutines (asmsub)
-For low-level assembly that gets arguments via registers and returns values in registers.
-- **Syntax**: `[private] [inline] asmsub subname(params) [clobbers(regs)] [-> returns] { %asm {{ ... }} }`
-- **Body Restriction**: The body of an `asmsub` **must only contain** a single `%asm {{ ... }}` node. Regular Prog8 statements or nested blocks are NOT allowed.
-- **Parameters**: `type name @register` (e.g., `ubyte val @A`, `uword addr @AX`, `float f @FAC1`).
-- **Return Values**: `-> type @register` (e.g., `-> ubyte @A`, `-> bool @Pz` for immediate branch use).
-- **Registers**:
-    - **8-bit**: `@A`, `@X`, `@Y`
-    - **16-bit**: `@AX`, `@AY`, `@XY` (register pairs)
-    - **Float**: `@FAC1`, `@FAC2` (Floating Point Accumulators)
-    - **Virtual**: `@R0`–`@R15` (16-bit), `@R0R1`–`@R14R15` (32-bit combined)
-    - **Status Flags (for returns)**: `@Pc` (Carry), `@Pz` (Zero), `@Pv` (Overflow), `@Pn` (Negative)
-- **Clobbers**: `clobbers (A, X, Y)` — list all hardware registers modified by the routine.
-- **Parameter names** are for documentation and type checking only. Use the registers in your assembly code.
-- **Inlining**: `inline asmsub` will paste the assembly code directly at the call site, avoiding `jsr`/`rts` overhead.
+- Subroutines can return zero, one, or multiple values:
+  `a, b = routine()` or `void routine()`.
+- On M68K, ordinary calls preserve the caller's active frame. A `goto` or branch
+  that leaves the current subroutine must tear that frame down before transferring
+  control. The built-in `call(address)` is a zero-argument indirect call; it does
+  not marshal stack arguments.
+- Avoid `private` unless requested. Prog8 symbols are public by default.
+- `asmsub` bodies contain only one `%asm {{ ... }}` node. Parameters are type
+  checked and documented, but assembly must use the mapped registers. Declare
+  every modified hardware register in `clobbers`.
+- On M68K targets never clobber `A5`, the stack frame pointer. On the Amiga
+  targets never clobber `A6` either, it holds the AmigaOS library base.
+- `extsub` maps a signature to a fixed external address and has no body. Use
+  it for ROM, kernel, drivers, or binary-library routines.
+- For inline assembly, load `asm6502-coder` for 64tass syntax, target-specific
+  instructions, zero-page rules, and symbol references. Load `m68k-coder` for
+  M68000 assembly.
 
-## External Subroutines (extsub)
-Used to call routines at fixed memory addresses (like ROM KERNAL routines or third-party drivers).
-- **Syntax**: `[private] extsub [@bank <value>] address = subname(params) [clobbers(regs)] [-> returns]`
-- **Address**: Can be a hex literal (`$C000`) or a constant expression.
-- **Bank (optional)**: `@bank <integer>` (constant bank) or `@bank <identifier>` (variable bank).
-- **No Body**: Unlike `asmsub`, `extsub` has no `{ }` body; it just maps a signature to an address.
-- **Example**:
-  ```prog8
-  ; CX16 KERNAL CHROUT
-  extsub $FFD2 = chrout(ubyte char @A) clobbers(A, X, Y)
+## Interrupts and Hardware
 
-  ; Routine in a specific RAM bank
-  extsub @bank 10 $C09F = audio_init() clobbers(A, X, Y) -> bool @Pc
-  ```
+- Keep IRQ handlers short. Set a flag and do substantial work in the main loop.
+- Virtual registers `R0-R15` are not preserved across IRQ handlers. Avoid them
+  or save and restore them with the library routines.
+- CX16 IRQ handlers that touch VERA registers must save and restore VERA
+  context.
+- Use only symbolic scratch names such as `P8ZP_SCRATCH_W1` and
+  `cx16.r0`; never hardcode zeropage addresses.
 
-## Inline Assembly
-When writing inline assembly (`%asm {{ }}` blocks or `asmsub` routines), load the **asm6502-coder** skill for formatting rules, symbol references, anonymous labels, and 64tass syntax details.
+## Verification
 
-
-
-## Standard Library
-- Find routines, functions, variables, modules and signatures in the symbol dump file for the given compilation target. 
-  - Online location: https://prog8.readthedocs.io/en/latest/libraries.html#low-fi-variable-and-subroutine-definitions-in-all-available-library-modules  they are linked there 1 for each compilation target
-  - Structure: builtin functions, then module sections with variables/constants (`type name`) and subroutines (`name (params) -> returntype`)
-- Text output: `textio` module (`txt.print`, `txt.chrout`, `txt.print_b`/`_ub`/`_w`/`_uw`/`_l`/`_bool`, `txt.print_f` for floats, `txt.spc()`, `txt.nl()`). **Note: Prog8 has no function overloading**, so you cannot use `txt.print(number)` — you must call the specific routine for the type (e.g., `txt.print_ub(val)` for an unsigned byte).
-- Math: `math` module — integer trig (`sin8`, `cos8`) via fast LUTs; `math.rnd()` for random numbers
-- String conversion: `conv` module (`str_uword`, `str2word`, etc.) — for printing numbers use txt routines instead
-- Char operations: `strings` module (`isdigit`, `isxdigit`, `isupper`, `islower`, `isletter`, `isspace`, `isprint`) — use these instead of manual ASCII/PETSCII arithmetic
-- String functions return useful lengths — capture them: `len = strings.copy(dest, src)`, `len = strings.append(buf, text)`, `len = strings.upper(mystr)`
-
-## Syntax & Formatting
-- Hex: `$FF` (not `0xFF`); Binary: `%1010` (not `0b1010`). Underscores for readability: `25_000_000`
-- 4-digit hex `$0000` = uword. No type suffixes (no `0L`). Cast: `expr as type`
-- Augmented assignment: `+=`, `-=`, `*=`, `/=`, `%=`, `&=`, `|=`, `^=`, `<<=`, `>>=`
-- **`;` starts a comment to end of line** — NOT a statement separator. One statement per line
-- **No `elif`**: use nested `else { if ... }`
-- Type casting: `expression as type` (e.g., `bytevar as word`). `as` has very low precedence (lower than arithmetic)
-- **No automatic type widening**: `byte*byte=byte` (overflow possible!). Cast explicitly
-- **No bare `{ }` blocks** like C/Java
-- Indentation: 4 spaces for .p8 and .asm files (no tabs)
-- Character encoding: 6502 targets use PETSCII by default (call `txt.lowercase()` at start for lowercase). Virtual target uses ISO (`%encoding iso` + `txt.iso()`)
-- **Test programs for CBM targets must only use lowercase text in string literals** (the C64/C128/PET32/CX16 targets all use PETSCII). In PETSCII, uppercase letters A-Z are the *graphics* symbols, so any uppercase text you put in a string literal is rendered as unpredictable graphics characters instead of readable letters. Use only lowercase (a-z) for any printed test output, labels, or expected-match strings on these targets; reserve uppercase for intentional PETSCII graphics. The `virtual` target (ISO) is unaffected and may use either case.
-- Array size inferred from initializer: `str[] types = ["a", "b", "c"]`
-- Enums: `Enum::Value` syntax (double colon), declared inside a block. They are syntactic sugar for a list of `const` declarations, not a type. Use enums for related values, `const` for standalone
-- Avoid `globals.XXXX` — move constants closer to where they're used
-- Member access through pointers: use `.` for both direct and pointer access. The compiler infers the type. For complex assignment targets, `^^` may be needed: `ptr^^.field = value`
-- Qualified names: must use full path from top level (e.g., `cx16.r0`, not relative)
+- Use the `virtual` target for behavioral tests when possible:
+  `prog8c -target virtual -emu program.p8`.
+- Use `-check` for syntax and semantic checks without output generation.
+- Use `-noopt` to determine whether a failure is optimizer-related.
+- For IR execution, use `-vmtrace` when control flow needs inspection.

.aiignore

@@ -14,6 +14,7 @@ build/
 out/
 output/
 gradle/
+Libs/
 .gradle/
 .kotlin/
 .git/

.github/workflows/all-ci.yml

@@ -31,7 +31,7 @@ jobs:
         run: ./gradlew compileKotlin compileTestKotlin --console=plain
 
       - name: Test
-        run: ./gradlew test --continue --console=plain
+        run: ./gradlew test --continue --max-workers=3 -PtestConcurrency=2 -PtestMaxHeap=8g --console=plain
 
       - name: Package
         run: |
@@ -52,4 +52,3 @@ jobs:
         with:
           name: test-results
           path: compiler/build/test-results/test/
-

.gitignore

@@ -15,6 +15,8 @@ out-old/
 .*cache/
 *.directory
 *.prg
+*.elf
+*.o
 *.bin
 *.p8ir
 *.bankedcalls
@@ -43,4 +45,7 @@ compiler/lib/
 .gradle
 sd*.img
 *.d64
-
+AsyncIO/
+AsyncIO.info
+m68k-amigaos-ahi/
+Libs/

.idea/compiler.xml

@@ -14,4 +14,4 @@
       <module name="prog8.codeGenNew6502.test" target="17" />
     </bytecodeTargetLevel>
   </component>
-</project>
+</project>

.idea/gradle.xml

@@ -4,7 +4,6 @@
   <component name="GradleSettings">
     <option name="linkedExternalProjectsSettings">
       <GradleProjectSettings>
-        <option name="distributionType" value="LOCAL" />
         <option name="externalProjectPath" value="$PROJECT_DIR$" />
         <option name="modules">
           <set>
@@ -12,6 +11,8 @@
             <option value="$PROJECT_DIR$/codeCore" />
             <option value="$PROJECT_DIR$/codeGenCpu6502" />
             <option value="$PROJECT_DIR$/codeGenIntermediate" />
+            <option value="$PROJECT_DIR$/codeGenM68k" />
+            <option value="$PROJECT_DIR$/codeGenNew6502" />
             <option value="$PROJECT_DIR$/codeOptimizers" />
             <option value="$PROJECT_DIR$/compiler" />
             <option value="$PROJECT_DIR$/compilerAst" />

.idea/kotlinc.xml

@@ -15,6 +15,6 @@
   </component>
   <component name="KotlinJpsPluginSettings">
     <option name="externalSystemId" value="Gradle" />
-    <option name="version" value="2.4.10" />
+    <option name="version" value="2.4.21" />
   </component>
-</project>
+</project>

.qwen/settings.json

@@ -1,31 +0,0 @@
-{
-  "context": {
-    "fileName": "AGENTS.md"
-  },
-  "$version": 3,
-  "permissions": {
-    "allow": [
-      "Bash(prog8c *)",
-      "Bash(gradle *)",
-      "Bash(mkdir *)",
-      "Bash(ls *)",
-      "Bash(cat *)",
-      "Bash(find *)",
-      "Bash(sed *)",
-      "Bash(grep *)",
-      "Bash(xargs *)",
-      "Bash(wc *)",
-      "Bash(diff *)",
-      "Bash(paste *)",
-      "Bash(echo *)",
-      "Bash(cp *)",
-      "Bash(time)",
-      "Bash(kotlinc *)",
-      "Bash(python3 *)",
-      "Bash(jar *)",
-      "Bash(od *)",
-      "Bash(xxd *)",
-      "Bash(gcc *)"
-    ]
-  }
-}

AGENTS.md

@@ -13,17 +13,23 @@
 
 ## CHANGE / EDIT MODE
 
-- Never implement features yourself when possible - use sub-agents!
+- Use a subagent for larger chunks of work
 - Identify changes from the plan that can be implemented in parallel, and use sub-agents to implement the features efficiently
 - When using sub-agents to implement features, act as a coordinator only
 - Use the best model for the task - premium models for complex tasks (like coding) and mid-tier models for simpler tasks, like documentation
-- After completing features (large or small), always run commands like lint, type check and next build to check code quality
+- After completing features (large or small), run the relevant quality commands (lint, type check, build) to check code quality.
+  Only compiler/standard-library changes require a build: that means Kotlin sources (`.kt`), the grammar (`.g4`), or embedded
+  standard library files (`compiler/res/prog8lib/**/*.p8` and `**/*.asm`). For changes limited to `examples/**`, `docs/**`, or
+  other non-compiler files, verify with the target compiler directly instead (for example `prog8c -target <target> <file>`), and
+  skip the build entirely
+  (exception: if the change is only whitespace or code comments, a build or test run is not necessary)
+- `gradle build` already runs the full test suite; don't run a separate full `gradle test` before it, to avoid lengthy double test runs (targeted `--tests` filters to find failing tests are fine)
 
 ## Project Overview
 - This project is a compiler for the Prog8 programming language.
 - Prog8 is a programming language primarily targeting 8-bit retro systems with the 6502 CPU, such as the Commodore 64, Commodore 128, and Commander X16.
 - The compiler has a 6502 code generator backend, and an IR code generator.
-- The IR code is meant to be used in a new machine specific code generator backend (primarily 6502 but maybe 68000 as well later)
+- The IR code is used by machine-specific code generator backends, including 6502 and M68K.
 - The compiler includes a simple 'virtual machine' that can execute the IR code directly via interpretation.
 - Prog8 source files have .p8 extension – these are *not* LUA or PICO-8 source files in this case!
 - Prog8 source files are a "module" that can contain one or more "blocks". They can also import other modules, from internal library files or from source files on the filesystem.
@@ -35,6 +41,7 @@
 - ANTLR4 version 4.13 is used for the parser implementation.
 - Dependent library versions can be found in 'build.gradle.kts' and in the IntelliJ IDEA configuration files in .idea/libraries
 - The compiler main entrypoint is in the "compiler" module, in src/prog8/CompilerMain.kt
+- **Codebase knowledge graph**: For a structured overview of modules, their dependencies, key entities, the end-to-end compilation pipeline, the standard library layout, and the test structure, consult `docs/CODEBASE-KNOWLEDGE-GRAPH.md`. Read it when you need to understand how the compiler fits together before diving into specific files.
 
 ## Compilation Flow (High-Level)
 ```
@@ -53,7 +60,9 @@ Source → parseMainModule() → processAst() → optimizeAst() → postprocessA
 
 ### Target Differences
 - **CPU instruction set differences**: Only the CommanderX16 target (cx16) can use 65C02 instructions such as STZ. The other targets (C64, C128, PET32) can only use original 6502 instructions.
-- When writing or understanding assembly code, load the `asm6502-coder` skill for 64tass syntax and conventions.
+- When writing or understanding 6502/65C02 assembly code, load the `asm6502-coder` skill for 64tass syntax and conventions.
+- When writing or understanding M68K assembly code (amiga500, amiga1200, qemu68k targets), load the `m68k-coder` skill for vasm mot syntax and conventions.
+- M68K ordinary subroutine locals and parameters use `A5`-based stack frames. Normal calls pass word-packed argument slots on the stack (2-byte slots for byte/word args, 4-byte for wide args, byte values at the slot base), and the caller cleans up; see `docs/source/technical.rst` for the complete convention and frame eligibility rules.
 
 
 ## DEBUGGING TIP: Use `-noopt` to isolate problems
@@ -125,7 +134,19 @@ Steps when encountering a compiler crash:
 - **NO EMOJI in user documentation**. Do not use emoji or decorative unicode symbols in documentation files. Functional unicode symbols are acceptable when they serve a clear purpose (e.g., → for arrows, ± for plus-minus, × for multiplication, ° for degrees). Avoid decorative emoji like ❌ ✅ ⚠️ 🎉 etc.
 
 ### Code Style Guidelines
-**Minimal comments when making changes**: When modifying existing code, add only essential comments that explain *why* a change was made or document non-obvious behavior. **Do not add verbose comments** that restate what the code does; let the code speak for itself. Existing extensive comments should be preserved, but new changes should have minimal commentary.
+**Code comments**: Comments are the exception, not the norm. A comment earns its place only by recording
+something the code itself cannot say: the reason for a non-obvious choice, or a constraint that
+a well-meaning future edit would otherwise violate.
+
+- Inline comment inside a function body: 1 line, never more.
+- KDoc on a declaration: 3 lines, unless a genuine non-obvious contract cannot be stated shorter.
+- Never restate the identifier, the type, or what the following line does.
+- Prefer a good name or a well-named function over a comment that explains it.
+
+Before keeping any comment, delete it mentally and ask what the reader loses. If nothing, it goes.
+
+Keep existing comments as they are; do not expand, reword, or reformat them while editing
+nearby lines - unless that would cause inconsistency. 
 
 **Kotlin string `$` handling**: The `$` character is special in Kotlin string literals (string interpolation). If you need a literal `$` character, use Kotlin's `$$"..."` interpolation prefix instead of backslash escaping (`\$`) or `${'$'}`. The `$$"..."` syntax is a fairly recent Kotlin feature that avoids nasty escaping. For example:
 ```kotlin
@@ -196,10 +217,12 @@ When the task involves writing or understanding `.p8` (Prog8 source) or `.p8ir`
 - This compiles AND installs, but still skips running tests (faster than `gradle build`)
 
 ### When to run full build with tests
-- **Before final verification**: Always run `gradle build` to ensure all tests pass
+- **Before final verification**: Run `gradle build` to ensure all tests pass - but only if the change actually touched the
+  compiler or standard library (`.kt`, `.g4`, or `compiler/res/prog8lib/**`). For `examples/**` or `docs/**`-only changes,
+  verify with `prog8c` directly instead and skip the build
 - **After major refactoring**: Run `gradle build` to catch regressions
 - **When debugging test failures**: Use `gradle test --tests "*TestName*"` for specific tests
-- **When working on a standalone codegen module** (one that reads `.p8ir` files): Run `gradle :<module>:test --console=plain` instead of full `gradle build`. Compiler unit tests are irrelevant for IR-level changes. Do run full `gradle build` before final verification.
+- **When working on a standalone code gen module** (one that reads `.p8ir` files): Run `gradle :<module>:test --console=plain` instead of full `gradle build`. Compiler unit tests are irrelevant for IR-level changes. Do run full `gradle build` before final verification if the change touched compiler sources.
 
 ### Build command summary
 
@@ -216,13 +239,15 @@ When the task involves writing or understanding `.p8` (Prog8 source) or `.p8ir`
 - the prog8c compiler executable can be found in the compiler/build/install/prog8c/bin folder (this is already added to the shell's path)
 - **the `-check` switch performs a quick syntax/semantic check only; it will NOT produce any output files (no .prg, .asm, etc.)**. Use it only for fast error checking during development.
 - **the `-noopt` switch DISABLES all optimizations** - useful for debugging to determine if a problem is caused by the optimizer. **Optimizations are ENABLED by default** (no flag needed).
-- **prog8c uses single-dash command line options** (e.g., `-target`, `-noopt`, `-check`), NOT double-dash (`--target` is invalid).
+- **prog8c options accept single-dash or double-dash** (e.g., `-target` and `--target` are equivalent).
 - **the `-printast1` switch prints out the internal Compiler AST** after parsing and semantic analysis.
 - **the `-printast2` switch prints out the optimized Simple AST** just before it goes to the code generator. This is useful for debugging optimizer issues.
-- **the `-out outdir` switch sets an alternative output directory** for compiled files (.prg, .asm, .list, etc.). **By default, output files are written to the same directory as the source file**.
+- **the `-out outdir` switch sets an alternative output directory** for compiled files (.prg, .asm, .list, etc.). **By default, output files are written to the current working directory**.
+- **the `-o <file>` switch names the exact final program artifact** (also spelled `--output`); all other output files go in the same directory with the same stem. Cannot be combined with `-out` or with multiple source files.
 - **Other useful flags**: `-quiet` (suppress messages), `-warnimplicitcasts` (warn on implicit type widening), `-daemon` (keep a background compiler process alive — must be passed on every invocation)
 
 ### Compilation Output Files
+Names follow the source file name, or the stem of the `-o` file when that option is used.
 - `*.prg` - The final compiled program file for the target system (e.g., Commander X16)
 - `*.asm` - Generated assembly code from the Prog8 source
 - `*.list` - Generated full assembly listing file from the Prog8 source
@@ -242,15 +267,25 @@ When the task involves writing or understanding `.p8` (Prog8 source) or `.p8ir`
 
 - Never assume your changes simply work, always test!
 - If the project does not have any testing tools, scripts, MCP tools, skills, etc. available for testing, ask the user whether testing should be skipped.
+- in prog8 code being tested, use '@shared' variables to avoid const-folding
 
 ### Automated Tests (gradle)
 Always use the `--console=plain` flag with these commands to avoid messing up the agent's text interface.
-- `gradle build --console=plain` - Full build including all tests (about 50s)
-- `gradle :compiler:test --console=plain` - Run only compiler tests (faster)
-- `gradle :compiler:test --tests "prog8tests.compiler.TestOptimization" --console=plain` - Run specific test class
+- `ionice -c 3 nice gradle build --console=plain` - Full build including all tests (about 50s)
+- `ionice -c 3 nice gradle :compiler:test --console=plain` - Run only compiler tests (faster)
+- `ionice -c 3 nice gradle :compiler:test --tests "prog8tests.compiler.TestOptimization" --console=plain` - Run specific test class
 
 Note: By default, Gradle only shows failed tests. Passed and skipped tests are silent.
 
+On Unix-like systems, prefix **every** gradle invocation with `nice` to lower its CPU priority,
+because gradle is CPU-heavy. This applies to all gradle commands (build, compile, test, install), not
+just `gradle build`. Use the same prefix for the quick compile checks and install commands in
+"Building and Installing the Compiler" above.
+
+On **Linux** additionally put `ionice -c 3` in front (idle-class I/O scheduling), so the build doesn't
+compete for disk with whatever else is running: `ionice -c 3 nice gradle ...`. `ionice` is
+Linux-only (util-linux); on macOS/BSD just use `nice`.
+
 **⚠️ CRITICAL: Test Filtering Patterns - Read This First!**
 
 Gradle's `--tests` filter has **strict rules** that are easy to get wrong:
@@ -281,9 +316,17 @@ gradle :compiler:compileTestKotlin --info 2>&1 | grep "^e:"
 - Unit tests use KoTest (FunSpec style)
 - Tests are in the "compiler" module's `test` directory (and some other modules)
 - Test config is in root `build.gradle.kts`; tests run in parallel
-- **When writing test programs**, add at the top: `%zeropage basicsafe` and `%option no_sysinit`
+- **When writing test programs**, add at the top: `%zeropage basicsafe` and `%option no_sysinit` - the latter ONLY when NOT writing for the amiga500 or qemu68k targets
 - When a test fails, the output shows "There were failing tests. See the report at:" - **read that HTML report**
 
+### TestCompilerOnExamples expected sizes
+`TestCompilerOnExamples` asserts exact output sizes per example. A code-generator change shifts them,
+so expect a handful of `expected:<N> but was:<M>` failures with no stack trace pointing at the cause.
+Attribute each one before editing: rebuild with the change reverted and compare the two generated
+`.asm` files, so the size delta is demonstrably yours and not a side effect. Then update the expected
+size. A delta of a few bytes from removed `ldy`/`lda` instructions is normal; a large or erratic delta
+means something else changed and needs investigating.
+
 ### Compiler Unit Test Snippets
 Prog8 code snippets embedded in the compiler's own Kotlin unit tests (e.g., in `TestAstChecks.kt`, `TestOptimization.kt`) follow different rules from standalone programs:
 - **Keep self-contained** — avoid `%import` directives (test setup may not have library search path configured)
@@ -291,6 +334,11 @@ Prog8 code snippets embedded in the compiler's own Kotlin unit tests (e.g., in `
 - **Set `writeAssembly=false`** if the test only checks generated AST (much faster)
 - **Set `optimize=false`** by default for these snippets
 - **Do NOT use** the `%encoding iso` / `txt.iso()` / `sys.poweroff_system()` pattern — that's only for real CX16 emulator runs
+- **Do NOT** write compiler tests that depend on the `vasm` assembler: it is not always installed
+  (CI only installs `64tass`). For m68k codegen tests (`amiga500`, `qemu68k` targets) pass
+  `writeAssembly = true, assemble = false` to the `compileText`/`compileFile` helpers. This still
+  generates the codegen AST to assert on but skips assembling, so no vasm is needed. If a test's
+  assertions cannot be made from the AST alone, drop the test.
 
 ### Testing Assembly Output
 When testing generated assembly in Kotlin tests:
@@ -307,6 +355,8 @@ When testing generated assembly in Kotlin tests:
 
 **Testing tip**: When writing and testing Prog8 programs, **use the `virtual` target** (e.g., `prog8c -target virtual -emu input.p8` or `prog8c -vm input.p8ir`). This is the preferred way to test because the virtual target can easily write output to stdout, making it simple to verify program behavior and check results.
 
+For the amiga targets, `-emu` launches FS-UAE (GUI) while `-emu2` launches Vamos (headless CLI). For automated/non-interactive testing, prefer `-emu2` (vamos) to avoid opening a GUI window.
+
 **CX16 output verification**: Use `x16emu -echo iso -run -prg input.prg` to echo screen output to stdout **and auto-start the program**. The `-run` flag is **critical**: without it, the program loads but doesn't execute, so you'll see no output. Pipe through `strings` or `grep` to filter: `x16emu -echo iso -run -prg input.prg 2>&1 | grep -E "(PASS|FAIL)"`.
 **IMPORTANT**: Always add `%encoding iso` at the top of your source file and call `txt.iso()` in `start()`. This prevents PETSCII→ISO charset translation errors that garble uppercase/special characters and make output unreadable:
 ```prog8

README.md

@@ -4,13 +4,14 @@ PayPal: [![PayPal](https://www.paypalobjects.com/en_US/i/btn/btn_donate_LG.gif)]
 
 [![Documentation](https://readthedocs.org/projects/prog8/badge/?version=latest)](https://prog8.readthedocs.io/)
 
-Prog8 - Structured Programming Language for 8-bit 6502/65c02 microprocessors
-============================================================================
+Prog8 - Structured Programming Language for 6502/65c02 (8-bit) and 68000 (32-bit) microprocessors
+=================================================================================================
 
 *Written by Irmen de Jong (irmen@razorvine.net)*
 
 This is a structured programming language for the 8-bit 6502/6510/65c02 microprocessor from the late 1970's and 1980's
-as used in many home computers from that era. It is a medium to low level programming language,
+as used in many home computers from that era. It also supports the 32-bit Motorola 68000 CPU, for which the
+Commodore Amiga is the intended machine. It is a medium to low level programming language,
 which aims to provide many conveniences over raw assembly code (even when using a macro assembler).
 
 **Want to buy me a coffee or a pizza perhaps?**
@@ -104,9 +105,14 @@ What does Prog8 provide?
 - "c64": Commodore-64  (6502 like CPU)
 - "c128": Commodore-128  (6502 like CPU - the Z80 cpu mode is not supported)
 - "pet32": Commodore PET (limited support)
+- "amiga500": Commodore Amiga 500  (16/32-bit 68000 CPU; experimental)
+- "amiga1200": Commodore Amiga 1200  (32-bit 68020 CPU with optional FPU; experimental)
+- "qemu68k": Qemu M68k simulator (experimental)
 - via external configurable targets: Atari 800 XL, Neo6502, NES, C64 OS, Foenix F256, ...
 - If you only use standard kernal and prog8 library routines, it is possible to compile the *exact same program* for different machines (just change the compiler target flag)
 
+The m68k targets (amiga500, amiga1200, qemu68k) and the optional `-newcodegen` 6502 code generator are experimental and not yet suitable for production code.
+
 
 
 Additional required tools

benchmark-program/b_3d.p8

@@ -41,7 +41,7 @@ rotate3d {
         for i in 0 to len(matrix_math.xcoor)-1 {
             rz = matrix_math.rotatedz[i]
             if rz >= 10 {
-                persp = 600 + rz/64
+                persp = 600 + (rz>>5)
                 sx = matrix_math.rotatedx[i] / persp as byte + WIDTH/2
                 sy = matrix_math.rotatedy[i] / persp as byte + HEIGHT/2
                 ;; txt.setcc(sx as ubyte, sy as ubyte, 46, 7)
@@ -51,7 +51,7 @@ rotate3d {
         for i in 0 to len(matrix_math.xcoor)-1 {
             rz = matrix_math.rotatedz[i]
             if rz < 10 {
-                persp = 600 + rz/64
+                persp = 600 + (rz>>5)
                 sx = matrix_math.rotatedx[i] / persp as byte + WIDTH/2
                 sy = matrix_math.rotatedy[i] / persp as byte + HEIGHT/2
                 ;; txt.setcc(sx as ubyte, sy as ubyte, 81, 7)
@@ -84,18 +84,19 @@ matrix_math {
         word wcosc = math.cos8(az)
         word wsinc = math.sin8(az)
 
-        word wcosa_sinb = wcosa*wsinb / 128
-        word wsina_sinb = wsina*wsinb / 128
+        ; instead of (slow) /128, we simply shift 7 bits (and take the roundoff error)
+        word wcosa_sinb = wcosa*wsinb >> 7
+        word wsina_sinb = wsina*wsinb >> 7
 
-        word Axx = wcosa*wcosb / 128
-        word Axy = (wcosa_sinb*wsinc - wsina*wcosc) / 128
-        word Axz = (wcosa_sinb*wcosc + wsina*wsinc) / 128
-        word Ayx = wsina*wcosb / 128
-        word Ayy = (wsina_sinb*wsinc + wcosa*wcosc) / 128
-        word Ayz = (wsina_sinb*wcosc - wcosa*wsinc) / 128
+        word Axx = wcosa*wcosb >> 7
+        word Axy = (wcosa_sinb*wsinc - wsina*wcosc) >> 7
+        word Axz = (wcosa_sinb*wcosc + wsina*wsinc) >> 7
+        word Ayx = wsina*wcosb >> 7
+        word Ayy = (wsina_sinb*wsinc + wcosa*wcosc) >> 7
+        word Ayz = (wsina_sinb*wcosc - wcosa*wsinc) >> 7
         word Azx = -wsinb
-        word Azy = wcosb*wsinc / 128
-        word Azz = wcosb*wcosc / 128
+        word Azy = wcosb*wsinc >> 7
+        word Azz = wcosb*wcosc >> 7
 
         ubyte @zp i
         for i in 0 to len(xcoor)-1 {

benchmark-program/b_btree.p8

@@ -220,10 +220,10 @@ btree {
 
 arena {
     ; extremely trivial arena allocator (that never frees)
-    uword buffer = memory("arena", 2000, 0)
-    uword next = buffer
+    pointer buffer = memory("arena", 2000, 0)
+    pointer next = buffer
 
-    sub alloc(ubyte size) -> uword {
+    sub alloc(ubyte size) -> pointer {
         defer next += size
         return next
     }

benchmark-program/benchmark.p8

@@ -4,6 +4,12 @@
 ;
 ;  As the X16 computer is a more or less fixed system, it's not very useful
 ;  to benchmark the computer itself with.
+;
+;  The individual scores are iteration counts, and they differ by more than an
+;  order of magnitude between benchmarks (maze does ~3000, game of life ~180).
+;  Summing the raw counts therefore let one or two benchmarks dominate the total
+;  and hide regressions in all the others, so the total is a weighted sum - see
+;  benchmark_weight below.
 
 
 %import textio
@@ -26,7 +32,15 @@ main {
 
     str[20] benchmark_names
     uword[20] benchmark_score
-    str version = "12.1"
+    ;  Weight per benchmark, in the same order as the benchmarks are run below. The total is the
+    ;  weighted sum of the raw iteration counts.
+    ;  The raw counts differ by more than an order of magnitude (maze ~3064, game of life ~180), so
+    ;  in an unweighted sum the two heaviest - maze solver and text-elite - were 57% of the total
+    ;  while game of life was 2%. A regression in a small benchmark was therefore nearly invisible.
+    ;  Their weights are scaled down so that every benchmark contributes roughly comparably; the
+    ;  other eight are left at 256 (i.e. unweighted).
+    uword[10] benchmark_weight = [ 64, 256, 256, 256, 256, 256, 256, 64, 256, 256 ]
+    str version = "14.0"
 
 
     sub start() {
@@ -96,11 +110,14 @@ main {
         txt.print(")\n\n")
 
         do {
+            ;  Weighted contribution. Done in long arithmetic on purpose: score*weight overflows a
+            ;  uword (3064*256 is 784384, which wraps to 63488 and would corrupt the total).
+            uword weighted = ((benchmark_score[benchmark_number] as long) * benchmark_weight[benchmark_number] / 256) as uword
             txt.spc()
-            txt.print_uw(benchmark_score[benchmark_number])
+            txt.print_uw(weighted)
             txt.column(6)
             txt.print(benchmark_names[benchmark_number])
-            total_score += benchmark_score[benchmark_number]
+            total_score += weighted
             txt.nl()
             benchmark_number++
         } until benchmark_names[benchmark_number]==0

build.gradle.kts

@@ -3,7 +3,7 @@ import org.jetbrains.kotlin.gradle.dsl.JvmTarget
 
 
 plugins {
-    kotlin("jvm") version "2.4.10"
+    kotlin("jvm") version "2.4.21"
 }
 
 allprojects {
@@ -52,8 +52,8 @@ subprojects {
     // Common test dependencies via Kotest BOM (Bill of Materials)
     // This manages all Kotest module versions centrally
     dependencies {
-        testImplementation(platform("io.kotest:kotest-bom:6.2.3"))
-        // implementation("com.github.irmen:ksim65:v2.1")
+        testImplementation(platform("io.kotest:kotest-bom:6.2.4"))
+        // implementation("org.codeberg.irmen:ksim65:v2.2")
     }
 
     // Common sourceSets configuration
@@ -72,16 +72,22 @@ subprojects {
         // Enable JUnit 5 (required for Kotest)
         useJUnitPlatform()
 
+        // Forked test JVMs each reserve up to -PtestMaxHeap, so keep the fork count
+        // bounded even on many-core machines to avoid swapping during parallel builds.
+        val testConcurrency = (project.findProperty("testConcurrency")?.toString()?.toIntOrNull()
+            ?: (Runtime.getRuntime().availableProcessors() / 2)).coerceIn(1, 4)
+        project.findProperty("testMaxHeap")?.toString()?.let { maxHeapSize = it }
+
         // Enable concurrent test execution for Kotest
-        // Set parallelism to number of CPU cores
-        jvmArgs("-Dkotest.framework.parallelism=${Runtime.getRuntime().availableProcessors()}")
+        // Allow CI to override parallelism for memory-constrained runners
+        jvmArgs("-Dkotest.framework.parallelism=$testConcurrency")
 
         // Disable Kotest autoscan warning
         jvmArgs("-Dkotest.framework.classpath.scanning.autoscan.disable=true")
 
         // Enable Gradle's parallel test execution (runs multiple test classes concurrently)
         // Use 50% of available processors to avoid over-subscription
-        maxParallelForks = (Runtime.getRuntime().availableProcessors() / 2).coerceAtLeast(1)
+        maxParallelForks = testConcurrency
 
         // Allow filtering tests via -PtestFilter="*TestName*" from command line
         // Example: gradle test -PtestFilter="*TestLookup*"

codeCore/src/prog8/code/Globals.kt

@@ -40,12 +40,10 @@ object SymbolNames {
      * See https://github.com/irmen/prog8/issues/198
      */
     fun stripPrefixes(scopedName: String): String {
-        return scopedName.split('.')
-            .map { part ->
-                if(part.length > 4 && part.startsWith("p8") && part[2].isLetter() && part[3] == '_')
-                    part.drop(4)
-                else part
-            }
-            .joinToString(".")
+        return scopedName.split('.').joinToString(".") { part ->
+            if (part.length > 4 && part.startsWith("p8") && part[2].isLetter() && part[3] == '_')
+                part.drop(4)
+            else part
+        }
     }
 }

codeCore/src/prog8/code/assembly/AssemblyProgram6502.kt

@@ -1,38 +1,29 @@
-package prog8.codegen.cpu6502
+package prog8.code.assembly
 
 import prog8.code.GENERATED_LABEL_PREFIX
-import prog8.code.IAssemblyProgram
-import prog8.code.core.CompilationOptions
-import prog8.code.core.ICompilationTarget
-import prog8.code.core.IErrorReporter
-import prog8.code.core.OutputType
+import prog8.code.core.*
 import prog8.code.target.C128Target
 import prog8.code.target.C64Target
 import prog8.code.target.PETTarget
 import java.nio.file.Path
 
 
-internal class AssemblyProgram(
-        override val name: String,
-        outputDir: Path,
-        private val compTarget: ICompilationTarget) : IAssemblyProgram {
+class AssemblyProgram6502(
+    override val name: String,
+    private val options: CompilationOptions,
+    private val compTarget: ICompilationTarget) : IAssemblyProgram {
 
     override val irInstructionCount: Int = 0
     override val irChunkCount: Int = 0
     override val irRegisterCount: Int = 0
 
-    private val assemblyFile = outputDir.resolve("$name.asm")
-    private val prgFile = outputDir.resolve("$name.prg")        // CBM prg executable program
-    private val xexFile = outputDir.resolve("$name.xex")        // Atari xex executable program
-    private val binFile = outputDir.resolve("$name.bin")
-    private val viceMonListFile = outputDir.resolve(C64Target.viceMonListName(name))
-    private val listFile = outputDir.resolve("$name.list")
-
     override fun assemble(options: CompilationOptions, errors: IErrorReporter): Boolean {
 
         val assemblerCommand: List<String>
+        val outs = OutputFiles.of(options, name)
 
         fun addRemainingOptions(command: MutableList<String>, program: Path, assembly: Path): List<String> {
+            // keep additionalAssemblerOptions BEFORE --output: a target config may contain its own --output
             if(options.compTarget.additionalAssemblerOptions.isNotEmpty())
                 command.addAll(options.compTarget.additionalAssemblerOptions)
 
@@ -45,7 +36,7 @@ internal class AssemblyProgram(
                 // CBM machines .prg generation.
 
                 val command = mutableListOf("64tass", "--cbm-prg", "--ascii", "--case-sensitive", "--long-branch",
-                    "-Wall", "-Wno-implied-reg", "--no-monitor", "--dump-labels", "--vice-labels", "--labels=$viceMonListFile")
+                    "-Wall", "-Wno-implied-reg", "--no-monitor", "--dump-labels", "--vice-labels", "--labels=${outs.viceMonList()}")
 
                 if(options.warnSymbolShadowing)
                     command.add("-Wshadow")
@@ -56,10 +47,10 @@ internal class AssemblyProgram(
                     command.add("--quiet")
 
                 if(options.asmListfile) {
-                    command.add("--list=$listFile")
+                    command.add("--list=${outs.list()}")
                 }
 
-                assemblerCommand = addRemainingOptions(command, prgFile, assemblyFile)
+                assemblerCommand = addRemainingOptions(command, outs.programFile, outs.asm())
                 if(!options.quiet)
                     println("\nCreating prg for target ${compTarget.name}.")
             }
@@ -67,7 +58,7 @@ internal class AssemblyProgram(
                 // Atari800XL .xex generation.
 
                 val command = mutableListOf("64tass", "--atari-xex", "--case-sensitive", "--long-branch",
-                    "-Wall", "-Wno-implied-reg", "--no-monitor", "--dump-labels", "--vice-labels", "--labels=$viceMonListFile")
+                    "-Wall", "-Wno-implied-reg", "--no-monitor", "--dump-labels", "--vice-labels", "--labels=${outs.viceMonList()}")
 
                 if(options.warnSymbolShadowing)
                     command.add("-Wshadow")
@@ -78,16 +69,16 @@ internal class AssemblyProgram(
                     command.add("--quiet")
 
                 if(options.asmListfile)
-                    command.add("--list=$listFile")
+                    command.add("--list=${outs.list()}")
 
-                assemblerCommand = addRemainingOptions(command,xexFile, assemblyFile)
+                assemblerCommand = addRemainingOptions(command, outs.programFile, outs.asm())
                 if(!options.quiet)
                     println("\nCreating xex for target ${compTarget.name}.")
             }
             OutputType.RAW -> {
                 // Neo6502/headerless raw program generation.
                 val command = mutableListOf("64tass", "--nostart", "--case-sensitive", "--long-branch",
-                    "-Wall", "-Wno-implied-reg", "--no-monitor", "--dump-labels", "--vice-labels", "--labels=$viceMonListFile")
+                    "-Wall", "-Wno-implied-reg", "--no-monitor", "--dump-labels", "--vice-labels", "--labels=${outs.viceMonList()}")
 
                 if(options.warnSymbolShadowing)
                     command.add("-Wshadow")
@@ -98,9 +89,9 @@ internal class AssemblyProgram(
                     command.add("--quiet")
 
                 if(options.asmListfile)
-                    command.add("--list=$listFile")
+                    command.add("--list=${outs.list()}")
 
-                assemblerCommand = addRemainingOptions(command, binFile, assemblyFile)
+                assemblerCommand = addRemainingOptions(command, outs.programFile, outs.asm())
                 if(!options.quiet)
                     println("\nCreating raw binary for target ${compTarget.name}.")
             }
@@ -108,7 +99,7 @@ internal class AssemblyProgram(
                 // CBM machines library (.bin) generation (with or without 2 byte load address header depending on the compilation target machine)
 
                 val command = mutableListOf("64tass", "--ascii", "--case-sensitive", "--long-branch",
-                    "-Wall", "-Wno-implied-reg", "--no-monitor", "--dump-labels", "--vice-labels", "--labels=$viceMonListFile")
+                    "-Wall", "-Wno-implied-reg", "--no-monitor", "--dump-labels", "--vice-labels", "--labels=${outs.viceMonList()}")
 
                 if(options.warnSymbolShadowing)
                     command.add("-Wshadow")
@@ -119,7 +110,7 @@ internal class AssemblyProgram(
                     command.add("--quiet")
 
                 if(options.asmListfile)
-                    command.add("--list=$listFile")
+                    command.add("--list=${outs.list()}")
 
                 if(compTarget.name in listOf(C64Target.NAME, C128Target.NAME, PETTarget.NAME)) {
                     if(!options.quiet)
@@ -131,8 +122,9 @@ internal class AssemblyProgram(
                     command.add("--nostart")       // should be headerless bin, because basic has problems doing a normal LOAD"lib",8,1 - need to use BLOAD
                 }
 
-                assemblerCommand = addRemainingOptions(command, binFile, assemblyFile)
+                assemblerCommand = addRemainingOptions(command, outs.programFile, outs.asm())
             }
+            else -> error("Unsupported output type: ${compTarget.defaultOutputType}")
         }
 
         val proc = ProcessBuilder(assemblerCommand)
@@ -159,16 +151,16 @@ internal class AssemblyProgram(
 
         val result = process.waitFor()
         if (result == 0) {
-            removeGeneratedLabelsFromMonlist()
-            generateBreakpointList()
+            removeGeneratedLabelsFromMonlist(outs)
+            generateBreakpointList(outs)
         }
         return result==0
     }
 
-    private fun removeGeneratedLabelsFromMonlist() {
+    private fun removeGeneratedLabelsFromMonlist(outs: OutputFiles) {
         val pattern = Regex("""al (\w+) \S+$GENERATED_LABEL_PREFIX.+?""")
-        val lines = viceMonListFile.toFile().readLines()
-        viceMonListFile.toFile().outputStream().bufferedWriter().use {
+        val lines = outs.viceMonList().toFile().readLines()
+        outs.viceMonList().toFile().outputStream().bufferedWriter().use {
             for (line in lines) {
                 if(pattern.matchEntire(line)==null)
                     it.write(line+"\n")
@@ -176,11 +168,11 @@ internal class AssemblyProgram(
         }
     }
 
-    private fun generateBreakpointList() {
+    private fun generateBreakpointList(outs: OutputFiles) {
         // builds list of breakpoints, appends to monitor list file
         val breakpoints = mutableListOf<String>()
         val pattern = Regex("""al (\w+) \S+_prog8_breakpoint_\d+.?""")      // gather breakpoints by the source label that's generated for them
-        for (line in viceMonListFile.toFile().readLines()) {
+        for (line in outs.viceMonList().toFile().readLines()) {
             val match = pattern.matchEntire(line)
             if (match != null)
                 breakpoints.add("break $" + match.groupValues[1])
@@ -189,6 +181,7 @@ internal class AssemblyProgram(
         breakpoints.add(0, "; breakpoint list now follows")
         breakpoints.add(1, "; $num breakpoints have been defined")
         breakpoints.add(2, "del")
-        viceMonListFile.toFile().appendText(breakpoints.joinToString("\n") + "\n")
+        outs.viceMonList().toFile().appendText(breakpoints.joinToString("\n") + "\n")
     }
 }
+

codeCore/src/prog8/code/assembly/IAssemblyProgram.kt

@@ -0,0 +1,12 @@
+package prog8.code.assembly
+
+import prog8.code.core.CompilationOptions
+import prog8.code.core.IErrorReporter
+
+interface IAssemblyProgram {
+    val name: String
+    val irInstructionCount: Int
+    val irChunkCount: Int
+    val irRegisterCount: Int
+    fun assemble(options: CompilationOptions, errors: IErrorReporter): Boolean
+}

codeCore/src/prog8/code/core/BuiltinFunctions.kt

@@ -33,7 +33,7 @@ class FSignature(val pure: Boolean,      // does it have side effects?
                  val returnTypes: Array<BaseDataType>,
                  vararg val parameters: FParam) {
 
-    fun callConvention(actualParamTypes: List<BaseDataType>): CallConvention {
+    fun callConventionFor6502(actualParamTypes: List<BaseDataType>): CallConvention {
         val returns: ReturnConvention
         if(returnTypes.isEmpty())
             returns = ReturnConvention(null, null)
@@ -41,7 +41,7 @@ class FSignature(val pure: Boolean,      // does it have side effects?
             val returnType = returnTypes[0]
             returns = when (returnType) {
                 BaseDataType.UBYTE, BaseDataType.BYTE -> ReturnConvention(returnType, RegisterOrPair.A)
-                BaseDataType.UWORD, BaseDataType.WORD -> ReturnConvention(returnType, RegisterOrPair.AY)
+                BaseDataType.UWORD, BaseDataType.WORD, BaseDataType.POINTER -> ReturnConvention(returnType, RegisterOrPair.AY)
                 BaseDataType.LONG -> ReturnConvention(returnType, RegisterOrPair.R14R15)
                 BaseDataType.FLOAT -> ReturnConvention(returnType, RegisterOrPair.FAC1)
                 in IterableDatatypes -> ReturnConvention(returnType, RegisterOrPair.AY)
@@ -49,7 +49,7 @@ class FSignature(val pure: Boolean,      // does it have side effects?
                     // return type depends on arg type
                     when (val paramType = actualParamTypes.first()) {
                         BaseDataType.UBYTE, BaseDataType.BYTE -> ReturnConvention(paramType, RegisterOrPair.A)
-                        BaseDataType.UWORD, BaseDataType.WORD -> ReturnConvention(paramType, RegisterOrPair.AY)
+                        BaseDataType.UWORD, BaseDataType.WORD, BaseDataType.POINTER -> ReturnConvention(paramType, RegisterOrPair.AY)
                         BaseDataType.LONG -> ReturnConvention(returnType, RegisterOrPair.R14R15)
                         BaseDataType.FLOAT -> ReturnConvention(paramType, RegisterOrPair.FAC1)
                         in IterableDatatypes -> ReturnConvention(paramType, RegisterOrPair.AY)
@@ -109,8 +109,8 @@ val BuiltinFunctions: Map<String, FSignature> = mapOf(
     "prog8_lib_square_word"       to FSignature(true, arrayOf(BaseDataType.UWORD), FParam("value", BaseDataType.WORD, BaseDataType.UWORD)),
     "prog8_lib_square_long"       to FSignature(true, arrayOf(BaseDataType.LONG), FParam("value", BaseDataType.LONG)),
     "prog8_lib_structalloc"       to FSignature(true, arrayOf(BaseDataType.UWORD)),
-    "prog8_lib_copylong"          to FSignature(false, emptyArray(), FParam("pointer1", BaseDataType.UWORD), FParam("pointer2", BaseDataType.UWORD)),
-    "prog8_lib_copyfloat"         to FSignature(false, emptyArray(), FParam("pointer1", BaseDataType.UWORD), FParam("pointer2", BaseDataType.UWORD)),
+    "prog8_lib_copylong"          to FSignature(false, emptyArray(), FParam("pointer1", BaseDataType.POINTER, BaseDataType.UWORD, BaseDataType.LONG), FParam("pointer2", BaseDataType.POINTER, BaseDataType.UWORD, BaseDataType.LONG)),
+    "prog8_lib_copyfloat"         to FSignature(false, emptyArray(), FParam("pointer1", BaseDataType.POINTER, BaseDataType.UWORD, BaseDataType.LONG), FParam("pointer2", BaseDataType.POINTER, BaseDataType.UWORD, BaseDataType.LONG)),
     "abs"           to FSignature(true, arrayOf(BaseDataType.UNDEFINED), FParam("value", *NumericDatatypes)),
     "abs__byte"     to FSignature(true, arrayOf(BaseDataType.UBYTE), FParam("value", BaseDataType.BYTE)),
     "abs__word"     to FSignature(true, arrayOf(BaseDataType.UWORD), FParam("value", BaseDataType.WORD)),
@@ -158,33 +158,37 @@ val BuiltinFunctions: Map<String, FSignature> = mapOf(
     "max__word"     to FSignature(true, arrayOf(BaseDataType.WORD), FParam("val1", BaseDataType.WORD), FParam("val2", BaseDataType.WORD)),
     "max__uword"    to FSignature(true, arrayOf(BaseDataType.UWORD), FParam("val1", BaseDataType.UWORD), FParam("val2", BaseDataType.UWORD)),
     "max__long"     to FSignature(true, arrayOf(BaseDataType.LONG), FParam("val1", BaseDataType.LONG), FParam("val2", BaseDataType.LONG)),
-    "peek"          to FSignature(true, arrayOf(BaseDataType.UBYTE), FParam("address", BaseDataType.UWORD)),
-    "peekbool"      to FSignature(true, arrayOf(BaseDataType.BOOL), FParam("address", BaseDataType.UWORD)),
-    "peekw"         to FSignature(true, arrayOf(BaseDataType.UWORD), FParam("address", BaseDataType.UWORD)),
-    "peekl"         to FSignature(true, arrayOf(BaseDataType.LONG), FParam("address", BaseDataType.UWORD)),
-    "peekf"         to FSignature(true, arrayOf(BaseDataType.FLOAT), FParam("address", BaseDataType.UWORD)),
-    "poke"          to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD), FParam("value", BaseDataType.UBYTE, BaseDataType.BYTE)),
-    "pokebool"      to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD), FParam("value", BaseDataType.BOOL)),
-    "pokebowl"      to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD), FParam("value", BaseDataType.BOOL)),
-    "pokew"         to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD), FParam("value", BaseDataType.UWORD, BaseDataType.WORD)),
-    "pokel"         to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD), FParam("value", BaseDataType.LONG)),
-    "pokef"         to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD), FParam("value", BaseDataType.FLOAT)),
-    "pokemon"       to FSignature(false, arrayOf(BaseDataType.UBYTE), FParam("address", BaseDataType.UWORD), FParam("value", BaseDataType.UBYTE)),
+    "peek"          to FSignature(true, arrayOf(BaseDataType.UBYTE), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER)),
+    "peekbool"      to FSignature(true, arrayOf(BaseDataType.BOOL), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER)),
+    "peekw"         to FSignature(true, arrayOf(BaseDataType.UWORD), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER)),
+    "peekl"         to FSignature(true, arrayOf(BaseDataType.LONG), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER)),
+    "peekf"         to FSignature(true, arrayOf(BaseDataType.FLOAT), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER)),
+    "peekp"         to FSignature(true, arrayOf(BaseDataType.POINTER), FParam("address", BaseDataType.POINTER, BaseDataType.UWORD, BaseDataType.LONG)),
+    "poke"          to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER), FParam("value", BaseDataType.UBYTE, BaseDataType.BYTE)),
+    "pokebool"      to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER), FParam("value", BaseDataType.BOOL)),
+    "pokebowl"      to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER), FParam("value", BaseDataType.BOOL)),
+    "pokew"         to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER), FParam("value", BaseDataType.UWORD, BaseDataType.WORD)),
+    "pokel"         to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER), FParam("value", BaseDataType.LONG)),
+    "pokef"         to FSignature(false, emptyArray(), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER), FParam("value", BaseDataType.FLOAT)),
+    "pokep"         to FSignature(false, emptyArray(), FParam("address", BaseDataType.POINTER, BaseDataType.UWORD, BaseDataType.LONG), FParam("value", BaseDataType.POINTER)),
+    "pokemon"       to FSignature(false, arrayOf(BaseDataType.UBYTE), FParam("address", BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.POINTER), FParam("value", BaseDataType.UBYTE)),
     "rsave"         to FSignature(false, emptyArray()),
     "rrestore"      to FSignature(false, emptyArray()),
-    "memory"        to FSignature(true,  arrayOf(BaseDataType.UWORD), FParam("name", BaseDataType.STR), FParam("size", BaseDataType.UWORD), FParam("alignment", BaseDataType.UWORD)),
-    "memory__ref"   to FSignature(true, arrayOf(BaseDataType.UWORD), FParam("name", BaseDataType.STR)),
-    "callfar"       to FSignature(false, arrayOf(BaseDataType.UWORD), FParam("bank", BaseDataType.UBYTE), FParam("address", BaseDataType.UWORD), FParam("arg", BaseDataType.UWORD)),
-    "callfar2"      to FSignature(false, arrayOf(BaseDataType.UWORD), FParam("bank", BaseDataType.UBYTE), FParam("address", BaseDataType.UWORD), FParam("argA", BaseDataType.UBYTE), FParam("argX", BaseDataType.UBYTE), FParam("argY", BaseDataType.UBYTE), FParam("argC", BaseDataType.BOOL)),
-    "call"          to FSignature(false, arrayOf(BaseDataType.UWORD), FParam("address", BaseDataType.UWORD)),
+    "memory"        to FSignature(true,  arrayOf(BaseDataType.POINTER), FParam("name", BaseDataType.STR), FParam("size", BaseDataType.UWORD), FParam("alignment", BaseDataType.UWORD)),
+    "memory__ref"   to FSignature(true, arrayOf(BaseDataType.POINTER), FParam("name", BaseDataType.STR)),
+    "callfar"       to FSignature(false, arrayOf(BaseDataType.UWORD), FParam("bank", BaseDataType.UBYTE), FParam("address", BaseDataType.POINTER), FParam("arg", BaseDataType.UWORD)),
+    "callfar2"      to FSignature(false, arrayOf(BaseDataType.UWORD), FParam("bank", BaseDataType.UBYTE), FParam("address", BaseDataType.POINTER), FParam("argA", BaseDataType.UBYTE), FParam("argX", BaseDataType.UBYTE), FParam("argY", BaseDataType.UBYTE), FParam("argC", BaseDataType.BOOL)),
+    "call"          to FSignature(false, arrayOf(BaseDataType.UWORD), FParam("address", BaseDataType.POINTER)),
     "push"          to FSignature(false, emptyArray(), FParam("value", BaseDataType.UBYTE, BaseDataType.BYTE, BaseDataType.BOOL)),
     "pushw"         to FSignature(false, emptyArray(), FParam("value", BaseDataType.UWORD, BaseDataType.WORD, BaseDataType.POINTER)),
     "pushl"         to FSignature(false, emptyArray(), FParam("value", BaseDataType.LONG)),
     "pushf"         to FSignature(false, emptyArray(), FParam("value", BaseDataType.FLOAT)),
+    "pushp"         to FSignature(false, emptyArray(), FParam("value", BaseDataType.POINTER)),
     "pop"           to FSignature(false, arrayOf(BaseDataType.UBYTE)),
     "popw"          to FSignature(false, arrayOf(BaseDataType.UWORD)),
     "popl"          to FSignature(false, arrayOf(BaseDataType.LONG)),
     "popf"          to FSignature(false, arrayOf(BaseDataType.FLOAT)),
+    "popp"          to FSignature(false, arrayOf(BaseDataType.POINTER)),
 )
 
 val InplaceModifyingBuiltinFunctions = setOf(

codeCore/src/prog8/code/core/CompilationOptions.kt

@@ -4,6 +4,7 @@ import java.nio.file.Path
 import kotlin.io.path.Path
 
 
+@Suppress("unused")
 class CompilationOptions(val output: OutputType,
                          val launcher: CbmPrgLauncherType,
                          val zeropage: ZeropageType,
@@ -17,6 +18,7 @@ class CompilationOptions(val output: OutputType,
                          // these are set later, based on command line arguments or options in the source code:
                          var loadAddress: UInt,
                          var memtopAddress: UInt,
+                         var varsAddress: UInt? = null,
                          var warnSymbolShadowing: Boolean = false,
                          var warnImplicitTypeCast: Boolean = false,
                          var optimize: Boolean = false,
@@ -25,16 +27,17 @@ class CompilationOptions(val output: OutputType,
                          var includeSourcelines: Boolean = false,
                          var dumpVariables: Boolean = false,
                          var dumpSymbols: Boolean = false,
+                         var newCodegen: Boolean = false,
                          var varsHighBank: Int? = null,
                          var varsGolden: Boolean = false,
-                         var slabsHighBank: Int? = null,
-                         var slabsGolden: Boolean = false,
                          var breakpointCpuInstruction: String? = null,
                          var ignoreFootguns: Boolean = false,
                          var outputDir: Path = Path(""),
                          var quiet: Boolean = false,
                          var profilingInstrumentation: Boolean = false,
-                         var symbolDefs: Map<String, String> = emptyMap()
+                         var symbolDefs: Map<String, String> = emptyMap(),
+                         var privateSymbols: Boolean = false,
+                         var outputFile: Path? = null
 ) {
     init {
         compTarget.initializeMemoryAreas(this)
@@ -57,6 +60,7 @@ class CompilationOptions(val output: OutputType,
             private var compilerVersion: String = "unknown"
             private var loadAddress: UInt = compTarget.PROGRAM_LOAD_ADDRESS
             private var memtopAddress: UInt = compTarget.PROGRAM_MEMTOP_ADDRESS
+            private var varsAddress: UInt? = null
             private var warnSymbolShadowing: Boolean = false
             private var warnImplicitTypeCast: Boolean = false
             private var optimize: Boolean = false
@@ -67,14 +71,14 @@ class CompilationOptions(val output: OutputType,
             private var dumpSymbols: Boolean = false
             private var varsHighBank: Int? = null
             private var varsGolden: Boolean = false
-            private var slabsHighBank: Int? = null
-            private var slabsGolden: Boolean = false
             private var breakpointCpuInstruction: String? = null
             private var ignoreFootguns: Boolean = false
             private var outputDir: Path = Path("")
             private var quiet: Boolean = false
             private var profilingInstrumentation: Boolean = false
+            private var newCodegen: Boolean = false
             private var symbolDefs: Map<String, String> = emptyMap()
+            private var outputFile: Path? = null
 
             fun output(output: OutputType) = apply { this.output = output }
             fun launcher(launcher: CbmPrgLauncherType) = apply { this.launcher = launcher }
@@ -87,6 +91,7 @@ class CompilationOptions(val output: OutputType,
             fun compilerVersion(compilerVersion: String) = apply { this.compilerVersion = compilerVersion }
             fun loadAddress(loadAddress: UInt) = apply { this.loadAddress = loadAddress }
             fun memtopAddress(memtopAddress: UInt) = apply { this.memtopAddress = memtopAddress }
+            fun varsAddress(varsAddress: UInt?) = apply { this.varsAddress = varsAddress }
             fun warnSymbolShadowing(warnSymbolShadowing: Boolean) = apply { this.warnSymbolShadowing = warnSymbolShadowing }
             fun warnImplicitTypeCast(warnImplicitTypeCast: Boolean) = apply { this.warnImplicitTypeCast = warnImplicitTypeCast }
             fun optimize(optimize: Boolean) = apply { this.optimize = optimize }
@@ -97,23 +102,23 @@ class CompilationOptions(val output: OutputType,
             fun dumpSymbols(dumpSymbols: Boolean) = apply { this.dumpSymbols = dumpSymbols }
             fun varsHighBank(varsHighBank: Int?) = apply { this.varsHighBank = varsHighBank }
             fun varsGolden(varsGolden: Boolean) = apply { this.varsGolden = varsGolden }
-            fun slabsHighBank(slabsHighBank: Int?) = apply { this.slabsHighBank = slabsHighBank }
-            fun slabsGolden(slabsGolden: Boolean) = apply { this.slabsGolden = slabsGolden }
             fun breakpointCpuInstruction(breakpointCpuInstruction: String?) = apply { this.breakpointCpuInstruction = breakpointCpuInstruction }
             fun ignoreFootguns(ignoreFootguns: Boolean) = apply { this.ignoreFootguns = ignoreFootguns }
             fun outputDir(outputDir: Path) = apply { this.outputDir = outputDir }
             fun quiet(quiet: Boolean) = apply { this.quiet = quiet }
             fun profilingInstrumentation(profilingInstrumentation: Boolean) = apply { this.profilingInstrumentation = profilingInstrumentation }
             fun symbolDefs(symbolDefs: Map<String, String>) = apply { this.symbolDefs = symbolDefs }
+            fun newCodegen(newCodegen: Boolean) = apply { this.newCodegen = newCodegen }
+            fun outputFile(outputFile: Path?) = apply { this.outputFile = outputFile }
 
             fun build(): CompilationOptions {
                 return CompilationOptions(
                     output, launcher, zeropage, zpReserved, zpAllowed, floats, noSysInit, romable, compTarget, compilerVersion,
-                    loadAddress, memtopAddress, warnSymbolShadowing, warnImplicitTypeCast, optimize, asmQuiet, asmListfile,
-                    includeSourcelines, dumpVariables, dumpSymbols, varsHighBank, varsGolden,
-                    slabsHighBank, slabsGolden, breakpointCpuInstruction, ignoreFootguns, outputDir, quiet,
+                    loadAddress, memtopAddress, varsAddress, warnSymbolShadowing, warnImplicitTypeCast, optimize, asmQuiet, asmListfile,
+                    includeSourcelines, dumpVariables, dumpSymbols, newCodegen, varsHighBank, varsGolden,
+                    breakpointCpuInstruction, ignoreFootguns, outputDir, quiet,
                     profilingInstrumentation, symbolDefs
-                )
+                ).apply { this.outputFile = this@Builder.outputFile }
             }
         }
     }

codeCore/src/prog8/code/core/DataTypes.kt

@@ -1,6 +1,8 @@
 package prog8.code.core
 
-import java.util.*
+import java.util.Objects
+
+
 
 
 /**
@@ -20,7 +22,6 @@ enum class BaseDataType {
     ARRAY_SPLITW,       // pass by reference, split word layout, subtype is the element type (restricted to word types)
     POINTER,            // typed pointer, subtype is whatever type is pointed to
     STRUCT_INSTANCE,    // the actual instance of a struct (not directly supported in the language yet, but we need its type)
-    ARRAY_POINTER,      // array of pointers (uwords), subtype is whatever type each element points to
     UNDEFINED;
 
 
@@ -30,7 +31,7 @@ enum class BaseDataType {
             this.isByteOrBool -> false
             this.isWord -> other.isByteOrBool
             this == LONG -> other.isByteOrBool || other.isWord
-            this == STR && other == UWORD || this == UWORD && other == STR -> false
+            this == STR && (other == UWORD || other == LONG) || this == UWORD && other == STR -> false
             this.isArray && other.isArray -> false
             this.isArray -> other.isByteOrBool
             this == STR -> other.isByteOrBool
@@ -38,14 +39,20 @@ enum class BaseDataType {
             else -> true
         }
 
-    fun equalsSize(other: BaseDataType) =
+    fun equalsSize(other: BaseDataType): Boolean = equalsSize(other, null)
+    fun equalsSize(other: BaseDataType, memsizer: IMemSizer?): Boolean =
         when {
             this == other -> true
             this.isArray && other.isArray -> true
             this.isByteOrBool -> other.isByteOrBool
-            this.isWord -> other.isWord || other.isPointer
-            this.isPointer -> other.isWord
-            this == STR && other== UWORD || this== UWORD && other== STR -> true
+            this.isWord -> other.isWord || (other.isPointer && (memsizer == null || memsizer.POINTER_MEM_SIZE == 2u))
+            this.isLong -> other.isLong || (other.isPointer && memsizer?.POINTER_MEM_SIZE != null && memsizer.POINTER_MEM_SIZE != 2u)
+            this.isPointer -> when(memsizer?.POINTER_MEM_SIZE) {
+                null -> other.isWord || other.isPointer
+                2u -> other.isWord || other.isPointer
+                else -> other.isLong || other.isPointer
+            }
+            this == STR && (other == UWORD || other == LONG) || (this == UWORD || this == LONG) && other == STR -> true
             this == STR && other.isArray -> true
             this.isArray && other == STR -> true
             else -> false
@@ -68,6 +75,15 @@ val BaseDataType.isSignedWord get() = this == BaseDataType.WORD
 val BaseDataType.isLong get() = this == BaseDataType.LONG
 val BaseDataType.isFloat get() = this == BaseDataType.FLOAT
 val BaseDataType.isInteger get() = this in setOf(BaseDataType.UBYTE, BaseDataType.BYTE, BaseDataType.UWORD, BaseDataType.WORD, BaseDataType.LONG)
+val BaseDataType.maxUnsignedValue: Double
+    get() = when (this) {
+        BaseDataType.UBYTE -> 255.0
+        BaseDataType.BYTE -> 127.0
+        BaseDataType.UWORD -> 65535.0
+        BaseDataType.WORD -> 32767.0
+        BaseDataType.LONG -> 2147483647.0
+        else -> Double.MAX_VALUE
+    }
 val BaseDataType.isIntegerOrBool get() = this in setOf(BaseDataType.UBYTE, BaseDataType.BYTE, BaseDataType.UWORD, BaseDataType.WORD, BaseDataType.LONG, BaseDataType.BOOL)
 val BaseDataType.isWordOrByteOrBool get() = this in setOf(BaseDataType.UBYTE, BaseDataType.BYTE, BaseDataType.UWORD, BaseDataType.WORD, BaseDataType.BOOL)
 val BaseDataType.isNumeric get() = this == BaseDataType.FLOAT || this.isInteger
@@ -76,12 +92,10 @@ val BaseDataType.isSigned get() = this in setOf(BaseDataType.BYTE, BaseDataType.
 val BaseDataType.isUnsigned get() = !this.isSigned
 val BaseDataType.isSignedInteger get() = this.isSigned && this.isInteger
 val BaseDataType.isUnsignedInteger get() = this.isUnsigned && this.isInteger
-val BaseDataType.isArray get() = this == BaseDataType.ARRAY || this == BaseDataType.ARRAY_SPLITW || this == BaseDataType.ARRAY_POINTER
+val BaseDataType.isArray get() = this == BaseDataType.ARRAY || this == BaseDataType.ARRAY_SPLITW
 val BaseDataType.isPointer get() = this == BaseDataType.POINTER
 val BaseDataType.isStructInstance get() = this == BaseDataType.STRUCT_INSTANCE
-val BaseDataType.isPointerArray get() = this == BaseDataType.ARRAY_POINTER
-val BaseDataType.isSplitWordArray get() = this == BaseDataType.ARRAY_SPLITW || this == BaseDataType.ARRAY_POINTER       // pointer arrays are also always stored as split uwords
-val BaseDataType.isIterable get() =  this in setOf(BaseDataType.STR, BaseDataType.ARRAY, BaseDataType.ARRAY_SPLITW, BaseDataType.ARRAY_POINTER)
+val BaseDataType.isIterable get() =  this in setOf(BaseDataType.STR, BaseDataType.ARRAY, BaseDataType.ARRAY_SPLITW)
 val BaseDataType.isPassByRef get() = this.isIterable && !this.isPointer
 val BaseDataType.isPassByValue get() = !this.isIterable || this.isPointer
 
@@ -112,18 +126,21 @@ interface ISubType {
 class DataType private constructor(
     val base: BaseDataType,
     val sub: BaseDataType?,
+    val pointeeSub: BaseDataType?,
     var subType: ISubType?,
     var subTypeFromAntlr: List<String>? = null
 ) {
 
     init {
         when {
-            base.isPointerArray -> {
-                require(sub!=null || subType!=null || subTypeFromAntlr!=null)
+            base.isArray && sub == BaseDataType.POINTER -> {
+                require(pointeeSub != null || subType != null || subTypeFromAntlr != null)
             }
             base.isArray -> {
-                require(sub != null && subType==null && subTypeFromAntlr==null)
-                if(base.isSplitWordArray)
+                require(sub != null)
+                // subType and subTypeFromAntlr allowed for arrays with struct element types
+                // (from parser creating ARRAY(LONG) directly for 32-bit targets, or from VariousCleanups conversion)
+                if(base == BaseDataType.ARRAY_SPLITW)
                     require(sub == BaseDataType.UWORD || sub == BaseDataType.WORD)
             }
             base==BaseDataType.STR -> require(sub==BaseDataType.UBYTE) { "string subtype should be ubyte" }
@@ -139,10 +156,15 @@ class DataType private constructor(
     override fun equals(other: Any?): Boolean {
         if (this === other) return true
         if (other !is DataType) return false
-        return base == other.base && sub == other.sub && (subType==other.subType || subType!!.sameas(other.subType!!))
+        if (base != other.base || sub != other.sub || pointeeSub != other.pointeeSub) return false
+        val st1 = subType
+        val st2 = other.subType
+        if (st1 == st2) return true
+        if (st1 == null || st2 == null) return false
+        return st1.sameas(st2)
     }
 
-    override fun hashCode(): Int = Objects.hash(base, sub, subType)
+    override fun hashCode(): Int = Objects.hash(base, sub, pointeeSub, subType)
 
     fun setActualSubType(actualSubType: ISubType) {
         subType = actualSubType
@@ -155,61 +177,72 @@ class DataType private constructor(
     
     companion object {
 
-        val UBYTE = DataType(BaseDataType.UBYTE, null, null)
-        val BYTE = DataType(BaseDataType.BYTE, null, null)
-        val UWORD = DataType(BaseDataType.UWORD, null, null)
-        val WORD = DataType(BaseDataType.WORD, null, null)
-        val LONG = DataType(BaseDataType.LONG, null, null)
-        val FLOAT = DataType(BaseDataType.FLOAT, null, null)
-        val BOOL = DataType(BaseDataType.BOOL, null, null)
-        val STR = DataType(BaseDataType.STR, BaseDataType.UBYTE, null)
-        val UNDEFINED = DataType(BaseDataType.UNDEFINED, null, null)
+        val UBYTE = DataType(BaseDataType.UBYTE, null, null, null)
+        val BYTE = DataType(BaseDataType.BYTE, null, null, null)
+        val UWORD = DataType(BaseDataType.UWORD, null, null, null)
+        val WORD = DataType(BaseDataType.WORD, null, null, null)
+        val LONG = DataType(BaseDataType.LONG, null, null, null)
+        val FLOAT = DataType(BaseDataType.FLOAT, null, null, null)
+        val BOOL = DataType(BaseDataType.BOOL, null, null, null)
+        val STR = DataType(BaseDataType.STR, BaseDataType.UBYTE, null, null)
+        val UNDEFINED = DataType(BaseDataType.UNDEFINED, null, null, null)
 
         private val simpletypes = mapOf(
-            BaseDataType.UBYTE to DataType(BaseDataType.UBYTE, null, null),
-            BaseDataType.BYTE to DataType(BaseDataType.BYTE, null, null),
-            BaseDataType.UWORD to DataType(BaseDataType.UWORD, null, null),
-            BaseDataType.WORD to DataType(BaseDataType.WORD, null, null),
-            BaseDataType.LONG to DataType(BaseDataType.LONG, null, null),
-            BaseDataType.FLOAT to DataType(BaseDataType.FLOAT, null, null),
-            BaseDataType.BOOL to DataType(BaseDataType.BOOL, null, null),
-            BaseDataType.STR to DataType(BaseDataType.STR, BaseDataType.UBYTE, null),
-            BaseDataType.UNDEFINED to DataType(BaseDataType.UNDEFINED, null, null)
+            BaseDataType.UBYTE to DataType(BaseDataType.UBYTE, null, null, null),
+            BaseDataType.BYTE to DataType(BaseDataType.BYTE, null, null, null),
+            BaseDataType.UWORD to DataType(BaseDataType.UWORD, null, null, null),
+            BaseDataType.WORD to DataType(BaseDataType.WORD, null, null, null),
+            BaseDataType.LONG to DataType(BaseDataType.LONG, null, null, null),
+            BaseDataType.FLOAT to DataType(BaseDataType.FLOAT, null, null, null),
+            BaseDataType.BOOL to DataType(BaseDataType.BOOL, null, null, null),
+            BaseDataType.STR to DataType(BaseDataType.STR, BaseDataType.UBYTE, null, null),
+            BaseDataType.POINTER to DataType(BaseDataType.POINTER, null, null, null),
+            BaseDataType.UNDEFINED to DataType(BaseDataType.UNDEFINED, null, null, null)
         )
 
         fun forDt(dt: BaseDataType): DataType {
             if(dt.isStructInstance)
-                TODO("cannot use struct instance as a data type (yet) - use a pointer instead")
+                return DataType(BaseDataType.STRUCT_INSTANCE, null, null, null)
             return simpletypes.getValue(dt)
         }
 
-        fun arrayFor(elementDt: BaseDataType, splitwordarray: Boolean=true): DataType {
-            require(!elementDt.isPointer) { "use other array constructor for arrays of pointers" }
-            val actualElementDt = if(elementDt==BaseDataType.STR) BaseDataType.UWORD else elementDt      // array of strings is actually just an array of UWORD pointers
-            return if(splitwordarray && actualElementDt.isWord)
-                DataType(BaseDataType.ARRAY_SPLITW, actualElementDt, null)
-            else {
-                if(actualElementDt.isNumericOrBool)
-                    DataType(BaseDataType.ARRAY, actualElementDt, null)
-                else
-                    throw NoSuchElementException("invalid basic element dt $elementDt")
-            }
+        fun arrayFor(elementDt: BaseDataType, memsizer: IMemSizer): DataType {
+            require(elementDt != BaseDataType.POINTER) { "use arrayOfPointersTo for arrays of pointers" }
+            val actualElementDt = if(elementDt==BaseDataType.STR) {
+                memsizer.pointerBaseType
+            } else elementDt
+            if(actualElementDt.isNumericOrBool)
+                return DataType(BaseDataType.ARRAY, actualElementDt, null, null)
+            else
+                throw NoSuchElementException("invalid basic element dt $elementDt")
         }
 
-        fun arrayOfPointersTo(sub: BaseDataType): DataType = DataType(BaseDataType.ARRAY_POINTER, sub, null)
-        fun arrayOfPointersTo(structType: ISubType?): DataType = DataType(BaseDataType.ARRAY_POINTER, null, structType)
-        fun arrayOfPointersFromAntlrTo(sub: BaseDataType?, identifier: List<String>?): DataType =
-            DataType(BaseDataType.ARRAY_POINTER, sub, null, identifier)
+        fun splitWordArrayFor(elementDt: BaseDataType): DataType {
+            require(elementDt.isWord) { "split word array element type must be word" }
+            return DataType(BaseDataType.ARRAY_SPLITW, elementDt, null, null)
+        }
 
-        fun pointer(base: BaseDataType): DataType = DataType(BaseDataType.POINTER, base, null)
+        fun arrayOfPointersTo(sub: BaseDataType): DataType =
+            DataType(BaseDataType.ARRAY, BaseDataType.POINTER, sub, null)
+        fun arrayOfPointersTo(structType: ISubType?): DataType =
+            DataType(BaseDataType.ARRAY, BaseDataType.POINTER, BaseDataType.STRUCT_INSTANCE, structType)
+        fun arrayOfPointersFromAntlrTo(sub: BaseDataType?, identifier: List<String>?): DataType =
+            DataType(BaseDataType.ARRAY, BaseDataType.POINTER, sub ?: BaseDataType.STRUCT_INSTANCE, null, identifier)
+
+        // array of struct instances: sub is STRUCT_INSTANCE, subType carries the resolved struct
+        fun arrayOfStructs(structType: ISubType): DataType = DataType(BaseDataType.ARRAY, BaseDataType.STRUCT_INSTANCE, null, structType)
+        fun arrayOfStructsFromAntlr(struct: List<String>): DataType =
+            DataType(BaseDataType.ARRAY, BaseDataType.STRUCT_INSTANCE, null, null, subTypeFromAntlr = struct)
+
+        fun pointer(base: BaseDataType): DataType = DataType(BaseDataType.POINTER, base, null, null)
         fun pointer(dt: DataType): DataType = if(dt.isBasic)
-                DataType(BaseDataType.POINTER, dt.base, null)
+                DataType(BaseDataType.POINTER, dt.base, null, null)
             else
-                DataType(BaseDataType.POINTER, null, dt.subType, dt.subTypeFromAntlr)
-        fun pointer(structType: ISubType): DataType = DataType(BaseDataType.POINTER, null, structType)
-        fun pointerFromAntlr(identifier: List<String>): DataType = DataType(BaseDataType.POINTER, null, null, identifier)
-        fun structInstance(type: ISubType?): DataType = DataType(BaseDataType.STRUCT_INSTANCE, sub=null, type)
-        fun structInstanceFromAntlr(struct: List<String>): DataType = DataType(BaseDataType.STRUCT_INSTANCE, null, null, subTypeFromAntlr = struct)
+                DataType(BaseDataType.POINTER, null, null, dt.subType, dt.subTypeFromAntlr)
+        fun pointer(structType: ISubType): DataType = DataType(BaseDataType.POINTER, null, null, structType)
+        fun pointerFromAntlr(identifier: List<String>): DataType = DataType(BaseDataType.POINTER, null, null, null, identifier)
+        fun structInstance(type: ISubType?): DataType = DataType(BaseDataType.STRUCT_INSTANCE, sub=null, pointeeSub=null, type)
+        fun structInstanceFromAntlr(struct: List<String>): DataType = DataType(BaseDataType.STRUCT_INSTANCE, null, null, null, subTypeFromAntlr = struct)
     }
 
 
@@ -217,53 +250,85 @@ class DataType private constructor(
     // DataType Methods
     // ============================================================================
 
-    fun elementToArray(splitwords: Boolean = true): DataType {
-        return if (base == BaseDataType.UWORD || base == BaseDataType.WORD || base == BaseDataType.STR) arrayFor(base, splitwords)
-        else if(base.isPointer) {
-            if(subType!=null) arrayOfPointersTo(subType)
-            else if(subTypeFromAntlr!=null) arrayOfPointersFromAntlrTo(sub, subTypeFromAntlr)
-            else arrayOfPointersTo(sub!!)
+    fun elementToArray(target: ICompilationTarget, splitwords: Boolean = true): DataType {
+        fun pointerArray(): DataType =
+            if (subType != null)
+                arrayOfPointersTo(subType)
+            else if (subTypeFromAntlr != null)
+                arrayOfPointersFromAntlrTo(sub, subTypeFromAntlr)
+            else
+                arrayOfPointersTo(sub!!)
+
+        if(base == BaseDataType.STRUCT_INSTANCE) {
+            val st = subType
+            return if(st != null)
+                arrayOfStructs(st)
+            else
+                arrayOfStructsFromAntlr(subTypeFromAntlr!!)
         }
-        else arrayFor(base, false)
+
+        if(base.isPointer)
+            return pointerArray()
+
+        if(!target.cpu.is6502)
+            return arrayFor(base, target)
+
+        return if (splitwords && (base == BaseDataType.UWORD || base == BaseDataType.WORD || base == BaseDataType.STR))
+            splitWordArrayFor(base)
+        else
+            arrayFor(base, target)
     }
 
     fun elementType(): DataType =
         when {
-            isPointerArray -> DataType(BaseDataType.POINTER, sub, subType)
-            base.isArray || base==BaseDataType.STR -> forDt(sub!!)
+            isPointerArray -> {
+                if (subType != null) DataType(BaseDataType.POINTER, null, null, subType)
+                else if (subTypeFromAntlr != null) DataType(BaseDataType.POINTER, null, null, null, subTypeFromAntlr)
+                else DataType(BaseDataType.POINTER, pointeeSub!!, null, null)
+            }
+            base.isArray || base==BaseDataType.STR -> {
+                if (sub == BaseDataType.STRUCT_INSTANCE && subType != null)
+                    return DataType(BaseDataType.STRUCT_INSTANCE, null, null, subType)
+                val dt = forDt(sub!!)
+                if (subType != null) DataType(BaseDataType.POINTER, null, null, subType) else dt
+            }
             else -> throw IllegalArgumentException("not an array")
         }
 
-    fun typeForAddressOf(msb: Boolean): DataType {
+    fun typeForUntypedAddressOf(msb: Boolean, mem: IMemSizer): DataType {
+        val untypedPointerType = if(mem.memorySize(BaseDataType.POINTER)>2) LONG else UWORD
         if (isUndefined)
-            return if(msb) pointer(BaseDataType.UBYTE) else UWORD
+            return if(msb) pointer(BaseDataType.UBYTE) else untypedPointerType
         else {
             if (isBasic)
                 return pointer(base)
             if (isString)
                 return pointer(BaseDataType.UBYTE)
             if (isPointer)
-                return UWORD
+                return untypedPointerType
             if (isArray) {
-                if (msb || isSplitWordArray)
+                val isSplit = base == BaseDataType.ARRAY_SPLITW ||
+                    (isPointerArray && mem.memorySize(BaseDataType.POINTER) <= 2)
+                if (msb || isSplit)
                     return pointer(BaseDataType.UBYTE)
                 val elementDt = elementType()
-                require(elementDt.isBasic)
-                return pointer(elementDt)
+                if (elementDt.isBasic)
+                    return pointer(elementDt)
+                return untypedPointerType
             }
             if (subType != null)
                 return pointer(this)
-            return UWORD
+            return untypedPointerType
         }
     }
 
     fun dereference(): DataType {
-        require(isPointer || isUnsignedWord) { "cannot dereference non-pointer type ${this}"}
+        require(isPointer || isUnsignedWord) { "cannot dereference non-pointer type $this"}
         return when {
             isUnsignedWord -> forDt(BaseDataType.UBYTE)
             sub!=null -> forDt(sub)
-            subType!=null -> DataType(BaseDataType.STRUCT_INSTANCE, null, subType)
-            subTypeFromAntlr!=null -> DataType(BaseDataType.STRUCT_INSTANCE, null, null, subTypeFromAntlr)
+            subType!=null -> DataType(BaseDataType.STRUCT_INSTANCE, null, null, subType)
+            subTypeFromAntlr!=null -> DataType(BaseDataType.STRUCT_INSTANCE, null, null, null, subTypeFromAntlr)
             else -> throw IllegalArgumentException("cannot dereference this pointer type")
         }
     }
@@ -278,6 +343,13 @@ class DataType private constructor(
                 BaseDataType.UBYTE -> "ubyte[]"
                 BaseDataType.UWORD -> "uword[]"
                 BaseDataType.LONG -> "long[]"
+                BaseDataType.POINTER -> {
+                    val pointee = if (subType != null) subType!!.scopedNameString
+                        else if (subTypeFromAntlr != null) "$subTypeFromAntlr"
+                        else pointeeSub!!.name.lowercase()
+                    "^^$pointee[]"
+                }
+                BaseDataType.STRUCT_INSTANCE -> if (subType != null) "${subType!!.scopedNameString}[]" else "$subTypeFromAntlr[]"
                 else -> throw IllegalArgumentException("invalid sub type")
             }
         }
@@ -291,9 +363,6 @@ class DataType private constructor(
         BaseDataType.POINTER -> {
             if(sub!=null) "^^${sub.name.lowercase()}" else if(subType!=null) "^^${subType!!.scopedNameString}" else "^^${subTypeFromAntlr}"
         }
-        BaseDataType.ARRAY_POINTER -> {
-            if(sub!=null) "^^${sub.name.lowercase()}[] (split)" else if (subType!=null) "^^${subType!!.scopedNameString}[] (split)" else "^^${subTypeFromAntlr}[] (split)"
-        }
         BaseDataType.STRUCT_INSTANCE -> {
             sub?.name?.lowercase() ?: if (subType!=null) subType!!.scopedNameString else "$subTypeFromAntlr"
         }
@@ -325,14 +394,6 @@ class DataType private constructor(
                 else -> "?????"
             }
         }
-        BaseDataType.ARRAY_POINTER -> {
-            when {
-                sub!=null -> "^^${sub.name.lowercase()}["
-                subType!=null -> "^^${subType!!.scopedNameString}["
-                subTypeFromAntlr!=null -> "^^${subTypeFromAntlr!!.joinToString(".")}["
-                else -> "????? ["
-            }
-        }
         BaseDataType.ARRAY -> {
             when(sub) {
                 BaseDataType.UBYTE -> "ubyte["
@@ -342,6 +403,13 @@ class DataType private constructor(
                 BaseDataType.WORD -> "@nosplit word["
                 BaseDataType.LONG -> "long["
                 BaseDataType.FLOAT -> "float["
+                BaseDataType.POINTER -> {
+                    val pointee = if (subType != null) "^^${subType!!.scopedNameString}["
+                        else if (subTypeFromAntlr != null) "^^${subTypeFromAntlr!!.joinToString(".")}["
+                        else "^^${pointeeSub!!.name.lowercase()}["
+                    pointee
+                }
+                BaseDataType.STRUCT_INSTANCE -> if (subType != null) "${subType!!.scopedNameString}[" else "$subTypeFromAntlr["
                 else -> throw IllegalArgumentException("invalid sub type")
             }
         }
@@ -364,12 +432,20 @@ class DataType private constructor(
             BaseDataType.BOOL -> targetType.base == BaseDataType.BOOL
             BaseDataType.UBYTE -> targetType.base in setOf(BaseDataType.UBYTE, BaseDataType.WORD, BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.FLOAT)
             BaseDataType.BYTE -> targetType.base in setOf(BaseDataType.BYTE, BaseDataType.WORD, BaseDataType.LONG, BaseDataType.FLOAT)
-            BaseDataType.UWORD -> targetType.base in setOf(BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.FLOAT, BaseDataType.POINTER, BaseDataType.ARRAY_POINTER)
+            BaseDataType.UWORD -> targetType.base in setOf(BaseDataType.UWORD, BaseDataType.LONG, BaseDataType.FLOAT, BaseDataType.POINTER) || targetType.isPointerArray
             BaseDataType.WORD -> targetType.base in setOf(BaseDataType.WORD, BaseDataType.LONG, BaseDataType.FLOAT)
-            BaseDataType.LONG -> targetType.base in setOf(BaseDataType.LONG, BaseDataType.FLOAT)
+            BaseDataType.LONG -> targetType.base in setOf(BaseDataType.LONG, BaseDataType.FLOAT, BaseDataType.POINTER) || targetType.isPointerArray
             BaseDataType.FLOAT -> targetType.base in arrayOf(BaseDataType.FLOAT)
-            BaseDataType.STR -> targetType.base in setOf(BaseDataType.STR, BaseDataType.UWORD) || (targetType.isPointer && targetType.sub==BaseDataType.UBYTE)
-            BaseDataType.ARRAY, BaseDataType.ARRAY_SPLITW -> targetType.base in setOf(BaseDataType.ARRAY, BaseDataType.ARRAY_SPLITW) && targetType.sub == sub
+            BaseDataType.STR -> targetType.base in setOf(BaseDataType.STR, BaseDataType.UWORD, BaseDataType.LONG) || (targetType.isPointer && targetType.sub==BaseDataType.UBYTE)
+            BaseDataType.ARRAY, BaseDataType.ARRAY_SPLITW -> {
+                if (isPointerArray) {
+                    targetType.isPointerArray && if (subType != null && targetType.subType != null) subType!!.sameas(targetType.subType!!)
+                    else if (subType == null && targetType.subType == null) pointeeSub == targetType.pointeeSub
+                    else false
+                } else {
+                    targetType.base in setOf(BaseDataType.ARRAY, BaseDataType.ARRAY_SPLITW) && targetType.sub == sub
+                }
+            }
             BaseDataType.POINTER -> {
                 when {
                     targetType.base == BaseDataType.UWORD || targetType.base == BaseDataType.LONG -> true
@@ -378,25 +454,24 @@ class DataType private constructor(
                 }
             }
             BaseDataType.STRUCT_INSTANCE -> false        // we cannot deal with actual struct instances yet in any shape or form (only getting fields from it)
-            BaseDataType.ARRAY_POINTER -> false
             BaseDataType.UNDEFINED -> false
         }
 
     fun largerSizeThan(other: DataType): Boolean = base.largerSizeThan(other.base)
     fun equalsSize(other: DataType): Boolean = base.equalsSize(other.base)
+    fun equalsSize(other: DataType, memsizer: IMemSizer): Boolean = base.equalsSize(other.base, memsizer)
 
     /**
      * Returns the memory size in bytes.
      * Note: for pointer types, size() doesn't return the size of the pointer itself
      * but the size of the thing it points to.
      */
-    fun size(memsizer: IMemSizer): Int = if(sub!=null) {
-            memsizer.memorySize(sub)
-        } else if(subType!=null) {
-            subType!!.memsize(memsizer)
-        } else {
-            memsizer.memorySize(base)
-        }
+    fun size(memsizer: IMemSizer): Int = when {
+        base == BaseDataType.ARRAY && sub == BaseDataType.POINTER -> memsizer.memorySize(BaseDataType.POINTER)
+        sub!=null -> memsizer.memorySize(sub)
+        subType!=null -> subType!!.memsize(memsizer)
+        else -> memsizer.memorySize(base)
+    }
 
     // ============================================================================
     // DataType Properties
@@ -425,24 +500,34 @@ class DataType private constructor(
     val isPointerToByte = base.isPointer && sub?.isByteOrBool==true
     val isPointerToWord = base.isPointer && sub?.isWord==true
     val isStructInstance = base.isStructInstance
-    val isPointerArray = base.isPointerArray
-    val isBoolArray = base.isArray && !base.isPointerArray && sub == BaseDataType.BOOL
-    val isByteArray = base.isArray && !base.isPointerArray && (sub == BaseDataType.UBYTE || sub == BaseDataType.BYTE)
-    val isUnsignedByteArray = base.isArray && !base.isPointerArray && sub == BaseDataType.UBYTE
-    val isSignedByteArray = base.isArray && !base.isPointerArray && sub == BaseDataType.BYTE
-    val isWordArray = base.isArray && !base.isPointerArray && (sub == BaseDataType.UWORD || sub == BaseDataType.WORD)
-    val isUnsignedWordArray = base.isArray && !base.isPointerArray && sub == BaseDataType.UWORD
-    val isSignedWordArray = base.isArray && !base.isPointerArray && sub == BaseDataType.WORD
-    val isLongArray = base.isArray && sub == BaseDataType.LONG
-    val isFloatArray = base.isArray && !base.isPointerArray && sub == BaseDataType.FLOAT
+    val isPointerArray = base == BaseDataType.ARRAY && sub == BaseDataType.POINTER
+    val isBoolArray = base == BaseDataType.ARRAY && sub == BaseDataType.BOOL
+    val isByteArray = base == BaseDataType.ARRAY && (sub == BaseDataType.UBYTE || sub == BaseDataType.BYTE)
+    val isUnsignedByteArray = base == BaseDataType.ARRAY && sub == BaseDataType.UBYTE
+    val isSignedByteArray = base == BaseDataType.ARRAY && sub == BaseDataType.BYTE
+    val isWordArray = (base == BaseDataType.ARRAY || base == BaseDataType.ARRAY_SPLITW) && (sub == BaseDataType.UWORD || sub == BaseDataType.WORD)
+    val isUnsignedWordArray = (base == BaseDataType.ARRAY || base == BaseDataType.ARRAY_SPLITW) && sub == BaseDataType.UWORD
+    val isSignedWordArray = (base == BaseDataType.ARRAY || base == BaseDataType.ARRAY_SPLITW) && sub == BaseDataType.WORD
+    val isLongArray = base == BaseDataType.ARRAY && sub == BaseDataType.LONG
+    val isFloatArray = base == BaseDataType.ARRAY && sub == BaseDataType.FLOAT
     val isString = base == BaseDataType.STR
     val isBool = base == BaseDataType.BOOL
     val isFloat = base == BaseDataType.FLOAT
     val isLong = base == BaseDataType.LONG
-    val isStringly = base == BaseDataType.STR || base == BaseDataType.UWORD || (base == BaseDataType.ARRAY && (sub == BaseDataType.UBYTE || sub == BaseDataType.BYTE))
-    val isSplitWordArray = base.isSplitWordArray
-    val isSplitUnsignedWordArray = base.isSplitWordArray && !base.isPointerArray && sub == BaseDataType.UWORD
-    val isSplitSignedWordArray = base.isSplitWordArray && !base.isPointerArray && sub == BaseDataType.WORD
+    fun isStringly(memsizer: IMemSizer): Boolean =
+        base == BaseDataType.STR ||
+        (memsizer.POINTER_MEM_SIZE == 2u && base == BaseDataType.UWORD) ||
+        (memsizer.POINTER_MEM_SIZE > 2u && base == BaseDataType.LONG) ||
+        (base == BaseDataType.ARRAY && (sub == BaseDataType.UBYTE || sub == BaseDataType.BYTE))
+    fun isSplitWordArray(memsizer: IMemSizer): Boolean  {
+        return when (base) {
+            BaseDataType.ARRAY_SPLITW -> true
+            BaseDataType.ARRAY if sub == BaseDataType.POINTER -> memsizer.POINTER_MEM_SIZE <= 2u
+            else -> false
+        }
+    }
+    val isSplitUnsignedWordArray = base == BaseDataType.ARRAY_SPLITW && sub == BaseDataType.UWORD
+    val isSplitSignedWordArray = base == BaseDataType.ARRAY_SPLITW && sub == BaseDataType.WORD
     val isIterable =  base.isIterable
     val isPassByRef = base.isPassByRef
     val isPassByValue = base.isPassByValue

codeCore/src/prog8/code/core/Enumerations.kt

@@ -4,7 +4,11 @@ package prog8.code.core
 enum class CpuRegister {
     A,
     X,
-    Y
+    Y,
+    // m68k:
+    D0, D1, D2, D3, D4, D5, D6, D7,
+    A0, A1, A2, A3, A4, A5, A6,
+    FP0, FP1, FP2, FP3, FP4, FP5, FP6, FP7
 }
 
 enum class RegisterOrPair {
@@ -20,7 +24,12 @@ enum class RegisterOrPair {
     R0, R1, R2, R3, R4, R5, R6, R7,
     R8, R9, R10, R11, R12, R13, R14, R15,
     // combined virtual registers to store 32 bits longs:
-    R0R1, R2R3, R4R5, R6R7, R8R9, R10R11, R12R13, R14R15;
+    R0R1, R2R3, R4R5, R6R7, R8R9, R10R11, R12R13, R14R15,
+    // m68k data & address registers:
+    D0, D1, D2, D3, D4, D5, D6, D7,
+    A0, A1, A2, A3, A4, A5, A6,
+    // m68k FPU registers:
+    FP0, FP1, FP2, FP3, FP4, FP5, FP6, FP7;
 
     companion object {
         val names: Set<String> = entries.map { it.toString() }.toSet()
@@ -29,6 +38,29 @@ enum class RegisterOrPair {
                 CpuRegister.A -> A
                 CpuRegister.X -> X
                 CpuRegister.Y -> Y
+                CpuRegister.D0 -> D0
+                CpuRegister.D1 -> D1
+                CpuRegister.D2 -> D2
+                CpuRegister.D3 -> D3
+                CpuRegister.D4 -> D4
+                CpuRegister.D5 -> D5
+                CpuRegister.D6 -> D6
+                CpuRegister.D7 -> D7
+                CpuRegister.A0 -> A0
+                CpuRegister.A1 -> A1
+                CpuRegister.A2 -> A2
+                CpuRegister.A3 -> A3
+                CpuRegister.A4 -> A4
+                CpuRegister.A5 -> A5
+                CpuRegister.A6 -> A6
+                CpuRegister.FP0 -> FP0
+                CpuRegister.FP1 -> FP1
+                CpuRegister.FP2 -> FP2
+                CpuRegister.FP3 -> FP3
+                CpuRegister.FP4 -> FP4
+                CpuRegister.FP5 -> FP5
+                CpuRegister.FP6 -> FP6
+                CpuRegister.FP7 -> FP7
             }
         }
     }
@@ -118,6 +150,15 @@ val CombinedLongRegisters = setOf(
     RegisterOrPair.R14R15
 )
 
+val M68kRegisters = setOf(
+    RegisterOrPair.D0, RegisterOrPair.D1, RegisterOrPair.D2, RegisterOrPair.D3,
+    RegisterOrPair.D4, RegisterOrPair.D5, RegisterOrPair.D6, RegisterOrPair.D7,
+    RegisterOrPair.A0, RegisterOrPair.A1, RegisterOrPair.A2, RegisterOrPair.A3,
+    RegisterOrPair.A4, RegisterOrPair.A5, RegisterOrPair.A6,
+    RegisterOrPair.FP0, RegisterOrPair.FP1, RegisterOrPair.FP2, RegisterOrPair.FP3,
+    RegisterOrPair.FP4, RegisterOrPair.FP5, RegisterOrPair.FP6, RegisterOrPair.FP7
+)
+
 val CpuRegisters = setOf(
     RegisterOrPair.A, RegisterOrPair.X, RegisterOrPair.Y,
     RegisterOrPair.AX, RegisterOrPair.AY, RegisterOrPair.XY
@@ -128,6 +169,8 @@ enum class OutputType {
     RAW,
     PRG,
     XEX,
+    ELF,
+    AMIGAHUNK,
     LIBRARY
 }
 

codeCore/src/prog8/code/core/ICompilationTarget.kt

@@ -35,6 +35,12 @@ enum class CpuType(val statusBitsOnMultiByteOps: Boolean) {
 
     /** 8-bit CMOS 65C02. Same status-bit issue as plain 6502. */
     CPU65C02(false),
+    
+    /** 16/32 bit Motorola 68000. has single instructions that set status bits when dealing with multi byte operands */
+    M68000(true),
+
+    /** 32 bit Motorola 68020. has single instructions that set status bits when dealing with multi byte operands */
+    M68020(true),
 
     /**
      * Virtual machine target. We use `false` here so the IR generator always
@@ -49,6 +55,14 @@ enum class CpuType(val statusBitsOnMultiByteOps: Boolean) {
     /** True for the 6502-family targets (plain 6502 and 65C02). */
     val is6502: Boolean
         get() = this == CPU6502 || this == CPU65C02
+    val is68k: Boolean
+        get() = this == M68000 || this == M68020
+    /** True for targets that use the m68k-style hardware-register calling convention (D0-D7, A0-A6, FP0-FP7). */
+    val usesM68kConvention: Boolean
+        get() = is68k || this == VIRTUAL
+    /** True for big-endian targets. The 6502 family and the VM are little-endian. */
+    val isBigEndian: Boolean
+        get() = this == M68000 || this == M68020
 }
 
 interface ICompilationTarget: IStringEncoding, IMemSizer {
@@ -56,7 +70,7 @@ interface ICompilationTarget: IStringEncoding, IMemSizer {
 
     val FLOAT_MAX_NEGATIVE: Double
     val FLOAT_MAX_POSITIVE: Double
-    val FLOAT_MEM_SIZE: UInt
+    val ARRAY_SIZE_LIMIT: UInt
     val PROGRAM_LOAD_ADDRESS : UInt
     val PROGRAM_MEMTOP_ADDRESS: UInt
     val BSSHIGHRAM_START: UInt
@@ -73,13 +87,25 @@ interface ICompilationTarget: IStringEncoding, IMemSizer {
     val defaultLauncherType: CbmPrgLauncherType
     val supportsBankedCalls: Boolean
 
+    /**
+     * True when the code generator allocates subroutine locals and per-activation intermediate values
+     * in a stack frame instead of static memory, so a subroutine on a call-graph cycle can have more
+     * than one live activation. False means two live activations would share the same storage.
+     */
+    val supportsStackFrames: Boolean
+
     fun initializeMemoryAreas(compilerOptions: CompilationOptions)
     fun getFloatAsmBytes(num: Number): String
 
     fun convertFloatToBytes(num: Double): List<UByte>
     fun convertBytesToFloat(bytes: List<UByte>): Double
 
-    fun launchEmulator(selectedEmulator: Int, programNameWithPath: Path, quiet: Boolean)
+    /**
+     * Launches an emulator to run the compiled program artifact.
+     * [programFile] is the exact artifact path including its extension; launchers must use it
+     * verbatim, without stripping or appending one.
+     */
+    fun launchEmulator(selectedEmulator: Int, programFile: Path, quiet: Boolean)
     fun isIOAddress(address: UInt): Boolean
 
     override fun encodeString(str: String, encoding: Encoding): List<UByte>

codeCore/src/prog8/code/core/IMemSizer.kt

@@ -1,15 +1,28 @@
 package prog8.code.core
 
 interface IMemSizer {
+    val FLOAT_MEM_SIZE: UInt
+    val POINTER_MEM_SIZE: UInt
+
+    val pointerType: DataType
+        get() = if(POINTER_MEM_SIZE > 2u) DataType.LONG else DataType.UWORD
+
+    val pointerBaseType: BaseDataType
+        get() = if(POINTER_MEM_SIZE > 2u) BaseDataType.LONG else BaseDataType.UWORD
+
     fun memorySize(dt: DataType, numElements: Int?): Int
 
     fun memorySize(dt: BaseDataType): Int {
         if(dt.isPassByRef)
-            return memorySize(DataType.UWORD, null)      // a pointer size
+            return POINTER_MEM_SIZE.toInt() 
         try {
             return memorySize(DataType.forDt(dt), null)
         } catch (x: NoSuchElementException) {
             throw IllegalArgumentException(x.message)
         }
     }
+
+    fun isSplitWordArray(dt: DataType): Boolean {
+        return dt.isSplitWordArray(this)
+    }
 }

codeCore/src/prog8/code/core/MemoryRegions.kt

@@ -33,6 +33,26 @@ abstract class Zeropage(options: CompilationOptions): MemoryAllocator(options) {
     val allocatedVariables = mutableMapOf<String, VarAllocation>()
     val free = mutableSetOf<UInt>()     // subclasses must set this to the appropriate free locations.
 
+
+    /**
+     * Report a warning for every scratch byte that is excluded from allocation by a %zpreserved
+     * or %zpallowed directive. The compiler's code generators always use these fixed scratch
+     * locations internally (P8ZP_SCRATCH_*), so a %zpreserved/%zpallowed range that excludes one
+     * of them would let the compiler overwrite memory the user explicitly wants to protect.
+     */
+    fun checkScratchConflicts(errors: IErrorReporter) {
+        if(options.zeropage== ZeropageType.DONTUSE)
+            return
+
+        val scratchBytes = listOf(SCRATCH_B1, SCRATCH_REG, SCRATCH_W1, SCRATCH_W1 + 1u, SCRATCH_W2, SCRATCH_W2 + 1u, SCRATCH_PTR, SCRATCH_PTR + 1u)
+        for(byte in scratchBytes) {
+            val preserved = options.zpReserved.any { byte in it }
+            val allowed = options.zpAllowed.any { byte in it }
+            if(preserved || !allowed)
+                errors.warn("zero-page address $${byte.toString(16)} is a fixed scratch location used by the compiler (P8ZP_SCRATCH_*), but it is excluded by a %zpreserved or %zpallowed directive", Position.DUMMY)
+        }
+    }
+
     fun removeReservedFromFreePool() {
         for (reserved in options.zpReserved)
             reserve(reserved)
@@ -46,12 +66,9 @@ abstract class Zeropage(options: CompilationOptions): MemoryAllocator(options) {
     }
 
     fun availableBytes() = if(options.zeropage== ZeropageType.DONTUSE) 0 else free.size
-    fun hasByteAvailable() = if(options.zeropage== ZeropageType.DONTUSE) false else free.isNotEmpty()
+    fun hasByteAvailable() = options.zeropage != ZeropageType.DONTUSE && free.isNotEmpty()
     fun hasWordAvailable(): Boolean {
-        if(options.zeropage== ZeropageType.DONTUSE)
-            return false
-
-        return free.windowed(2).any { it[0] == it[1] - 1u }
+        return options.zeropage != ZeropageType.DONTUSE && free.windowed(2).any { it[0] == it[1] - 1u }
     }
 
     override fun allocate(name: String,

codeCore/src/prog8/code/core/OutputFiles.kt

@@ -0,0 +1,80 @@
+package prog8.code.core
+
+import prog8.code.target.VMTarget
+import java.nio.file.Path
+import kotlin.io.path.extension
+import kotlin.io.path.name
+
+
+/** VICE/box16 monitor (symbol) list file name for an artifact stem. */
+fun viceMonListName(stem: String) = "$stem.vice-mon-list"
+
+
+/**
+ * Artifact stem of a user-supplied `-o` output file name.
+ * A dot at index 0 is not an extension separator: ".pgz" keeps its whole name.
+ * Only applied to `-o`; the legacy program stem (source name without extension) is used verbatim.
+ */
+fun outputStemOf(name: Path): String {
+    val filename = name.name
+    val dotIndex = filename.lastIndexOf('.')
+    return if (dotIndex <= 0) filename else filename.substring(0, dotIndex)
+}
+
+/**
+ * The final program artifact path: [CompilationOptions.outputFile] verbatim when set,
+ * otherwise [defaultName] plus the extension for [CompilationOptions.output], in [CompilationOptions.outputDir].
+ * RAW and LIBRARY both use ".bin"; AMIGAHUNK has no extension.
+ * On the virtual target the artifact is the IR file, so ".p8ir" is always enforced on -o.
+ */
+fun CompilationOptions.programFile(defaultName: String): Path {
+    // VMTarget's defaultOutputType = PRG is misleading: its artifact is the .p8ir file, never PRG.
+    // The VM identifies the IR by its extension, so -o is not verbatim here: ".p8ir" is enforced.
+    // Only the exact lowercase extension is recognized (same rule as -vm in CompilerMain).
+    if(compTarget is VMTarget) {
+        val ir = outputFile ?: return outputDir.resolve("$defaultName.p8ir")
+        return if (ir.extension=="p8ir") ir else ir.resolveSibling("${ir.name}.p8ir")
+    }
+    outputFile?.let { return it }
+    val filename = when(output) {
+        OutputType.PRG -> "$defaultName.prg"
+        OutputType.XEX -> "$defaultName.xex"
+        OutputType.RAW -> "$defaultName.bin"
+        OutputType.LIBRARY -> "$defaultName.bin"
+        OutputType.ELF -> "$defaultName.elf"
+        OutputType.AMIGAHUNK -> defaultName
+    }
+    return outputDir.resolve(filename)
+}
+
+/**
+ * Output artifact naming.
+ * Invariant: [asm] must resolve into the same directory as `CompilationOptions.outputDir`,
+ * because 64tass resolves `.binary` includes relative to the generated `.asm`.
+ */
+class OutputFiles(val programFile: Path, val stem: String, val dir: Path) {
+    fun asm() = dir.resolve("$stem.asm")
+    fun list() = dir.resolve("$stem.list")
+    fun viceMonList() = dir.resolve(viceMonListName(stem))
+    fun binFile() = dir.resolve("$stem.bin")
+    fun ir() = dir.resolve("$stem.p8ir")
+    fun bankedCalls() = dir.resolve("$stem.bankedcalls")
+    fun obj() = dir.resolve("$stem.o")
+    fun linkScript() = dir.resolve("$stem.link.ld")
+
+    companion object {
+        /**
+         * Builds the output file set. Without `-o` the stem is [defaultName] verbatim
+         * (outputStemOf is only applied to the `-o` name). With `-o` the stem is derived
+         * from the resolved artifact name, so it also covers the virtual target appending
+         * ".p8ir". [dir] is the program file's parent, or [CompilationOptions.outputDir]
+         * for a bare -o filename.
+         */
+        fun of(options: CompilationOptions, defaultName: String): OutputFiles {
+            val programFile = options.programFile(defaultName)
+            val stem = options.outputFile?.let { outputStemOf(programFile) } ?: defaultName
+            val dir = programFile.parent ?: options.outputDir
+            return OutputFiles(programFile, stem, dir)
+        }
+    }
+}

codeCore/src/prog8/code/cpu6502/ConstMultiply.kt

@@ -0,0 +1,80 @@
+package prog8.code.cpu6502
+
+// Horner ("shift-add") expansion of a multiply by a small compile-time constant, for the 6502 and
+// 65C02 backends. The runtime routines prog8_math.multiply_words / multiply_bytes cost the same
+// whatever the constant is (~442 and ~130 cycles), while this expansion costs O(bit length), so
+// for a small constant inlining wins. It generalises the hand-written prog8_math.mul_word_N /
+// mul_byte_N routines, which are themselves unrolled Horner chains (mul_word_3 is "AY = AY*2 + AY",
+// one Horner step for 11).
+//
+// The sequences live here rather than in either backend because the register bookkeeping is easy to
+// get subtly wrong and the two backends must not drift apart.
+//
+// The gate is on generated size, not on the constant's magnitude: cycles alone would argue for
+// expanding nearly every multiply and bloating every program. Budgets are the max instruction count
+// the expansion may take; the call sequence each replaces is 6 instructions.
+
+private const val WORD_BUDGET = 32
+private const val BYTE_BUDGET = 24
+
+// --- byte ---
+// Contract: A holds the multiplicand. Result in A. Only A and one scratch byte are live, so there
+// is no second register to keep in sync (unlike the word case, where a bare doubling leaves the
+// high byte stale in Y).
+
+fun byteShiftAddExpansion(value: Int): List<String>? {
+    if (value !in 3..255)
+        return null
+    val bitlen = 31 - Integer.numberOfLeadingZeros(value)
+    val popcount = Integer.bitCount(value)
+    if (1 + (bitlen-1) + 2*(popcount-1) > BYTE_BUDGET)
+        return null
+    val result = ArrayList<String>(1 + 3*bitlen)
+    result.add("sta  P8ZP_SCRATCH_REG")                     // keep x for the conditional adds
+    for (bit in (bitlen-1) downTo 0) {
+        result.add("asl  a")                                // r = r*2
+        if (value and (1 shl bit) != 0) {                   // r += x
+            result.add("clc")
+            result.add("adc  P8ZP_SCRATCH_REG")
+        }
+    }
+    return result
+}
+
+// --- word ---
+// Contract on entry: A = r.lo, Y = r.hi, P8ZP_SCRATCH_W1 = r, P8ZP_SCRATCH_W2 = x (all 16-bit,
+// little-endian: byte 0 is the low byte). On exit: A = result.lo, Y = result.hi.
+//
+// Two things this sequence must get right, both learned the hard way:
+//  * the add step must store the new high byte back to W1+1 as well as to Y. Writing it only to Y
+//    leaves W1+1 stale, and the next step's 'rol P8ZP_SCRATCH_W1+1' would then double the wrong
+//    high byte.
+//  * a trailing bare doubling (any even constant) updates A and W1+1 but not Y, so Y needs an
+//    explicit reload at the end. It cannot be 'tay', because at that point A holds the low byte.
+
+fun wordShiftAddExpansion(value: Int): List<String>? {
+    if (value < 3)
+        return null
+    val bitlen = 31 - Integer.numberOfLeadingZeros(value)
+    val popcount = Integer.bitCount(value)
+    if (1 + 2*(bitlen-1) + 7*(popcount-1) > WORD_BUDGET)
+        return null
+    val result = ArrayList<String>(2 + 9*bitlen)
+    for (bit in (bitlen-1) downTo 0) {
+        result.add("asl  a")                                // r = r*2
+        result.add("rol  P8ZP_SCRATCH_W1+1")
+        if (value and (1 shl bit) != 0) {                   // r += x
+            result.add("clc")
+            result.add("adc  P8ZP_SCRATCH_W2")
+            result.add("sta  P8ZP_SCRATCH_W1")
+            result.add("lda  P8ZP_SCRATCH_W1+1")
+            result.add("adc  P8ZP_SCRATCH_W2+1")
+            result.add("sta  P8ZP_SCRATCH_W1+1")
+            result.add("tay")
+            result.add("lda  P8ZP_SCRATCH_W1")
+        }
+    }
+    if (value and 1 == 0)
+        result.add("ldy  P8ZP_SCRATCH_W1+1")
+    return result
+}

codeCore/src/prog8/code/target/Amiga1200Target.kt

@@ -0,0 +1,86 @@
+package prog8.code.target
+
+import prog8.code.core.*
+import prog8.code.target.encodings.Encoder
+import prog8.code.target.zp.M68kZeropage
+import java.nio.file.Path
+
+
+class Amiga1200Target: ICompilationTarget,
+    IStringEncoding by Encoder(false),
+    IMemSizer by NormalMemSizer(4u, 4u) {
+
+    override val name = NAME
+    override val supportsBankedCalls = true
+    override val supportsStackFrames = true
+    override val defaultEncoding = Encoding.ISO
+    override val libraryPath = null
+    override val customLauncher = emptyList<String>()
+    override val additionalAssemblerOptions = emptyList<String>()
+    override val defaultOutputType = OutputType.AMIGAHUNK
+    override val defaultLauncherType = CbmPrgLauncherType.NONE
+
+    companion object {
+        const val NAME = "amiga1200"
+
+        /** needed to map bank numbers to libraries in extsub/CALLFAR library calls */
+        val LibraryNumbers = Amiga500Target.LibraryNumbers
+    }
+
+    override val cpu = CpuType.M68020
+
+    override val FLOAT_MAX_POSITIVE = Float.MAX_VALUE.toDouble()
+    override val FLOAT_MAX_NEGATIVE = -Float.MAX_VALUE.toDouble()
+    override val FLOAT_MEM_SIZE = 4u
+    override val POINTER_MEM_SIZE = 4u
+    override val ARRAY_SIZE_LIMIT = 32768u
+    override val PROGRAM_LOAD_ADDRESS = 0u      // amiga executables are relocatable
+    override val PROGRAM_MEMTOP_ADDRESS = 0xfffffffeu       // not actually used
+
+    override val BSSHIGHRAM_START = 0u          // not used
+    override val BSSHIGHRAM_END = 0u            // not used
+    override val BSSGOLDENRAM_START = 0u        // not used
+    override val BSSGOLDENRAM_END = 0u          // not used
+    override lateinit var zeropage: Zeropage
+
+    override fun getFloatAsmBytes(num: Number): String {
+        TODO("float asm bytes")
+    }
+
+    override fun convertFloatToBytes(num: Double): List<UByte> {
+        TODO("convert float to bytes")
+    }
+
+    override fun convertBytesToFloat(bytes: List<UByte>): Double {
+        require(bytes.size==4) { "need 4 bytes" }
+        TODO("convert bytes to float")
+    }
+
+    override fun launchEmulator(selectedEmulator: Int, programFile: Path, quiet: Boolean) {
+        if (selectedEmulator == 1) {
+            CopperlineEmulator.launch(programFile, quiet, "A1200", "68020", listOf("--fast", "2M", "--fpu"))
+            return
+        }
+        if (selectedEmulator != 2) {
+            System.err.println("The amiga1200 target supports FS-UAE (-emu) and Amitools's vamos (-emu2).")
+            return
+        }
+        val cpuStr = when (this.cpu) {
+            CpuType.M68000 -> "68000"
+            CpuType.M68020 -> "68020"
+            else -> error("invalid cpu type")
+        }
+        VamosEmulator.launch(programFile, quiet, cpuStr)
+        return
+    }
+
+    override fun isIOAddress(address: UInt): Boolean =
+        address in 0xbfd000u..0xbfdfffu ||             // CIA-B
+        address in 0xbfe000u..0xbfefffu ||             // CIA-A
+        address in 0xdff000u..0xdff1ffu ||             // Custom chipset (Paula, Alice, Lisa, Denise)
+        address in 0xe80000u..0xefffffu                // Zorro II autoconfig space
+
+    override fun initializeMemoryAreas(compilerOptions: CompilationOptions) {
+        zeropage = M68kZeropage(compilerOptions)
+    }
+}

codeCore/src/prog8/code/target/Amiga500Target.kt

@@ -0,0 +1,126 @@
+package prog8.code.target
+
+import prog8.code.core.*
+import prog8.code.target.encodings.Encoder
+import prog8.code.target.zp.M68kZeropage
+import java.nio.file.Path
+
+
+class Amiga500Target: ICompilationTarget,
+    IStringEncoding by Encoder(false),
+    IMemSizer by NormalMemSizer(4u, 4u) {
+
+    override val name = NAME
+    override val supportsBankedCalls = true
+    override val supportsStackFrames = true
+    override val defaultEncoding = Encoding.ISO
+    override val libraryPath = null
+    override val customLauncher = emptyList<String>()
+    override val additionalAssemblerOptions = emptyList<String>()
+    override val defaultOutputType = OutputType.AMIGAHUNK
+    override val defaultLauncherType = CbmPrgLauncherType.NONE
+
+    companion object {
+        const val NAME = "amiga500"
+        
+        /** needed to map bank numbers to libraries in extsub/CALLFAR library calls */
+        val LibraryNumbers = mapOf(
+             1 to "Sys",        // exec
+             2 to "DOS",
+             3 to "Gfx",
+             4 to "Intuition",
+             5 to "GadTools",
+             6 to "Layers",
+             7 to "Asl",
+             8 to "ConsoleDevice",
+             9 to "Utility",
+            10 to "Expansion",
+            11 to "Icon",
+            12 to "Workbench",
+            13 to "Diskfont",
+            14 to "IFFParse",
+            15 to "Input",
+            16 to "Keymap",
+            17 to "Locale",
+            18 to "Timer",
+            19 to "LowLevel",
+            20 to "Math",
+            21 to "MathIeeeSingBas",
+            22 to "MathIeeeSingTrans",
+            23 to "MathIeeeDoubBas",
+            24 to "MathIeeeDoubTrans",
+            25 to "MathTrans",
+            26 to "NV",
+            27 to "RealTime",
+            28 to "Translator",
+            29 to "RexxSys",
+            30 to "Cx",
+            31 to "DataTypes",
+            32 to "Disk",
+            33 to "AmigaGuide",
+            34 to "ARexx",
+            35 to "BattClock",
+            36 to "BattMem",
+            // from here, non-standard libraries are listed
+            100 to "AsyncIO",
+            101 to "AHI",        // ahi.device base (amiga1200 target only, 68020+, used by the low-level ahi.* LVOs)
+        )
+    }
+
+    override val cpu = CpuType.M68000
+
+    override val FLOAT_MAX_POSITIVE = Float.MAX_VALUE.toDouble()
+    override val FLOAT_MAX_NEGATIVE = -Float.MAX_VALUE.toDouble()
+    override val FLOAT_MEM_SIZE = 4u
+    override val POINTER_MEM_SIZE = 4u
+    override val ARRAY_SIZE_LIMIT = 32768u
+    override val PROGRAM_LOAD_ADDRESS = 0u      // amiga executables are relocatable
+    override val PROGRAM_MEMTOP_ADDRESS = 0xfffffffeu       // not actually used
+
+    override val BSSHIGHRAM_START = 0u          // not used
+    override val BSSHIGHRAM_END = 0u            // not used
+    override val BSSGOLDENRAM_START = 0u        // not used    
+    override val BSSGOLDENRAM_END = 0u          // not used
+    override lateinit var zeropage: Zeropage
+
+    override fun getFloatAsmBytes(num: Number): String {
+        TODO("float asm bytes")
+    }
+
+    override fun convertFloatToBytes(num: Double): List<UByte> {
+        TODO("convert float to bytes")
+    }
+
+    override fun convertBytesToFloat(bytes: List<UByte>): Double {
+        require(bytes.size==4) { "need 4 bytes" }
+        TODO("convert bytes to float")
+    }
+
+    override fun launchEmulator(selectedEmulator: Int, programFile: Path, quiet: Boolean) {
+        if (selectedEmulator == 1) {
+            CopperlineEmulator.launch(programFile, quiet, "A500", "68000", listOf("--slow", "512K"))
+            return
+        }
+        if (selectedEmulator != 2) {
+            System.err.println("The amiga500 target supports FS-UAE (-emu) and Amitools's vamos (-emu2).")
+            return
+        }
+        val cpuStr = when (this.cpu) {
+            CpuType.M68000 -> "68000"
+            CpuType.M68020 -> "68020"
+            else -> error("invalid cpu type")
+        }
+        VamosEmulator.launch(programFile, quiet, cpuStr)
+        return
+    }
+
+    override fun isIOAddress(address: UInt): Boolean =
+        address in 0xbfd000u..0xbfdfffu ||             // CIA-B
+        address in 0xbfe000u..0xbfefffu ||             // CIA-A
+        address in 0xdff000u..0xdff1ffu ||             // Custom chipset (Paula, Alice, Lisa, Denise)
+        address in 0xe80000u..0xefffffu                // Zorro II autoconfig space
+
+    override fun initializeMemoryAreas(compilerOptions: CompilationOptions) {
+        zeropage = M68kZeropage(compilerOptions)
+    }
+}

codeCore/src/prog8/code/target/C128Target.kt

@@ -8,10 +8,11 @@ import java.nio.file.Path
 
 class C128Target: ICompilationTarget,
     IStringEncoding by Encoder(true),
-    IMemSizer by NormalMemSizer(Mflpt5.FLOAT_MEM_SIZE) {
+    IMemSizer by NormalMemSizer(Mflpt5.FLOAT_MEM_SIZE, 2u) {
 
     override val name = NAME
     override val supportsBankedCalls = true
+    override val supportsStackFrames = false
     override val defaultEncoding = Encoding.PETSCII
     override val libraryPath = null
     override val customLauncher = emptyList<String>()
@@ -28,7 +29,9 @@ class C128Target: ICompilationTarget,
 
     override val FLOAT_MAX_POSITIVE = Mflpt5.FLOAT_MAX_POSITIVE
     override val FLOAT_MAX_NEGATIVE = Mflpt5.FLOAT_MAX_NEGATIVE
-    override val FLOAT_MEM_SIZE = Mflpt5.FLOAT_MEM_SIZE.toUInt()
+    override val FLOAT_MEM_SIZE = Mflpt5.FLOAT_MEM_SIZE
+    override val POINTER_MEM_SIZE = 2u
+    override val ARRAY_SIZE_LIMIT = 256u
     override val PROGRAM_LOAD_ADDRESS = 0x1c01u
     override val PROGRAM_MEMTOP_ADDRESS = 0xc000u
 
@@ -52,7 +55,7 @@ class C128Target: ICompilationTarget,
         return m5.toDouble()
     }
 
-    override fun launchEmulator(selectedEmulator: Int, programNameWithPath: Path, quiet: Boolean) {
+    override fun launchEmulator(selectedEmulator: Int, programFile: Path, quiet: Boolean) {
         if(selectedEmulator!=1) {
             System.err.println("The c128 target only supports the main emulator (Vice).")
             return
@@ -61,9 +64,9 @@ class C128Target: ICompilationTarget,
         if(!quiet)
             println("\nStarting C-128 emulator x128...")
 
-        val viceMonlist = C64Target.viceMonListName(programNameWithPath.toString())
-        val cmdline = listOf("x128", "-silent", "-moncommands", viceMonlist,
-            "-autostartprgmode", "1", "-autostart-warp", "-autostart", "${programNameWithPath}.prg")
+        val viceMonlist = programFile.resolveSibling(viceMonListName(outputStemOf(programFile)))
+        val cmdline = listOf("x128", "-silent", "-moncommands", viceMonlist.toString(),
+            "-autostartprgmode", "1", "-autostart-warp", "-autostart", programFile.toString())
         val processb = ProcessBuilder(cmdline)
         if(!quiet)
             processb.inheritIO()

codeCore/src/prog8/code/target/C64Target.kt

@@ -9,10 +9,11 @@ import java.nio.file.Path
 
 class C64Target: ICompilationTarget,
     IStringEncoding by Encoder(true),
-    IMemSizer by NormalMemSizer(Mflpt5.FLOAT_MEM_SIZE) {
+    IMemSizer by NormalMemSizer(Mflpt5.FLOAT_MEM_SIZE, 2u) {
 
     override val name = NAME
     override val supportsBankedCalls = true
+    override val supportsStackFrames = false
     override val defaultEncoding = Encoding.PETSCII
     override val libraryPath = null
     override val customLauncher = emptyList<String>()
@@ -22,8 +23,6 @@ class C64Target: ICompilationTarget,
 
     companion object {
         const val NAME = "c64"
-
-        fun viceMonListName(baseFilename: String) = "$baseFilename.vice-mon-list"
     }
 
 
@@ -31,7 +30,9 @@ class C64Target: ICompilationTarget,
 
     override val FLOAT_MAX_POSITIVE = Mflpt5.FLOAT_MAX_POSITIVE
     override val FLOAT_MAX_NEGATIVE = Mflpt5.FLOAT_MAX_NEGATIVE
-    override val FLOAT_MEM_SIZE = Mflpt5.FLOAT_MEM_SIZE.toUInt()
+    override val FLOAT_MEM_SIZE = Mflpt5.FLOAT_MEM_SIZE
+    override val POINTER_MEM_SIZE = 2u
+    override val ARRAY_SIZE_LIMIT = 256u
     override val PROGRAM_LOAD_ADDRESS = 0x0801u
     override val PROGRAM_MEMTOP_ADDRESS = 0xcfe0u      // $a000  if floats are used
     // note that at $cfe0-$cfff are the 16 'virtual registers' R0-R15
@@ -56,7 +57,7 @@ class C64Target: ICompilationTarget,
         return m5.toDouble()
     }
 
-    override fun launchEmulator(selectedEmulator: Int, programNameWithPath: Path, quiet: Boolean) {
+    override fun launchEmulator(selectedEmulator: Int, programFile: Path, quiet: Boolean) {
         if(selectedEmulator!=1) {
             System.err.println("The c64 target only supports the main emulator (Vice).")
             return
@@ -66,9 +67,9 @@ class C64Target: ICompilationTarget,
             if(!quiet)
                 println("\nStarting C-64 emulator $emulator...")
 
-            val viceMonlist = viceMonListName(programNameWithPath.toString())
-            val cmdline = listOf(emulator, "-silent", "-moncommands", viceMonlist,
-                "-autostartprgmode", "1", "-autostart-warp", "-autostart", "${programNameWithPath}.prg")
+            val viceMonlist = programFile.resolveSibling(viceMonListName(outputStemOf(programFile)))
+            val cmdline = listOf(emulator, "-silent", "-moncommands", viceMonlist.toString(),
+                "-autostartprgmode", "1", "-autostart-warp", "-autostart", programFile.toString())
             val processb = ProcessBuilder(cmdline)
             if(!quiet)
                 processb.inheritIO()
@@ -97,6 +98,9 @@ val CompilationTargets = listOf(
     C128Target.NAME,
     Cx16Target.NAME,
     PETTarget.NAME,
+    Qemu68kTarget.NAME,
+    Amiga500Target.NAME,
+    Amiga1200Target.NAME,
     VMTarget.NAME
 )
 
@@ -105,6 +109,9 @@ fun getCompilationTargetByName(name: String) = when(name.lowercase()) {
     C128Target.NAME -> C128Target()
     Cx16Target.NAME -> Cx16Target()
     PETTarget.NAME -> PETTarget()
+    Qemu68kTarget.NAME -> Qemu68kTarget()
+    Amiga500Target.NAME -> Amiga500Target()
+    Amiga1200Target.NAME -> Amiga1200Target()
     VMTarget.NAME -> VMTarget()
     else -> throw IllegalArgumentException("invalid compilation target")
 }

codeCore/src/prog8/code/target/ConfigFileTarget.kt

@@ -4,6 +4,7 @@ import prog8.code.core.*
 import prog8.code.source.ImportFileSystem.expandTilde
 import prog8.code.target.encodings.Encoder
 import prog8.code.target.zp.ConfigurableZeropage
+import prog8.code.target.zp.M68kZeropage
 import java.io.IOException
 import java.nio.file.Path
 import java.util.*
@@ -18,6 +19,7 @@ class ConfigFileTarget(
     override val supportsBankedCalls: Boolean,
     override val defaultEncoding: Encoding,
     override val cpu: CpuType,
+    override val supportsStackFrames: Boolean,
     override val PROGRAM_LOAD_ADDRESS: UInt,
     override val PROGRAM_MEMTOP_ADDRESS: UInt,
     override val BSSHIGHRAM_START: UInt,
@@ -29,6 +31,7 @@ class ConfigFileTarget(
     override val libraryPath: Path,
     override val customLauncher: List<String>,
     override val additionalAssemblerOptions: List<String>,
+    val varsAddress: UInt? = null,
     val ioAddresses: List<UIntRange>,
     val zpScratchB1: UInt,
     val zpScratchReg: UInt,
@@ -38,8 +41,13 @@ class ConfigFileTarget(
     val virtualregistersStart: UInt,
     val zpFullsafe: List<UIntRange>,
     val zpKernalsafe: List<UIntRange>,
-    val zpBasicsafe: List<UIntRange>
-): ICompilationTarget, IStringEncoding by Encoder(true), IMemSizer by NormalMemSizer(8) {
+    val zpBasicsafe: List<UIntRange>,
+    val pointerSize: Int = 2
+): ICompilationTarget, IStringEncoding by Encoder(true), IMemSizer by NormalMemSizer(8u, pointerSize.toUInt()) {
+
+    init {
+        require(pointerSize == 2 || pointerSize == 4) { "unsupported pointer size: $pointerSize" }
+    }
 
     companion object {
 
@@ -66,6 +74,12 @@ class ConfigFileTarget(
             return value.toUInt()
         }
 
+        private fun parsePointerSize(props: Properties): Int {
+            val pointerSize = parseInt(props.getProperty("pointer_size", "2")).toInt()
+            require(pointerSize == 2 || pointerSize == 4) { "unsupported pointer size: $pointerSize" }
+            return pointerSize
+        }
+
         private fun parseAddressRanges(key: String, props: Properties): List<UIntRange> {
             val rangesStr = props.getString(key)
             if(rangesStr.isBlank())
@@ -97,9 +111,12 @@ class ConfigFileTarget(
                 CpuType.valueOf("CPU$cpuString")
             }
             val ioAddresses = parseAddressRanges("io_regions", props)
-            val zpFullsafe = parseAddressRanges("zp_fullsafe", props)
-            val zpKernalsafe = parseAddressRanges("zp_kernalsafe", props)
-            val zpBasicsafe = parseAddressRanges("zp_basicsafe", props)
+            // The m68k has no zero page, so the zeropage settings below don't apply to an m68k
+            // target and are not required in its config file.
+            fun zpSetting(name: String) = if(cpuType.is68k) 0u else props.getInteger(name)
+            val zpFullsafe = if(cpuType.is68k) emptyList() else parseAddressRanges("zp_fullsafe", props)
+            val zpKernalsafe = if(cpuType.is68k) emptyList() else parseAddressRanges("zp_kernalsafe", props)
+            val zpBasicsafe = if(cpuType.is68k) emptyList() else parseAddressRanges("zp_basicsafe", props)
 
             val libraryPath = expandTilde(Path(props.getString("library")))
             if(!libraryPath.isDirectory())
@@ -111,6 +128,7 @@ class ConfigFileTarget(
                     (customLauncherStr+"\n").lines().map { it.trimEnd() }
                 else emptyList()
             val assemblerOptionsStr = props.getProperty("assembler_options", "").trim()
+            val pointerSize = parsePointerSize(props)
             val outputTypeString = props.getProperty("output_type", "PRG")
             val defaultOutputType = OutputType.valueOf(outputTypeString.uppercase())
             val launcherString = props.getProperty("launcher", "BASIC")
@@ -120,11 +138,14 @@ class ConfigFileTarget(
                 CbmPrgLauncherType.BASIC
             }
 
+            val varsAddress = props.getProperty("vars_address")?.let { parseInt(it) }
+
             return ConfigFileTarget(
                 configfile.nameWithoutExtension,
                 props.getProperty("supports_banked_calls", "false").toBoolean(),
                 Encoding.entries.first { it.prefix==props.getString("encoding") },
                 cpuType,
+                props.getProperty("supports_stack_frames", cpuType.is68k.toString()).toBoolean(),
                 props.getInteger("load_address"),
                 props.getInteger("memtop"),
                 props.getInteger("bss_highram_start"),
@@ -136,16 +157,18 @@ class ConfigFileTarget(
                 libraryPath,
                 customLauncher,
                 if(assemblerOptionsStr=="") emptyList() else assemblerOptionsStr.split(" "),
+                varsAddress,
                 ioAddresses,
-                props.getInteger("zp_scratch_b1"),
-                props.getInteger("zp_scratch_reg"),
-                props.getInteger("zp_scratch_w1"),
-                props.getInteger("zp_scratch_w2"),
-                props.getInteger("zp_scratch_ptr"),
-                props.getInteger("virtual_registers"),
+                zpSetting("zp_scratch_b1"),
+                zpSetting("zp_scratch_reg"),
+                zpSetting("zp_scratch_w1"),
+                zpSetting("zp_scratch_w2"),
+                zpSetting("zp_scratch_ptr"),
+                zpSetting("virtual_registers"),
                 zpFullsafe,
                 zpKernalsafe,
                 zpBasicsafe,
+                pointerSize,
             )
         }
     }
@@ -154,20 +177,22 @@ class ConfigFileTarget(
     override val FLOAT_MAX_POSITIVE = 9.999999999e97
     override val FLOAT_MAX_NEGATIVE = -9.999999999e97
     override val FLOAT_MEM_SIZE = 8u
-
+    override val POINTER_MEM_SIZE = pointerSize.toUInt()
+    override val ARRAY_SIZE_LIMIT = 256u
     override lateinit var zeropage: Zeropage
 
     override fun getFloatAsmBytes(num: Number) = TODO("floats")
     override fun convertFloatToBytes(num: Double): List<UByte> = TODO("floats")
     override fun convertBytesToFloat(bytes: List<UByte>): Double = TODO("floats")
-    override fun launchEmulator(selectedEmulator: Int, programNameWithPath: Path, quiet: Boolean) {
+    override fun launchEmulator(selectedEmulator: Int, programFile: Path, quiet: Boolean) {
         throw IllegalArgumentException("Custom compiler target cannot automatically launch an emulator. Do this manually.")
     }
 
     override fun isIOAddress(address: UInt): Boolean = ioAddresses.any { address in it }
 
     override fun initializeMemoryAreas(compilerOptions: CompilationOptions) {
-        zeropage = ConfigurableZeropage(
+        // the m68k has no zero page, so the zp scratch addresses in the config file don't apply there
+        zeropage = if(cpu.is68k) M68kZeropage(compilerOptions) else ConfigurableZeropage(
             zpScratchB1, zpScratchReg, zpScratchW1, zpScratchW2, zpScratchPtr,
             virtualregistersStart,
             zpBasicsafe,

codeCore/src/prog8/code/target/CopperlineEmulator.kt

@@ -0,0 +1,71 @@
+package prog8.code.target
+
+import java.io.File
+import java.nio.file.Path
+import java.nio.file.Paths
+import kotlin.io.path.copyTo
+import kotlin.io.path.createDirectories
+
+object CopperlineEmulator {
+    fun launch(programFile: Path, quiet: Boolean, model: String, cpu: String, memoryArgs: List<String>) {
+        val home = System.getProperty("user.home")
+        val sourceHdName = if (model == "A500") "Harddisk500" else "Harddisk1200"
+        val cmdDir = if (model == "A500") "c" else "C"
+        val libsDir = if (model == "A500") "libs" else "Libs"
+        val libs = listOf("diskfont.library", "iffparse.library", "lowlevel.library", "rexxsyslib.library", "asyncio.library", "icon.library")
+
+        val tempDir = Paths.get(System.getProperty("java.io.tmpdir"), "prog8-copperline-$model")
+        tempDir.createDirectories()
+        val progName = programFile.fileName.toString()
+        val stagedProg = tempDir.resolve(progName)
+        programFile.copyTo(stagedProg, overwrite = true)
+
+        val srcHd = File("$home/Documents/FS-UAE/Hard Drives/$sourceHdName")
+        if (!srcHd.isDirectory) System.err.println("Warning: source hard disk folder not found: ${srcHd}; skipping file staging.")
+        for (lib in libs) {
+            val src = File(srcHd, "$libsDir/$lib")
+            if (src.exists()) src.toPath().copyTo(tempDir.resolve(lib), overwrite = true)
+        }
+        val commands = if (model == "A500") listOf("Dir","List","Makedir","Assign","SetPatch","Mount","CPU","Echo","Avail") else listOf("Dir","List","MakeDir","Assign","SetPatch","Mount","CPU","Echo","Avail")
+        for (cmd in commands) {
+            val src = File(srcHd, "$cmdDir/$cmd")
+            if (src.exists()) src.toPath().copyTo(tempDir.resolve(cmd), overwrite = true)
+        }
+        val ramHandlerSrc = File(srcHd, "l/Ram-Handler")
+        if (model == "A500" && ramHandlerSrc.exists()) ramHandlerSrc.toPath().copyTo(tempDir.resolve("Ram-Handler"), overwrite = true)
+        val sysCfgSrc = File(srcHd, "Devs/system-configuration")
+        if (sysCfgSrc.exists()) sysCfgSrc.toPath().copyTo(tempDir.resolve("system-configuration"), overwrite = true)
+
+        val ksDirs = listOf(
+            File("${System.getProperty("user.home")}/Documents/FS-UAE/Kickstarts"),
+            File("${System.getProperty("user.home")}/Documents/Kickstarts")
+        )
+        val ksDir = ksDirs.firstOrNull { it.isDirectory } ?: ksDirs.first()
+        val candidates = ksDirs.filter { it.isDirectory }.flatMap { d -> d.listFiles()?.filter { it.name.endsWith(".rom", ignoreCase = true) && it.name.contains(model, ignoreCase = true) } ?: emptyList() }
+        val preferredVersion = if (model == "A500") "v1.3" else "v3.1"
+        val rom = candidates.firstOrNull { it.name.contains(preferredVersion, ignoreCase = true) && it.name.contains("[!", ignoreCase = true) }
+            ?: candidates.firstOrNull { it.name.contains(preferredVersion, ignoreCase = true) && !it.name.contains("[b", ignoreCase = true) && !it.name.contains("[h", ignoreCase = true) }
+            ?: candidates.firstOrNull { it.name.contains(preferredVersion, ignoreCase = true) }
+        if (rom == null) {
+            System.err.println("Warning: no Kickstart ROM for $model found in ${ksDir}; using bundled AROS.")
+        }
+        val cmd = mutableListOf(
+            "copperline",
+            "--factory",
+            "--model", model,
+            "--cpu", cpu
+        )
+        cmd.addAll(memoryArgs)
+        if (rom != null) cmd.add(rom.absolutePath)
+        cmd.addAll(listOf(
+            "--run", stagedProg.toString()
+        ))
+        if (!quiet) println("Launching Copperline...")
+        val pb = ProcessBuilder(cmd).inheritIO()
+        try {
+            pb.start().waitFor()
+        } catch (_: Exception) {
+            System.err.println("Cannot find 'copperline'. Install it from https://copperline.dev/")
+        }
+    }
+}

codeCore/src/prog8/code/target/Cx16Target.kt

@@ -8,10 +8,11 @@ import java.nio.file.Path
 
 class Cx16Target: ICompilationTarget,
     IStringEncoding by Encoder(true),
-    IMemSizer by NormalMemSizer(Mflpt5.FLOAT_MEM_SIZE) {
+    IMemSizer by NormalMemSizer(Mflpt5.FLOAT_MEM_SIZE, 2u) {
 
     override val name = NAME
     override val supportsBankedCalls = true
+    override val supportsStackFrames = false
     override val defaultEncoding = Encoding.PETSCII
     override val libraryPath = null
     override val customLauncher = emptyList<String>()
@@ -28,7 +29,9 @@ class Cx16Target: ICompilationTarget,
 
     override val FLOAT_MAX_POSITIVE = Mflpt5.FLOAT_MAX_POSITIVE
     override val FLOAT_MAX_NEGATIVE = Mflpt5.FLOAT_MAX_NEGATIVE
-    override val FLOAT_MEM_SIZE = Mflpt5.FLOAT_MEM_SIZE.toUInt()
+    override val FLOAT_MEM_SIZE = Mflpt5.FLOAT_MEM_SIZE
+    override val POINTER_MEM_SIZE = 2u
+    override val ARRAY_SIZE_LIMIT = 256u
     override val PROGRAM_LOAD_ADDRESS = 0x0801u
     override val PROGRAM_MEMTOP_ADDRESS = 0x9f00u
 
@@ -52,7 +55,7 @@ class Cx16Target: ICompilationTarget,
         return m5.toDouble()
     }
 
-    override fun launchEmulator(selectedEmulator: Int, programNameWithPath: Path, quiet: Boolean) {
+    override fun launchEmulator(selectedEmulator: Int, programFile: Path, quiet: Boolean) {
         val emulator: String
         val extraArgs: List<String>
 
@@ -63,7 +66,7 @@ class Cx16Target: ICompilationTarget,
             }
             2 -> {
                 emulator = "box16"
-                extraArgs = listOf("-sym", C64Target.viceMonListName(programNameWithPath.toString()))
+                extraArgs = listOf("-sym", programFile.resolveSibling(viceMonListName(outputStemOf(programFile))).toString())
             }
             else -> {
                 System.err.println("Cx16 target only supports x16emu and box16 emulators.")
@@ -74,7 +77,7 @@ class Cx16Target: ICompilationTarget,
         if(!quiet)
             println("\nStarting Commander X16 emulator $emulator...")
 
-        val cmdline = listOf(emulator, "-scale", "2", "-rtc", "-run", "-prg", "${programNameWithPath}.prg") + extraArgs
+        val cmdline = listOf(emulator, "-scale", "2", "-rtc", "-run", "-prg", programFile.toString()) + extraArgs
         val processb = ProcessBuilder(cmdline)
         if(!quiet)
             processb.inheritIO()

codeCore/src/prog8/code/target/FsUaeEmulator.kt

@@ -0,0 +1,86 @@
+package prog8.code.target
+
+import java.io.File
+import java.nio.file.Path
+import java.nio.file.Paths
+import kotlin.io.path.copyTo
+import kotlin.io.path.createDirectories
+import kotlin.io.path.writeText
+
+object FsUaeEmulator {
+    fun launch(
+        programFile: Path,
+        quiet: Boolean,
+        model: String,
+        preferredVersion: String,
+        romHint: String,
+        sourceHdName: String,
+        cmdDir: String,
+        libsDir: String,
+        commands: List<String>,
+        libs: List<String>,
+        extraFiles: List<Pair<String, String>>,
+        startupSequence: String
+    ) {
+        val home = System.getProperty("user.home")
+        val ksDir = File("$home/Documents/FS-UAE/Kickstarts")
+        val candidates = ksDir.listFiles()?.filter { it.name.endsWith(".rom", ignoreCase = true) && it.name.contains(model, ignoreCase = true) } ?: emptyList()
+        val rom = candidates.firstOrNull { it.name.contains(preferredVersion, ignoreCase = true) && it.name.contains("[!", ignoreCase = true) }
+            ?: candidates.firstOrNull { it.name.contains(preferredVersion, ignoreCase = true) && !it.name.contains("[b", ignoreCase = true) && !it.name.contains("[h", ignoreCase = true) }
+            ?: candidates.firstOrNull { it.name.contains(preferredVersion, ignoreCase = true) }
+            ?: candidates.firstOrNull { !it.name.contains("[b", ignoreCase = true) && !it.name.contains("[h", ignoreCase = true) }
+            ?: candidates.firstOrNull()
+            ?: ksDir.listFiles()?.firstOrNull { it.name.contains(romHint, ignoreCase = true) && it.name.endsWith(".rom", ignoreCase = true) }
+        if (rom == null) {
+            System.err.println("Cannot find a Kickstart ROM for $model in ${ksDir}")
+            return
+        }
+        val tempDir = Paths.get(System.getProperty("java.io.tmpdir"), "prog8-fsuae-$model")
+        tempDir.createDirectories()
+        val hdDir = tempDir.resolve("hd")
+        hdDir.resolve("S").createDirectories()
+        hdDir.resolve("C").createDirectories()
+        hdDir.resolve("L").createDirectories()
+        hdDir.resolve("Libs").createDirectories()
+        hdDir.resolve("Devs").createDirectories()
+        val progName = programFile.fileName.toString()
+        programFile.copyTo(hdDir.resolve(progName), overwrite = true)
+        val srcHd = File("$home/Documents/FS-UAE/Hard Drives/$sourceHdName")
+        for (cmd in commands) {
+            val src = File(srcHd, "$cmdDir/$cmd")
+            if (src.exists()) src.toPath().copyTo(hdDir.resolve("C/$cmd"), overwrite = true)
+        }
+        for (lib in libs) {
+            val src = File(srcHd, "$libsDir/$lib")
+            if (src.exists()) src.toPath().copyTo(hdDir.resolve("Libs/$lib"), overwrite = true)
+        }
+        for ((srcRel, dstRel) in extraFiles) {
+            val src = File(srcHd, srcRel)
+            if (src.exists()) {
+                val dst = hdDir.resolve(dstRel)
+                dst.parent.createDirectories()
+                src.toPath().copyTo(dst, overwrite = true)
+            }
+        }
+        hdDir.resolve("S/Startup-Sequence").writeText(startupSequence.replace("\$progName", progName))
+        val memoryArgs = if (model == "A500") listOf("--slow_memory=512") else listOf("--fast_memory=2048")
+        val cmd = listOf(
+            "fs-uae",
+            "--amiga_model=$model",
+            "--kickstart_file=${rom.absolutePath}",
+            "--hard_drive_0=${hdDir}",
+            "--automatic_input_grab=0",
+            "--window_width=1280",
+            "--window_height=1024",
+            "--floppy_drive_volume=8",
+            "--floppy_drive_volume_empty=8"
+        ) + memoryArgs
+        if (!quiet) println("Launching FS-UAE...")
+        val pb = ProcessBuilder(cmd).inheritIO()
+        try {
+            pb.start().waitFor()
+        } catch (_: Exception) {
+            System.err.println("Cannot find 'fs-uae'. Install it from https://fs-uae.net/")
+        }
+    }
+}

codeCore/src/prog8/code/target/Mflpt5.kt

@@ -9,7 +9,7 @@ data class Mflpt5(val b0: UByte, val b1: UByte, val b2: UByte, val b3: UByte, va
     companion object {
         const val FLOAT_MAX_POSITIVE = 1.7014118345e+38         // bytes: 255,127,255,255,255
         const val FLOAT_MAX_NEGATIVE = -1.7014118345e+38        // bytes: 255,255,255,255,255
-        const val FLOAT_MEM_SIZE = 5
+        const val FLOAT_MEM_SIZE = 5u
 
         val zero = Mflpt5(0u, 0u, 0u, 0u, 0u)
         fun fromNumber(num: Number): Mflpt5 {
@@ -73,4 +73,4 @@ data class Mflpt5(val b0: UByte, val b1: UByte, val b2: UByte, val b3: UByte, va
         val b4 = "$" + b4.toString(16).padStart(2, '0')
         return "$b0, $b1, $b2, $b3, $b4"
     }
-}
+}

codeCore/src/prog8/code/target/NormalMemSizer.kt

@@ -4,35 +4,41 @@ import prog8.code.core.BaseDataType
 import prog8.code.core.DataType
 import prog8.code.core.IMemSizer
 
-internal class NormalMemSizer(val floatsize: Int): IMemSizer {
+internal class NormalMemSizer(override val FLOAT_MEM_SIZE: UInt, override val POINTER_MEM_SIZE: UInt): IMemSizer {
 
+    private val pointerSize = POINTER_MEM_SIZE.toInt()
+    private val floatSize = FLOAT_MEM_SIZE.toInt()
+    
     override fun memorySize(dt: DataType, numElements: Int?): Int {
         if(dt.isPointerArray)
-            return 2 * numElements!!        // array of pointers is just array of uwords
+            return pointerSize * numElements!!        // array of pointers
         else if(dt.isArray) {
-            if(numElements==null) return 2      // treat it as a pointer size
+            if(numElements==null) return pointerSize      // treat it as a pointer size
             return when(dt.sub) {
                 BaseDataType.BOOL, BaseDataType.UBYTE, BaseDataType.BYTE -> numElements
-                BaseDataType.UWORD, BaseDataType.WORD, BaseDataType.STR -> numElements * 2
+                BaseDataType.UWORD, BaseDataType.WORD  -> numElements * 2
+                BaseDataType.STR, BaseDataType.POINTER -> numElements * pointerSize
                 BaseDataType.LONG -> numElements * 4
-                BaseDataType.FLOAT-> numElements * floatsize
+                BaseDataType.FLOAT-> numElements * floatSize
+                BaseDataType.STRUCT_INSTANCE -> numElements * (dt.subType?.memsize(this) ?: throw IllegalArgumentException("struct instance array has unresolved subtype"))
                 BaseDataType.UNDEFINED -> throw IllegalArgumentException("undefined has no memory size")
                 else -> throw IllegalArgumentException("invalid sub type")
             }
         }
         else if (dt.isString) {
             return numElements        // treat it as the size of the given string with the length
-                ?: 2    // treat it as the size to store a string pointer
+                ?: pointerSize    // treat it as the size to store a string pointer
         }
 
         return when {
             dt.isByteOrBool -> 1 * (numElements ?: 1)
-            dt.isFloat -> floatsize * (numElements ?: 1)
+            dt.isFloat -> floatSize * (numElements ?: 1)
             dt.isLong -> 4 * (numElements ?: 1)
-            dt.isPointer -> 2  // pointer is just a uword
-            dt.isStructInstance -> dt.subType?.memsize(this) ?: 0
+            dt.isWord -> 2 * (numElements ?: 1)
+            dt.isPointer -> pointerSize * (numElements ?: 1)
+            dt.isStructInstance -> dt.subType?.memsize(this) ?: throw IllegalArgumentException("struct instance has unresolved subtype")
             dt.isUndefined -> throw IllegalArgumentException("undefined has no memory size")
-            else -> 2 * (numElements ?: 1)
+            else -> throw IllegalArgumentException("invalid dt $dt")
         }
     }
 

codeCore/src/prog8/code/target/PETTarget.kt

@@ -8,10 +8,11 @@ import java.nio.file.Path
 
 class PETTarget: ICompilationTarget,
     IStringEncoding by Encoder(true),
-    IMemSizer by NormalMemSizer(Mflpt5.FLOAT_MEM_SIZE) {
+    IMemSizer by NormalMemSizer(Mflpt5.FLOAT_MEM_SIZE, 2u) {
 
     override val name = NAME
     override val supportsBankedCalls = false
+    override val supportsStackFrames = false
     override val defaultEncoding = Encoding.PETSCII
     override val libraryPath = null
     override val customLauncher = emptyList<String>()
@@ -27,7 +28,9 @@ class PETTarget: ICompilationTarget,
 
     override val FLOAT_MAX_POSITIVE = Mflpt5.FLOAT_MAX_POSITIVE
     override val FLOAT_MAX_NEGATIVE = Mflpt5.FLOAT_MAX_NEGATIVE
-    override val FLOAT_MEM_SIZE = Mflpt5.FLOAT_MEM_SIZE.toUInt()
+    override val FLOAT_MEM_SIZE = Mflpt5.FLOAT_MEM_SIZE
+    override val POINTER_MEM_SIZE = 2u
+    override val ARRAY_SIZE_LIMIT = 256u
     override val PROGRAM_LOAD_ADDRESS = 0x0401u
     override val PROGRAM_MEMTOP_ADDRESS = 0x8000u
 
@@ -51,7 +54,7 @@ class PETTarget: ICompilationTarget,
         return m5.toDouble()
     }
 
-    override fun launchEmulator(selectedEmulator: Int, programNameWithPath: Path, quiet: Boolean) {
+    override fun launchEmulator(selectedEmulator: Int, programFile: Path, quiet: Boolean) {
         if(selectedEmulator!=1) {
             System.err.println("The pet target only supports the main emulator (Vice).")
             return
@@ -60,9 +63,9 @@ class PETTarget: ICompilationTarget,
         if(!quiet)
             println("\nStarting PET emulator...")
 
-        val viceMonlist = C64Target.viceMonListName(programNameWithPath.toString())
-        val cmdline = listOf("xpet", "-model", "4032", "-ramsize", "32", "-videosize", "40", "-silent", "-moncommands", viceMonlist,
-            "-autostartprgmode", "1", "-autostart-warp", "-autostart", "${programNameWithPath}.prg")
+        val viceMonlist = programFile.resolveSibling(viceMonListName(outputStemOf(programFile)))
+        val cmdline = listOf("xpet", "-model", "4032", "-ramsize", "32", "-videosize", "40", "-silent", "-moncommands", viceMonlist.toString(),
+            "-autostartprgmode", "1", "-autostart-warp", "-autostart", programFile.toString())
         val processb = ProcessBuilder(cmdline)
         if(!quiet)
             processb.inheritIO()

codeCore/src/prog8/code/target/Qemu68kTarget.kt

@@ -0,0 +1,99 @@
+package prog8.code.target
+
+import prog8.code.core.*
+import prog8.code.target.encodings.Encoder
+import prog8.code.target.zp.M68kZeropage
+import java.nio.file.Path
+
+/**
+ * For now target a m68020 cpu
+ * Eventually the goal is to be able to create programs for the Amiga A1200 (68020 cpu) or A500 (68000 cpu)
+ */
+
+class Qemu68kTarget: ICompilationTarget,
+    IStringEncoding by Encoder(false),
+    IMemSizer by NormalMemSizer(4u, 4u) {
+
+    override val name = NAME
+    override val supportsBankedCalls = false
+    override val supportsStackFrames = true
+    override val defaultEncoding = Encoding.ISO
+    override val libraryPath = null
+    override val customLauncher = emptyList<String>()
+    override val additionalAssemblerOptions = emptyList<String>()
+    override val defaultOutputType = OutputType.ELF
+    override val defaultLauncherType = CbmPrgLauncherType.NONE
+
+    companion object {
+        const val NAME = "qemu68k"
+    }
+
+    override val cpu = CpuType.M68020
+
+    override val FLOAT_MAX_POSITIVE = Float.MAX_VALUE.toDouble()
+    override val FLOAT_MAX_NEGATIVE = -Float.MAX_VALUE.toDouble()
+    override val FLOAT_MEM_SIZE = 4u
+    override val POINTER_MEM_SIZE = 4u
+    override val ARRAY_SIZE_LIMIT = 32768u
+    override val PROGRAM_LOAD_ADDRESS = 0x10000u      
+    override val PROGRAM_MEMTOP_ADDRESS = 0x00100000u       // TODO hardcoded at 1 Mb of RAM for now... it starts at $0
+
+    override val BSSHIGHRAM_START = 0u          // not actually used
+    override val BSSHIGHRAM_END = 0u            // not actually used
+    override val BSSGOLDENRAM_START = 0u        // not actually used
+    override val BSSGOLDENRAM_END = 0u          // not actually used
+    override lateinit var zeropage: Zeropage    // not actually used
+
+    override fun getFloatAsmBytes(num: Number): String {
+        TODO("float asm bytes")
+    }
+
+    override fun convertFloatToBytes(num: Double): List<UByte> {
+        TODO("convert float to bytes")
+    }
+
+    override fun convertBytesToFloat(bytes: List<UByte>): Double {
+        require(bytes.size==4) { "need 4 bytes" }
+        TODO("convert bytes to float")
+    }
+
+    override fun launchEmulator(selectedEmulator: Int, programFile: Path, quiet: Boolean) {
+        if(selectedEmulator!=1) {
+            System.err.println("The qemu68k target only supports the main emulator (Qemu).")
+            return
+        }
+        if(!programFile.toFile().exists()) {
+            System.err.println("No program file found: $programFile")
+            return
+        }
+        val cpuStr = this.cpu.toString().lowercase()
+
+        val cmd = listOf(
+            "qemu-system-m68k",
+            "-M", "virt",
+            "-cpu", cpuStr,
+            "-m", "1M",
+            "-kernel", programFile.toString(),
+            "-nographic"
+        )
+        if(!quiet) {
+            println("Launching QEMU (press Ctrl-A X to exit)...")
+        }
+        val pb = ProcessBuilder(cmd).inheritIO()
+        try {
+            pb.start().waitFor()
+        } catch (_: java.io.IOException) {
+            System.err.println("Cannot launch qemu-system-m68k. Install it via your package manager, e.g.:")
+            System.err.println("  sudo apt install qemu-system-m68k       # Debian/Ubuntu")
+            System.err.println("  sudo pacman -S qemu-system-m68k         # Arch Linux")
+            System.err.println("or build it from source: https://www.qemu.org/download/")
+        }
+    }
+
+    override fun isIOAddress(address: UInt): Boolean = address>=0xff000000u
+
+    override fun initializeMemoryAreas(compilerOptions: CompilationOptions) {
+        zeropage = M68kZeropage(compilerOptions)
+    }
+}
+

codeCore/src/prog8/code/target/VMTarget.kt

@@ -3,17 +3,16 @@ package prog8.code.target
 import prog8.code.core.*
 import prog8.code.target.encodings.Encoder
 import java.nio.file.Path
-import kotlin.io.path.extension
 import kotlin.io.path.isReadable
-import kotlin.io.path.name
 import kotlin.io.path.readText
 
 class VMTarget: ICompilationTarget,
     IStringEncoding by Encoder(false),
-    IMemSizer by NormalMemSizer(FLOAT_MEM_SIZE) {
+    IMemSizer by NormalMemSizer(8u, 4u) {
 
     override val name = NAME
     override val supportsBankedCalls = false
+    override val supportsStackFrames = false
     override val defaultEncoding = Encoding.ISO
     override val libraryPath = null
     override val customLauncher = emptyList<String>()
@@ -23,16 +22,19 @@ class VMTarget: ICompilationTarget,
 
     companion object {
         const val NAME = "virtual"
-        const val FLOAT_MEM_SIZE = 8             // 64-bits double
+        const val RAM_SIZE = 16 * 1024 * 1024   // 16 MB
     }
 
     override val cpu = CpuType.VIRTUAL
 
     override val FLOAT_MAX_POSITIVE = Double.MAX_VALUE
     override val FLOAT_MAX_NEGATIVE = -Double.MAX_VALUE
-    override val FLOAT_MEM_SIZE = VMTarget.FLOAT_MEM_SIZE.toUInt()
-    override val PROGRAM_LOAD_ADDRESS = 0u      // not actually used
-    override val PROGRAM_MEMTOP_ADDRESS = 0xffffu  // not actually used
+    override val FLOAT_MEM_SIZE = 8u
+    override val POINTER_MEM_SIZE = 4u
+    override val ARRAY_SIZE_LIMIT = 65535u
+
+    override val PROGRAM_LOAD_ADDRESS = 0u
+    override val PROGRAM_MEMTOP_ADDRESS = 0xffffffu
 
     override val BSSHIGHRAM_START = 0u          // not actually used
     override val BSSHIGHRAM_END = 0u            // not actually used
@@ -68,21 +70,24 @@ class VMTarget: ICompilationTarget,
         return Double.fromBits(b0 or b1 or b2 or b3 or b4 or b5 or b6 or b7)
     }
 
-    override fun launchEmulator(selectedEmulator: Int, programNameWithPath: Path, quiet: Boolean) {
-        launchEmulatorWithTrace(programNameWithPath, quiet, traceEnabled = false)
+    override fun launchEmulator(selectedEmulator: Int, programFile: Path, quiet: Boolean) {
+        launchEmulatorWithTrace(programFile, quiet, traceEnabled = false)
     }
 
-    fun launchEmulatorWithTrace(programNameWithPath: Path, quiet: Boolean, traceEnabled: Boolean) {
+    /**
+     * Runs the given .p8ir file in the embedded virtual machine.
+     * The caller must pass the exact path; no fallback probing happens here, the -vm CLI flow resolves a bare name itself.
+     */
+    fun launchEmulatorWithTrace(programFile: Path, quiet: Boolean, traceEnabled: Boolean) {
         if(!quiet)
             println("\nStarting Virtual Machine...")
 
         // to not have external module dependencies in our own module, we launch the virtual machine via reflection
         val vm = Class.forName("prog8.vm.VmRunner").getDeclaredConstructor().newInstance() as IVirtualMachineRunner
-        val withExt = if(programNameWithPath.extension=="p8ir") programNameWithPath else programNameWithPath.resolveSibling("${programNameWithPath.name}.p8ir")
-        if(withExt.isReadable())
-            vm.runProgram(withExt.readText(), quiet, traceEnabled)
+        if(programFile.isReadable())
+            vm.runProgram(programFile.readText(), quiet, traceEnabled)
         else
-            throw java.nio.file.NoSuchFileException(withExt.name, null, "not a .p8ir file")
+            throw java.nio.file.NoSuchFileException(programFile.toString(), null, "not a .p8ir file")
     }
 
     override fun isIOAddress(address: UInt): Boolean = false

codeCore/src/prog8/code/target/VamosEmulator.kt

@@ -0,0 +1,16 @@
+package prog8.code.target
+
+import java.nio.file.Path
+
+object VamosEmulator {
+    fun launch(programFile: Path, quiet: Boolean, cpu: String) {
+        val cmd = listOf("vamos", "--cpu", cpu, programFile.toString())
+        if (!quiet) println("Launching Amitools's Vamos...")
+        val pb = ProcessBuilder(cmd).inheritIO()
+        try {
+            pb.start().waitFor()
+        } catch (_: Exception) {
+            System.err.println("Cannot find 'vamos' (Amitools's Amiga emulator). Install it from: https://github.com/cnvogelg/amitools")
+        }
+    }
+}

codeCore/src/prog8/code/target/zp/M68kZeropage.kt

@@ -0,0 +1,21 @@
+package prog8.code.target.zp
+
+import prog8.code.core.CompilationOptions
+import prog8.code.core.Zeropage
+
+/**
+ * The m68k has no zero page, so there are no scratch locations to hand out.
+ * Any attempt to use one is a bug in the code generator.
+ */
+class M68kZeropage(options: CompilationOptions): Zeropage(options) {
+    override val SCRATCH_B1: UInt
+        get() = throw IllegalStateException("m68k shouldn't use this zeropage variable")
+    override val SCRATCH_REG: UInt
+        get() = throw IllegalStateException("m68k shouldn't use this zeropage variable")
+    override val SCRATCH_W1: UInt
+        get() = throw IllegalStateException("m68k shouldn't use this zeropage variable")
+    override val SCRATCH_W2: UInt
+        get() = throw IllegalStateException("m68k shouldn't use this zeropage variable")
+    override val SCRATCH_PTR: UInt
+        get() = throw IllegalStateException("m68k shouldn't use this zeropage variable")
+}

codeCore/src/testFixtures/kotlin/prog8tests/helpers/Dummies.kt

@@ -8,6 +8,9 @@ import prog8.code.core.*
  * Returns standard memory sizes for all data types.
  */
 object DummyMemsizer : IMemSizer {
+    override val FLOAT_MEM_SIZE: UInt = 4u
+    override val POINTER_MEM_SIZE: UInt = 2u
+
     override fun memorySize(dt: DataType, numElements: Int?): Int {
         if (dt.isPointerArray)
             return 2 * numElements!!
@@ -18,6 +21,8 @@ object DummyMemsizer : IMemSizer {
                 BaseDataType.UWORD, BaseDataType.WORD -> numElements * 2
                 BaseDataType.LONG -> numElements * 4
                 BaseDataType.FLOAT -> numElements * 5
+                BaseDataType.POINTER -> numElements * 2
+                BaseDataType.STRUCT_INSTANCE -> numElements * (dt.subType?.memsize(this) ?: throw IllegalArgumentException("struct instance array has unresolved subtype"))
                 else -> throw IllegalArgumentException("invalid sub type")
             }
         }

codeCore/test/prog8tests/codecore/TestConfigFileTarget.kt

@@ -0,0 +1,66 @@
+package prog8tests.codecore
+
+import io.kotest.assertions.throwables.shouldThrow
+import io.kotest.core.spec.style.FunSpec
+import io.kotest.matchers.shouldBe
+import prog8.code.core.BaseDataType
+import prog8.code.target.ConfigFileTarget
+import java.nio.file.Files
+import kotlin.io.path.writeText
+
+class TestConfigFileTarget: FunSpec({
+    fun loadTarget(extraProperties: String = ""): ConfigFileTarget {
+        val directory = Files.createTempDirectory("prog8-config-target")
+        val library = Files.createDirectories(directory.resolve("library"))
+        val config = directory.resolve("target.properties")
+        config.writeText(
+            """
+            cpu = 65C02
+            encoding = iso
+            load_address = 2048
+            memtop = 65535
+            bss_highram_start = 0
+            bss_highram_end = 0
+            bss_goldenram_start = 0
+            bss_goldenram_end = 0
+            zp_scratch_b1 = 240
+            zp_scratch_reg = 241
+            zp_scratch_w1 = 242
+            zp_scratch_w2 = 244
+            zp_scratch_ptr = 246
+            virtual_registers = 2
+            zp_fullsafe = 34-239
+            zp_kernalsafe = 34-239
+            zp_basicsafe = 34-239
+            io_regions =
+            """.trimIndent() + "\nlibrary = $library\n$extraProperties"
+        )
+        return try {
+            ConfigFileTarget.fromConfigFile(config)
+        } finally {
+            directory.toFile().deleteRecursively()
+        }
+    }
+
+    test("legacy config defaults to 2-byte pointers and 256 element arrays") {
+        val target = loadTarget()
+
+        target.POINTER_MEM_SIZE shouldBe 2u
+        target.memorySize(BaseDataType.POINTER) shouldBe 2
+        target.ARRAY_SIZE_LIMIT shouldBe 256u
+    }
+
+    test("config supports 4-byte pointers") {
+        val target = loadTarget("pointer_size = 4")
+
+        target.POINTER_MEM_SIZE shouldBe 4u
+        target.memorySize(BaseDataType.POINTER) shouldBe 4
+        target.ARRAY_SIZE_LIMIT shouldBe 256u
+    }
+
+    test("config rejects unsupported pointer sizes") {
+        shouldThrow<IllegalArgumentException> {
+            loadTarget("pointer_size = 3")
+        }
+    }
+})

codeCore/test/prog8tests/codecore/TestConversions.kt

@@ -23,31 +23,31 @@ class TestConversions: FunSpec({
     }
 
     test("toHex for 16-255 (hex with $)") {
-        16.toHex() shouldBe "\$10"
-        255.toHex() shouldBe "\$ff"
+        16.toHex() shouldBe $$"$10"
+        255.toHex() shouldBe $$"$ff"
     }
 
     test("toHex for 256-65535 (4 digit hex)") {
-        256.toHex() shouldBe "\$0100"
-        65535.toHex() shouldBe "\$ffff"
+        256.toHex() shouldBe $$"$0100"
+        65535.toHex() shouldBe $$"$ffff"
     }
 
     test("toHex for larger numbers (8 digit hex)") {
-        65536.toHex() shouldBe "\$00010000"
-        0x12345678.toHex() shouldBe "\$12345678"
+        65536.toHex() shouldBe $$"$00010000"
+        0x12345678.toHex() shouldBe $$"$12345678"
     }
 
     test("toHex for negative numbers") {
         (-1).toHex() shouldBe "-1"
-        (-255).toHex() shouldBe "-\$ff"
-        Int.MIN_VALUE.toHex() shouldBe "\$80000000"
+        (-255).toHex() shouldBe $$"-$ff"
+        Int.MIN_VALUE.toHex() shouldBe $$"$80000000"
     }
 
     test("UInt.toHex") {
         0u.toHex() shouldBe "0"
         15u.toHex() shouldBe "15"
-        255u.toHex() shouldBe "\$ff"
-        65535u.toHex() shouldBe "\$ffff"
+        255u.toHex() shouldBe $$"$ff"
+        65535u.toHex() shouldBe $$"$ffff"
     }
 
     // ============================================================================

codeCore/test/prog8tests/codecore/TestDataType.kt

@@ -2,28 +2,36 @@ package prog8tests.codecore
 
 import io.kotest.assertions.throwables.shouldThrow
 import io.kotest.core.spec.style.FunSpec
-import io.kotest.matchers.booleans.shouldBeFalse
-import io.kotest.matchers.booleans.shouldBeTrue
 import io.kotest.matchers.shouldBe
 import io.kotest.matchers.shouldNotBe
 import prog8.code.core.*
+import prog8.code.target.Amiga500Target
+import prog8.code.target.Cx16Target
+import prog8.code.target.VMTarget
+
 
 /**
  * Unit tests for DataType and BaseDataType classes in codeCore.
  * Focus on type predicates, relationships, and key operations.
  */
 class TestDataType: FunSpec({
+    val tgt = Cx16Target()
 
     val dummyMemSizer = object : IMemSizer {
+        override val FLOAT_MEM_SIZE: UInt = 4u
+        override val POINTER_MEM_SIZE: UInt = 2u
+
         override fun memorySize(dt: DataType, numElements: Int?): Int {
             if (dt.isPointerArray) return 2 * numElements!!
-            if (dt.isArray || dt.isSplitWordArray) {
+            if (dt.isArray || dt.isSplitWordArray(this)) {
                 require(numElements != null)
                 return when (dt.sub) {
                     BaseDataType.BOOL, BaseDataType.BYTE, BaseDataType.UBYTE -> numElements
                     BaseDataType.UWORD, BaseDataType.WORD -> numElements * 2
                     BaseDataType.LONG -> numElements * 4
                     BaseDataType.FLOAT -> numElements * 5
+                    BaseDataType.POINTER -> numElements * 2
+                    BaseDataType.STRUCT_INSTANCE -> numElements * (dt.subType?.memsize(this) ?: throw IllegalArgumentException("struct instance array has unresolved subtype"))
                     else -> throw IllegalArgumentException("invalid sub type")
                 }
             }
@@ -45,61 +53,74 @@ class TestDataType: FunSpec({
     // ============================================================================
 
     test("isByte extension") {
-        BaseDataType.UBYTE.isByte.shouldBeTrue()
-        BaseDataType.BYTE.isByte.shouldBeTrue()
-        BaseDataType.UWORD.isByte.shouldBeFalse()
-        BaseDataType.WORD.isByte.shouldBeFalse()
+        BaseDataType.UBYTE.isByte shouldBe true
+        BaseDataType.BYTE.isByte shouldBe true
+        BaseDataType.UWORD.isByte shouldBe false
+        BaseDataType.WORD.isByte shouldBe false
     }
 
     test("isByteOrBool extension") {
-        BaseDataType.UBYTE.isByteOrBool.shouldBeTrue()
-        BaseDataType.BYTE.isByteOrBool.shouldBeTrue()
-        BaseDataType.BOOL.isByteOrBool.shouldBeTrue()
-        BaseDataType.UWORD.isByteOrBool.shouldBeFalse()
+        BaseDataType.UBYTE.isByteOrBool shouldBe true
+        BaseDataType.BYTE.isByteOrBool shouldBe true
+        BaseDataType.BOOL.isByteOrBool shouldBe true
+        BaseDataType.UWORD.isByteOrBool shouldBe false
     }
 
     test("isWord extension") {
-        BaseDataType.UWORD.isWord.shouldBeTrue()
-        BaseDataType.WORD.isWord.shouldBeTrue()
-        BaseDataType.UBYTE.isWord.shouldBeFalse()
+        BaseDataType.UWORD.isWord shouldBe true
+        BaseDataType.WORD.isWord shouldBe true
+        BaseDataType.UBYTE.isWord shouldBe false
     }
 
     test("isInteger extension") {
-        BaseDataType.UBYTE.isInteger.shouldBeTrue()
-        BaseDataType.WORD.isInteger.shouldBeTrue()
-        BaseDataType.LONG.isInteger.shouldBeTrue()
-        BaseDataType.FLOAT.isInteger.shouldBeFalse()
+        BaseDataType.UBYTE.isInteger shouldBe true
+        BaseDataType.WORD.isInteger shouldBe true
+        BaseDataType.LONG.isInteger shouldBe true
+        BaseDataType.FLOAT.isInteger shouldBe false
     }
 
     test("isNumeric extension") {
-        BaseDataType.UBYTE.isNumeric.shouldBeTrue()
-        BaseDataType.FLOAT.isNumeric.shouldBeTrue()
-        BaseDataType.BOOL.isNumeric.shouldBeFalse()
+        BaseDataType.UBYTE.isNumeric shouldBe true
+        BaseDataType.FLOAT.isNumeric shouldBe true
+        BaseDataType.BOOL.isNumeric shouldBe false
     }
 
     test("isSigned extension") {
-        BaseDataType.BYTE.isSigned.shouldBeTrue()
-        BaseDataType.WORD.isSigned.shouldBeTrue()
-        BaseDataType.LONG.isSigned.shouldBeTrue()
-        BaseDataType.UBYTE.isSigned.shouldBeFalse()
+        BaseDataType.BYTE.isSigned shouldBe true
+        BaseDataType.WORD.isSigned shouldBe true
+        BaseDataType.LONG.isSigned shouldBe true
+        BaseDataType.UBYTE.isSigned shouldBe false
     }
 
     test("isArray extension") {
-        BaseDataType.ARRAY.isArray.shouldBeTrue()
-        BaseDataType.ARRAY_SPLITW.isArray.shouldBeTrue()
-        BaseDataType.ARRAY_POINTER.isArray.shouldBeTrue()
-        BaseDataType.UBYTE.isArray.shouldBeFalse()
+        BaseDataType.ARRAY.isArray shouldBe true
+        BaseDataType.ARRAY_SPLITW.isArray shouldBe true
+        BaseDataType.UBYTE.isArray shouldBe false
+    }
+
+    test("isPointerArray property") {
+        DataType.arrayOfPointersTo(BaseDataType.UBYTE).isPointerArray shouldBe true
+        DataType.arrayOfPointersTo(BaseDataType.WORD).isPointerArray shouldBe true
+        DataType.arrayFor(BaseDataType.UBYTE, tgt).isPointerArray shouldBe false
     }
 
     test("isPointer extension") {
-        BaseDataType.POINTER.isPointer.shouldBeTrue()
-        BaseDataType.UBYTE.isPointer.shouldBeFalse()
+        BaseDataType.POINTER.isPointer shouldBe true
+        BaseDataType.UBYTE.isPointer shouldBe false
     }
 
     test("isSplitWordArray extension") {
-        BaseDataType.ARRAY_SPLITW.isSplitWordArray.shouldBeTrue()
-        BaseDataType.ARRAY_POINTER.isSplitWordArray.shouldBeTrue()
-        BaseDataType.ARRAY.isSplitWordArray.shouldBeFalse()
+        DataType.splitWordArrayFor(BaseDataType.UWORD).isSplitWordArray(Cx16Target()) shouldBe true
+        DataType.splitWordArrayFor(BaseDataType.UWORD).isSplitWordArray(VMTarget()) shouldBe true
+        DataType.splitWordArrayFor(BaseDataType.UWORD).isSplitWordArray(Amiga500Target()) shouldBe true
+
+        DataType.arrayOfPointersTo(BaseDataType.UWORD).isSplitWordArray(Cx16Target()) shouldBe true
+        DataType.arrayOfPointersTo(BaseDataType.UWORD).isSplitWordArray(VMTarget()) shouldBe false
+        DataType.arrayOfPointersTo(BaseDataType.UWORD).isSplitWordArray(Amiga500Target()) shouldBe false
+
+        DataType.arrayFor(BaseDataType.UWORD, Cx16Target()).isSplitWordArray(Cx16Target()) shouldBe false
+        DataType.arrayFor(BaseDataType.UWORD, VMTarget()).isSplitWordArray(VMTarget()) shouldBe false
+        DataType.arrayFor(BaseDataType.UWORD, Amiga500Target()).isSplitWordArray(Amiga500Target()) shouldBe false
     }
 
     // ============================================================================
@@ -113,20 +134,20 @@ class TestDataType: FunSpec({
         DataType.forDt(BaseDataType.FLOAT) shouldBe DataType.FLOAT
     }
 
-    test("DataType.forDt throws for struct instance") {
-        shouldThrow<NotImplementedError> {
-            DataType.forDt(BaseDataType.STRUCT_INSTANCE)
-        }
+    test("DataType.forDt supports struct instance") {
+        val dt = DataType.forDt(BaseDataType.STRUCT_INSTANCE)
+        dt.base shouldBe BaseDataType.STRUCT_INSTANCE
+        dt.isStructInstance shouldBe true
     }
 
     test("DataType.arrayFor creates array types") {
-        val ubyteArray = DataType.arrayFor(BaseDataType.UBYTE)
+        val ubyteArray = DataType.arrayFor(BaseDataType.UBYTE, tgt)
         ubyteArray.base shouldBe BaseDataType.ARRAY
         ubyteArray.sub shouldBe BaseDataType.UBYTE
 
-        val wordArraySplit = DataType.arrayFor(BaseDataType.WORD)
-        wordArraySplit.base shouldBe BaseDataType.ARRAY_SPLITW
-        wordArraySplit.sub shouldBe BaseDataType.WORD
+        val wordArray = DataType.arrayFor(BaseDataType.WORD, tgt)
+        wordArray.base shouldBe BaseDataType.ARRAY
+        wordArray.sub shouldBe BaseDataType.WORD
     }
 
     test("DataType.pointer creates pointer types") {
@@ -140,45 +161,45 @@ class TestDataType: FunSpec({
     // ============================================================================
 
     test("DataType.isBasic property") {
-        DataType.UBYTE.isBasic.shouldBeTrue()
-        DataType.WORD.isBasic.shouldBeTrue()
-        DataType.STR.isBasic.shouldBeFalse()
-        DataType.arrayFor(BaseDataType.UBYTE).isBasic.shouldBeFalse()
+        DataType.UBYTE.isBasic shouldBe true
+        DataType.WORD.isBasic shouldBe true
+        DataType.STR.isBasic shouldBe false
+        DataType.arrayFor(BaseDataType.UBYTE, tgt).isBasic shouldBe false
     }
 
     test("DataType.isByte property") {
-        DataType.UBYTE.isByte.shouldBeTrue()
-        DataType.BYTE.isByte.shouldBeTrue()
-        DataType.WORD.isByte.shouldBeFalse()
+        DataType.UBYTE.isByte shouldBe true
+        DataType.BYTE.isByte shouldBe true
+        DataType.WORD.isByte shouldBe false
     }
 
     test("DataType.isNumeric property") {
-        DataType.UBYTE.isNumeric.shouldBeTrue()
-        DataType.FLOAT.isNumeric.shouldBeTrue()
-        DataType.BOOL.isNumeric.shouldBeFalse()
-        DataType.STR.isNumeric.shouldBeFalse()
+        DataType.UBYTE.isNumeric shouldBe true
+        DataType.FLOAT.isNumeric shouldBe true
+        DataType.BOOL.isNumeric shouldBe false
+        DataType.STR.isNumeric shouldBe false
     }
 
     test("DataType.isArray property") {
-        DataType.arrayFor(BaseDataType.UBYTE).isArray.shouldBeTrue()
-        DataType.UBYTE.isArray.shouldBeFalse()
+        DataType.arrayFor(BaseDataType.UBYTE, tgt).isArray shouldBe true
+        DataType.UBYTE.isArray shouldBe false
     }
 
     test("DataType.isPointer property") {
-        DataType.pointer(BaseDataType.UBYTE).isPointer.shouldBeTrue()
-        DataType.UBYTE.isPointer.shouldBeFalse()
+        DataType.pointer(BaseDataType.UBYTE).isPointer shouldBe true
+        DataType.UBYTE.isPointer shouldBe false
     }
 
     test("DataType.isString property") {
-        DataType.STR.isString.shouldBeTrue()
-        DataType.UBYTE.isString.shouldBeFalse()
+        DataType.STR.isString shouldBe true
+        DataType.UBYTE.isString shouldBe false
     }
 
     test("DataType.isPassByRef property") {
-        DataType.STR.isPassByRef.shouldBeTrue()
-        DataType.arrayFor(BaseDataType.UBYTE).isPassByRef.shouldBeTrue()
-        DataType.UBYTE.isPassByRef.shouldBeFalse()
-        DataType.pointer(BaseDataType.UBYTE).isPassByRef.shouldBeFalse()
+        DataType.STR.isPassByRef shouldBe true
+        DataType.arrayFor(BaseDataType.UBYTE, tgt).isPassByRef shouldBe true
+        DataType.UBYTE.isPassByRef shouldBe false
+        DataType.pointer(BaseDataType.UBYTE).isPassByRef shouldBe false
     }
 
     // ============================================================================
@@ -186,8 +207,8 @@ class TestDataType: FunSpec({
     // ============================================================================
 
     test("DataType.elementType returns element type for arrays") {
-        DataType.arrayFor(BaseDataType.UBYTE).elementType() shouldBe DataType.UBYTE
-        DataType.arrayFor(BaseDataType.WORD).elementType() shouldBe DataType.WORD
+        DataType.arrayFor(BaseDataType.UBYTE, tgt).elementType() shouldBe DataType.UBYTE
+        DataType.arrayFor(BaseDataType.WORD, tgt).elementType() shouldBe DataType.WORD
         DataType.STR.elementType() shouldBe DataType.UBYTE
     }
 
@@ -226,31 +247,31 @@ class TestDataType: FunSpec({
     }
 
     test("BaseDataType.largerSizeThan") {
-        BaseDataType.WORD.largerSizeThan(BaseDataType.UBYTE).shouldBeTrue()
-        BaseDataType.LONG.largerSizeThan(BaseDataType.WORD).shouldBeTrue()
-        BaseDataType.UBYTE.largerSizeThan(BaseDataType.WORD).shouldBeFalse()
-        BaseDataType.STR.largerSizeThan(BaseDataType.UBYTE).shouldBeTrue()
-        BaseDataType.STR.largerSizeThan(BaseDataType.BOOL).shouldBeTrue()
-        BaseDataType.STR.largerSizeThan(BaseDataType.UWORD).shouldBeFalse()
-        BaseDataType.STR.largerSizeThan(BaseDataType.WORD).shouldBeFalse()
-        BaseDataType.STR.largerSizeThan(BaseDataType.POINTER).shouldBeFalse()
-        BaseDataType.STR.largerSizeThan(BaseDataType.LONG).shouldBeFalse()
-        BaseDataType.POINTER.largerSizeThan(BaseDataType.UBYTE).shouldBeTrue()
-        BaseDataType.POINTER.largerSizeThan(BaseDataType.BOOL).shouldBeTrue()
-        BaseDataType.POINTER.largerSizeThan(BaseDataType.UWORD).shouldBeFalse()
-        BaseDataType.POINTER.largerSizeThan(BaseDataType.STR).shouldBeFalse()
+        BaseDataType.WORD.largerSizeThan(BaseDataType.UBYTE) shouldBe true
+        BaseDataType.LONG.largerSizeThan(BaseDataType.WORD) shouldBe true
+        BaseDataType.UBYTE.largerSizeThan(BaseDataType.WORD) shouldBe false
+        BaseDataType.STR.largerSizeThan(BaseDataType.UBYTE) shouldBe true
+        BaseDataType.STR.largerSizeThan(BaseDataType.BOOL) shouldBe true
+        BaseDataType.STR.largerSizeThan(BaseDataType.UWORD) shouldBe false
+        BaseDataType.STR.largerSizeThan(BaseDataType.WORD) shouldBe false
+        BaseDataType.STR.largerSizeThan(BaseDataType.POINTER) shouldBe false
+        BaseDataType.STR.largerSizeThan(BaseDataType.LONG) shouldBe false
+        BaseDataType.POINTER.largerSizeThan(BaseDataType.UBYTE) shouldBe true
+        BaseDataType.POINTER.largerSizeThan(BaseDataType.BOOL) shouldBe true
+        BaseDataType.POINTER.largerSizeThan(BaseDataType.UWORD) shouldBe false
+        BaseDataType.POINTER.largerSizeThan(BaseDataType.STR) shouldBe false
     }
 
     test("BaseDataType.equalsSize") {
-        BaseDataType.UBYTE.equalsSize(BaseDataType.UBYTE).shouldBeTrue()
-        BaseDataType.UBYTE.equalsSize(BaseDataType.BYTE).shouldBeTrue()
-        BaseDataType.WORD.equalsSize(BaseDataType.UWORD).shouldBeTrue()
-        BaseDataType.UBYTE.equalsSize(BaseDataType.WORD).shouldBeFalse()
-        BaseDataType.STR.equalsSize(BaseDataType.STR).shouldBeTrue()
-        BaseDataType.STR.equalsSize(BaseDataType.UWORD).shouldBeTrue()
-        BaseDataType.STR.equalsSize(BaseDataType.BYTE).shouldBeFalse()
-        BaseDataType.POINTER.equalsSize(BaseDataType.UWORD).shouldBeTrue()
-        BaseDataType.POINTER.equalsSize(BaseDataType.STR).shouldBeFalse()
+        BaseDataType.UBYTE.equalsSize(BaseDataType.UBYTE) shouldBe true
+        BaseDataType.UBYTE.equalsSize(BaseDataType.BYTE) shouldBe true
+        BaseDataType.WORD.equalsSize(BaseDataType.UWORD) shouldBe true
+        BaseDataType.UBYTE.equalsSize(BaseDataType.WORD) shouldBe false
+        BaseDataType.STR.equalsSize(BaseDataType.STR) shouldBe true
+        BaseDataType.STR.equalsSize(BaseDataType.UWORD) shouldBe true
+        BaseDataType.STR.equalsSize(BaseDataType.BYTE) shouldBe false
+        BaseDataType.POINTER.equalsSize(BaseDataType.UWORD) shouldBe true
+        BaseDataType.POINTER.equalsSize(BaseDataType.STR) shouldBe false
     }
 
     // ============================================================================
@@ -258,39 +279,39 @@ class TestDataType: FunSpec({
     // ============================================================================
 
     test("bool assignability") {
-        DataType.BOOL.isAssignableTo(DataType.BOOL).shouldBeTrue()
-        DataType.BOOL.isAssignableTo(DataType.UBYTE).shouldBeFalse()
+        DataType.BOOL.isAssignableTo(DataType.BOOL) shouldBe true
+        DataType.BOOL.isAssignableTo(DataType.UBYTE) shouldBe false
     }
 
     test("ubyte assignability") {
-        DataType.UBYTE.isAssignableTo(DataType.UBYTE).shouldBeTrue()
-        DataType.UBYTE.isAssignableTo(DataType.UWORD).shouldBeTrue()
-        DataType.UBYTE.isAssignableTo(DataType.LONG).shouldBeTrue()
-        DataType.UBYTE.isAssignableTo(DataType.FLOAT).shouldBeTrue()
+        DataType.UBYTE.isAssignableTo(DataType.UBYTE) shouldBe true
+        DataType.UBYTE.isAssignableTo(DataType.UWORD) shouldBe true
+        DataType.UBYTE.isAssignableTo(DataType.LONG) shouldBe true
+        DataType.UBYTE.isAssignableTo(DataType.FLOAT) shouldBe true
     }
 
     test("byte assignability") {
-        DataType.BYTE.isAssignableTo(DataType.BYTE).shouldBeTrue()
-        DataType.BYTE.isAssignableTo(DataType.WORD).shouldBeTrue()
-        DataType.BYTE.isAssignableTo(DataType.LONG).shouldBeTrue()
+        DataType.BYTE.isAssignableTo(DataType.BYTE) shouldBe true
+        DataType.BYTE.isAssignableTo(DataType.WORD) shouldBe true
+        DataType.BYTE.isAssignableTo(DataType.LONG) shouldBe true
     }
 
     test("uword assignability") {
-        DataType.UWORD.isAssignableTo(DataType.UWORD).shouldBeTrue()
-        DataType.UWORD.isAssignableTo(DataType.LONG).shouldBeTrue()
-        DataType.UWORD.isAssignableTo(DataType.FLOAT).shouldBeTrue()
+        DataType.UWORD.isAssignableTo(DataType.UWORD) shouldBe true
+        DataType.UWORD.isAssignableTo(DataType.LONG) shouldBe true
+        DataType.UWORD.isAssignableTo(DataType.FLOAT) shouldBe true
     }
 
     test("string assignability") {
-        DataType.STR.isAssignableTo(DataType.STR).shouldBeTrue()
-        DataType.STR.isAssignableTo(DataType.UWORD).shouldBeTrue()
+        DataType.STR.isAssignableTo(DataType.STR) shouldBe true
+        DataType.STR.isAssignableTo(DataType.UWORD) shouldBe true
     }
 
     test("pointer assignability") {
         val ptr = DataType.pointer(BaseDataType.UBYTE)
-        ptr.isAssignableTo(DataType.UWORD).shouldBeTrue()
-        ptr.isAssignableTo(DataType.LONG).shouldBeTrue()
-        ptr.isAssignableTo(DataType.pointer(BaseDataType.UBYTE)).shouldBeTrue()
+        ptr.isAssignableTo(DataType.UWORD) shouldBe true
+        ptr.isAssignableTo(DataType.LONG) shouldBe true
+        ptr.isAssignableTo(DataType.pointer(BaseDataType.UBYTE)) shouldBe true
     }
 
     // ============================================================================
@@ -305,7 +326,7 @@ class TestDataType: FunSpec({
     }
 
     test("DataType.toString for arrays") {
-        DataType.arrayFor(BaseDataType.UBYTE).toString() shouldBe "ubyte[]"
-        DataType.arrayFor(BaseDataType.WORD).toString() shouldBe "word[] (split)"
+        DataType.arrayFor(BaseDataType.UBYTE, tgt).toString() shouldBe "ubyte[]"
+        DataType.arrayFor(BaseDataType.WORD, tgt).toString() shouldBe "word[]"
     }
 })

codeCore/test/prog8tests/codecore/TestZeropageFreeList.kt

@@ -19,7 +19,7 @@ import prog8.code.target.zp.PETZeropage
 
 
 /**
- * Comprehensive tests for Zeropage free list configurations across all targets.
+ * Tests for Zeropage free list configurations across all targets.
  * 
  * Tests verify that for each target and ZeropageType combination:
  * - The exact list of free locations is correct

codeGenCpu6502/src/prog8/codegen/cpu6502/AsmGen.kt

@@ -2,8 +2,12 @@ package prog8.codegen.cpu6502
 
 import com.github.michaelbull.result.fold
 import prog8.code.*
+import prog8.code.assembly.AssemblyProgram6502
+import prog8.code.assembly.IAssemblyProgram
 import prog8.code.ast.*
 import prog8.code.core.*
+import prog8.code.cpu6502.byteShiftAddExpansion
+import prog8.code.cpu6502.wordShiftAddExpansion
 import prog8.code.source.ImportFileSystem
 import prog8.code.source.SourceCode
 import prog8.code.target.Cx16Target
@@ -15,8 +19,7 @@ import kotlin.io.path.writeLines
 internal const val subroutineFloatEvalResultVar1 = "prog8_float_eval_result1"
 internal const val subroutineFloatEvalResultVar2 = "prog8_float_eval_result2"
 
-class AsmGen6502(val prefixSymbols: Boolean,
-                 private val lastGeneratedLabelSequenceNr: Int,
+class AsmGen6502(private val lastGeneratedLabelSequenceNr: Int,
                  private val preassignedCallSiteIds: Map<PtAsmSub, UByte> = emptyMap()
 ): ICodeGeneratorBackend {
     override fun generate(
@@ -25,7 +28,7 @@ class AsmGen6502(val prefixSymbols: Boolean,
         options: CompilationOptions,
         errors: IErrorReporter,
     ): IAssemblyProgram? {
-        val st = if(prefixSymbols) prefixSymbols(program, options, symbolTable) else symbolTable
+        val st = prefixSymbols(program, options, symbolTable)
         val asmgen = AsmGen6502Internal(program, st, options, errors, lastGeneratedLabelSequenceNr, preassignedCallSiteIds)
         return asmgen.compileToAssembly()
     }
@@ -44,8 +47,36 @@ class AsmGen6502Internal (
     preassignedCallSiteIds: Map<PtAsmSub, UByte> = emptyMap()
 ) {
 
-    internal val optimizedByteMultiplications = setOf(3,5,6,7,9,10,11,12,13,14,15,20,25,40,50,80,100)
-    internal val optimizedWordMultiplications = setOf(3,5,6,7,9,10,12,15,20,25,40,50,80,100,320,640)
+    // 40, 50, 80 and 100 are deliberately absent: prog8_math's mul_byte_40/50/80/100 are wrong.
+    // They mask the multiplicand with `and #7` or `and #3` and index a small table, which only
+    // works if x*k mod 256 has that period. The true period is 256/gcd(k,256), i.e. 32, 128, 16
+    // and 64 entries respectively, so the tables are far too small (e.g. mul_byte_100 returns
+    // (x&3)*100, giving 0 for x=4 where the answer is 144). The generic expansion below computes
+    // them correctly in 8-11 instructions.
+    internal val optimizedByteMultiplications = setOf(3,5,6,7,9,10,11,12,13,14,15,20,25)
+    // 320 and 640 are deliberately absent: mul_word_320/640 compute (low byte of x)*320/*640 and
+    // ignore the multiplicand's high byte, so they are wrong for any x >= 256 (the comment
+    // "msb in Y doesn't matter" is false). The generic expansion handles them in 27-29 instructions.
+    internal val optimizedWordMultiplications = setOf(3,5,6,7,9,10,12,15,20,25,40,50,80,100)
+
+    // Thin wrappers over the shared expansion in codeCore, so this backend and codeGenNew6502 emit
+    // an identical sequence. Works on the raw bit pattern, exactly like the hand-written
+    // prog8_math.mul_word_N / mul_byte_N routines, so signed and unsigned agree because both
+    // truncate mod 2^n and no sign extension is applied afterwards.
+    internal fun emitWordShiftAdd(value: Int): Boolean {
+        val expansion = wordShiftAddExpansion(value) ?: return false
+        out("  sta  P8ZP_SCRATCH_W2 |  sty  P8ZP_SCRATCH_W2+1")   // x
+        out("  sta  P8ZP_SCRATCH_W1 |  sty  P8ZP_SCRATCH_W1+1")   // r = x
+        expansion.forEach { out("  $it") }
+        return true
+    }
+
+    internal fun emitByteShiftAdd(value: Int): Boolean {
+        val expansion = byteShiftAddExpansion(value) ?: return false
+        expansion.forEach { out("  $it") }
+        return true
+    }
+
     internal val loopEndLabels = ArrayDeque<String>()
     private val zeropage = options.compTarget.zeropage
     private val allocator = VariableAllocator(symbolTable, options, errors, program)
@@ -79,7 +110,7 @@ class AsmGen6502Internal (
         programGen.generate()
 
         if(errors.noErrors()) {
-            val output = options.outputDir.resolve("${program.name}.asm")
+            val output = OutputFiles.of(options, program.name).asm()
             val asmLines = assembly.toMutableList()
             if(options.compTarget.name==Cx16Target.NAME) {
                 scanInvalid65816instructions(asmLines)
@@ -99,7 +130,7 @@ class AsmGen6502Internal (
             if(options.dumpVariables)
                 dumpVariables()
 
-            return AssemblyProgram(program.name, options.outputDir, options.compTarget)
+            return AssemblyProgram6502(program.name, options, options.compTarget)
         } else {
             errors.report()
             return null
@@ -286,16 +317,16 @@ class AsmGen6502Internal (
                 return Pair("#<$label", "#>$label")
             } else if (e.value != null) {
                 val value = e.value!!.toInt()
-                return Pair("#<${(value and 0xff).toHex()}", "#>${((value shr 8) and 0xff).toHex()}")
+                return Pair("#<${value.toHex()}", "#>${value.toHex()}")
             }
         }
         if (e is PtNumber) {
             val value = e.number.toInt()
-            return Pair("#<${(value and 0xff).toHex()}", "#>${((value shr 8) and 0xff).toHex()}")
+            return Pair("#<${value.toHex()}", "#>${value.toHex()}")
         }
         if (e is PtAddressOf && e.identifier != null && !e.isFromArrayElement && e.dereference == null) {
             var symbol = asmVariableName(e.identifier!!)
-            if (e.identifier!!.type.isSplitWordArray) {
+            if (e.identifier!!.type.isSplitWordArray(options.compTarget)) {
                 symbol = if (e.isMsbForSplitArray) symbol + "_msb" else symbol + "_lsb"
             }
             return Pair("#<$symbol", "#>$symbol")
@@ -441,6 +472,7 @@ class AsmGen6502Internal (
                         out("  tya |  pha")
                 }
             }
+            else -> throw AssemblyError("register not available on this target")
         }
     }
 
@@ -469,6 +501,7 @@ class AsmGen6502Internal (
                         out("  pla |  tay")
                 }
             }
+            else -> throw AssemblyError("register not available on this target")
         }
     }
 
@@ -549,6 +582,7 @@ class AsmGen6502Internal (
                     CpuRegister.A -> {}
                     CpuRegister.X -> out(" tax")
                     CpuRegister.Y -> out(" tay")
+                    else -> throw AssemblyError("register not available on this target")
                 }
             }
             expr.type.isLong -> {
@@ -558,6 +592,7 @@ class AsmGen6502Internal (
                     CpuRegister.A -> {}
                     CpuRegister.X -> out(" tax")
                     CpuRegister.Y -> out(" tay")
+                    else -> throw AssemblyError("register not available on this target")
                 }
             }
             expr.type.isFloat -> {
@@ -575,6 +610,7 @@ class AsmGen6502Internal (
                     CpuRegister.A -> {}
                     CpuRegister.X -> out(" tax")
                     CpuRegister.Y -> out(" tay")
+                    else -> throw AssemblyError("register not available on this target")
                 }
             }
             else -> throw AssemblyError("weird dt")
@@ -701,7 +737,7 @@ class AsmGen6502Internal (
                         Statusflag.Pc -> out("  lda  #0 |  rol  a |  sta  $tempReg")
                         else -> throw AssemblyError("unsupported statusflag ${fromReg.statusflag}")
                     }
-                    else -> throw AssemblyError("weird CPU register ${fromReg}")
+                    else -> throw AssemblyError("weird CPU register $fromReg")
                 }
                 
                 // Store from temp to destination register
@@ -714,7 +750,7 @@ class AsmGen6502Internal (
                         Statusflag.Pc -> out("  lda  $tempReg |  asl  a")
                         else -> throw AssemblyError("unsupported statusflag ${toReg.statusflag}")
                     }
-                    else -> throw AssemblyError("weird CPU register ${toReg}")
+                    else -> throw AssemblyError("weird CPU register $toReg")
                 }
             }
         }
@@ -1403,23 +1439,35 @@ $repeatLabel""")
                         pla
                         sta  ($var2ZpPtrVar),y""")
                 } else {
-                    if (v1.address is PtNumber) {
-                        out("  lda  ${v1.address.asConstInteger()!!.toHex()} |  pha")
-                    } else if (v1.address is PtIdentifier) {
-                        var1ZpPtrVar = loadByteFromPointerIntoA(v1.address as PtIdentifier)
-                        out("  pha")
-                    } else {
-                        TODO("swap bytes not supported for this expression. Use a simpler expression, or even just a temporary variable and assignments for now. ${v1.position}")
+                    when (v1.address) {
+                        is PtNumber -> {
+                            out("  lda  ${v1.address.asConstInteger()!!.toHex()} |  pha")
+                        }
+
+                        is PtIdentifier -> {
+                            var1ZpPtrVar = loadByteFromPointerIntoA(v1.address as PtIdentifier)
+                            out("  pha")
+                        }
+
+                        else -> {
+                            TODO("swap bytes not supported for this expression. Use a simpler expression, or even just a temporary variable and assignments for now. ${v1.position}")
+                        }
                     }
-                    if (v2.address is PtNumber) {
-                        out("  lda  ${v2.address.asConstInteger()!!.toHex()}")
-                    } else if (v2.address is PtIdentifier) {
-                        var2ZpPtrVar = loadByteFromPointerIntoA(
-                            v2.address as PtIdentifier,
-                            tempZpPtrVar = "P8ZP_SCRATCH_W1"
-                        )
-                    } else {
-                        TODO("swap bytes not supported for this expression. Use a simpler expression, or even just a temporary variable and assignments for now. ${v2.position}")
+                    when (v2.address) {
+                        is PtNumber -> {
+                            out("  lda  ${v2.address.asConstInteger()!!.toHex()}")
+                        }
+
+                        is PtIdentifier -> {
+                            var2ZpPtrVar = loadByteFromPointerIntoA(
+                                v2.address as PtIdentifier,
+                                tempZpPtrVar = "P8ZP_SCRATCH_W1"
+                            )
+                        }
+
+                        else -> {
+                            TODO("swap bytes not supported for this expression. Use a simpler expression, or even just a temporary variable and assignments for now. ${v2.position}")
+                        }
                     }
 
                     if (v1.address is PtNumber) {
@@ -1937,7 +1985,7 @@ $repeatLabel""")
     private fun translate(ret: PtReturn) {
         val returnvalue = ret.children.singleOrNull() as? PtExpression      // could be a multi-value returning functioncall
         val sub = ret.definingSub()!!
-        val returnRegs = sub.returnsWhatWhere()
+        val returnRegs = sub.returnsWhatWhere(options.compTarget)
 
         if(returnvalue!=null) {
 
@@ -1948,9 +1996,8 @@ $repeatLabel""")
                     ?: throw AssemblyError("expected function call for multi-value return ${ret.position}")
                 
                 // Get the return register specs for the called function
-                val calledSub = symbolTable.lookup(fcall.name)
-                val calledReturnRegs = when(calledSub) {
-                    is StSub -> (calledSub.astNode!! as IPtSubroutine).returnsWhatWhere()
+                val calledReturnRegs = when(val calledSub = symbolTable.lookup(fcall.name)) {
+                    is StSub -> (calledSub.astNode!! as IPtSubroutine).returnsWhatWhere(options.compTarget)
                     is StExtSub -> calledSub.returns.map { it.register to it.type }
                     else -> throw AssemblyError("unexpected subroutine type for multi-value return ${ret.position}")
                 }
@@ -1976,13 +2023,23 @@ $repeatLabel""")
                 return
             }
 
+            if(sub.signature.returns.isEmpty()) {
+                // a void subroutine may return the result of a void function call (e.g. "return voidfunc(args)"):
+                // perform the call for its side effects, then return
+                val fcall = ret.children.single() as? PtFunctionCall
+                    ?: throw AssemblyError("void subroutine can only return the result of a void function call ${ret.position}")
+                functioncallAsmGen.translateFunctionCall(fcall)
+                out("  rts")
+                return
+            }
+
             val returnDt = sub.signature.returns.single()
             if (returnDt.isNumericOrBool || returnDt.isPointer) {
                 assignExpressionToRegister(returnvalue, returnRegs.single().first.registerOrPair!!, returnDt.isSigned)
             }
             else {
                 // all else take its address and assign that also to AY register pair
-                val addrOfDt = returnvalue.type.typeForAddressOf(false)
+                val addrOfDt = returnvalue.type.typeForUntypedAddressOf(false, program.memsizer)
                 val addrofValue = PtAddressOf(addrOfDt, false, returnvalue.position)
                 addrofValue.add(returnvalue)
                 addrofValue.parent = ret.parent
@@ -2010,14 +2067,19 @@ $repeatLabel""")
                 val tgt = AsmAssignTarget(TargetStorageKind.REGISTER, this, targetDt, null, it.first.position, register = it.second.first.registerOrPair!!)
                 assignExpressionTo(it.first as PtExpression, tgt)
             }
+            // if the float is not the first value, assign it now: the float load (MOVFM) clobbers A/Y,
+            // and the first non-float value is passed in AY, so it must be loaded after the float
+            if(floatIdx > 0 && floatAssign != null) {
+                assignExpressionToRegister(floatAssign.first as PtExpression, floatAssign.second.first.registerOrPair!!, false)
+            }
             if(nonFloatAssigns.isNotEmpty()) {
                 nonFloatAssigns.first().also {
                     assignExpressionToRegister(it.first as PtExpression, it.second.first.registerOrPair!!, (it.first as PtExpression).type.isSigned)
                 }
             }
-            
+
             // Assign float return LAST (so FAC1 has the correct value when we RTS)
-            if(floatAssign != null) {
+            if(floatIdx == 0 && floatAssign != null) {
                 assignExpressionToRegister(floatAssign.first as PtExpression, floatAssign.second.first.registerOrPair!!, false)
             }
         }
@@ -2042,11 +2104,17 @@ $repeatLabel""")
             throw AssemblyError("%asmbinary inside non-library/non-filesystem module not yet supported")
         val sourcePath = Path(incbin.definingBlock()!!.source.origin)
         val includedPath = sourcePath.resolveSibling(incbin.file)
-        val pathForAssembler = options.outputDir // #54: 64tass needs the path *relative to the .asm file*
-            .sanitize()
-            .relativize(includedPath.sanitize())
-            .normalize() // avoid assembler warnings (-Wportable; only some, not all)
-            .toString().replace('\\', '/')
+        val pathForAssembler = try {
+            options.outputDir // #54: 64tass needs the path *relative to the .asm file*
+                .sanitize()
+                .relativize(includedPath.sanitize())
+                .normalize() // avoid assembler warnings (-Wportable; only some, not all)
+                .toString().replace('\\', '/')
+        } catch (_: IllegalArgumentException) {
+            // Path.relativize throws when the two paths have different roots (different drive/volume)
+            errors.err("cannot determine relative path for %asmbinary file '${incbin.file}' (at ${includedPath}): it must be on the same drive/volume as the generated assembly in ${options.outputDir}", incbin.position)
+            return
+        }
         out("  .binary \"$pathForAssembler\" $offset $length")
     }
 
@@ -2161,6 +2229,13 @@ $repeatLabel""")
 
     internal fun jmp(asmLabel: String, indirect: Boolean=false, indexedX: Boolean=false) {
         if(indirect) {
+            // NOTE: `jmp (ptr)` / `jmp (ptr,x)` have the 6502 page-wrap bug
+            // on plain 6502: if the pointer's address ends in $FF, the
+            // high byte of the target is read from the same page instead
+            // of the next. 65C02 is fine. Latent: only misbehaves when the
+            // pointer variable (the operand, e.g. a computed-goto target)
+            // is placed at a $xxFF address. The new6502 codegen has the
+            // same hazard for its JUMPI/CALLI emitters.
             if(indexedX)
                 out("  jmp  ($asmLabel,x)")
             else
@@ -2216,9 +2291,7 @@ $repeatLabel""")
         // optimize pointer,indexregister if possible
 
         fun evalBytevalueWillClobberA(expr: PtExpression): Boolean {
-            if(!expr.type.isByte)
-                return true
-            return when(expr) {
+            return !expr.type.isByte || when(expr) {
                 is PtIdentifier -> false
                 is PtNumber -> false
                 is PtBool -> false
@@ -2238,12 +2311,22 @@ $repeatLabel""")
                 val constOffset = (ptrAndIndex.second as? PtNumber)?.number?.toInt()
                 if(addrOf!=null && constOffset!=null) {
                     if(addrOf.isFromArrayElement) {
+                        val arrayId = addrOf.identifier
+                        val arrayDt = arrayId?.type
+                        if(arrayDt?.sub == BaseDataType.STRUCT_INSTANCE) {
+                            val base = asmSymbolName(arrayId)
+                            out("  pha")
+                            assignmentAsmGen.computeStructArrayOffsetY(arrayDt, addrOf.arrayIndexExpr!!, constOffset)
+                            out("  pla")
+                            out("  sta  $base,y")
+                            return true
+                        }
                         TODO("address-of array element ${addrOf.position}")
                     } else if(addrOf.dereference!=null) {
                         throw AssemblyError("write &dereference, makes no sense at ${addrOf.position}")
                     } else {
                         var symbolName = asmSymbolName(addrOf.identifier!!)
-                        if(addrOf.identifier!!.type.isSplitWordArray) {
+                        if(addrOf.identifier!!.type.isSplitWordArray(options.compTarget)) {
                             symbolName = if(addrOf.isMsbForSplitArray) symbolName+"_msb" else symbolName+"_lsb"
                         }
                         out("  sta  $symbolName+${constOffset}")
@@ -2260,18 +2343,18 @@ $repeatLabel""")
                     out("  sta  (${asmSymbolName(pointervar)}),y")
                 } else {
                     // copy the pointer var to zp first
-                    val saveA = evalBytevalueWillClobberA(ptrAndIndex.first) || evalBytevalueWillClobberA(ptrAndIndex.second)
-                    if(saveA) out("  pha")
+                    // assignExpressionToVariable always clobbers A, so always save it
+                    out("  pha")
                     if(ptrAndIndex.second.isSimple()) {
                         assignExpressionToVariable(ptrAndIndex.first, "P8ZP_SCRATCH_W2", DataType.UWORD)
                         assignExpressionToRegister(ptrAndIndex.second, RegisterOrPair.Y)
-                        if(saveA) out("  pla")
+                        out("  pla")
                         out("  sta  (P8ZP_SCRATCH_W2),y")
                     } else {
                         pushCpuStack(BaseDataType.UBYTE,  ptrAndIndex.second)
                         assignExpressionToVariable(ptrAndIndex.first, "P8ZP_SCRATCH_W2", DataType.UWORD)
                         restoreRegisterStack(CpuRegister.Y, true)
-                        if(saveA) out("  pla")
+                        out("  pla")
                         out("  sta  (P8ZP_SCRATCH_W2),y")
                     }
                 }
@@ -2284,13 +2367,21 @@ $repeatLabel""")
             val constOffset = (ptrAndIndex.second as? PtNumber)?.number?.toInt()
             if(addrOf!=null && constOffset!=null) {
                 if(addrOf.isFromArrayElement) {
+                    val arrayId = addrOf.identifier
+                    val arrayDt = arrayId?.type
+                    if(arrayDt?.sub == BaseDataType.STRUCT_INSTANCE) {
+                        val base = asmSymbolName(arrayId)
+                        assignmentAsmGen.computeStructArrayOffsetY(arrayDt, addrOf.arrayIndexExpr!!, constOffset)
+                        out("  lda  $base,y")
+                        return true
+                    }
                     TODO("address-of array element ${addrOf.position}")
                 } else if(addrOf.dereference!=null) {
                     if(pointerGen.readByteByAddressOfDereference(addrOf, constOffset))
                         return true
                 } else {
                     var symbolName = asmSymbolName(addrOf.identifier!!)
-                    if(addrOf.identifier!!.type.isSplitWordArray) {
+                    if(addrOf.identifier!!.type.isSplitWordArray(options.compTarget)) {
                         symbolName = if(addrOf.isMsbForSplitArray) symbolName+"_msb" else symbolName+"_lsb"
                     }
                     out("  lda  $symbolName+${constOffset}")
@@ -2344,7 +2435,7 @@ $repeatLabel""")
                         throw AssemblyError("write &dereference, makes no sense at ${addrOf.position}")
                     } else {
                         var symbolName = asmSymbolName(addrOf.identifier!!)
-                        if(addrOf.identifier!!.type.isSplitWordArray) {
+                        if(addrOf.identifier!!.type.isSplitWordArray(options.compTarget)) {
                             symbolName = if(addrOf.isMsbForSplitArray) symbolName+"_msb" else symbolName+"_lsb"
                         }
                         out("  sta  $symbolName-${constOffset}")
@@ -2381,7 +2472,7 @@ $repeatLabel""")
                         TODO("read &dereference ${addrOf.position}")
                     } else {
                         var symbolName = asmSymbolName(addrOf.identifier!!)
-                        if(addrOf.identifier!!.type.isSplitWordArray) {
+                        if(addrOf.identifier!!.type.isSplitWordArray(options.compTarget)) {
                             symbolName = if(addrOf.isMsbForSplitArray) symbolName+"_msb" else symbolName+"_lsb"
                         }
                         out("  lda  $symbolName-${constOffset}")
@@ -2609,8 +2700,7 @@ $repeatLabel""")
             out("  $compare  P8ZP_SCRATCH_REG")
         }
 
-        val e = unwrapCasts(value)
-        when(e) {
+        when(val e = unwrapCasts(value)) {
             is PtArrayIndexer -> {
                 val constIndex = e.index.asConstInteger()
                 if(constIndex!=null) {
@@ -2991,6 +3081,7 @@ $repeatLabel""")
                 CpuRegister.A -> {}
                 CpuRegister.X -> out("  txa")
                 CpuRegister.Y -> out("  tya")
+                else -> throw AssemblyError("register not available on this target")
             }
             signExtendAXlsb(if(signed) BaseDataType.BYTE else BaseDataType.UBYTE)
             out("""
@@ -3007,6 +3098,7 @@ $repeatLabel""")
                 CpuRegister.A -> out("  ldy  #$offset |  sta  ($zpPtrVar),y")
                 CpuRegister.X -> out("  txa |  ldy  #$offset |  sta  ($zpPtrVar),y")
                 CpuRegister.Y -> out("  tya |  ldy  #$offset |  sta  ($zpPtrVar),y")
+                else -> throw AssemblyError("register not available on this target")
             }
             return
         }
@@ -3023,6 +3115,7 @@ $repeatLabel""")
                 out("  tya")
                 if(isTargetCpu(CpuType.CPU65C02)) out("  sta  ($zpPtrVar)") else out("  ldy  #0 |  sta  ($zpPtrVar),y")
             }
+            else -> throw AssemblyError("register not available on this target")
         }
     }
 

codeGenCpu6502/src/prog8/codegen/cpu6502/AsmOptimizer.kt

@@ -1,9 +1,6 @@
 package prog8.codegen.cpu6502
 
-import prog8.code.GENERATED_LABEL_PREFIX
-import prog8.code.StConstant
-import prog8.code.StMemVar
-import prog8.code.SymbolTable
+import prog8.code.*
 import prog8.code.core.ICompilationTarget
 
 // note: see https://wiki.nesdev.org/w/index.php/6502_assembly_optimisations
@@ -41,9 +38,18 @@ internal fun optimizeAssembly(lines: MutableList<String>, machine: ICompilationT
         linesByFour = runPass(optimizeUnneededTempvarInAdd(linesByFour), 4, linesByFour)
         linesByFour = runPass(optimizeTSBtoRegularOr(linesByFour), 4, linesByFour)
 
+        var linesByFive = getLinesBy(pretrimmed, lines, 5)
+        linesByFive = runPass(optimizeFoldIndexOffset(linesByFive, symbolTable), 5, linesByFive)
+
+        var linesBySix = getLinesBy(pretrimmed, lines, 6)
+        linesBySix = runPass(optimizeFoldIndexOffsetSplitWord(linesBySix, symbolTable), 6, linesBySix)
+
+        var linesBySeven = getLinesBy(pretrimmed, lines, 7)
+        linesBySeven = runPass(optimizeFoldIndexOffsetNonSplitWord(linesBySeven, symbolTable), 7, linesBySeven)
+
         var linesByFourteen = getLinesBy(pretrimmed, lines, 14)
         linesByFourteen = runPass(optimizeSameAssignments(linesByFourteen, machine, symbolTable), 14, linesByFourteen)
-        linesByFourteen = runPass(optimizeSamePointerIndexingAndUselessBeq(linesByFourteen), 14, linesByFourteen)
+        linesByFourteen = runPass(optimizeSamePointerIndexingAndUselessBeq(linesByFourteen, machine, symbolTable), 14, linesByFourteen)
         linesByFourteen = runPass(optimizeAddWordToSameVariableOrExtraRegisterLoadInWordStore(linesByFourteen), 14, linesByFourteen)
 
         if(!modified)
@@ -88,7 +94,7 @@ private fun apply(modifications: List<Modification>, lines: MutableList<String>,
                 lines.removeAt(modification.lineIndex)
             else {
                 val pretrim = pretrimmed.getOrNull(modification.lineIndex)
-                if (pretrim == null || pretrim.isBlank() || pretrim[0] == ';')
+                if (pretrim.isNullOrBlank() || pretrim[0] == ';')
                     lines.removeAt(modification.lineIndex)
                 else if (haslabelPretrimmed(pretrim)) {
                     val label = keeplabelPretrimmed(pretrim)
@@ -406,7 +412,95 @@ private fun optimizeSameAssignments(
     return mods
 }
 
-private fun optimizeSamePointerIndexingAndUselessBeq(linesByFourteen: Sequence<List<TrimmedLine>>): List<Modification> {
+private const val NO_MEMORY_WRITE = ""
+private const val UNKNOWN_WRITE_TARGET = "\u0000"
+
+/** Mnemonics that can change memory (direct stores plus read-modify-write). */
+private val memoryWritingMnemonics = setOf(
+    "sta", "stx", "sty", "stz", "inc", "dec", "asl", "lsr", "rol", "ror",
+    "and", "ora", "eor", "adc", "sbc", "bit"
+)
+
+/** Control transfers that can reach the code after them with a different Y, or change it. */
+private val flowMnemonics = setOf(
+    "jsr", "rts", "rti", "jmp", "bra", "brk",
+    "bcc", "bcs", "beq", "bmi", "bne", "bpl", "bvc", "bvs", "bbs", "bbr"
+)
+
+private fun TrimmedLine.mnemonic(): String =
+    instruction.trimStart().takeWhile { !it.isWhitespace() }.lowercase()
+
+/** True if the line carries a label, with or without an instruction following it. */
+private fun TrimmedLine.hasLabel(): Boolean {
+    return haslabelPretrimmed(trimmed) || trimmed.isNotEmpty() && (trimmed[0] == '+' || trimmed[0] == '-')
+}
+
+/** The memory operand this instruction writes, with any +offset and ,index suffix stripped.
+ *  Returns [NO_MEMORY_WRITE] for immediate and implied forms, and [UNKNOWN_WRITE_TARGET] for
+ *  indirect forms, where the destination cannot be determined statically. */
+private fun TrimmedLine.memoryWriteBase(): String {
+    val op = instruction.extractOperandTrimmed()
+    if (op.isEmpty() || op.startsWith("#") || op == "a") return NO_MEMORY_WRITE
+    if (op.startsWith("(")) return UNKNOWN_WRITE_TARGET
+    return op.removeSuffix(",y").removeSuffix(",x").substringBefore('+')
+}
+
+/** True if this line may write to [base]. Conservative: an indirect store is assumed to. */
+private fun TrimmedLine.mayWriteBase(base: String): Boolean {
+    if (mnemonic() !in memoryWritingMnemonics) return false
+    val target = memoryWriteBase()
+    if (target == NO_MEMORY_WRITE) return false
+    if (target == UNKNOWN_WRITE_TARGET) return true
+    return target.equals(base, ignoreCase = true)
+}
+
+/** True if operand is safe to read more than once: an immediate, or memory that is not an
+ *  IO location (an IO read can return a different value without the CPU having written it). */
+private fun isNonVolatileIndexOperand(indexLine: TrimmedLine, machine: ICompilationTarget, symbolTable: SymbolTable): Boolean {
+    val operand = indexLine.instruction.extractOperandTrimmed()
+    if (operand.isEmpty() || operand.startsWith("#")) return true
+    val addr = getAddressArg(indexLine.instruction, symbolTable)
+    return if (addr == null) !looksLikeIOAddress(operand, machine)
+           else !machine.isIOAddress(addr)
+}
+
+/** Decides whether the index reload at [reloadIdx] inside [window] can be deleted.
+ *
+ *  The reload is redundant only if Y still holds the value it would have loaded. That requires
+ *  the reload to load the very same operand as the original index load, and none of the
+ *  following in between:
+ *  - modifies Y
+ *  - writes the index operand itself (so `ldy i / lda arr,y / inc i / ldy i / sta arr,y` keeps
+ *    its reload: without it the store lands one element earlier)
+ *  - is a call or control transfer, since a callee can clobber Y and another path can arrive
+ *    with a different Y
+ *  - carries a label, for the same reason
+ *  The index operand must also not be a volatile IO location.
+ */
+private fun canDropIndexReload(
+    window: List<TrimmedLine>,
+    reloadIdx: Int,
+    indexLine: TrimmedLine,
+    machine: ICompilationTarget,
+    symbolTable: SymbolTable
+): Boolean {
+    if (reloadIdx <= 1) return false
+    val operand = indexLine.instruction.extractOperandTrimmed()
+    // the reload has to load the same index, or Y holds a different one and is not redundant
+    if (window[reloadIdx].instruction.extractOperandTrimmed() != operand) return false
+    val base = if (operand.startsWith("#") || operand.isEmpty()) NO_MEMORY_WRITE
+               else operand.removeSuffix(",y").removeSuffix(",x").substringBefore('+')
+    for (idx in 1 until reloadIdx) {
+        val line = window[idx]
+        if (line.instruction.modifiesYRegister()) return false
+        if (line.mnemonic() in flowMnemonics) return false
+        if (line.hasLabel()) return false
+        if (base != NO_MEMORY_WRITE && line.mayWriteBase(base)) return false
+    }
+    return isNonVolatileIndexOperand(indexLine, machine, symbolTable)
+}
+
+private fun optimizeSamePointerIndexingAndUselessBeq(linesByFourteen: Sequence<List<TrimmedLine>>, machine: ICompilationTarget, symbolTable: SymbolTable): List<Modification> {
 
     // Optimize same pointer indexing where for instance we load and store to the same ptr index in Y
     // if Y isn't modified in between we can omit the second LDY:
@@ -420,30 +514,48 @@ private fun optimizeSamePointerIndexingAndUselessBeq(linesByFourteen: Sequence<L
     for (lines in linesByFourteen) {
         val f1 = lines[0].instruction
         val f2 = lines[1].instruction
-        val f3 = lines[2].instruction
+        //val f3 = lines[2].instruction
         val f4 = lines[3].instruction
         val f5 = lines[4].instruction
         val f6 = lines[5].instruction
 
         if(f1.startsWith("ldy ") && f2.startsWith("lda ") && f4.startsWith("ldy ") && f5.startsWith("sta ")) {
-            val firstvalue = f1.extractOperandTrimmed()
+            //val firstvalue = f1.extractOperandTrimmed()
             val secondvalue = f2.extractOperandTrimmed()
-            val fourthvalue = f4.extractOperandTrimmed()
+            //val fourthvalue = f4.extractOperandTrimmed()
             val fifthvalue = f5.extractOperandTrimmed()
-            if(!f3.modifiesYRegister() && firstvalue==fourthvalue && secondvalue==fifthvalue && secondvalue.endsWith(",y") && fifthvalue.endsWith(",y")) {
+            if(secondvalue==fifthvalue && secondvalue.endsWith(",y") && fifthvalue.endsWith(",y") &&
+                    canDropIndexReload(lines, 3, lines[0], machine, symbolTable)) {
                 mods.add(Modification(lines[3].index, true, null))
             }
         }
         if(f1.startsWith("ldy ") && f2.startsWith("lda ") && f5.startsWith("ldy ") && f6.startsWith("sta ")) {
-            val firstvalue = f1.extractOperandTrimmed()
             val secondvalue = f2.extractOperandTrimmed()
-            val fifthvalue = f5.extractOperandTrimmed()
             val sixthvalue = f6.extractOperandTrimmed()
-            if(!f3.modifiesYRegister() && !f4.modifiesYRegister() && firstvalue==fifthvalue && secondvalue==sixthvalue && secondvalue.endsWith(",y") && sixthvalue.endsWith(",y")) {
+            if(secondvalue==sixthvalue && secondvalue.endsWith(",y") && sixthvalue.endsWith(",y") &&
+                    canDropIndexReload(lines, 4, lines[0], machine, symbolTable)) {
                 mods.add(Modification(lines[4].index, true, null))
             }
         }
 
+        // Relaxed variant: two consecutive indexed accesses, not necessarily a matched load/store
+        // pair. Only the index reload has to be the same, and nothing in between may disturb it.
+        //    ldy  V
+        //    lda  A,y
+        //    ...            <-- does not modify Y, does not write V, no call or label
+        //    ldy  V        <-- can be removed
+        //    lda  B,y
+        if(f1.startsWith("ldy ") && f2.startsWith("lda ") && f4.startsWith("ldy ") && f5.startsWith("lda ") &&
+                f2.endsWith(",y") && f5.endsWith(",y") &&
+                canDropIndexReload(lines, 3, lines[0], machine, symbolTable)) {
+            mods.add(Modification(lines[3].index, true, null))
+        }
+        if(f1.startsWith("ldy ") && f2.startsWith("lda ") && f5.startsWith("ldy ") && f6.startsWith("lda ") &&
+                f2.endsWith(",y") && f6.endsWith(",y") &&
+                canDropIndexReload(lines, 4, lines[0], machine, symbolTable)) {
+            mods.add(Modification(lines[4].index, true, null))
+        }
+
 
         /*
     beq  +
@@ -631,11 +743,9 @@ private fun looksLikeIOAddress(operand: String, machine: ICompilationTarget): Bo
     }
     // Try to parse decimal address (e.g. "53280" = $d000) and delegate to machine
     val decVal = operand.toUIntOrNull()
-    if(decVal != null)
-        return machine.isIOAddress(decVal)
     // For unresolved symbolic names we can't determine - assume NOT IO
     // (the symbol table lookup already failed, so we have no better info)
-    return false
+    return decVal != null && machine.isIOAddress(decVal)
 }
 
 private fun optimizeIncDec(linesByFour: Sequence<List<TrimmedLine>>): List<Modification> {
@@ -889,6 +999,205 @@ private fun optimizeUnneededTempvarInAdd(linesByFour: Sequence<List<TrimmedLine>
     return mods
 }
 
+private fun withLabel(line: TrimmedLine, newInstruction: String): String {
+    val label = keeplabelPretrimmed(line.value)
+    return if (label.isEmpty()) newInstruction else "$label ${newInstruction.trimStart()}"
+}
+
+private fun scaleConstant(constStr: String, scale: Int): String {
+    val value = if (constStr.startsWith("$"))
+        constStr.substring(1).toInt(16)
+    else
+        constStr.toInt()
+    val scaled = value * scale
+    return if (constStr.startsWith("$"))
+        "$${scaled.toString(16).padStart(2, '0')}"
+    else
+        scaled.toString()
+}
+
+private fun isArrayByteSizedAndSmall(name: String, symbolTable: SymbolTable): Boolean {
+    return when (val symbol = symbolTable.flat[name]) {
+        is StStaticVariable -> {
+            val len = symbol.length
+            (symbol.dt.isByteArray || symbol.dt.isString) && len != null && len <= 256u
+        }
+        is StMemVar -> {
+            val len = symbol.length
+            symbol.dt.isByteArray && len != null && len <= 256u
+        }
+        else -> false
+    }
+}
+
+private fun isArrayWordSizedAndSmall(name: String, symbolTable: SymbolTable): Boolean {
+    return when (val symbol = symbolTable.flat[name]) {
+        is StStaticVariable -> {
+            val len = symbol.length
+            symbol.dt.isWordArray && len != null && len <= 256u
+        }
+        is StMemVar -> {
+            val len = symbol.length
+            symbol.dt.isWordArray && len != null && len <= 256u
+        }
+        else -> false
+    }
+}
+
+private fun optimizeFoldIndexOffset(
+    linesByFive: Sequence<List<TrimmedLine>>,
+    symbolTable: SymbolTable
+): List<Modification> {
+    val mods = mutableListOf<Modification>()
+
+    for (lines in linesByFive) {
+        val first = lines[0].instruction
+        val second = lines[1].instruction
+        val third = lines[2].instruction
+        val fourth = lines[3].instruction
+        val fifth = lines[4].instruction
+
+        if (!first.startsWith("lda ") || second != "clc" || !third.startsWith("adc "))
+            continue
+        if (fourth != "tay" && fourth != "tax")
+            continue
+
+        val reg = fourth[2]
+        val constOperand = third.extractOperandTrimmed()
+        if (!constOperand.startsWith("#"))
+            continue
+        val constValue = constOperand.substring(1)
+
+        val fifthOperand = fifth.extractOperandTrimmed()
+        if (!fifthOperand.endsWith(",$reg") || fifthOperand.startsWith("("))
+            continue
+
+        val addrPart = fifthOperand.dropLast(2)
+        if (!isArrayByteSizedAndSmall(addrPart, symbolTable)) continue
+
+        val mnemonic = fifth.take(3)
+        val varName = first.extractOperandTrimmed()
+        mods.add(Modification(lines[0].index, false, withLabel(lines[0], "  ld$reg  $varName")))
+        mods.add(Modification(lines[1].index, true, null))
+        mods.add(Modification(lines[2].index, true, null))
+        mods.add(Modification(lines[3].index, true, null))
+        mods.add(Modification(lines[4].index, false, withLabel(lines[4], "  $mnemonic  $addrPart+$constValue,$reg")))
+    }
+
+    return mods
+}
+
+private fun optimizeFoldIndexOffsetSplitWord(
+    linesBySix: Sequence<List<TrimmedLine>>,
+    symbolTable: SymbolTable
+): List<Modification> {
+    val mods = mutableListOf<Modification>()
+
+    for (lines in linesBySix) {
+        val first = lines[0].instruction
+        val second = lines[1].instruction
+        val third = lines[2].instruction
+        val fourth = lines[3].instruction
+        val fifth = lines[4].instruction
+        val sixth = lines[5].instruction
+
+        if (!first.startsWith("lda ") || second != "clc" || !third.startsWith("adc "))
+            continue
+        if (fourth != "tay" && fourth != "tax")
+            continue
+
+        val reg = fourth[2]
+        val constOperand = third.extractOperandTrimmed()
+        if (!constOperand.startsWith("#"))
+            continue
+        val constValue = constOperand.substring(1)
+
+        val fifthOp = fifth.extractOperandTrimmed()
+        val sixthOp = sixth.extractOperandTrimmed()
+        if (!fifthOp.endsWith(",$reg") || fifthOp.startsWith("("))
+            continue
+        if (!sixthOp.endsWith(",$reg") || sixthOp.startsWith("("))
+            continue
+
+        val lsbAddr = fifthOp.dropLast(2)
+        val msbAddr = sixthOp.dropLast(2)
+        if (!lsbAddr.endsWith("_lsb") || !msbAddr.endsWith("_msb"))
+            continue
+        val baseName = lsbAddr.dropLast(4)
+        if (msbAddr != baseName + "_msb")
+            continue
+
+        if (!isArrayWordSizedAndSmall(baseName, symbolTable))
+            continue
+
+        val lsbMnemonic = fifth.take(3)
+        val msbMnemonic = sixth.take(3)
+        val varName = first.extractOperandTrimmed()
+
+        mods.add(Modification(lines[0].index, false, withLabel(lines[0], "  ld$reg  $varName")))
+        mods.add(Modification(lines[1].index, true, null))
+        mods.add(Modification(lines[2].index, true, null))
+        mods.add(Modification(lines[3].index, true, null))
+        mods.add(Modification(lines[4].index, false, withLabel(lines[4], "  $lsbMnemonic  $lsbAddr+$constValue,$reg")))
+        mods.add(Modification(lines[5].index, false, withLabel(lines[5], "  $msbMnemonic  $msbAddr+$constValue,$reg")))
+    }
+
+    return mods
+}
+
+private fun optimizeFoldIndexOffsetNonSplitWord(
+    linesBySeven: Sequence<List<TrimmedLine>>,
+    symbolTable: SymbolTable
+): List<Modification> {
+    val mods = mutableListOf<Modification>()
+
+    for (lines in linesBySeven) {
+        val first = lines[0].instruction
+        val second = lines[1].instruction
+        val third = lines[2].instruction
+        val fourth = lines[3].instruction
+        val fifth = lines[4].instruction
+        val sixth = lines[5].instruction
+        val seventh = lines[6].instruction
+
+        if (!first.startsWith("lda ") || second != "clc" || !third.startsWith("adc "))
+            continue
+        if (fourth != "asl  a" || (fifth != "tay" && fifth != "tax"))
+            continue
+
+        val reg = fifth[2]
+        val constOperand = third.extractOperandTrimmed()
+        if (!constOperand.startsWith("#"))
+            continue
+        val constValue = constOperand.substring(1)
+        val scaledConst = scaleConstant(constValue, 2)
+
+        val sixthOp = sixth.extractOperandTrimmed()
+        val seventhOp = seventh.extractOperandTrimmed()
+        if (!sixthOp.endsWith(",$reg") || sixthOp.startsWith("("))
+            continue
+        if (!seventhOp.endsWith(",$reg") || seventhOp.startsWith("("))
+            continue
+
+        val baseAddr = sixthOp.dropLast(2)
+        if (seventhOp != "$baseAddr+1,$reg")
+            continue
+
+        if (!isArrayWordSizedAndSmall(baseAddr, symbolTable))
+            continue
+
+        val lowMnemonic = sixth.take(3)
+        val highMnemonic = seventh.take(3)
+
+        mods.add(Modification(lines[1].index, true, null))
+        mods.add(Modification(lines[2].index, true, null))
+        mods.add(Modification(lines[5].index, false, withLabel(lines[5], "  $lowMnemonic  $baseAddr+$scaledConst,$reg")))
+        mods.add(Modification(lines[6].index, false, withLabel(lines[6], "  $highMnemonic  $baseAddr+1+$scaledConst,$reg")))
+    }
+
+    return mods
+}
+
 private fun optimizeAddWordToSameVariableOrExtraRegisterLoadInWordStore(linesByFourteen: Sequence<List<TrimmedLine>>): List<Modification> {
     /*
         ; FIRST SEQUYENCE: P8ZP_SCRATCH_PTR += AY :

codeGenCpu6502/src/prog8/codegen/cpu6502/BuiltinFunctionsAsmGen.kt

@@ -20,6 +20,25 @@ import prog8.codegen.cpu6502.assignment.*
         translateFunctioncall(fcall, null, discardResult = true)
     }
 
+    /**
+     * Several builtins can only produce their result in a CPU register, and which one that is
+     * depends on the width of the result: the byte-returning ones (lsb/msb) take a single register,
+     * the word-returning ones (lsw/msw) and popw take a register pair, and the builtin picks the
+     * specific register. [hardwareReg] is the register to fall back on when the caller has no
+     * preference, or when it asked for something the builtin cannot deliver to directly.
+     *
+     * The caller's requested register is honoured when it is a real CPU register. A CX16 virtual
+     * register (for example an asmsub parameter declared as @R0) is not: those are memory locations,
+     * so the builtin produces its result in [hardwareReg] and the caller's assignment then stores it
+     * into the virtual register. Passing the virtual register through would make the builtin reject
+     * it as an invalid result register.
+     *
+     * (funcPeekW and funcMkword already handle virtual registers themselves.)
+     */
+    private fun withResultRegister(requested: RegisterOrPair?, hardwareReg: RegisterOrPair, compute: (RegisterOrPair) -> Array<RegisterOrPair>): Array<RegisterOrPair> =
+        if (requested != null && requested !in Cx16VirtualRegisters) compute(requested)
+        else compute(hardwareReg)
+
     private fun translateFunctioncall(fcall: PtFunctionCall, firstReturnRegister: RegisterOrPair?, discardResult: Boolean): Array<RegisterOrPair> {
         
         // Note: returns the actual register(s) that the return value is in, so the caller should still make sure to place it into the actual correct destination.
@@ -31,12 +50,12 @@ import prog8.codegen.cpu6502.assignment.*
         val sscope = fcall.definingISub()
 
         return when (fcall.name) {
-            "msw" -> funcMsw(fcall, firstReturnRegister ?: RegisterOrPair.AY)
-            "lsw" -> funcLsw(fcall, firstReturnRegister ?: RegisterOrPair.AY)
-            "msb" -> funcMsb(fcall, false, firstReturnRegister ?: RegisterOrPair.A)
-            "msb__long" -> funcMsb(fcall, true, firstReturnRegister ?: RegisterOrPair.A)
-            "lsb" -> funcLsb(fcall, false, firstReturnRegister ?: RegisterOrPair.A)
-            "lsb__long" -> funcLsb(fcall,true, firstReturnRegister ?: RegisterOrPair.A)
+            "msw" -> withResultRegister(firstReturnRegister, RegisterOrPair.AY) { funcMsw(fcall, it) }
+            "lsw" -> withResultRegister(firstReturnRegister, RegisterOrPair.AY) { funcLsw(fcall, it) }
+            "msb" -> withResultRegister(firstReturnRegister, RegisterOrPair.A) { funcMsb(fcall, false, it) }
+            "msb__long" -> withResultRegister(firstReturnRegister, RegisterOrPair.A) { funcMsb(fcall, true, it) }
+            "lsb" -> withResultRegister(firstReturnRegister, RegisterOrPair.A) { funcLsb(fcall, false, it) }
+            "lsb__long" -> withResultRegister(firstReturnRegister, RegisterOrPair.A) { funcLsb(fcall, true, it) }
             "mkword" -> funcMkword(fcall, firstReturnRegister ?: RegisterOrPair.AY)
             "mklong", "mklong2" -> funcMklong(fcall, firstReturnRegister ?: RegisterOrPair.R14R15)
             "clamp__byte", "clamp__ubyte" -> funcClamp(fcall)
@@ -98,8 +117,8 @@ import prog8.codegen.cpu6502.assignment.*
             "pushw" -> funcPushW(fcall)
             "pushl" -> funcPushL(fcall)
             "pushf" -> funcPushF(fcall)
-            "pop" -> funcPop(firstReturnRegister ?: RegisterOrPair.A)
-            "popw" -> funcPopW(firstReturnRegister ?: RegisterOrPair.AY)
+            "pop" -> withResultRegister(firstReturnRegister, RegisterOrPair.A) { funcPop(it) }
+            "popw" -> withResultRegister(firstReturnRegister, RegisterOrPair.AY) { funcPopW(it) }
             "popl" -> funcPopL()
             "popf" -> funcPopF()
             else -> throw AssemblyError("sizeof must have been replaced with a constant  ${fcall.position}")
@@ -247,11 +266,15 @@ import prog8.codegen.cpu6502.assignment.*
             }
             BaseDataType.UWORD -> {
                 asmgen.assignExpressionToRegister(fcall.args[0], RegisterOrPair.AY)
+                // prog8_math.square does its own range check and falls back to the general
+                // multiply_words for values outside -255..255, so the fast path stays valid.
                 asmgen.out("  jsr  prog8_math.square")     // result is in AY
                 return arrayOf(RegisterOrPair.AY)
             }
             BaseDataType.LONG -> {
                 asmgen.assignExpressionToRegister(fcall.args[0], RegisterOrPair.R14R15, true)
+                // prog8_math.square_long does its own range check and falls back to the general
+                // multiply_longs when the long doesn't fit in a signed 16-bit word.
                 asmgen.out("  jsr  prog8_math.square_long")     // result is in R14:R15
                 return arrayOf(RegisterOrPair.R14R15)
             }
@@ -376,6 +399,11 @@ import prog8.codegen.cpu6502.assignment.*
 
         val identifier = fcall.args[0] as? PtIdentifier
         if(identifier!=null) {
+            // NOTE: `jmp (ptr)` below has the 6502 page-wrap bug on plain
+            // 6502 if the pointer (the identifier, a variable holding a
+            // routine address) lands at $xxFF. 65C02 is fine. Latent; only
+            // misbehaves for variables placed at a $xxFF address. The
+            // new6502 codegen's CALLI has the same hazard.
             asmgen.out("""
                 ; push a return address so the jmp becomes indirect jsr
                 lda  #>((+)-1)
@@ -387,6 +415,9 @@ import prog8.codegen.cpu6502.assignment.*
             return arrayOf(RegisterOrPair.AY)
         }
 
+        // NOTE: same 6502 page-wrap pitfall as above: the pointer is
+        // staged through P8ZP_SCRATCH_W2 before the indirect jmp, so the
+        // scratch word must not land at $xxFF on plain 6502. 65C02 is fine.
         asmgen.assignExpressionToVariable(fcall.args[0], asmgen.asmVariableName("P8ZP_SCRATCH_W2"), DataType.UWORD)     // jump address
         asmgen.out("""
                 ; push a return address so the jmp becomes indirect jsr
@@ -401,8 +432,7 @@ import prog8.codegen.cpu6502.assignment.*
     }
 
     private fun funcCallFar(fcall: PtFunctionCall): Array<RegisterOrPair> {
-        val targetName = asmgen.options.compTarget.name
-        val jsrfar = when(targetName) {
+        val jsrfar = when(val targetName = asmgen.options.compTarget.name) {
             "cx16" -> "cx16.JSRFAR"
             "c64" -> "c64.x16jsrfar"
             "c128" -> "c128.x16jsrfar"
@@ -418,7 +448,7 @@ import prog8.codegen.cpu6502.assignment.*
                 .byte  $constBank""")
         } else {
             if(asmgen.options.romable)
-                TODO("non-const callfar (jsrfar) yet that's usable in ROM  ${fcall.position}")
+                asmgen.errors.err("non-const callfar (jsrfar) has no ROMable code generation", fcall.position)
             // self-modifying code: set jsrfar arguments
             asmgen.assignExpressionToRegister(fcall.args[0], RegisterOrPair.A)      // bank
             asmgen.out("  sta  (++)+0")
@@ -475,7 +505,7 @@ import prog8.codegen.cpu6502.assignment.*
                 .byte  $constBank""")
         } else {
             if(asmgen.options.romable)
-                TODO("non-const callfar2 (jsrfar) yet that's usable in ROM  ${fcall.position}")
+                asmgen.errors.err("non-const callfar2 (jsrfar) has no ROMable code generation", fcall.position)
             // self-modifying code: set jsrfar arguments
             asmgen.assignExpressionToRegister(fcall.args[0], RegisterOrPair.A)      // bank
             asmgen.out("  sta  (++)+0")
@@ -1137,7 +1167,7 @@ import prog8.codegen.cpu6502.assignment.*
                 val msbAdd: Int
                 if(indexer.splitWords) {
                     val arrayVariable = indexer.variable ?: TODO("ptr indexing ${indexer.position}")
-                    indexer.setChild(0, PtIdentifier(arrayVariable.name + (if(msb) "_msb" else "_lsb"), DataType.arrayFor(BaseDataType.UBYTE, false), arrayVariable.position))
+                    indexer.setChild(0, PtIdentifier(arrayVariable.name + (if(msb) "_msb" else "_lsb"), DataType.arrayFor(BaseDataType.UBYTE, asmgen.options.compTarget), arrayVariable.position))
                     elementSize = 1
                     msbAdd = 0
                 } else {
@@ -1432,7 +1462,7 @@ import prog8.codegen.cpu6502.assignment.*
                     return emptyArray()
                 } else if(addressOfIdentifier!=null && (addressOfIdentifier.type.isWord || addressOfIdentifier.type.isPointer || addressOfIdentifier.type.isByteArray)) {
                     var varname = asmgen.asmVariableName(addressOfIdentifier)
-                    if(addressOfIdentifier.type.isSplitWordArray) {
+                    if(addressOfIdentifier.type.isSplitWordArray(asmgen.options.compTarget)) {
                         val addrOf = result.first as PtAddressOf
                         varname = if(addrOf.isMsbForSplitArray) varname+"_msb" else varname+"_lsb"
                     }
@@ -1814,7 +1844,7 @@ import prog8.codegen.cpu6502.assignment.*
                     }
                 } else if(addressOfIdentifier!=null && (addressOfIdentifier.type.isWord || addressOfIdentifier.type.isPointer || addressOfIdentifier.type.isByteArray)) {
                     var varname = asmgen.asmVariableName(addressOfIdentifier)
-                    if(addressOfIdentifier.type.isSplitWordArray) {
+                    if(addressOfIdentifier.type.isSplitWordArray(asmgen.options.compTarget)) {
                         val addrOf = result.first as PtAddressOf
                         varname = if(addrOf.isMsbForSplitArray) varname+"_msb" else varname+"_lsb"
                     }
@@ -1973,11 +2003,252 @@ import prog8.codegen.cpu6502.assignment.*
         return arrayOf(RegisterOrPair.R14R15)
     }
 
+    private fun inlineClampByte(value: PtExpression, minVal: Int, maxVal: Int, signed: Boolean) {
+        val typeMin = if (signed) -128 else 0
+        val typeMax = if (signed) 127 else 255
+        assignAsmGen.assignExpressionToRegister(value, RegisterOrPair.A, signed)
+        when {
+            minVal == typeMin && maxVal == typeMax -> {
+                // no-op: the value in A is already the clamped result
+            }
+            minVal == typeMin -> {
+                // upper clamp only: min(x, maxVal)
+                emitByteUpperClamp(maxVal, signed)
+            }
+            maxVal == typeMax -> {
+                // lower clamp only: max(x, minVal)
+                emitByteLowerClamp(minVal, signed)
+            }
+            else -> {
+                // two-sided clamp: upper then lower, sequentially on the running value in A
+                emitByteUpperClamp(maxVal, signed)
+                emitByteLowerClamp(minVal, signed)
+            }
+        }
+    }
+
+    private fun emitByteUpperClamp(maxVal: Int, signed: Boolean) {
+        // result = min(x, maxVal); x is in A on entry
+        // mirrors the proven func_clamp_byte sequence (sec/sbc + bmi means "x < max")
+        val imm = "#${(maxVal and 0xff).toHex()}"
+        if (signed) {
+            val keep = asmgen.makeLabel("clampkeep")
+            val done = asmgen.makeLabel("clampdone")
+            asmgen.out(listOf(
+                "            tay",
+                "            sec",
+                "            sbc  $imm",
+                "            bvc  +",
+                "            eor  #128",
+                "+           bmi  $keep",
+                "            lda  $imm",
+                "            jmp  $done",
+                "$keep       tya",
+                done
+            ).joinToString("\n"))
+        } else {
+            val keep = asmgen.makeLabel("clampkeep")
+            asmgen.out(listOf(
+                "            cmp  $imm",
+                "            bcc  $keep",
+                "            lda  $imm",
+                keep
+            ).joinToString("\n"))
+        }
+    }
+
+    private fun emitByteLowerClamp(minVal: Int, signed: Boolean) {
+        // result = max(x, minVal); x is in A on entry
+        // mirrors the proven func_clamp_byte sequence (sec/sbc + bpl means "x >= min")
+        val imm = "#${(minVal and 0xff).toHex()}"
+        if (signed) {
+            val keep = asmgen.makeLabel("clampkeep")
+            val done = asmgen.makeLabel("clampdone")
+            asmgen.out(listOf(
+                "            tay",
+                "            sec",
+                "            sbc  $imm",
+                "            bvc  +",
+                "            eor  #128",
+                "+           bpl  $keep",
+                "            lda  $imm",
+                "            jmp  $done",
+                "$keep       tya",
+                done
+            ).joinToString("\n"))
+        } else {
+            val keep = asmgen.makeLabel("clampkeep")
+            asmgen.out(listOf(
+                "            cmp  $imm",
+                "            bcs  $keep",
+                "            lda  $imm",
+                keep
+            ).joinToString("\n"))
+        }
+    }
+
+    private fun funcClampWordConst(minExpr: PtExpression, maxExpr: PtExpression, minOp: Pair<String, String>, maxOp: Pair<String, String>, signed: Boolean) {
+        // value is already in AY on entry, result in AY
+        val minVal = (asmgen.unwrapCasts(minExpr) as? PtNumber)?.number?.toInt()
+        val maxVal = (asmgen.unwrapCasts(maxExpr) as? PtNumber)?.number?.toInt()
+        val typeMin = if (signed) -32768 else 0
+        val typeMax = if (signed) 32767 else 65535
+        val oneSided = minVal != null && minVal == typeMin || maxVal != null && maxVal == typeMax
+        val fullRange = minVal != null && maxVal != null && minVal == typeMin && maxVal == typeMax
+        when {
+            fullRange -> {
+                // no-op: the value in AY is already the clamped result
+            }
+            signed && minVal == typeMin -> {
+                // upper clamp only
+                emitSignedWordClampProlog()
+                emitSignedWordUpperCompare(maxOp)
+                emitWordLoadValueFromW1()
+            }
+            signed && maxVal == typeMax -> {
+                // lower clamp only
+                emitSignedWordClampProlog()
+                emitSignedWordLowerCompare(minOp)
+            }
+            signed -> {
+                emitSignedWordClampProlog()
+                emitSignedWordUpperCompare(maxOp)
+                emitSignedWordLowerCompare(minOp)
+            }
+            !oneSided -> {
+                // uword, two-sided clamp
+                emitWordUpperClamp(maxOp)
+                emitWordLowerClamp(minOp)
+            }
+            minVal == typeMin -> {
+                // uword, upper clamp only
+                emitWordUpperClamp(maxOp)
+            }
+            else -> {
+                // uword, lower clamp only
+                emitWordLowerClamp(minOp)
+            }
+        }
+    }
+
+    private fun emitSignedWordClampProlog() {
+        asmgen.out(listOf(
+            "            sta  P8ZP_SCRATCH_W1",
+            "            sty  P8ZP_SCRATCH_W1+1"
+        ).joinToString("\n"))
+    }
+
+    private fun emitWordLoadValueFromW1() {
+        asmgen.out(listOf(
+            "            lda  P8ZP_SCRATCH_W1",
+            "            ldy  P8ZP_SCRATCH_W1+1"
+        ).joinToString("\n"))
+    }
+
+    // signed compare of max/min (16-bit immediates) against the running value in P8ZP_SCRATCH_W1.
+    // The result of the two-sided compare is the high byte of (bound - value), with the
+    // overflow-compensated sign bit set, mirroring the func_clamp_word sequence.
+
+    private fun emitSignedWordUpperCompare(maxOp: Pair<String, String>) {
+        // if value > max: running value becomes max
+        val keep = asmgen.makeLabel("clampkeep")
+        val done = asmgen.makeLabel("clampdone")
+        asmgen.out(listOf(
+            "            ldy  ${maxOp.second}",
+            "            lda  ${maxOp.first}",
+            "            cmp  P8ZP_SCRATCH_W1",
+            "            tya",
+            "            sbc  P8ZP_SCRATCH_W1+1",
+            "            bvc  $keep",
+            "            eor  #$80",
+            keep,
+            "            bpl  $done",
+            "            lda  ${maxOp.first}",
+            "            ldy  ${maxOp.second}",
+            "            sta  P8ZP_SCRATCH_W1",
+            "            sty  P8ZP_SCRATCH_W1+1",
+            done
+        ).joinToString("\n"))
+    }
+
+    private fun emitSignedWordLowerCompare(minOp: Pair<String, String>) {
+        // if value < min: result becomes min, otherwise the running value (in W1)
+        val keep = asmgen.makeLabel("clampkeep")
+        val set = asmgen.makeLabel("clampset")
+        val done = asmgen.makeLabel("clampdone")
+        asmgen.out(listOf(
+            "            ldy  ${minOp.second}",
+            "            lda  ${minOp.first}",
+            "            cmp  P8ZP_SCRATCH_W1",
+            "            tya",
+            "            sbc  P8ZP_SCRATCH_W1+1",
+            "            bvc  $keep",
+            "            eor  #$80",
+            keep,
+            "            bpl  $set",
+            "            lda  P8ZP_SCRATCH_W1",
+            "            ldy  P8ZP_SCRATCH_W1+1",
+            "            jmp  $done",
+            set,
+            "            lda  ${minOp.first}",
+            "            ldy  ${minOp.second}",
+            done
+        ).joinToString("\n"))
+    }
+
+    private fun emitWordUpperClamp(maxOp: Pair<String, String>) {
+        // uword: clamp the value in AY down to maxOp (result = min(value, max))
+        val keep = asmgen.makeLabel("clampkeep")
+        val set = asmgen.makeLabel("clampset")
+        asmgen.out(listOf(
+            "            cpy  ${maxOp.second}",
+            "            bcc  $keep",
+            "            bne  $set",
+            "            cmp  ${maxOp.first}",
+            "            bcc  $keep",
+            "            beq  $keep",
+            set,
+            "            lda  ${maxOp.first}",
+            "            ldy  ${maxOp.second}",
+            keep
+        ).joinToString("\n"))
+    }
+
+    private fun emitWordLowerClamp(minOp: Pair<String, String>) {
+        // uword: clamp the value in AY up to minOp (result = max(value, min))
+        val keep = asmgen.makeLabel("clampkeep")
+        val set = asmgen.makeLabel("clampset")
+        val done = asmgen.makeLabel("clampdone")
+        asmgen.out(listOf(
+            "            cpy  ${minOp.second}",
+            "            bcc  $set",
+            "            bne  $keep",
+            "            cmp  ${minOp.first}",
+            "            bcc  $set",
+            "            beq  $keep",
+            keep,
+            "            jmp  $done",
+            set,
+            "            lda  ${minOp.first}",
+            "            ldy  ${minOp.second}",
+            done
+        ).joinToString("\n"))
+    }
+
     private fun funcClamp(fcall: PtFunctionCall): Array<RegisterOrPair> {
         val signed = fcall.type.isSigned
         val zpCheck = { name: String -> asmgen.isZpVar(name) }
         when {
             fcall.type.isByte -> {
+                if (!signed && fcall.args[1] is PtNumber && fcall.args[2] is PtNumber) {
+                    // unsigned (ubyte) clamp with constant bounds: emit a small inline clamp.
+                    // Only worthwhile for ubyte: the signed inline sequence is larger than the
+                    // shared jsr func_clamp_byte subroutine, so signed clamps keep using that.
+                    val minVal = (fcall.args[1] as PtNumber).number.toInt()
+                    val maxVal = (fcall.args[2] as PtNumber).number.toInt()
+                    inlineClampByte(fcall.args[0], minVal, maxVal, signed)
+                    return arrayOf(RegisterOrPair.A)
+                }
                 if (fcall.args[0].isSimple(zpCheck) && fcall.args[1].isSimple(zpCheck) && fcall.args[2].isSimple(zpCheck)) {
                     // simple values (number/variable) do not clobber W1, so min/max can be evaluated first
                     assignAsmGen.assignExpressionToVariable(fcall.args[1], "P8ZP_SCRATCH_W1", fcall.args[1].type)  // minimum
@@ -1998,6 +2269,17 @@ import prog8.codegen.cpu6502.assignment.*
                 return arrayOf(RegisterOrPair.A)
             }
             fcall.type.isWord -> {
+                val minOp = asmgen.getStaticAddressLowHigh(fcall.args[1])
+                val maxOp = asmgen.getStaticAddressLowHigh(fcall.args[2])
+                val minIsConst = asmgen.unwrapCasts(fcall.args[1]) !is PtIdentifier
+                val maxIsConst = asmgen.unwrapCasts(fcall.args[2]) !is PtIdentifier
+                if (minOp != null && maxOp != null && minIsConst && maxIsConst) {
+                    // both bounds are compile-time constants (numeric literals or fixed label addresses):
+                    // emit a small inline clamp instead of evaluating bounds into scratch + jsr func_clamp_word/uword
+                    assignAsmGen.assignExpressionToRegister(fcall.args[0], RegisterOrPair.AY, signed)    // value
+                    funcClampWordConst(fcall.args[1], fcall.args[2], minOp, maxOp, signed)
+                    return arrayOf(RegisterOrPair.AY)
+                }
                 if (fcall.args[0].isSimple(zpCheck) && fcall.args[1].isSimple(zpCheck) && fcall.args[2].isSimple(zpCheck)) {
                     assignAsmGen.assignExpressionToVariable(fcall.args[1], "P8ZP_SCRATCH_W1", fcall.args[1].type)  // minimum
                     assignAsmGen.assignExpressionToVariable(fcall.args[2], "P8ZP_SCRATCH_W2", fcall.args[2].type)  // maximum
@@ -2292,9 +2574,7 @@ import prog8.codegen.cpu6502.assignment.*
 
     internal fun funcMklong(fcall: PtFunctionCall, resultReg: RegisterOrPair, targetVarName: String? = null): Array<RegisterOrPair> {
         fun isArgRegister(expression: PtExpression, reg: RegisterOrPair): Boolean {
-            if(expression !is PtIdentifier)
-                return false
-            return expression.name.startsWith("cx16.${reg.name.lowercase()}")
+            return expression is PtIdentifier && expression.name.startsWith("cx16.${reg.name.lowercase()}")
         }
 
         val (r1, r2) = when(resultReg) {
@@ -2669,7 +2949,7 @@ import prog8.codegen.cpu6502.assignment.*
 
     private fun translateArguments(call: PtFunctionCall, funcname: String?, scope: IPtSubroutine?) {
         val signature = BuiltinFunctions.getValue(call.name)
-        val callConv = signature.callConvention(call.args.map {
+        val callConv = signature.callConventionFor6502(call.args.map {
             require(it.type.isNumericOrBool)
             it.type.base
         })
@@ -2722,7 +3002,7 @@ import prog8.codegen.cpu6502.assignment.*
                         conv.dt==BaseDataType.LONG -> getSourceForLong(value)
                         conv.dt.isPassByRef -> {
                             // put the address of the argument in AY
-                            val addr = PtAddressOf(DataType.forDt(conv.dt).typeForAddressOf(false), false, value.position)
+                            val addr = PtAddressOf(DataType.forDt(conv.dt).typeForUntypedAddressOf(false, program.memsizer), false, value.position)
                             addr.add(value)
                             addr.parent = call
                             AsmAssignSource.fromAstSource(addr, program, asmgen)
@@ -2741,7 +3021,7 @@ import prog8.codegen.cpu6502.assignment.*
                         conv.dt==BaseDataType.LONG -> AsmAssignSource.fromAstSource(value, program, asmgen)
                         conv.dt.isPassByRef -> {
                             // put the address of the argument in AY
-                            val addr = PtAddressOf(DataType.forDt(conv.dt).typeForAddressOf(false), false,value.position)
+                            val addr = PtAddressOf(DataType.forDt(conv.dt).typeForUntypedAddressOf(false, program.memsizer), false,value.position)
                             addr.add(value)
                             addr.parent = call
                             AsmAssignSource.fromAstSource(addr, program, asmgen)

codeGenCpu6502/src/prog8/codegen/cpu6502/ForLoopsAsmGen.kt

@@ -5,6 +5,7 @@ import prog8.code.StMemVar
 import prog8.code.StStaticVariable
 import prog8.code.ast.PtForLoop
 import prog8.code.ast.PtIdentifier
+import prog8.code.ast.PtNumber
 import prog8.code.ast.PtRange
 import prog8.code.core.*
 import kotlin.math.absoluteValue
@@ -33,15 +34,21 @@ internal class ForLoopsAsmGen(
     }
 
     private fun translateForOverNonconstRange(stmt: PtForLoop, iterableDt: DataType, range: PtRange) {
-        if(range.step.asConstInteger()!! < -1) {
+        val constStep = range.step.asConstInteger()
+        if(constStep != null && constStep < -1) {
             val limit = range.to.asConstInteger()
             if(limit==0)
                 throw AssemblyError("for unsigned loop variable it's not possible to count down with step != -1 from a non-const value to exactly zero due to value wrapping")
         }
 
+        if(range.step !is PtNumber) {
+            translateForOverNonconstStepRange(stmt, iterableDt, range)
+            return
+        }
+
         when {
             iterableDt.isByteArray -> forOverNonconstByteRange(stmt, iterableDt, range)
-            iterableDt.isWordArray && !iterableDt.isSplitWordArray -> forOverNonconstWordRange(stmt, iterableDt, range)
+            iterableDt.isWordArray && !iterableDt.isSplitWordArray(asmgen.options.compTarget) -> forOverNonconstWordRange(stmt, iterableDt, range)
             iterableDt.isLongArray -> forOverNonconstLongRange(stmt, iterableDt, range)
             else -> throw AssemblyError("range expression can only be byte, word or long")
         }
@@ -49,6 +56,369 @@ internal class ForLoopsAsmGen(
         asmgen.loopEndLabels.removeLast()
     }
 
+    private fun translateForOverNonconstStepRange(stmt: PtForLoop, iterableDt: DataType, range: PtRange) {
+        when {
+            iterableDt.isByteArray -> forOverNonconstStepByteRange(stmt, iterableDt, range)
+            iterableDt.isWordArray && !iterableDt.isSplitWordArray(asmgen.options.compTarget) -> forOverNonconstStepWordRange(stmt, iterableDt, range)
+            iterableDt.isLongArray -> throw AssemblyError("variable-step for loops over long ranges are not supported by the legacy 6502 code generator at ${stmt.position}")
+            else -> throw AssemblyError("range expression can only be byte, word or long")
+        }
+        asmgen.loopEndLabels.removeLast()
+    }
+
+    private fun forOverNonconstStepByteRange(stmt: PtForLoop, iterableDt: DataType, range: PtRange) {
+        val loopDt = iterableDt.elementType()
+        val loopvar = asmgen.asmVariableName(stmt.variable)
+        val toValueVar = asmgen.createTempVarReused(loopDt.base, false, range)
+        val stepVar = asmgen.createTempVarReused(range.step.type.base, false, range)
+        val nextVar = asmgen.createTempVarReused(loopDt.base, false, range)
+        val loopLabel = asmgen.makeLabel("for_loop")
+        val descendingLabel = if(range.step.type.isSigned) asmgen.makeLabel("for_desc") else null
+        val endLabel = asmgen.makeLabel("for_end")
+        asmgen.loopEndLabels.add(endLabel)
+
+        // Keep range expression evaluation in source order and preserve each value across the next one.
+        asmgen.assignExpressionToVariable(range.from, loopvar, loopDt)
+        asmgen.assignExpressionToVariable(range.to, toValueVar, loopDt)
+        asmgen.assignExpressionToVariable(range.step, stepVar, range.step.type)
+        if (range.step.type.isSigned)
+            asmgen.out("  lda  $stepVar  |  beq  $endLabel  |  bmi  $descendingLabel")
+        else
+            asmgen.out("  lda  $stepVar  |  beq  $endLabel")
+
+        emitByteRangePrecheck(loopvar, toValueVar, loopLabel, endLabel, loopDt.isSigned, true)
+        asmgen.out(loopLabel)
+        asmgen.translate(stmt.statements)
+        emitByteRangeUpdate(loopvar, toValueVar, stepVar, nextVar, endLabel, loopLabel, loopDt.isSigned, range.step.type.isSigned)
+
+        if(descendingLabel != null) {
+            asmgen.out(descendingLabel)
+            emitByteRangePrecheck(loopvar, toValueVar, loopLabel, endLabel, loopDt.isSigned, false)
+            asmgen.out("  jmp  $loopLabel")
+        }
+        asmgen.out(endLabel)
+    }
+
+    private fun emitByteRangePrecheck(loopvar: String, toValueVar: String, loopLabel: String, endLabel: String, signed: Boolean, ascending: Boolean) {
+        val notEqualLabel = asmgen.makeLabel("for_not_equal")
+        val compareDoneLabel = asmgen.makeLabel("for_compare_done")
+        if (signed) {
+            asmgen.out("""
+                lda  $loopvar
+                cmp  $toValueVar
+                bne  $notEqualLabel
+                jmp  $loopLabel
+$notEqualLabel
+                sec
+                lda  $loopvar
+                sbc  $toValueVar
+                bvc  $compareDoneLabel
+                eor  #$80
+$compareDoneLabel
+                ${if (ascending) "bpl" else "bmi"}  $endLabel
+            """)
+        } else if (ascending) {
+            asmgen.out("""
+                lda  $toValueVar
+                cmp  $loopvar
+                bcc  $endLabel
+                jmp  $loopLabel
+            """)
+        } else {
+            asmgen.out("""
+                lda  $loopvar
+                cmp  $toValueVar
+                bcc  $endLabel
+            """)
+        }
+    }
+
+    private fun emitByteRangeUpdate(
+        loopvar: String,
+        toValueVar: String,
+        stepVar: String,
+        nextVar: String,
+        endLabel: String,
+        loopLabel: String,
+        loopvarSigned: Boolean,
+        stepSigned: Boolean
+    ) {
+        val notLastLabel = asmgen.makeLabel("for_not_last")
+        val ascendingCompareDoneLabel = asmgen.makeLabel("for_next_compare_done")
+        val descendingCompareDoneLabel = if(stepSigned) asmgen.makeLabel("for_desc_compare_done") else null
+        val descendingUpdateLabel = if(stepSigned) asmgen.makeLabel("for_desc_update") else null
+        val storeNextLabel = asmgen.makeLabel("for_store_next")
+        val ascendingWrapBranch = if (loopvarSigned) "bvs" else "bcs"
+        val descendingWrapBranch = if (loopvarSigned) "bvs" else "bcc"
+        asmgen.out("""
+            lda  $loopvar
+            cmp  $toValueVar
+            bne  $notLastLabel
+            jmp  $endLabel
+$notLastLabel
+        """)
+        if (stepSigned)
+            asmgen.out("  lda  $stepVar  |  bmi  $descendingUpdateLabel")
+        asmgen.out("""
+            clc
+            lda  $loopvar
+            adc  $stepVar
+            $ascendingWrapBranch  $endLabel
+            sta  $nextVar
+        """)
+        asmgen.out("""
+            lda  $nextVar
+            cmp  $toValueVar
+            beq  $storeNextLabel
+            sec
+            lda  $nextVar
+            sbc  $toValueVar
+            bvc  $ascendingCompareDoneLabel
+            eor  #$80
+$ascendingCompareDoneLabel
+        """)
+        if (loopvarSigned)
+            asmgen.out("  bpl  $endLabel")
+        else
+            asmgen.out("  bcs  $endLabel")
+        if(stepSigned) {
+            asmgen.out("""
+            jmp  $storeNextLabel
+
+$descendingUpdateLabel
+            lda  $loopvar
+            cmp  $toValueVar
+            beq  $endLabel
+            clc
+            lda  $loopvar
+            adc  $stepVar
+            $descendingWrapBranch  $endLabel
+            sta  $nextVar
+        """)
+            if (loopvarSigned)
+            asmgen.out("""
+            sec
+            lda  $nextVar
+            sbc  $toValueVar
+            bvc  $descendingCompareDoneLabel
+            eor  #$80
+$descendingCompareDoneLabel
+            bmi  $endLabel
+            """)
+            else
+            asmgen.out("""
+            lda  $nextVar
+            cmp  $toValueVar
+            bcc  $endLabel
+                """)
+        }
+        asmgen.out("""
+            jmp  $storeNextLabel
+
+$storeNextLabel
+            lda  $nextVar
+            sta  $loopvar
+            jmp  $loopLabel
+        """)
+    }
+
+    private fun forOverNonconstStepWordRange(stmt: PtForLoop, iterableDt: DataType, range: PtRange) {
+        val loopDt = iterableDt.elementType()
+        val loopvar = asmgen.asmVariableName(stmt.variable)
+        val toValueVar = asmgen.createTempVarReused(loopDt.base, false, range)
+        val stepVar = asmgen.createTempVarReused(range.step.type.base, false, range)
+        val nextVar = asmgen.createTempVarReused(loopDt.base, false, range)
+        val loopLabel = asmgen.makeLabel("for_loop")
+        val descendingLabel = if(range.step.type.isSigned) asmgen.makeLabel("for_desc") else null
+        val endLabel = asmgen.makeLabel("for_end")
+        asmgen.loopEndLabels.add(endLabel)
+
+        // Keep range expression evaluation in source order and preserve each value across the next one.
+        asmgen.assignExpressionToVariable(range.from, loopvar, loopDt)
+        asmgen.assignExpressionToVariable(range.to, toValueVar, loopDt)
+        asmgen.assignExpressionToVariable(range.step, stepVar, range.step.type)
+        if (range.step.type.isSigned)
+            asmgen.out("  lda  $stepVar  |  ora  ${stepVar}+1  |  beq  $endLabel  |  lda  ${stepVar}+1  |  bmi  $descendingLabel")
+        else
+            asmgen.out("  lda  $stepVar  |  ora  ${stepVar}+1  |  beq  $endLabel")
+
+        emitWordRangePrecheck(loopvar, toValueVar, loopLabel, endLabel, loopDt.isSigned, true)
+        asmgen.out(loopLabel)
+        asmgen.translate(stmt.statements)
+        emitWordRangeUpdate(loopvar, toValueVar, stepVar, nextVar, endLabel, loopLabel, loopDt.isSigned, range.step.type.isSigned)
+
+        if(descendingLabel != null) {
+            asmgen.out(descendingLabel)
+            emitWordRangePrecheck(loopvar, toValueVar, loopLabel, endLabel, loopDt.isSigned, false)
+            asmgen.out("  jmp  $loopLabel")
+        }
+        asmgen.out(endLabel)
+    }
+
+    private fun emitWordRangePrecheck(loopvar: String, toValueVar: String, loopLabel: String, endLabel: String, signed: Boolean, ascending: Boolean) {
+        val notEqualLabel = asmgen.makeLabel("for_not_equal")
+        val compareDoneLabel = asmgen.makeLabel("for_compare_done")
+        if (signed) {
+            asmgen.out("""
+                lda  $loopvar+1
+                cmp  $toValueVar+1
+                bne  $notEqualLabel
+                lda  $loopvar
+                cmp  $toValueVar
+                bne  $notEqualLabel
+                jmp  $loopLabel
+$notEqualLabel
+                sec
+                lda  $loopvar
+                sbc  $toValueVar
+                lda  $loopvar+1
+                sbc  $toValueVar+1
+                bvc  $compareDoneLabel
+                eor  #$80
+$compareDoneLabel
+                ${if (ascending) "bpl" else "bmi"}  $endLabel
+            """)
+        } else if (ascending) {
+            asmgen.out("""
+                lda  $toValueVar+1
+                cmp  $loopvar+1
+                bcc  $endLabel
+                bne  $loopLabel
+                lda  $toValueVar
+                cmp  $loopvar
+                bcc  $endLabel
+                jmp  $loopLabel
+            """)
+        } else {
+            asmgen.out("""
+                lda  $loopvar+1
+                cmp  $toValueVar+1
+                bcc  $endLabel
+                bne  $loopLabel
+                lda  $loopvar
+                cmp  $toValueVar
+                bcc  $endLabel
+                jmp  $loopLabel
+            """)
+        }
+    }
+
+    private fun emitWordRangeUpdate(
+        loopvar: String,
+        toValueVar: String,
+        stepVar: String,
+        nextVar: String,
+        endLabel: String,
+        loopLabel: String,
+        loopvarSigned: Boolean,
+        stepSigned: Boolean
+    ) {
+        val notLastLabel = asmgen.makeLabel("for_not_last")
+        val ascendingCompareLabel = asmgen.makeLabel("for_next_compare")
+        val ascendingCompareDoneLabel = asmgen.makeLabel("for_next_compare_done")
+        val descendingCompareDoneLabel = if(stepSigned) asmgen.makeLabel("for_desc_compare_done") else null
+        val descendingUpdateLabel = if(stepSigned) asmgen.makeLabel("for_desc_update") else null
+        val descendingNotLastLabel = if(stepSigned) asmgen.makeLabel("for_desc_not_last") else null
+        val storeNextLabel = asmgen.makeLabel("for_store_next")
+        val ascendingWrapBranch = if (loopvarSigned) "bvs" else "bcs"
+        val descendingWrapBranch = if (loopvarSigned) "bvs" else "bcc"
+        asmgen.out("""
+            lda  $loopvar+1
+            cmp  $toValueVar+1
+            bne  $notLastLabel
+            lda  $loopvar
+            cmp  $toValueVar
+            beq  $endLabel
+$notLastLabel
+        """)
+        if (stepSigned)
+            asmgen.out("  lda  ${stepVar}+1  |  bmi  $descendingUpdateLabel")
+        asmgen.out("""
+            clc
+            lda  $loopvar
+            adc  $stepVar
+            sta  $nextVar
+            lda  $loopvar+1
+            adc  $stepVar+1
+            $ascendingWrapBranch  $endLabel
+            sta  $nextVar+1
+        """)
+        asmgen.out("""
+            lda  $nextVar+1
+            cmp  $toValueVar+1
+            bne  $ascendingCompareLabel
+            lda  $nextVar
+            cmp  $toValueVar
+            beq  $storeNextLabel
+$ascendingCompareLabel
+            sec
+            lda  $nextVar
+            sbc  $toValueVar
+            lda  $nextVar+1
+            sbc  $toValueVar+1
+            bvc  $ascendingCompareDoneLabel
+            eor  #$80
+$ascendingCompareDoneLabel
+        """)
+        if (loopvarSigned)
+            asmgen.out("  bpl  $endLabel")
+        else
+            asmgen.out("  bcs  $endLabel")
+        if(stepSigned) {
+            asmgen.out("""
+            jmp  $storeNextLabel
+
+$descendingUpdateLabel
+            lda  $loopvar+1
+            cmp  $toValueVar+1
+            bne  $descendingNotLastLabel
+            lda  $loopvar
+            cmp  $toValueVar
+            beq  $endLabel
+$descendingNotLastLabel
+            clc
+            lda  $loopvar
+            adc  $stepVar
+            sta  $nextVar
+            lda  $loopvar+1
+            adc  $stepVar+1
+            $descendingWrapBranch  $endLabel
+            sta  $nextVar+1
+        """)
+            if (loopvarSigned)
+            asmgen.out("""
+            sec
+            lda  $nextVar
+            sbc  $toValueVar
+            lda  $nextVar+1
+            sbc  $toValueVar+1
+            bvc  $descendingCompareDoneLabel
+            eor  #$80
+$descendingCompareDoneLabel
+            bmi  $endLabel
+            """)
+            else
+            asmgen.out("""
+            lda  $nextVar+1
+            cmp  $toValueVar+1
+            bcc  $endLabel
+            bne  $storeNextLabel
+            lda  $nextVar
+            cmp  $toValueVar
+            bcc  $endLabel
+            """)
+        }
+        else asmgen.out("  jmp  $storeNextLabel")
+        asmgen.out("""
+$storeNextLabel
+            lda  $nextVar
+            sta  $loopvar
+            lda  $nextVar+1
+            sta  $loopvar+1
+            jmp  $loopLabel
+        """)
+    }
+
     private fun forOverNonconstLongRange(stmt: PtForLoop, iterableDt: DataType, range: PtRange) {
         val stepsize = range.step.asConstInteger()!!
         val loopLabel = asmgen.makeLabel("for_loop")
@@ -157,12 +527,7 @@ internal class ForLoopsAsmGen(
     }
 
     private fun forOverLongsRangeStepGreaterOne(range: PtRange, varname: String, iterableDt: DataType, loopLabel: String, endLabel: String, forloop: PtForLoop) {
-        // TODO: implement 32-bit loop
-        asmgen.romableError("for loops over long ranges with step != 1 not yet fully implemented", forloop.position)
-        asmgen.out(loopLabel)
-        asmgen.translate(forloop.statements)
-        asmgen.jmp(loopLabel)
-        asmgen.out(endLabel)
+        TODO("legacy 6502 codegen does not support long FOR ranges with step other than 1 or -1 at ${forloop.position}")
     }
 
     private fun forOverNonconstByteRange(stmt: PtForLoop, iterableDt: DataType, range: PtRange) {
@@ -822,7 +1187,7 @@ $loopLabel          sty  $indexVar
         when {
             iterableDt.isString -> iterateStrings()
             iterableDt.isByteArray || iterableDt.isBoolArray -> iterateBytes()
-            iterableDt.isSplitWordArray -> iterateSplitWords()
+            iterableDt.isSplitWordArray(asmgen.options.compTarget) -> iterateSplitWords()
             iterableDt.isWordArray -> iterateWords()
             iterableDt.isLongArray -> iterateWords(true)
             iterableDt.isFloatArray -> throw AssemblyError("for loop with floating point variables is not supported")
@@ -920,7 +1285,7 @@ $loopLabel""")
                 }
                 asmgen.out(endLabel)
             }
-            iterableDt.isWordArray && !iterableDt.isSplitWordArray -> {
+            iterableDt.isWordArray && !iterableDt.isSplitWordArray(asmgen.options.compTarget) -> {
                 // loop over word range via loopvar, step >= 2 or <= -2
                 when (range.step) {
                     1, -1 -> {
@@ -1091,16 +1456,15 @@ $loopLabel""")
 $loopLabel""")
         asmgen.translate(stmt.statements)
         asmgen.out("""
-            lda  $varname
-            cmp  #<${range.last}
+            inc  $varname
             bne  +
-            lda  $varname+1
-            cmp  #>${range.last}
-            beq  $endLabel
-+           inc  $varname
+            inc  $varname+1
++           lda  $varname
+            cmp  #<${range.last+1}
             bne  $loopLabel
-            inc  $varname+1""")
-        asmgen.jmp(loopLabel)
+            lda  $varname+1
+            cmp  #>${range.last+1}
+            bne  $loopLabel""")
         asmgen.out(endLabel)
         asmgen.loopEndLabels.removeLast()
     }
@@ -1121,24 +1485,21 @@ $loopLabel""")
         if(range.last==0) {
             asmgen.out("""
                 lda  $varname
-                bne  ++
-                lda  $varname+1
-                beq  $endLabel""")
-        } else {
-            asmgen.out("""
-                lda  $varname
-                cmp  #<${range.last}
                 bne  +
                 lda  $varname+1
-                cmp  #>${range.last}
                 beq  $endLabel""")
         }
         asmgen.out("""
 +           lda  $varname
             bne  +
             dec  $varname+1
-+           dec  $varname""")
-        asmgen.jmp(loopLabel)
++           dec  $varname
+            lda  $varname
+            cmp  #<${range.last-1}
+            bne  $loopLabel
+            lda  $varname+1
+            cmp  #>${range.last-1}
+            bne  $loopLabel""")
         asmgen.out(endLabel)
         asmgen.loopEndLabels.removeLast()
     }
@@ -1150,7 +1511,7 @@ $loopLabel""")
         asmgen.loopEndLabels.add(endLabel)
         val varname = asmgen.asmVariableName(stmt.variable)
         val from = range.first.toLongHex()
-        val to = range.last.toLongHex()
+        val toPlus1 = (range.last + 1).toLongHex()
         asmgen.out("""
             lda  #$${from.substring(6, 8)}
             sta  $varname
@@ -1163,26 +1524,25 @@ $loopLabel""")
 $loopLabel""")
         asmgen.translate(stmt.statements)
         asmgen.out("""
-            lda  $varname
-            cmp  #$${to.substring(6, 8)}
+            inc  $varname
             bne  +
-            lda  $varname+1
-            cmp  #$${to.substring(4, 6)}
-            bne  +
-            lda  $varname+2
-            cmp  #$${to.substring(2, 4)}
-            bne  +
-            lda  $varname+3
-            cmp  #$${to.take(2)}
-            beq  $endLabel
-+           inc  $varname
-            bne  $loopLabel
             inc  $varname+1
-            bne  $loopLabel
+            bne  +
             inc  $varname+2
+            bne  +
+            inc  $varname+3
++           lda  $varname
+            cmp  #$${toPlus1.substring(6, 8)}
             bne  $loopLabel
-            inc  $varname+3""")
-        asmgen.jmp(loopLabel)
+            lda  $varname+1
+            cmp  #$${toPlus1.substring(4, 6)}
+            bne  $loopLabel
+            lda  $varname+2
+            cmp  #$${toPlus1.substring(2, 4)}
+            bne  $loopLabel
+            lda  $varname+3
+            cmp  #$${toPlus1.take(2)}
+            bne  $loopLabel""")
         asmgen.out(endLabel)
         asmgen.loopEndLabels.removeLast()
     }
@@ -1194,7 +1554,7 @@ $loopLabel""")
         asmgen.loopEndLabels.add(endLabel)
         val varname = asmgen.asmVariableName(stmt.variable)
         val from = range.first.toLongHex()
-        val to = range.last.toLongHex()
+        val toMinus1 = (range.last - 1).toLongHex()
         asmgen.out("""
             lda  #$${from.substring(6, 8)}
             sta  $varname
@@ -1216,20 +1576,6 @@ $loopLabel""")
                 bne  ++
                 lda  $varname+3
                 beq  $endLabel""")
-        } else {
-            asmgen.out("""
-                lda  $varname
-                cmp  #$${to.substring(6, 8)}
-                bne  +
-                lda  $varname+1
-                cmp  #$${to.substring(4, 6)}
-                bne  +
-                lda  $varname+2
-                cmp  #$${to.substring(2, 4)}
-                bne  +
-                lda  $varname+3
-                cmp  #$${to.take(2)}
-                beq  $endLabel""")
         }
         asmgen.out("""
 +           lda  $varname
@@ -1241,8 +1587,19 @@ $loopLabel""")
             dec  $varname+3
 +           dec  $varname+2
 +           dec  $varname+1
-+           dec  $varname""")
-        asmgen.jmp(loopLabel)
++           dec  $varname
+            lda  $varname
+            cmp  #$${toMinus1.substring(6, 8)}
+            bne  $loopLabel
+            lda  $varname+1
+            cmp  #$${toMinus1.substring(4, 6)}
+            bne  $loopLabel
+            lda  $varname+2
+            cmp  #$${toMinus1.substring(2, 4)}
+            bne  $loopLabel
+            lda  $varname+3
+            cmp  #$${toMinus1.take(2)}
+            bne  $loopLabel""")
         asmgen.out(endLabel)
         asmgen.loopEndLabels.removeLast()
     }

codeGenCpu6502/src/prog8/codegen/cpu6502/FunctionCallAsmGen.kt

@@ -388,7 +388,7 @@ $label .byte  0    ; modified"""
                     }
                 val src = if(value.type.isPassByRef) {
                     if(value is PtIdentifier || value is PtPointerDeref) {
-                        val addr = PtAddressOf(value.type.typeForAddressOf(false), false, value.position)
+                        val addr = PtAddressOf(value.type.typeForUntypedAddressOf(false, program.memsizer), false, value.position)
                         addr.add(value)
                         addr.parent = scope as PtNode
                         AsmAssignSource.fromAstSource(addr, program, asmgen).adjustSignedUnsigned(target)

codeGenCpu6502/src/prog8/codegen/cpu6502/IfElseAsmGen.kt

@@ -313,9 +313,7 @@ internal class IfElseAsmGen(private val program: PtProgram,
 
         val left = condition.left
         val right = condition.right
-        val operator = condition.operator
-
-        when (operator) {
+        when (val operator = condition.operator) {
             "==" -> {
                 // if X==value
                 asmgen.assignExpressionToRegister(left, RegisterOrPair.A, signed)
@@ -595,9 +593,7 @@ internal class IfElseAsmGen(private val program: PtProgram,
 
         val left = condition.left
         val right = condition.right
-        val operator = condition.operator
-
-        return when (operator) {
+        return when (val operator = condition.operator) {
             "==" -> longEqualsValue(left, right, false, jumpAfterIf, stmt)
             "!=" -> longEqualsValue(left, right, true, jumpAfterIf, stmt)
             "<", "<=", ">", ">=" -> compareLongValues(left, operator, right, jumpAfterIf, stmt)
@@ -932,19 +928,17 @@ internal class IfElseAsmGen(private val program: PtProgram,
             asmgen.out("  lda  $varname+$offset")
             return
         }
-        val e = asmgen.unwrapCasts(expr)
-        when (e) {
+        when (val e = asmgen.unwrapCasts(expr)) {
             is PtArrayIndexer -> {
                 if (e.variable == null)
                     throw AssemblyError("support for ptr indexing ${e.position}")
                 val baseVarname = asmgen.asmVariableName(e.variable!!)
                 val constIndex = e.index.asConstInteger()
                 if (constIndex != null) {
-                    val idx = constIndex
                     if (e.splitWords) {
-                        require(idx < 256)
+                        require(constIndex < 256)
                         val suffix = if (long) "_3" else "_msb"
-                        asmgen.out("  lda  $baseVarname$suffix+$idx")
+                        asmgen.out("  lda  $baseVarname$suffix+$constIndex")
                         return
                     }
                 }
@@ -964,7 +958,7 @@ internal class IfElseAsmGen(private val program: PtProgram,
                     asmgen.out("  cpy  #0")
                 } else {
                     var varname = asmgen.asmVariableName(e.identifier!!)
-                    if(e.identifier!!.type.isSplitWordArray) {
+                    if(e.identifier!!.type.isSplitWordArray(asmgen.options.compTarget)) {
                         varname += if(e.isMsbForSplitArray) "_msb" else "_lsb"
                     }
                     asmgen.out("  lda  #>$varname")
@@ -1831,24 +1825,30 @@ internal class IfElseAsmGen(private val program: PtProgram,
             block("#<${constValue}", "#>${constValue}")
         } else {
             val e = asmgen.unwrapCasts(expr)
-            if (e is PtIdentifier) {
-                val varname = asmgen.asmVariableName(e)
-                block(varname, "$varname+1")
-            } else if (e is PtAddressOf && !e.isFromArrayElement) {
-                val identifier = e.identifier!!
-                val varname = if (identifier.type.isSplitWordArray) {
-                    if (e.isMsbForSplitArray) identifier.name + "_msb" else identifier.name + "_lsb"
-                } else {
-                    identifier.name
+            when (e) {
+                is PtIdentifier -> {
+                    val varname = asmgen.asmVariableName(e)
+                    block(varname, "$varname+1")
+                }
+
+                is PtAddressOf if !e.isFromArrayElement && e.identifier != null -> {
+                    val identifier = e.identifier!!
+                    val varname = if (identifier.type.isSplitWordArray(asmgen.options.compTarget)) {
+                        if (e.isMsbForSplitArray) identifier.name + "_msb" else identifier.name + "_lsb"
+                    } else {
+                        identifier.name
+                    }
+                    block("#<$varname", "#>$varname")
+                }
+
+                else -> {
+                    asmgen.saveRegisterStack(CpuRegister.A, false)
+                    asmgen.saveRegisterStack(CpuRegister.Y, false)
+                    asmgen.assignExpressionToVariable(expr, "P8ZP_SCRATCH_W1", expr.type)
+                    asmgen.restoreRegisterStack(CpuRegister.Y, false)
+                    asmgen.restoreRegisterStack(CpuRegister.A, false)
+                    block("P8ZP_SCRATCH_W1", "P8ZP_SCRATCH_W1+1")
                 }
-                block("#<$varname", "#>$varname")
-            } else {
-                asmgen.saveRegisterStack(CpuRegister.A, false)
-                asmgen.saveRegisterStack(CpuRegister.Y, false)
-                asmgen.assignExpressionToVariable(expr, "P8ZP_SCRATCH_W1", expr.type)
-                asmgen.restoreRegisterStack(CpuRegister.Y, false)
-                asmgen.restoreRegisterStack(CpuRegister.A, false)
-                block("P8ZP_SCRATCH_W1", "P8ZP_SCRATCH_W1+1")
             }
         }
     }

codeGenCpu6502/src/prog8/codegen/cpu6502/IfExpressionAsmGen.kt

@@ -104,8 +104,7 @@ internal class IfExpressionAsmGen(private val asmgen: AsmGen6502Internal, privat
     }
 
     private fun evalIfExpressionConditonAndBranchWhenFalse(stmtCondition: PtExpression, falseLabel: String) {
-        val condition = asmgen.unwrapCasts(stmtCondition)
-        when (condition) {
+        when (val condition = asmgen.unwrapCasts(stmtCondition)) {
             is PtBinaryExpression -> {
                 val rightDt = condition.right.type
                 return when {
@@ -723,7 +722,7 @@ internal class IfExpressionAsmGen(private val asmgen: AsmGen6502Internal, privat
             is PtAddressOf -> {
                 val identifier = e.identifier!!
                 var varname = asmgen.asmVariableName(identifier)
-                if (identifier.type.isSplitWordArray) {
+                if (identifier.type.isSplitWordArray(asmgen.options.compTarget)) {
                     varname += if (e.isMsbForSplitArray) "_msb" else "_lsb"
                 }
                 asmgen.out("  lda  #<$varname | ora  #>$varname | bne  $falseLabel")
@@ -774,7 +773,7 @@ internal class IfExpressionAsmGen(private val asmgen: AsmGen6502Internal, privat
             is PtAddressOf -> {
                 val identifier = e.identifier!!
                 var varname = asmgen.asmVariableName(identifier)
-                if (identifier.type.isSplitWordArray) {
+                if (identifier.type.isSplitWordArray(asmgen.options.compTarget)) {
                     varname += if (e.isMsbForSplitArray) "_msb" else "_lsb"
                 }
                 asmgen.out("  lda  #<$varname | ora  #>$varname | beq  $falseLabel")
@@ -802,8 +801,7 @@ internal class IfExpressionAsmGen(private val asmgen: AsmGen6502Internal, privat
             translateWordExprIsZero(expr, falseLabel)
             return
         }
-        val e = asmgen.unwrapCasts(expr)
-        when (e) {
+        when (val e = asmgen.unwrapCasts(expr)) {
             is PtIdentifier -> {
                 val varname = asmgen.asmVariableName(e)
                 asmgen.out("""
@@ -853,8 +851,7 @@ internal class IfExpressionAsmGen(private val asmgen: AsmGen6502Internal, privat
             translateWordExprIsNotZero(expr, falseLabel)
             return
         }
-        val e = asmgen.unwrapCasts(expr)
-        when (e) {
+        when (val e = asmgen.unwrapCasts(expr)) {
             is PtIdentifier -> {
                 val varname = asmgen.asmVariableName(e)
                 asmgen.out("""
@@ -1092,19 +1089,17 @@ internal class IfExpressionAsmGen(private val asmgen: AsmGen6502Internal, privat
             asmgen.out("  lda  $varname+$offset")
             return
         }
-        val e = asmgen.unwrapCasts(expr)
-        when (e) {
+        when (val e = asmgen.unwrapCasts(expr)) {
             is PtArrayIndexer -> {
                 if (e.variable == null)
                     throw AssemblyError("support for ptr indexing ${e.position}")
                 val baseVarname = asmgen.asmVariableName(e.variable!!)
                 val constIndex = e.index.asConstInteger()
                 if (constIndex != null) {
-                    val idx = constIndex
                     if (e.splitWords) {
-                        require(idx < 256)
+                        require(constIndex < 256)
                         val suffix = if (long) "_3" else "_msb"
-                        asmgen.out("  lda  $baseVarname$suffix+$idx")
+                        asmgen.out("  lda  $baseVarname$suffix+$constIndex")
                         return
                     }
                 }
@@ -1124,7 +1119,7 @@ internal class IfExpressionAsmGen(private val asmgen: AsmGen6502Internal, privat
                     asmgen.out("  cpy  #0")
                 } else {
                     var varname = asmgen.asmVariableName(e.identifier!!)
-                    if(e.identifier!!.type.isSplitWordArray) {
+                    if(e.identifier!!.type.isSplitWordArray(asmgen.options.compTarget)) {
                         varname += if(e.isMsbForSplitArray) "_msb" else "_lsb"
                     }
                     asmgen.out("  lda  #>$varname")
@@ -1150,23 +1145,29 @@ internal class IfExpressionAsmGen(private val asmgen: AsmGen6502Internal, privat
             block("#<${constValue}", "#>${constValue}")
         } else {
             val e = asmgen.unwrapCasts(expr)
-            if (e is PtIdentifier) {
-                val varname = asmgen.asmVariableName(e)
-                block(varname, "$varname+1")
-            } else if (e is PtAddressOf && !e.isFromArrayElement) {
-                val identifier = e.identifier!!
-                var varname = asmgen.asmVariableName(identifier)
-                if (identifier.type.isSplitWordArray) {
-                    varname += if (e.isMsbForSplitArray) "_msb" else "_lsb"
+            when (e) {
+                is PtIdentifier -> {
+                    val varname = asmgen.asmVariableName(e)
+                    block(varname, "$varname+1")
+                }
+
+                is PtAddressOf if !e.isFromArrayElement -> {
+                    val identifier = e.identifier!!
+                    var varname = asmgen.asmVariableName(identifier)
+                    if (identifier.type.isSplitWordArray(asmgen.options.compTarget)) {
+                        varname += if (e.isMsbForSplitArray) "_msb" else "_lsb"
+                    }
+                    block("#<$varname", "#>$varname")
+                }
+
+                else -> {
+                    asmgen.saveRegisterStack(CpuRegister.A, false)
+                    asmgen.saveRegisterStack(CpuRegister.Y, false)
+                    asmgen.assignExpressionToVariable(expr, "P8ZP_SCRATCH_W1", expr.type)
+                    asmgen.restoreRegisterStack(CpuRegister.Y, false)
+                    asmgen.restoreRegisterStack(CpuRegister.A, false)
+                    block("P8ZP_SCRATCH_W1", "P8ZP_SCRATCH_W1+1")
                 }
-                block("#<$varname", "#>$varname")
-            } else {
-                asmgen.saveRegisterStack(CpuRegister.A, false)
-                asmgen.saveRegisterStack(CpuRegister.Y, false)
-                asmgen.assignExpressionToVariable(expr, "P8ZP_SCRATCH_W1", expr.type)
-                asmgen.restoreRegisterStack(CpuRegister.Y, false)
-                asmgen.restoreRegisterStack(CpuRegister.A, false)
-                block("P8ZP_SCRATCH_W1", "P8ZP_SCRATCH_W1+1")
             }
         }
     }

codeGenCpu6502/src/prog8/codegen/cpu6502/ProgramAndVarsGen.kt

@@ -173,6 +173,7 @@ internal class ProgramAndVarsGen(
                         asmgen.out("  jsr  p8_sys_startup.init_system")
                     asmgen.out("  jsr  p8_sys_startup.init_system_phase2")
                 }
+                else -> error("Unsupported output type: ${compTarget.defaultOutputType}")
             }
 
             if (options.zeropage !in arrayOf(ZeropageType.BASICSAFE, ZeropageType.DONTUSE)) {
@@ -223,20 +224,23 @@ internal class ProgramAndVarsGen(
     private fun footer() {
         asmgen.out("  .dsection STRUCTINSTANCES\n")
 
-        var relocateBssVars = false
-        var relocateBssSlabs = false
-        var relocatedBssStart = 0u
-        var relocatedBssEnd = 0u
+        var relocateVars = false
+        var relocatedStart = 0u
+        var relocatedEnd = 0u
 
-        if(options.varsGolden) {
+        if(options.varsAddress != null) {
+            relocateVars = true
+            relocatedStart = options.varsAddress!!
+            relocatedEnd = if(options.memtopAddress != 0u) options.memtopAddress else 0xFFFFu
+        } else if(options.varsGolden) {
             if(options.compTarget.BSSGOLDENRAM_START == 0u ||
                 options.compTarget.BSSGOLDENRAM_END == 0u ||
                 options.compTarget.BSSGOLDENRAM_END <= options.compTarget.BSSGOLDENRAM_START) {
                 throw AssemblyError("current compilation target hasn't got the golden ram area properly defined or it is simply not available")
             }
-            relocateBssVars = true
-            relocatedBssStart = options.compTarget.BSSGOLDENRAM_START
-            relocatedBssEnd = options.compTarget.BSSGOLDENRAM_END
+            relocateVars = true
+            relocatedStart = options.compTarget.BSSGOLDENRAM_START
+            relocatedEnd = options.compTarget.BSSGOLDENRAM_END
         }
         else if(options.varsHighBank!=null) {
             if(options.compTarget.BSSHIGHRAM_START == 0u ||
@@ -244,67 +248,34 @@ internal class ProgramAndVarsGen(
                 options.compTarget.BSSHIGHRAM_END <= options.compTarget.BSSHIGHRAM_START) {
                 throw AssemblyError("current compilation target hasn't got the high ram area properly defined or it is simply not available")
             }
-            if(options.slabsHighBank!=null && options.varsHighBank!=options.slabsHighBank)
-                throw AssemblyError("slabs and vars high bank must be the same")
-            relocateBssVars = true
-            relocatedBssStart = options.compTarget.BSSHIGHRAM_START
-            relocatedBssEnd = options.compTarget.BSSHIGHRAM_END
-        }
-
-        if(options.slabsGolden) {
-            if(options.compTarget.BSSGOLDENRAM_START == 0u ||
-                options.compTarget.BSSGOLDENRAM_END == 0u ||
-                options.compTarget.BSSGOLDENRAM_END <= options.compTarget.BSSGOLDENRAM_START) {
-                throw AssemblyError("current compilation target hasn't got the golden ram area properly defined or it is simply not available")
-            }
-            relocateBssSlabs = true
-            relocatedBssStart = options.compTarget.BSSGOLDENRAM_START
-            relocatedBssEnd = options.compTarget.BSSGOLDENRAM_END
-        }
-        else if(options.slabsHighBank!=null) {
-            if(options.compTarget.BSSHIGHRAM_START == 0u ||
-                options.compTarget.BSSHIGHRAM_END == 0u ||
-                options.compTarget.BSSHIGHRAM_END <= options.compTarget.BSSHIGHRAM_START) {
-                throw AssemblyError("current compilation target hasn't got the high ram area properly defined or it is simply not available")
-            }
-            if(options.varsHighBank!=null && options.varsHighBank!=options.slabsHighBank)
-                throw AssemblyError("slabs and vars high bank must be the same")
-            relocateBssSlabs = true
-            relocatedBssStart = options.compTarget.BSSHIGHRAM_START
-            relocatedBssEnd = options.compTarget.BSSHIGHRAM_END
+            relocateVars = true
+            relocatedStart = options.compTarget.BSSHIGHRAM_START
+            relocatedEnd = options.compTarget.BSSHIGHRAM_END
         }
 
         asmgen.out("; bss sections")
         asmgen.out("PROG8_VARSHIGH_RAMBANK = ${options.varsHighBank ?: 1}")
-        if(relocateBssVars) {
-            if(!relocateBssSlabs)
-                asmgen.out("  .dsection BSS_SLABS")
+        if(relocateVars) {
             asmgen.out("prog8_program_end\t; end of program label for progend()")
-            asmgen.out("  * = ${relocatedBssStart.toHex()}")
+            asmgen.out("  * = ${relocatedStart.toHex()}")
             asmgen.out("  .dsection BSS_NOCLEAR")
             asmgen.out("prog8_bss_section_start")
             asmgen.out("  .dsection BSS")
-            if(relocateBssSlabs)
-                asmgen.out("  .dsection BSS_SLABS")
-            asmgen.out("  .cerror * > ${relocatedBssEnd.toHex()}, \"too many variables/data for BSS section\"")
+            asmgen.out("  .dsection BSS_SLABS")
+            asmgen.out("  .cerror * > ${relocatedEnd.toHex()}, \"too many variables/data for BSS section\"")
             asmgen.out("prog8_bss_section_size = * - prog8_bss_section_start")
         } else {
             asmgen.out("  .dsection BSS_NOCLEAR")
             asmgen.out("prog8_bss_section_start")
             asmgen.out("  .dsection BSS")
             asmgen.out("prog8_bss_section_size = * - prog8_bss_section_start")
-            if(!relocateBssSlabs)
-                asmgen.out("  .dsection BSS_SLABS")
+            asmgen.out("  .dsection BSS_SLABS")
             asmgen.out("prog8_program_end\t; end of program label for progend()")
-            if(relocateBssSlabs) {
-                asmgen.out("  * = ${relocatedBssStart.toHex()}")
-                asmgen.out("  .dsection BSS_SLABS")
-                asmgen.out("  .cerror * > ${relocatedBssEnd.toHex()}, \"too many data for BSS_SLABS section\"")
-            }
         }
 
-        if(relocatedBssEnd >= options.memtopAddress)
-            options.memtopAddress = relocatedBssEnd+1u
+        val effectiveEnd = if(relocateVars) relocatedEnd else 0u
+        if(effectiveEnd >= options.memtopAddress && effectiveEnd != 0u)
+            options.memtopAddress = effectiveEnd+1u
 
         asmgen.out("  ; memtop check")
         asmgen.out("  .cerror * >= ${options.memtopAddress.toHex()}, \"Program too long by \", * - ${(options.memtopAddress-1u).toHex()}, \" bytes, memtop=${options.memtopAddress.toHex()}\"")
@@ -425,30 +396,33 @@ internal class ProgramAndVarsGen(
 
         asmgen.out("; struct types")
         symboltable.allStructTypes().distinctBy { it.name }.forEach { structtype ->
-            val paramFields = structtype.fields.filter { !it.isArray }
-            val structargs = paramFields.indices.joinToString(",") { "f$it" }
-            asmgen.out("${structtype.scopedNameString}    .struct $structargs\n")
-            var paramIdx = 0
-            for(field in structtype.fields) {
-                val type = asmTypeString(field.type)
-                if(field.isArray) {
-                    val anonymous = List(field.arraySize!!) { "?" }.joinToString(",")
-                    asmgen.out("p8v_${field.name}  $type  $anonymous")
-                } else {
-                    asmgen.out("p8v_${field.name}  $type  \\f$paramIdx")
-                    paramIdx++
-                }
+            if(structtype.isUnion) {
+                // NB: .union takes no constructor parameters, unlike .struct
+                asmgen.out("${structtype.scopedNameString}    .union\n")
+                emitStructDefFields(structtype) { "?" }
+                asmgen.out("    .endunion\n")
+                asmgen.out("${structtype.scopedNameString}_size=${structtype.size}\n")
+            } else {
+                val paramFields = structtype.fields.filter { !it.isArray }
+                val structargs = paramFields.indices.joinToString(",") { "f$it" }
+                asmgen.out("${structtype.scopedNameString}    .struct $structargs\n")
+                var paramIdx = 0
+                emitStructDefFields(structtype) { "\\f${paramIdx++}" }
+                asmgen.out("    .endstruct\n")
+                asmgen.out("${structtype.scopedNameString}_size=${structtype.size}\n")
             }
-            asmgen.out("    .endstruct\n")
         }
 
-        val (instancesNoInit, instances) = symboltable.allStructInstances().partition { it.initialValues.isEmpty() }
+        val (instancesNoInit, instances) = symboltable.allStructInstances().partition {
+            val structtype = symboltable.lookup(it.structName) as StStruct
+            it.initialValues.isEmpty() || structtype.isUnion
+        }
         asmgen.out("; struct instances without initialization values, as BSS zeroed at startup\n")
         asmgen.out("    .section BSS\n")
         asmgen.out("${StStructInstanceBlockName}_bss  .block\n")
         instancesNoInit.forEach {
             val structtype: StStruct = symboltable.lookup(it.structName) as StStruct
-            asmgen.out("${it.name}    .fill  ${structtype.size}\n")
+            asmgen.out("${it.name}    .fill  ${structtype.size}    ; struct ${it.structName}\n")
         }
         asmgen.out("    .endblock\n")
         asmgen.out("    .send BSS\n")
@@ -483,6 +457,27 @@ internal class ProgramAndVarsGen(
         asmgen.out("    .send STRUCTINSTANCES\n")
     }
 
+    private fun emitStructDefFields(structtype: StStruct, scalarValue: () -> String) {
+        for(field in structtype.fields) {
+            val effectiveDt = if(field.isArray) field.type.elementType() else field.type
+            if(effectiveDt.isFloat) {
+                val floatSize = compTarget.FLOAT_MEM_SIZE.toInt()
+                if(field.isArray)
+                    asmgen.out("p8v_${field.name}  .fill  ${floatSize*field.arraySize!!}, ?")
+                else
+                    asmgen.out("p8v_${field.name}  .fill $floatSize, ${scalarValue()}")
+            } else {
+                val type = asmTypeString(field.type)
+                if(field.isArray) {
+                    val anonymous = List(field.arraySize!!) { "?" }.joinToString(",")
+                    asmgen.out("p8v_${field.name}  $type  $anonymous")
+                } else {
+                    asmgen.out("p8v_${field.name}  $type  ${scalarValue()}")
+                }
+            }
+        }
+    }
+
     internal fun translateAsmSubroutine(sub: PtAsmSub) {
         if(sub.inline) {
             return      // subroutine gets inlined at call site.
@@ -644,7 +639,7 @@ internal class ProgramAndVarsGen(
             arrayVarsWithInitInZp.forEach {
                 val size = it.alloc.size
                 val name = asmgen.asmVariableName(it.name)
-                if(it.alloc.dt.isSplitWordArray) {
+                if(it.alloc.dt.isSplitWordArray(compTarget)) {
                     val halfSize = size / 2
                     if(halfSize>0) {
                         asmgen.out("""
@@ -723,7 +718,7 @@ internal class ProgramAndVarsGen(
             if (scopedName.startsWith("cx16.r"))
                 continue        // The 16 virtual registers of the cx16 are not actual variables in zp, they're memory mapped
             val variable = symboltable.flat.getValue(scopedName) as StStaticVariable
-            if(variable.dt.isSplitWordArray) {
+            if(variable.dt.isSplitWordArray(compTarget)) {
                 val lsbAddr = zpvar.address
                 val msbAddr = zpvar.address + (zpvar.size/2).toUInt()
                 asmgen.out("${scopedName.substringAfterLast('.')}_lsb \t= $lsbAddr \t; zp ${zpvar.dt} (lsbs)")
@@ -815,21 +810,19 @@ internal class ProgramAndVarsGen(
             dt.isSignedWord -> asmgen.out("${variable.name}\t.sint  ?")
             dt.isLong -> asmgen.out("${variable.name}\t.dint  ?")
             dt.isFloat -> asmgen.out("${variable.name}\t.fill  ${compTarget.FLOAT_MEM_SIZE}")
-            dt.isSplitWordArray -> {
+            dt.isSplitWordArray(compTarget) -> {
                 alignVar(variable.align)
                 val numbytesPerHalf = compTarget.memorySize(variable.dt, variable.length!!.toInt()) / 2
-                asmgen.out("${variable.name}_lsb\t.fill  $numbytesPerHalf")
-                asmgen.out("${variable.name}_msb\t.fill  $numbytesPerHalf")
+                asmgen.out("${variable.name}_lsb\t.fill  $numbytesPerHalf  ; split word array of ${variable.length} ${variable.dt.elementType()} (lsbs)")
+                asmgen.out("${variable.name}_msb\t.fill  $numbytesPerHalf  ; split word array of ${variable.length} ${variable.dt.elementType()} (msbs)")
             }
             dt.isArray -> {
                 alignVar(variable.align)
-                val numbytes = compTarget.memorySize(variable.dt, variable.length!!.toInt())
-                asmgen.out("${variable.name}\t.fill  $numbytes")
+                val amount = variable.length!!.toInt()
+                val elementSize = compTarget.memorySize(variable.dt.elementType(), null)
+                asmgen.out("${variable.name}\t.fill  $amount * $elementSize    ; array of $amount ${variable.dt.elementType()} ")
             }
             dt.isPointer -> asmgen.out("${variable.name}\t.word  ?")        // a pointer is just an uword address
-            dt.isPointerArray -> {
-                TODO("pointers are not supported yet for uninitialized array ${variable.astNode?.position}")
-            }
             else -> {
                 throw AssemblyError("weird dt")
             }
@@ -906,7 +899,7 @@ internal class ProgramAndVarsGen(
                         asmgen.out("  .char  " + chunk.joinToString())
                 }
             }
-            dt.isSplitWordArray -> {
+            dt.isSplitWordArray(compTarget) -> {
                 if(dt.elementType().isUnsignedWord || dt.elementType().isPointer) {
                     val data = makeArrayFillDataUnsigned(dt, value, orNumberOfZeros)
                     asmgen.out("${varname}_words := ${data.joinToString()}")
@@ -960,6 +953,50 @@ internal class ProgramAndVarsGen(
                 for (f in array.zip(floatFills))
                     asmgen.out("  .byte  ${f.second}  ; float ${f.first}")
             }
+            dt.base==BaseDataType.ARRAY && dt.sub==BaseDataType.STRUCT_INSTANCE -> {
+                val struct = dt.subType as? StStruct ?: throw AssemblyError("struct type not found for $varname")
+                val structSize = struct.size.toInt()
+                if(value==null) {
+                    val totalBytes = (orNumberOfZeros ?: 0) * structSize
+                    if(totalBytes==0) {
+                        asmgen.out("$varname: .byte 0")
+                    } else {
+                        asmgen.out("$varname: .fill $totalBytes, 0")
+                    }
+                } else {
+                    // initialized struct array: value contains StructInstance elements
+                    asmgen.out("$varname     ; array of ${value.size} ${struct.name} structs")
+                    // expand struct field types for per-element handling
+                    val expandedFieldTypes = struct.fields.flatMap { field ->
+                        val arraySz = field.arraySize
+                        if(arraySz!=null) List(arraySz) { field.type } else listOf(field.type)
+                    }
+                    for(elt in value) {
+                        val instanceName = (elt as? StArrayElement.StructInstance)?.name ?: throw AssemblyError("expected struct instance in array init for $varname")
+                        val instance = symboltable.lookup(instanceName) as? StStructInstance ?: throw AssemblyError("struct instance $instanceName not found")
+                        val fieldValues = instance.initialValues
+                        require(fieldValues.size==expandedFieldTypes.size) { "field count mismatch for $instanceName" }
+                        for((fieldType, fieldVal) in expandedFieldTypes.zip(fieldValues)) {
+                            val num = when(fieldVal) {
+                                is StArrayElement.Number -> fieldVal.value
+                                is StArrayElement.BoolValue -> if(fieldVal.value) 1.0 else 0.0
+                                else -> throw AssemblyError("struct field value must be number for $varname (got $fieldVal)")
+                            }
+                            when {
+                                fieldType.isByteOrBool -> asmgen.out("  .byte  ${num.toInt()}")
+                                fieldType.isWord -> asmgen.out("  .word  ${num.toInt()}")
+                                fieldType.isLong -> asmgen.out("  .dword  ${num.toInt()}")
+                                fieldType.isFloat -> {
+                                    val bytes = compTarget.getFloatAsmBytes(num)
+                                    asmgen.out("  .byte  $bytes  ; float $num")
+                                }
+                                fieldType.isPointer -> asmgen.out("  .word  ${num.toInt()}  ; pointer")
+                                else -> throw AssemblyError("unsupported struct field type $fieldType")
+                            }
+                        }
+                    }
+                }
+            }
             else -> throw AssemblyError("require array dt")
         }
     }
@@ -975,14 +1012,14 @@ internal class ProgramAndVarsGen(
         consts.sortedBy { it.name }.forEach {
             val nameWithoutColons = it.name.replace("::", "_")
             if(it.dt.isFloat)
-                asmgen.out("  ${nameWithoutColons} = ${it.value}")
+                asmgen.out("  $nameWithoutColons = ${it.value}")
             else {
                 if(it.value!=null)
-                    asmgen.out("  ${nameWithoutColons} = ${it.value!!.toHex()}")
+                    asmgen.out("  $nameWithoutColons = ${it.value!!.toHex()}")
                 else if(it.memorySlab!=null)
-                    asmgen.out("  ${nameWithoutColons} = ${StMemorySlabBlockName}.${it.memorySlab!!.name}")
+                    asmgen.out("  $nameWithoutColons = ${StMemorySlabBlockName}.${it.memorySlab!!.name}")
                 else
-                    TODO("asm gen for ${nameWithoutColons} = memory alloc?")
+                    TODO("asm gen for $nameWithoutColons = memory alloc?")
             }
         }
     }
@@ -1031,7 +1068,7 @@ internal class ProgramAndVarsGen(
                     is StArrayElement.Number -> "$" + it.value.toInt().toString(16).padStart(4, '0')
                     is StArrayElement.AddressOf -> {
                         val symbol = symboltable.lookup(it.symbol)!!
-                        if(symbol is StStaticVariable && symbol.dt.isSplitWordArray)
+                        if(symbol is StStaticVariable && symbol.dt.isSplitWordArray(compTarget))
                             asmgen.asmSymbolName(it.symbol+"_lsb")
                         else
                             asmgen.asmSymbolName(it.symbol)
@@ -1072,7 +1109,7 @@ internal class ProgramAndVarsGen(
                         is StArrayElement.Number -> "$" + it.value.toInt().toString(16).padStart(4, '0')
                         is StArrayElement.AddressOf -> {
                             val symbol = symboltable.lookup(it.symbol)!!
-                            if(symbol is StStaticVariable && symbol.dt.isSplitWordArray)
+                            if(symbol is StStaticVariable && symbol.dt.isSplitWordArray(compTarget))
                                 asmgen.asmSymbolName(it.symbol+"_lsb")
                             else
                                 asmgen.asmSymbolName(it.symbol)

codeGenCpu6502/src/prog8/codegen/cpu6502/VariableAllocator.kt

@@ -29,7 +29,7 @@ internal class VariableAllocator(private val symboltable: SymbolTable,
             return true
 
         val v = symboltable.lookup(scopedName)
-        return if(v is StMemVar) v.address <= 255u else false
+        return v is StMemVar && v.address <= 255u
     }
 
     internal fun getFloatAsmConst(number: Double): String {
@@ -123,11 +123,11 @@ internal class VariableAllocator(private val symboltable: SymbolTable,
     private fun collectAllVariables(st: SymbolTable): Collection<StStaticVariable> {
         val vars = mutableListOf<StStaticVariable>()
         fun collect(node: StNode) {
-            for(child in node.children) {
-                if(child.value.type == StNodeType.STATICVAR)
-                    vars.add(child.value as StStaticVariable)
+            for((_, value) in node.children) {
+                if(value.type == StNodeType.STATICVAR)
+                    vars.add(value as StStaticVariable)
                 else
-                    collect(child.value)
+                    collect(value)
             }
         }
         collect(st)
@@ -136,6 +136,12 @@ internal class VariableAllocator(private val symboltable: SymbolTable,
 
     private fun computeUsageScores(vars: Collection<StStaticVariable>): Map<String, Int> {
         val scores = mutableMapOf<String, Int>()
+        // Note: by the time codegen runs, the simple-AST identifier names are already fully scoped
+        // AND symbol-prefixed ("p8b_maze.p8s_solve.p8v_cx"), so a plain lookup by PtIdentifier.name
+        // correctly matches the scopedNameString keys used here. This looks like a name/scope
+        // mismatch bug but is not: resolving identifiers to their enclosing scope instead (or
+        // keying this map by the bare variable name) silently makes every score zero, which
+        // degenerates the sort in allocateZeropageVariables() into a stable no-op.
         val varInfo = vars.associateBy { it.scopedNameString }
         var loopDepth = 0
 

codeGenCpu6502/src/prog8/codegen/cpu6502/assignment/AsmAssignment.kt

@@ -3,7 +3,6 @@ package prog8.codegen.cpu6502.assignment
 import prog8.code.StConstant
 import prog8.code.ast.*
 import prog8.code.core.*
-import prog8.code.core.BaseDataType
 import prog8.codegen.cpu6502.AsmGen6502Internal
 
 
@@ -237,7 +236,7 @@ internal class AsmAssignSource(val kind: SourceStorageKind,
                             if(sub is PtSub && sub.signature.returns.size>1)
                                 DataType.UNDEFINED      // TODO list of types instead?
                             else
-                                sub.returnsWhatWhere().firstOrNull { rr -> rr.first.registerOrPair != null || rr.first.statusflag!=null }?.second
+                                sub.returnsWhatWhere(asmgen.options.compTarget).firstOrNull { rr -> rr.first.registerOrPair != null || rr.first.statusflag!=null }?.second
                                 ?: throw AssemblyError("can't translate zero return values in assignment")
                         AsmAssignSource(SourceStorageKind.EXPRESSION, program, asmgen, returnType, expression = value)
                     }

codeGenCpu6502/src/prog8/codegen/cpu6502/assignment/AugmentableAssignmentAsmGen.kt

@@ -2,6 +2,8 @@ package prog8.codegen.cpu6502.assignment
 
 import prog8.code.ast.*
 import prog8.code.core.*
+import prog8.code.cpu6502.byteShiftAddExpansion
+import prog8.code.cpu6502.wordShiftAddExpansion
 import prog8.codegen.cpu6502.AsmGen6502Internal
 import prog8.codegen.cpu6502.VariableAllocator
 import prog8.codegen.cpu6502.toLongHex
@@ -119,6 +121,96 @@ internal class AugmentableAssignmentAsmGen(private val program: PtProgram,
         }
     }
 
+    private fun inplaceModificationPointerField(target: AsmAssignTarget, operator: String, value: AsmAssignSource) {
+        // the pointer to the target field is materialized in the shared scratch pointer
+        // P8ZP_SCRATCH_PTR because the pointer variable isn't in the zeropage (or the
+        // deref chain is too complex). Evaluating the source value may clobber that
+        // scratch pointer, so we operate on a temporary variable and preserve the
+        // address on the hardware stack, mirroring inplaceModificationIndexedPointerField.
+        val elementDt = target.datatype
+        if(!elementDt.isByteOrBool && !elementDt.isWord && !elementDt.isLong)
+            throw AssemblyError("inplace modification of pointer field for type $elementDt not supported ${target.position}")
+        val ptrTarget = PtrTarget(target)
+        val (zpPtrVar, offset) = ptrgen.deref(ptrTarget.pointer)
+        require(zpPtrVar=="P8ZP_SCRATCH_PTR")
+
+        val elementTemp = asmgen.createTempVarReused(elementDt.base, false, target.origAstTarget ?: target.pointer!!)
+        if(elementDt.isByteOrBool) {
+            asmgen.out("""
+                ldy  #$offset
+                lda  ($zpPtrVar),y
+                sta  $elementTemp""")
+        } else if(elementDt.isWord) {
+            asmgen.out("""
+                ldy  #$offset
+                lda  ($zpPtrVar),y
+                sta  $elementTemp
+                iny
+                lda  ($zpPtrVar),y
+                sta  $elementTemp+1""")
+        } else {
+            asmgen.out("""
+                ldy  #$offset
+                lda  ($zpPtrVar),y
+                sta  $elementTemp
+                iny
+                lda  ($zpPtrVar),y
+                sta  $elementTemp+1
+                iny
+                lda  ($zpPtrVar),y
+                sta  $elementTemp+2
+                iny
+                lda  ($zpPtrVar),y
+                sta  $elementTemp+3""")
+        }
+
+        // save the computed address on the hardware stack so later code can't clobber it
+        asmgen.out("""
+            lda  $zpPtrVar
+            pha
+            lda  $zpPtrVar+1
+            pha""")
+
+        // reuse the existing variable-target logic to modify the temp
+        val tempTarget = AsmAssignTarget(TargetStorageKind.VARIABLE, asmgen, elementDt, null, target.position, variableAsmName = elementTemp)
+        inplaceModification(tempTarget, operator, value)
+
+        // restore the address and store the temp back
+        asmgen.out("""
+            pla
+            sta  $zpPtrVar+1
+            pla
+            sta  $zpPtrVar""")
+        if(elementDt.isByteOrBool) {
+            asmgen.out("""
+                ldy  #$offset
+                lda  $elementTemp
+                sta  ($zpPtrVar),y""")
+        } else if(elementDt.isWord) {
+            asmgen.out("""
+                ldy  #$offset
+                lda  $elementTemp
+                sta  ($zpPtrVar),y
+                iny
+                lda  $elementTemp+1
+                sta  ($zpPtrVar),y""")
+        } else {
+            asmgen.out("""
+                ldy  #$offset
+                lda  $elementTemp
+                sta  ($zpPtrVar),y
+                iny
+                lda  $elementTemp+1
+                sta  ($zpPtrVar),y
+                iny
+                lda  $elementTemp+2
+                sta  ($zpPtrVar),y
+                iny
+                lda  $elementTemp+3
+                sta  ($zpPtrVar),y""")
+        }
+    }
+
     private fun inplaceModification(target: AsmAssignTarget, operator: String, value: AsmAssignSource) {
 
         // the asm-gen code can deal with situations where you want to assign a byte into a word.
@@ -611,13 +703,47 @@ internal class AugmentableAssignmentAsmGen(private val program: PtProgram,
                     }
                 }
             }
-            TargetStorageKind.POINTER -> ptrgen.inplaceModification(PtrTarget(target), operator, value)
+            TargetStorageKind.POINTER -> {
+                // if the effective target address ends up in the shared scratch pointer
+                // P8ZP_SCRATCH_PTR, evaluating the source value could clobber it. In that
+                // case operate on a protected temporary value instead.
+                val usesOffsetPointer = target.datatype.isFloat || operator in setOf("*", "/", "%")
+                val needsProtectedPtr = ptrgen.derefUsesScratchPointer(target.pointer!!, usesOffsetPointer)
+                if (value.kind in setOf(SourceStorageKind.EXPRESSION, SourceStorageKind.REGISTER)
+                        && needsProtectedPtr
+                        && (target.datatype.isByteOrBool || target.datatype.isWord || target.datatype.isLong))
+                    inplaceModificationPointerField(target, operator, value)
+                else
+                    ptrgen.inplaceModification(PtrTarget(target), operator, value)
+            }
             TargetStorageKind.REGISTER -> throw AssemblyError("no asm gen for reg in-place modification")
             TargetStorageKind.VOID -> { /* do nothing */ }
         }
     }
 
     internal fun inplacemodificationLongWithExpression(targetVar: String, operator: String, value: PtExpression) {
+        // Adding an unsigned byte to a long: skip zero-extending the byte into a
+        // full 4-byte long and just add it with carry propagation through the upper bytes.
+        // The byte shows up here as a (widened) long-typed cast of an ubyte expression,
+        // so detect it via the cast child type.
+        if(operator=="+" && value is PtTypeCast && value.value.type==DataType.UBYTE) {
+            assignmentAsmGen.assignExpressionToRegister(value.value, RegisterOrPair.A, false)
+            asmgen.out("""
+                clc
+                adc  $targetVar+0
+                sta  $targetVar+0
+                lda  $targetVar+1
+                adc  #0
+                sta  $targetVar+1
+                lda  $targetVar+2
+                adc  #0
+                sta  $targetVar+2
+                lda  $targetVar+3
+                adc  #0
+                sta  $targetVar+3
+                """)
+            return
+        }
         // it's not an expression so no need to preserve R14-R15
         if(value.type.isSigned) {
             assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R14R15, true)
@@ -988,8 +1114,12 @@ internal class AugmentableAssignmentAsmGen(private val program: PtProgram,
                 }
                 else -> {
                     asmgen.out("""
-                        lda  #0
-                        sta  $variable
+                        lda  $variable+3
+                        bpl  +
+                        lda  #255
+                        bne  ++
++                       lda  #0
++                       sta  $variable
                         sta  $variable+1
                         sta  $variable+2
                         sta  $variable+3""")
@@ -1727,7 +1857,7 @@ $shortcutLabel:""")
             // use the already existing optimized codegen for regular assignments  x += array[index]
             val binexpr = PtBinaryExpression(operator, dt, value.position)
             binexpr.add(PtIdentifier(name, dt, value.position))
-            val arrayValue = PtArrayIndexer(value.type, value.position)
+            val arrayValue = PtArrayIndexer(value.type, valueVar.type.isSplitWordArray(program.memsizer), value.position)
             arrayValue.add(valueVar)
             arrayValue.add(value.index)
             binexpr.add(arrayValue)
@@ -2117,6 +2247,11 @@ $shortcutLabel:""")
             "*" -> {
                 if(value in asmgen.optimizedByteMultiplications)
                     asmgen.out("  lda  $name |  jsr  prog8_math.mul_byte_$value |  sta  $name")
+                else if(byteShiftAddExpansion(value)!=null) {
+                    asmgen.out("  lda  $name")
+                    asmgen.emitByteShiftAdd(value)
+                    asmgen.out("  sta  $name")
+                }
                 else
                     asmgen.out("  lda  $name |  ldy  #$value |  jsr  prog8_math.multiply_bytes |  sta  $name")
             }
@@ -2428,8 +2563,7 @@ $shortcutLabel:""")
                 if(value in asmgen.optimizedWordMultiplications) {
                     asmgen.out("  lda  $lsb |  ldy  $msb |  jsr  prog8_math.mul_word_$value |  sta  $lsb |  sty  $msb")
                 } else {
-                    if(block?.options?.veraFxMuls==true) {
-                        // cx16 verafx hardware mul
+                    if(block?.options?.veraFxMuls==true) {                        // cx16 verafx hardware mul
                         asmgen.out("""
                             lda  $lsb
                             ldy  $msb
@@ -2443,6 +2577,11 @@ $shortcutLabel:""")
                             sta  $lsb
                             sty  $msb""")
                     }
+                    else if(wordShiftAddExpansion(value)!=null) {
+                        asmgen.out("  lda  $lsb |  ldy  $msb")
+                        asmgen.emitWordShiftAdd(value)
+                        asmgen.out("  sta  $lsb |  sty  $msb")
+                    }
                     else
                         asmgen.out("""
                             lda  $lsb

codeGenCpu6502/src/prog8/codegen/cpu6502/assignment/PointerAssignmentsGen.kt

@@ -160,7 +160,7 @@ internal class PointerAssignmentsGen(private val asmgen: AsmGen6502Internal, pri
                 sta  ($zpPtrVar),y""")
     }
 
-    internal fun inplaceByteNegate(target: PtrTarget, scope: IPtSubroutine?) {
+    internal fun inplaceByteNegate(target: PtrTarget) {
         val (zpPtrVar, offset) = deref(target.pointer)
         if(asmgen.isTargetCpu(CpuType.CPU65C02)) {
             asmgen.out("""
@@ -180,7 +180,7 @@ internal class PointerAssignmentsGen(private val asmgen: AsmGen6502Internal, pri
         }
     }
 
-    internal fun inplaceWordNegate(target: PtrTarget, scope: IPtSubroutine?) {
+    internal fun inplaceWordNegate(target: PtrTarget) {
         val (zpPtrVar, offset) = deref(target.pointer)
         asmgen.out("""
                 ldy  #$offset
@@ -194,7 +194,7 @@ internal class PointerAssignmentsGen(private val asmgen: AsmGen6502Internal, pri
                 sta  ($zpPtrVar),y""")
     }
 
-    internal fun inplaceLongNegate(target: PtrTarget, scope: IPtSubroutine?) {
+    internal fun inplaceLongNegate(target: PtrTarget) {
         val (zpPtrVar, offset) = deref(target.pointer)
         asmgen.out("""
                 ldy  #$offset
@@ -216,23 +216,33 @@ internal class PointerAssignmentsGen(private val asmgen: AsmGen6502Internal, pri
                 sta  ($zpPtrVar),y""")
     }
 
-    internal fun inplaceFloatNegate(target: PtrTarget, scope: IPtSubroutine?) {
+    internal fun inplaceFloatNegate(target: PtrTarget) {
         val (zpPtrVar, offset) = deref(target.pointer)
         // flip sign bit in second byte of 5-byte float
-        asmgen.out("""
-                ldy  #$offset
+        asmgen.out(
+            $$"""
+                ldy  #$$offset
                 iny
-                lda  ($zpPtrVar),y
-                eor  #${'$'}80
-                sta  ($zpPtrVar),y""")
+                lda  ($$zpPtrVar),y
+                eor  #$80
+                sta  ($$zpPtrVar),y"""
+        )
     }
 
 
+    /**
+     * Walk the pointer deref chain. Always returns (zpVar, offset) such that one indirect
+     * access through zpVar at byte offset reaches the final thing:
+     *   derefLast=true  -> (zpVar),offset is the address of the value itself
+     *   derefLast=false -> (zpVar),offset is the location of the field content
+     * This contract holds for every chain length: the final pointer (if any) is already
+     * followed inside deref(), so callers never need a second dereference.
+     * zpVar is P8ZP_SCRATCH_PTR for non-empty chains or non-zeropage pointers; it may be
+     * the original (zeropage) pointer variable only for an empty chain with derefLast=false.
+     * loadPointerFieldValue: for addOffsetToPointer=true, also load the value if the final
+     * field is a pointer-typed field (rather than returning its address).
+     */
     internal fun deref(pointer: PtPointerDeref, addOffsetToPointer: Boolean=false, loadPointerFieldValue: Boolean=false): Pair<String, UByte> {
-        // walk the pointer deref chain and leaves the final pointer value in a ZP var
-        // this will often be the temp var P8ZP_SCRATCH_PTR but can also be the original pointer variable if it is already in zeropage and there is nothing to add to it
-        // returns the ZP var to use as a pointer, and a Y register offset (which can be zero)
-        // loadPointerFieldValue: for addOffsetToPointer=true, also load the value if the final field is a pointer-typed field (rather than returning its address).
 
         fun addFieldOffsetToScratchPointer(fieldoffset: UInt) {
             if(fieldoffset==0u)
@@ -285,20 +295,22 @@ internal class PointerAssignmentsGen(private val asmgen: AsmGen6502Internal, pri
                     if(pointer.derefLast) {
                         asmgen.assignExpressionToVariable(pointer.startpointer, "P8ZP_SCRATCH_PTR", DataType.UWORD)
                         addFieldOffsetToScratchPointer(field.second.toUInt())
+                        updateScratchPointer()
                         return "P8ZP_SCRATCH_PTR" to 0u
                     }
                     else
-                        return pointer.startpointer.name to field.second
+                        return pointer.startpointer.name to field.second.toUByte()
                 }
                 else {
                     // have to copy it to temp zp var
                     asmgen.assignExpressionToVariable(pointer.startpointer, "P8ZP_SCRATCH_PTR", DataType.UWORD)
                     if(pointer.derefLast) {
                         addFieldOffsetToScratchPointer(field.second.toUInt())
+                        updateScratchPointer()
                         return "P8ZP_SCRATCH_PTR" to 0u
                     }
                     else
-                        return "P8ZP_SCRATCH_PTR" to field.second
+                        return "P8ZP_SCRATCH_PTR" to field.second.toUByte()
                 }
             }
         }
@@ -331,10 +343,24 @@ internal class PointerAssignmentsGen(private val asmgen: AsmGen6502Internal, pri
                     updateScratchPointer()
                 return "P8ZP_SCRATCH_PTR" to 0u
             } else
-                return "P8ZP_SCRATCH_PTR" to fieldinfo.second
+                return "P8ZP_SCRATCH_PTR" to fieldinfo.second.toUByte()
         }
     }
 
+    internal fun derefUsesScratchPointer(pointer: PtPointerDeref, addOffsetToPointer: Boolean=false): Boolean {
+        // determine if deref() would need to keep the effective address in the shared
+        // scratch pointer P8ZP_SCRATCH_PTR, which a subsequent evaluation of a source
+        // expression could then clobber. Only if it doesn't, may the pointer itself (if
+        // located in the zeropage) keep being used for the read-modify-write.
+        if(pointer.chain.isEmpty())
+            return !allocator.isZpVar(pointer.startpointer.name)
+        if(pointer.chain.size > 1 || addOffsetToPointer)
+            return true
+        // chain==1 derefLast=true now fully follows into P8ZP_SCRATCH_PTR (see deref()),
+        // so it always uses the scratch pointer regardless of isZpVar.
+        return !(allocator.isZpVar(pointer.startpointer.name) && !pointer.derefLast)
+    }
+
     internal fun indexedEffectiveAddress(target: IndexedPtrTarget): String {
         val eltSize = asmgen.program.memsizer.memorySize(target.elementDt, null)
         // generic: evaluate index as word, scale by element size, add field-corrected base pointer
@@ -381,15 +407,7 @@ internal class PointerAssignmentsGen(private val asmgen: AsmGen6502Internal, pri
     }
 
     internal fun assignPointerDerefExpression(target: AsmAssignTarget, value: PtPointerDeref) {
-        val (zpPtrVar, offset1) = deref(value)
-        val offset: UByte
-        if(value.derefLast) {
-            asmgen.loadIndirectWordAY(zpPtrVar, offset1)
-            asmgen.out("  sta  $zpPtrVar |  sty  $zpPtrVar+1")
-            offset = 0.toUByte()
-        } else {
-            offset = offset1
-        }
+        val (zpPtrVar, offset) = deref(value)
         if(value.type.isByteOrBool) {
             asmgen.loadIndirectByte(zpPtrVar, offset)
             asmgen.assignRegister(RegisterOrPair.A, target)
@@ -944,7 +962,7 @@ internal class PointerAssignmentsGen(private val asmgen: AsmGen6502Internal, pri
         val right = binExpr.right as? PtIdentifier
         require(binExpr.operator=="." && left!=null && right!=null) {"invalid dereference expression ${binExpr.position}"}
 
-        val field: Pair<DataType, UByte>
+        val field: Pair<DataType, Int>
         var extraFieldOffset = 0
 
         if(left.type.isStructInstance) {
@@ -975,7 +993,7 @@ internal class PointerAssignmentsGen(private val asmgen: AsmGen6502Internal, pri
             field = struct.getField(right.name, asmgen.program.memsizer)
         }
 
-        val offset = extraFieldOffset + field.second.toInt()
+        val offset = extraFieldOffset + field.second
         // TODO avoid always using a zp scratch reg below if the only thing we dereferenced above is a pointer variable that is already in zeropage itself
         if(offset>=256) {
             // add field offset to pointer if it doesn't fit in the index register

codeGenCpu6502/src/prog8/codegen/cpu6502/assignment/TypeCastAssignmentsGen.kt

@@ -0,0 +1,860 @@
+package prog8.codegen.cpu6502.assignment
+
+import prog8.code.ast.*
+import prog8.code.core.*
+import prog8.codegen.cpu6502.AsmGen6502Internal
+
+internal class TypeCastAssignmentsGen(
+    private val program: PtProgram,
+    private val asmgen: AsmGen6502Internal,
+    private val pointergen: PointerAssignmentsGen
+) {
+    lateinit var assignmentAsmGen: AssignmentAsmGen
+
+    internal fun assignTypeCastedValue(target: AsmAssignTarget, targetDt: DataType, value: PtExpression, origTypeCastExpression: PtTypeCast) {
+        val valueDt = value.type
+        if(valueDt==targetDt)
+            throw AssemblyError("type cast to identical dt should have been removed: $valueDt at ${value.position}")
+
+        when(value) {
+            is PtIdentifier -> {
+                if(targetDt.isWord) {
+                    if(valueDt.isUnsignedByte || valueDt.isBool) {
+                        assignmentAsmGen.assignVariableUByteIntoWord(target, value)
+                        return
+                    }
+                    if(valueDt.isSignedByte) {
+                        assignmentAsmGen.assignVariableByteIntoWord(target, value)
+                        return
+                    }
+                }
+            }
+            is PtMemoryByte -> {
+                if(targetDt.isWord) {
+
+                    fun assignViaExprEval(addressExpression: PtExpression) {
+                        asmgen.assignExpressionToVariable(addressExpression, "P8ZP_SCRATCH_W2", DataType.UWORD)
+                        asmgen.loadAFromZpPointerVar("P8ZP_SCRATCH_W2")
+                        asmgen.out("  ldy  #0")
+                        assignmentAsmGen.assignRegisterpairWord(target, RegisterOrPair.AY)
+                    }
+
+                    when (value.address) {
+                        is PtNumber -> {
+                            val address = (value.address as PtNumber).number.toUInt()
+                            assignmentAsmGen.assignMemoryByteIntoWord(target, address, null)
+                        }
+                        is PtIdentifier -> {
+                            assignmentAsmGen.assignMemoryByteIntoWord(target, null, value.address as PtIdentifier)
+                        }
+                        is PtBinaryExpression -> {
+                            val addrExpr = value.address as PtBinaryExpression
+                            if(asmgen.tryOptimizedPointerAccessWithA(addrExpr, false)) {
+                                asmgen.out("  ldy  #0")
+                                assignmentAsmGen.assignRegisterpairWord(target, RegisterOrPair.AY)
+                            } else {
+                                assignViaExprEval(value.address)
+                            }
+                        }
+                        else -> {
+                            assignViaExprEval(value.address)
+                        }
+                    }
+                    return
+                }
+            }
+            is PtNumber -> {
+                if(targetDt.isPointer) {
+                    // assign a number to a pointer type
+                    require(valueDt.isInteger)
+                    return assignmentAsmGen.assignConstantWord(target, value.number.toInt())
+                }
+                else
+                    throw AssemblyError("literal value cast should have been const-folded away (target type=$targetDt)")
+            }
+            is PtBool -> throw AssemblyError("literal value cast should have been const-folded away (target type=$targetDt)")
+            is PtArrayIndexer -> {
+                if(targetDt.isByte && valueDt.isWord) {
+                    // just assign the lsb from the array value
+                    return assignCastWordViaLsbFunc(value, target)
+                } else if(targetDt.isBool && valueDt.isWord) {
+                    return assignWordToBool(value, target)
+                }
+            }
+            else -> {}
+        }
+
+
+        // special case optimizations
+        if(target.kind == TargetStorageKind.VARIABLE) {
+            if(value is PtIdentifier && !valueDt.isUndefined)
+                return assignTypeCastedIdentifier(target.asmVarname, targetDt, asmgen.asmVariableName(value), valueDt)
+
+            when {
+                valueDt.isBool -> {
+                    if(targetDt.isInteger) {
+                        // optimization to assign boolean expression to integer target (just assign the 0 or 1 directly)
+                        val assignDirect = AsmAssignment(
+                            AsmAssignSource.fromAstSource(value, program, asmgen),
+                            listOf(target),
+                            program.memsizer,
+                            target.position
+                        )
+                        assignmentAsmGen.assignExpression(assignDirect, target.scope)
+                    } else {
+                        TODO("assign bool to non-integer type $targetDt ${value.position}")
+                    }
+                }
+                valueDt.isByte -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.A, valueDt.isSigned)
+                    assignTypeCastedRegisters(target.asmVarname, targetDt.base, RegisterOrPair.A, valueDt.base, target.position)
+                }
+                valueDt.isLong -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R14R15, valueDt.isSigned)
+                    assignTypeCastedRegisters(target.asmVarname, targetDt.base, RegisterOrPair.R14R15, valueDt.base, target.position)
+                }
+                valueDt.isWord || valueDt.isPointer -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.AY, valueDt.isSigned)
+                    assignTypeCastedRegisters(target.asmVarname, targetDt.base, RegisterOrPair.AY, valueDt.base, target.position)
+                }
+                valueDt.isFloat -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.FAC1, true)
+                    assignTypeCastedFloatFAC1(target.asmVarname, targetDt.base)
+                }
+                valueDt.isPassByRef -> {
+                    // str/array value cast (most likely to UWORD, take address-of)
+                    assignmentAsmGen.assignExpressionToVariable(value, target.asmVarname, targetDt)
+                }
+                else -> throw AssemblyError("strange dt in typecast assign to var: $valueDt  -->  $targetDt")
+            }
+            return
+        }
+
+        if(valueDt.isWord && origTypeCastExpression.type.isUnsignedByte) {
+            val parentTc = origTypeCastExpression.parent as? PtTypeCast
+            if(parentTc!=null && parentTc.type.isUnsignedWord) {
+                // typecast a word value to ubyte and directly back to uword
+                // generate code for lsb(value) here instead of the ubyte typecast
+                return assignCastWordViaLsbFunc(value, target)
+            }
+        }
+
+        if(valueDt.isByteOrBool) {
+            when(target.register) {
+                RegisterOrPair.A,
+                RegisterOrPair.X,
+                RegisterOrPair.Y -> {
+                    // 'cast' an ubyte value to a byte register; no cast needed at all
+                    return assignmentAsmGen.assignExpressionToRegister(value, target.register, valueDt.isSigned)
+                }
+                RegisterOrPair.AX,
+                RegisterOrPair.AY,
+                RegisterOrPair.XY,
+                in Cx16VirtualRegisters -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.A, false)
+                    assignmentAsmGen.assignRegisterByte(target, CpuRegister.A, valueDt.isSigned, true)
+                    return
+                }
+                in CombinedLongRegisters -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.A, false)
+                    assignmentAsmGen.assignRegisterByte(target, CpuRegister.A, valueDt.isSigned, true)
+                    return
+                }
+                else -> {}
+            }
+        } else if(valueDt.isWord) {
+            when(target.register) {
+                RegisterOrPair.A,
+                RegisterOrPair.X,
+                RegisterOrPair.Y -> {
+                    return if(targetDt.isBool)
+                        assignWordToBool(value, target)
+                    else
+                        // cast an uword to a byte register, do this via lsb(value)
+                        // generate code for lsb(value) here instead of the ubyte typecast
+                        assignCastWordViaLsbFunc(value, target)
+                }
+                RegisterOrPair.AX,
+                RegisterOrPair.AY,
+                RegisterOrPair.XY,
+                in Cx16VirtualRegisters -> {
+                    // 'cast' uword into a 16 bits register, just assign it
+                    return assignmentAsmGen.assignExpressionToRegister(value, target.register!!, targetDt.isSigned)
+                }
+                RegisterOrPair.R0R1 -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R0, false)
+                    return asmgen.signExtendLongVariable("cx16.r0", valueDt.base)
+                }
+                RegisterOrPair.R2R3 -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R2, false)
+                    return asmgen.signExtendLongVariable("cx16.r2", valueDt.base)
+                }
+                RegisterOrPair.R4R5 -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R4, false)
+                    return asmgen.signExtendLongVariable("cx16.r4", valueDt.base)
+                }
+                RegisterOrPair.R6R7 -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R6, false)
+                    return asmgen.signExtendLongVariable("cx16.r6", valueDt.base)
+                }
+                RegisterOrPair.R8R9 -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R8, false)
+                    return asmgen.signExtendLongVariable("cx16.r8", valueDt.base)
+                }
+                RegisterOrPair.R10R11 -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R10, false)
+                    return asmgen.signExtendLongVariable("cx16.r10", valueDt.base)
+                }
+                RegisterOrPair.R12R13 -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R12, false)
+                    return asmgen.signExtendLongVariable("cx16.r12", valueDt.base)
+                }
+                RegisterOrPair.R14R15 -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R14, false)
+                    return asmgen.signExtendLongVariable("cx16.r14", valueDt.base)
+                }
+                else -> {}
+            }
+        }
+
+        if(target.kind==TargetStorageKind.REGISTER) {
+            if(valueDt.isFloat && !target.datatype.isFloat) {
+                // have to typecast the float number on the fly down to an integer
+                assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.FAC1, targetDt.isSigned)
+                assignTypeCastedFloatFAC1("P8ZP_SCRATCH_W1", targetDt.base)
+                assignmentAsmGen.assignVariableToRegister("P8ZP_SCRATCH_W1", target.register!!, targetDt.isSigned, origTypeCastExpression.definingISub(), target.position)
+                return
+            } else {
+                if(!(valueDt isAssignableTo targetDt)) {
+                    return if(valueDt.isWord && targetDt.isByte) {
+                        // word to byte, just take the lsb
+                        assignCastWordViaLsbFunc(value, target)
+                    } else if(valueDt.isWord && targetDt.isBool) {
+                        // word to bool
+                        assignWordToBool(value, target)
+                    } else if(valueDt.isWord && targetDt.isWord) {
+                        // word to word, just assign
+                        assignmentAsmGen.assignExpressionToRegister(value, target.register!!, valueDt.isSigned)
+                    } else if(valueDt.isByteOrBool && targetDt.isByte) {
+                        // byte to byte, just assign
+                        assignmentAsmGen.assignExpressionToRegister(value, target.register!!, valueDt.isSigned)
+                    } else if(valueDt.isByteOrBool && targetDt.isWord) {
+                        // byte to word, just assign
+                        assignmentAsmGen.assignExpressionToRegister(value, target.register!!, valueDt.isSigned)
+                    } else if(valueDt.isLong && targetDt.isByte) {
+                        // long to byte, just take the lsb
+                        assignCastLongToByte(value, target)
+                    } else if(valueDt.isLong && targetDt.isWord) {
+                        // long to word, just take the lsw
+                        assignCastViaLswFunc(value, target)
+                    } else if(valueDt.isPointer && targetDt.isPointer) {
+                        // pointer type A to pointer type B, just assign
+                        assignmentAsmGen.assignExpressionToRegister(value, target.register!!, valueDt.isSigned)
+                    } else
+                        throw AssemblyError("can't cast $valueDt to $targetDt, this should have been checked in the astchecker  ${value.position}")
+                }
+            }
+        }
+
+        if(targetDt.isInteger && valueDt.isByteOrBool && valueDt.isAssignableTo(targetDt)) {
+            require(targetDt.isWord || targetDt.isLong) {
+                "should be byte to word or long assignment ${origTypeCastExpression.position}"
+            }
+            when(target.kind) {
+//                TargetStorageKind.VARIABLE -> {
+//                    This has been handled already earlier on line 961.
+//                    // byte to word, just assign to registers first, then assign to variable
+//                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.AY, targetDt==BaseDataType.WORD)
+//                    assignTypeCastedRegisters(target.asmVarname, targetDt, RegisterOrPair.AY, targetDt)
+//                    return
+//                }
+                TargetStorageKind.ARRAY -> {
+                    // byte to word, just assign to registers first, then assign into array
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.AY, targetDt.isSigned)
+                    val deref = target.array!!.pointerderef
+                    if(deref!=null)
+                        pointergen.assignWordReg(IndexedPtrTarget(target), RegisterOrPair.AY)
+                    else
+                        assignmentAsmGen.assignRegisterpairWord(target, RegisterOrPair.AY)
+                    return
+                }
+                TargetStorageKind.REGISTER -> {
+                    // byte to word or long, just assign to registers
+                    assignmentAsmGen.assignExpressionToRegister(value, target.register!!, targetDt.isSigned)
+                    return
+                }
+                TargetStorageKind.POINTER -> {
+                    pointergen.assignByteToWord(PtrTarget(target), value)
+                    return
+                }
+                else -> throw AssemblyError("weird target at ${target.position}")
+            }
+        }
+
+        if(targetDt.isFloat && (target.register==RegisterOrPair.FAC1 || target.register==RegisterOrPair.FAC2)) {
+            if(target.register==RegisterOrPair.FAC2)
+                asmgen.pushFAC1()
+            when {
+                valueDt.isUnsignedByte -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.Y, false)
+                    asmgen.out("  jsr  floats.FREADUY")
+                }
+                valueDt.isSignedByte -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.A, true)
+                    asmgen.out("  jsr  floats.FREADSA")
+                }
+                valueDt.isUnsignedWord -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.AY, false)
+                    asmgen.out("  jsr  floats.GIVUAYFAY")
+                }
+                valueDt.isSignedWord -> {
+                    assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.AY, true)
+                    asmgen.out("  jsr  floats.GIVAYFAY")
+                }
+                valueDt.isLong -> {
+                    when(value) {
+                        is PtIdentifier -> {
+                            val varname = asmgen.asmVariableName(value)
+                            asmgen.out("""
+                                lda  #<$varname
+                                ldy  #>$varname
+                                jsr  floats.internal_long_AY_to_FAC""")
+                        }
+                        else -> {
+                            assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R14R15, true)
+                            asmgen.out("""
+                                lda  #<cx16.r14
+                                ldy  #>cx16.r14
+                                jsr  floats.internal_long_AY_to_FAC""")
+                        }
+                    }
+                }
+                else -> throw AssemblyError("invalid dt at ${target.position}")
+            }
+            if(target.register==RegisterOrPair.FAC2) {
+                asmgen.out("  jsr  floats.MOVEF")
+                asmgen.popFAC1()
+            }
+            return
+        }
+
+        if(targetDt.isUnsignedWord && valueDt.isPointer) {
+            assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.AY, false)
+            assignmentAsmGen.assignRegisterpairWord(target, RegisterOrPair.AY)
+            return
+        }
+
+        // No more special optimized cases yet. Do the rest via more complex evaluation
+        // note: cannot use assignTypeCastedValue because that is ourselves :P
+        // NOTE: THIS MAY TURN INTO A STACK OVERFLOW ERROR IF IT CAN'T SIMPLIFY THE TYPECAST..... :-/
+        asmgen.assignExpressionTo(origTypeCastExpression, target)
+    }
+
+
+    internal fun assignCastLongToByte(value: PtExpression, target: AsmAssignTarget) {
+        // long to byte, can't use lsb() because that only works on words
+        when(value) {
+            is PtIdentifier -> {
+                val longvar = asmgen.asmVariableName(value)
+                asmgen.out("  lda  $longvar")
+                assignmentAsmGen.assignRegisterByte(target, CpuRegister.A, true, false)
+            }
+            is PtNumber -> throw AssemblyError("casting a long number to byte should have been const-folded away ${value.position}")
+            else -> {
+                asmgen.pushLongRegisters(RegisterOrPair.R14R15, 1)
+                assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.R14R15, true)
+                asmgen.out("  lda  cx16.r14")
+                assignmentAsmGen.assignRegisterByte(target, CpuRegister.A, true, false)
+                asmgen.popLongRegisters(RegisterOrPair.R14R15, 1)
+            }
+        }
+    }
+
+
+    internal fun assignCastWordViaLsbFunc(value: PtExpression, target: AsmAssignTarget) {
+        val lsb = PtFunctionCall("lsb", true, true,arrayOf(DataType.UBYTE), value.position)
+        lsb.parent = value.parent
+        lsb.add(value)
+        val src = AsmAssignSource(SourceStorageKind.EXPRESSION, program, asmgen, DataType.UBYTE, expression = lsb)
+        val assign = AsmAssignment(src, listOf(target), program.memsizer, value.position)
+        assignmentAsmGen.translateNormalAssignment(assign, value.definingISub())
+    }
+
+
+    internal fun assignCastViaLswFunc(value: PtExpression, target: AsmAssignTarget) {
+        val lsb = PtFunctionCall("lsw", true, true, arrayOf(DataType.UWORD), value.position)
+        lsb.parent = value.parent
+        lsb.add(value)
+        val src = AsmAssignSource(SourceStorageKind.EXPRESSION, program, asmgen, DataType.UWORD, expression = lsb)
+        val assign = AsmAssignment(src, listOf(target), program.memsizer, value.position)
+        assignmentAsmGen.translateNormalAssignment(assign, value.definingISub())
+    }
+
+
+    internal fun assignWordToBool(value: PtExpression, target: AsmAssignTarget) {
+        assignmentAsmGen.assignExpressionToRegister(value, RegisterOrPair.AY, false)
+        asmgen.out("""
+            sty  P8ZP_SCRATCH_REG
+            ora  P8ZP_SCRATCH_REG
+            beq  +
+            lda  #1
++""")
+        assignmentAsmGen.assignRegisterByte(target, CpuRegister.A, false, false)
+    }
+
+
+    internal fun assignTypeCastedFloatFAC1(targetAsmVarName: String, targetDt: BaseDataType) {
+
+        if(targetDt==BaseDataType.FLOAT)
+            throw AssemblyError("typecast to identical type")
+
+        when(targetDt) {
+            BaseDataType.BOOL -> asmgen.out("  jsr  floats.cast_FAC1_as_bool_into_a |  sta  $targetAsmVarName")
+            BaseDataType.UBYTE -> asmgen.out("  jsr  floats.cast_FAC1_as_uw_into_ya |  sty  $targetAsmVarName")
+            BaseDataType.BYTE -> asmgen.out("  jsr  floats.cast_FAC1_as_w_into_ay |  sta  $targetAsmVarName")
+            BaseDataType.UWORD -> asmgen.out("  jsr  floats.cast_FAC1_as_uw_into_ya |  sty  $targetAsmVarName |  sta  $targetAsmVarName+1")
+            BaseDataType.WORD -> asmgen.out("  jsr  floats.cast_FAC1_as_w_into_ay |  sta  $targetAsmVarName |  sty  $targetAsmVarName+1")
+            BaseDataType.LONG -> {
+                asmgen.out("""
+                    jsr  floats.QINT
+                    lda  floats.FAC_ADDR+4
+                    sta  $targetAsmVarName
+                    lda  floats.FAC_ADDR+3
+                    sta  $targetAsmVarName+1
+                    lda  floats.FAC_ADDR+2
+                    sta  $targetAsmVarName+2
+                    lda  floats.FAC_ADDR+1
+                    sta  $targetAsmVarName+3
+                """.trimIndent())
+            }
+            else -> throw AssemblyError("weird type $targetDt")
+        }
+    }
+
+
+
+    internal fun assignTypeCastedIdentifier(targetAsmVarName: String, targetDt: DataType,
+                                           sourceAsmVarName: String, sourceDt: DataType) {
+        if(sourceDt == targetDt)
+            throw AssemblyError("typecast to identical type")
+
+        // also see: PtExpressionAsmGen,   fun translateExpression(typecast: PtTypeCast)
+        when {
+            sourceDt.isUnsignedByte || sourceDt.isBool -> {
+                when(targetDt.base) {
+                    BaseDataType.BOOL -> {
+                        asmgen.out("""
+                        lda  $sourceAsmVarName
+                        beq  +
+                        lda  #1
++                       sta  $targetAsmVarName""")
+                    }
+                    BaseDataType.UBYTE, BaseDataType.BYTE -> {
+                        asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName")
+                    }
+                    BaseDataType.UWORD, BaseDataType.WORD -> {
+                        if(asmgen.isTargetCpu(CpuType.CPU65C02))
+                            asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName |  stz  $targetAsmVarName+1")
+                        else
+                            asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName |  lda  #0  |  sta  $targetAsmVarName+1")
+                    }
+                    BaseDataType.POINTER -> TODO("byte should not be cast to pointer")
+                    BaseDataType.LONG -> {
+                        asmgen.out("""
+                            lda  $sourceAsmVarName
+                            sta  $targetAsmVarName
+                            lda  #0
+                            sta  $targetAsmVarName+1
+                            sta  $targetAsmVarName+2
+                            sta  $targetAsmVarName+3""")
+                    }
+                    BaseDataType.FLOAT -> {
+                        asmgen.out("""
+                            lda  #<$targetAsmVarName
+                            ldy  #>$targetAsmVarName
+                            sta  P8ZP_SCRATCH_W2
+                            sty  P8ZP_SCRATCH_W2+1
+                            ldy  $sourceAsmVarName
+                            jsr  floats.cast_from_ub""")
+                    }
+                    else -> throw AssemblyError("weird type $targetDt")
+                }
+            }
+            sourceDt.isSignedByte -> {
+                when(targetDt.base) {
+                    BaseDataType.UBYTE, BaseDataType.BOOL -> {
+                        asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName")
+                    }
+                    BaseDataType.UWORD -> {
+                        if(asmgen.isTargetCpu(CpuType.CPU65C02))
+                            asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName |  stz  $targetAsmVarName+1")
+                        else
+                            asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName |  lda  #0  |  sta  $targetAsmVarName+1")
+                    }
+                    BaseDataType.WORD -> {
+                        asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName")
+                        asmgen.signExtendVariableLsb(targetAsmVarName, BaseDataType.BYTE)
+                    }
+                    BaseDataType.POINTER -> TODO("byte should not be cast to pointer")
+                    BaseDataType.LONG -> {
+                        asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName")
+                        asmgen.signExtendVariableLsb(targetAsmVarName, BaseDataType.BYTE)
+                        asmgen.signExtendLongVariable(targetAsmVarName, BaseDataType.WORD)
+                    }
+                    BaseDataType.FLOAT -> {
+                        asmgen.out("""
+                            lda  #<$targetAsmVarName
+                            ldy  #>$targetAsmVarName
+                            sta  P8ZP_SCRATCH_W2
+                            sty  P8ZP_SCRATCH_W2+1
+                            lda  $sourceAsmVarName
+                            jsr  floats.cast_from_b""")
+                    }
+                    else -> throw AssemblyError("weird type")
+                }
+            }
+            sourceDt.isUnsignedWord -> {
+                when(targetDt.base) {
+                    BaseDataType.BOOL -> {
+                        asmgen.out("""
+                            lda  $sourceAsmVarName
+                            ora  $sourceAsmVarName+1
+                            beq  +
+                            lda  #1
++                           sta  $targetAsmVarName""")
+                    }
+                    BaseDataType.BYTE, BaseDataType.UBYTE -> {
+                        asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName")
+                    }
+                    BaseDataType.WORD -> {
+                        asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName |  lda  $sourceAsmVarName+1 |  sta  $targetAsmVarName+1")
+                    }
+                    BaseDataType.POINTER -> asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName |  lda  $sourceAsmVarName+1 |  sta  $targetAsmVarName+1")
+                    BaseDataType.LONG -> {
+                        asmgen.out("""
+                            lda  $sourceAsmVarName
+                            sta  $targetAsmVarName
+                            lda  $sourceAsmVarName+1
+                            sta  $targetAsmVarName+1
+                            lda  #0
+                            sta  $targetAsmVarName+2
+                            sta  $targetAsmVarName+3""")
+                    }
+                    BaseDataType.FLOAT -> {
+                        asmgen.out("""
+                            lda  #<$targetAsmVarName
+                            ldy  #>$targetAsmVarName
+                            sta  P8ZP_SCRATCH_W2
+                            sty  P8ZP_SCRATCH_W2+1
+                            lda  $sourceAsmVarName
+                            ldy  $sourceAsmVarName+1
+                            jsr  floats.cast_from_uw""")
+                    }
+                    else -> throw AssemblyError("weird type")
+                }
+            }
+            sourceDt.isSignedWord || sourceDt.isPointer -> {
+                when(targetDt.base) {
+                    BaseDataType.BOOL -> {
+                        asmgen.out("""
+                            lda  $sourceAsmVarName
+                            ora  $sourceAsmVarName+1
+                            beq  +
+                            lda  #1
++                           sta  $targetAsmVarName""")
+                    }
+                    BaseDataType.BYTE, BaseDataType.UBYTE -> {
+                        asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName")
+                    }
+                    BaseDataType.UWORD, BaseDataType.POINTER -> {
+                        asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName |  lda  $sourceAsmVarName+1 |  sta  $targetAsmVarName+1")
+                    }
+                    BaseDataType.LONG -> {
+                        asmgen.out("""
+                            lda  $sourceAsmVarName
+                            sta  $targetAsmVarName
+                            lda  $sourceAsmVarName+1
+                            sta  $targetAsmVarName+1""")
+                        asmgen.signExtendLongVariable(targetAsmVarName, BaseDataType.WORD)
+                    }
+                    BaseDataType.FLOAT -> {
+                        asmgen.out("""
+                            lda  #<$targetAsmVarName
+                            ldy  #>$targetAsmVarName
+                            sta  P8ZP_SCRATCH_W2
+                            sty  P8ZP_SCRATCH_W2+1
+                            lda  $sourceAsmVarName
+                            ldy  $sourceAsmVarName+1
+                            jsr  floats.cast_from_w""")
+                    }
+                    else -> throw AssemblyError("weird type")
+                }
+            }
+            sourceDt.isFloat -> {
+                asmgen.out("  lda  #<$sourceAsmVarName |  ldy  #>$sourceAsmVarName")
+                when(targetDt.base) {
+                    BaseDataType.BOOL -> asmgen.out("  jsr  floats.cast_as_bool_into_a |  sta  $targetAsmVarName")
+                    BaseDataType.UBYTE -> asmgen.out("  jsr  floats.cast_as_uw_into_ya |  sty  $targetAsmVarName")
+                    BaseDataType.BYTE -> asmgen.out("  jsr  floats.cast_as_w_into_ay |  sta  $targetAsmVarName")
+                    BaseDataType.UWORD -> asmgen.out("  jsr  floats.cast_as_uw_into_ya |  sty  $targetAsmVarName |  sta  $targetAsmVarName+1")
+                    BaseDataType.WORD -> asmgen.out("  jsr  floats.cast_as_w_into_ay |  sta  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                    BaseDataType.LONG -> {
+                        asmgen.out("""
+                            sta  cx16.r0L
+                            sty  cx16.r0H
+                            lda  #<$targetAsmVarName
+                            ldy  #>$targetAsmVarName
+                            jsr  floats.cast_as_long""")
+                    }
+                    else -> throw AssemblyError("weird type")
+                }
+            }
+            sourceDt.isString -> throw AssemblyError("cannot typecast a string value")
+            sourceDt.isLong -> {
+                if(targetDt.isByteOrBool) {
+                    asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName")
+                } else if(targetDt.isWord || targetDt.isPointer) {
+                    asmgen.out("  lda  $sourceAsmVarName |  sta  $targetAsmVarName |  lda  $sourceAsmVarName+1 |  sta  $targetAsmVarName+1")
+                } else if(targetDt.isFloat) {
+                    asmgen.out("""
+                        lda  #<$sourceAsmVarName
+                        ldy  #>$sourceAsmVarName
+                        sta  cx16.r1L
+                        sty  cx16.r1H
+                        lda  #<$targetAsmVarName
+                        ldy  #>$targetAsmVarName
+                        jsr  floats.internal_long_R1_to_float_AY""")
+                } else
+                    throw AssemblyError("weird type")
+            }
+            else -> throw AssemblyError("weird type")
+        }
+    }
+
+
+
+    internal fun assignTypeCastedRegisters(targetAsmVarName: String, targetDt: BaseDataType,
+                                          regs: RegisterOrPair, sourceDt: BaseDataType, position: Position) {
+        if(sourceDt == targetDt && sourceDt != BaseDataType.POINTER)
+            throw AssemblyError("typecast to identical type")
+
+        // also see: PtExpressionAsmGen,   fun translateExpression(typecast: PtTypeCast)
+        when(sourceDt) {
+            BaseDataType.BOOL -> {
+                if (targetDt.isByteOrBool) asmgen.out("  st${regs.toString().lowercase()}  $targetAsmVarName")
+                else throw AssemblyError("assign bool to non-byte variable")
+            }
+            BaseDataType.UBYTE -> {
+                when(targetDt) {
+                    BaseDataType.BOOL -> {
+                        val compare = if(regs==RegisterOrPair.A) "cmp" else "cp${regs.toString().lowercase()}"
+                        asmgen.out("""
+                            $compare  #0
+                            beq  +
+                            ld${regs.toString().lowercase()}  #1
++                           st${regs.toString().lowercase()}  $targetAsmVarName""")
+                    }
+                    BaseDataType.BYTE -> {
+                        asmgen.out("  st${regs.toString().lowercase()}  $targetAsmVarName")
+                    }
+                    BaseDataType.UWORD, BaseDataType.WORD, BaseDataType.POINTER -> {
+                        if(asmgen.isTargetCpu(CpuType.CPU65C02))
+                            asmgen.out("  st${regs.toString().lowercase()}  $targetAsmVarName |  stz  $targetAsmVarName+1")
+                        else
+                            asmgen.out("  st${regs.toString().lowercase()}  $targetAsmVarName |  lda  #0  |  sta  $targetAsmVarName+1")
+                    }
+                    BaseDataType.LONG -> {
+                        asmgen.out("""
+                            st${regs.toString().lowercase()}  $targetAsmVarName
+                            lda  #0
+                            sta  $targetAsmVarName+1
+                            sta  $targetAsmVarName+2
+                            sta  $targetAsmVarName+3""")
+                    }
+                    BaseDataType.FLOAT -> {
+                        when(regs) {
+                            RegisterOrPair.A -> asmgen.out("  tay")
+                            RegisterOrPair.X -> asmgen.out("  txa |  tay")
+                            RegisterOrPair.Y -> {}
+                            else -> throw AssemblyError("non-byte regs")
+                        }
+                        asmgen.out("""
+                            lda  #<$targetAsmVarName
+                            sta  P8ZP_SCRATCH_W2
+                            lda  #>$targetAsmVarName
+                            sta  P8ZP_SCRATCH_W2+1
+                            jsr  floats.cast_from_ub""")
+                    }
+                    else -> throw AssemblyError("weird type")
+                }
+            }
+            BaseDataType.BYTE -> {
+                when(targetDt) {
+                    BaseDataType.BOOL -> TODO("assign byte to bool $position")
+                    BaseDataType.UBYTE -> {
+                        asmgen.out("  st${regs.toString().lowercase()}  $targetAsmVarName")
+                    }
+                    BaseDataType.UWORD -> {
+                        if(asmgen.isTargetCpu(CpuType.CPU65C02))
+                            asmgen.out("  st${regs.toString().lowercase()}  $targetAsmVarName |  stz  $targetAsmVarName+1")
+                        else
+                            asmgen.out("  st${regs.toString().lowercase()}  $targetAsmVarName |  lda  #0  |  sta  $targetAsmVarName+1")
+                    }
+                    BaseDataType.WORD, BaseDataType.POINTER -> {
+                        when(regs) {
+                            RegisterOrPair.A -> {}
+                            RegisterOrPair.X -> asmgen.out("  txa")
+                            RegisterOrPair.Y -> asmgen.out("  tya")
+                            else -> throw AssemblyError("non-byte regs")
+                        }
+                        asmgen.signExtendAYlsb(sourceDt)
+                        asmgen.out("  sta  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                    }
+                    BaseDataType.LONG -> {
+                        asmgen.out("  st${regs.toString().lowercase()}  $targetAsmVarName")
+                        asmgen.signExtendLongVariable(targetAsmVarName, sourceDt)
+                    }
+                    BaseDataType.FLOAT -> {
+                        when(regs) {
+                            RegisterOrPair.A -> {}
+                            RegisterOrPair.X -> asmgen.out("  txa")
+                            RegisterOrPair.Y -> asmgen.out("  tya")
+                            else -> throw AssemblyError("non-byte regs")
+                        }
+                        asmgen.out("""
+                            ldy  #<$targetAsmVarName
+                            sty  P8ZP_SCRATCH_W2
+                            ldy  #>$targetAsmVarName
+                            sty  P8ZP_SCRATCH_W2+1
+                            jsr  floats.cast_from_b""")
+                    }
+                    else -> throw AssemblyError("weird type")
+                }
+            }
+            BaseDataType.UWORD -> {
+                when(targetDt) {
+                    BaseDataType.BOOL -> TODO("assign uword to bool $position")
+                    BaseDataType.BYTE, BaseDataType.UBYTE -> {
+                        asmgen.out("  st${regs.toString().lowercase().first()}  $targetAsmVarName")
+                    }
+                    BaseDataType.WORD, BaseDataType.POINTER -> {
+                        when(regs) {
+                            RegisterOrPair.AX -> asmgen.out("  sta  $targetAsmVarName |  stx  $targetAsmVarName+1")
+                            RegisterOrPair.AY -> asmgen.out("  sta  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                            RegisterOrPair.XY -> asmgen.out("  stx  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                            else -> throw AssemblyError("non-word regs")
+                        }
+                    }
+                    BaseDataType.LONG -> {
+                        when(regs) {
+                            RegisterOrPair.AX -> asmgen.out("  sta  $targetAsmVarName |  stx  $targetAsmVarName+1")
+                            RegisterOrPair.AY -> asmgen.out("  sta  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                            RegisterOrPair.XY -> asmgen.out("  stx  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                            else -> throw AssemblyError("non-word regs")
+                        }
+                        asmgen.out("  lda  #0 |  sta  $targetAsmVarName+2 |  sta  $targetAsmVarName+3")
+                    }
+                    BaseDataType.FLOAT -> {
+                        if(regs!=RegisterOrPair.AY)
+                            throw AssemblyError("only supports AY here")
+                        asmgen.out("""
+                            tax
+                            lda  #<$targetAsmVarName
+                            sta  P8ZP_SCRATCH_W2
+                            lda  #>$targetAsmVarName
+                            sta  P8ZP_SCRATCH_W2+1
+                            txa
+                            jsr  floats.cast_from_uw""")
+                    }
+                    else -> throw AssemblyError("weird type")
+                }
+            }
+            BaseDataType.WORD -> {
+                when(targetDt) {
+                    BaseDataType.BOOL -> TODO("assign word to bool $position")
+                    BaseDataType.BYTE, BaseDataType.UBYTE -> {
+                        asmgen.out("  st${regs.toString().lowercase().first()}  $targetAsmVarName")
+                    }
+                    BaseDataType.UWORD, BaseDataType.POINTER -> {
+                        when(regs) {
+                            RegisterOrPair.AX -> asmgen.out("  sta  $targetAsmVarName |  stx  $targetAsmVarName+1")
+                            RegisterOrPair.AY -> asmgen.out("  sta  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                            RegisterOrPair.XY -> asmgen.out("  stx  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                            else -> throw AssemblyError("non-word regs")
+                        }
+                    }
+                    BaseDataType.LONG -> {
+                        when(regs) {
+                            RegisterOrPair.AX -> asmgen.out("  sta  $targetAsmVarName |  stx  $targetAsmVarName+1")
+                            RegisterOrPair.AY -> asmgen.out("  sta  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                            RegisterOrPair.XY -> asmgen.out("  stx  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                            else -> throw AssemblyError("non-word regs")
+                        }
+                        asmgen.signExtendLongVariable(targetAsmVarName, BaseDataType.WORD)
+                    }
+                    BaseDataType.FLOAT -> {
+                        if(regs!=RegisterOrPair.AY)
+                            throw AssemblyError("only supports AY here")
+                        asmgen.out("""
+                            tax
+                            lda  #<$targetAsmVarName
+                            sta  P8ZP_SCRATCH_W2
+                            lda  #>$targetAsmVarName
+                            sta  P8ZP_SCRATCH_W2+1
+                            txa
+                            jsr  floats.cast_from_w""")
+                    }
+                    else -> throw AssemblyError("weird type")
+                }
+            }
+            BaseDataType.STR -> throw AssemblyError("cannot typecast a string value")
+            BaseDataType.POINTER -> {
+                if(targetDt.isWord || targetDt.isPointer) {
+                    when(regs) {
+                        RegisterOrPair.AX -> asmgen.out("  sta  $targetAsmVarName |  stx  $targetAsmVarName+1")
+                        RegisterOrPair.AY -> asmgen.out("  sta  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                        RegisterOrPair.XY -> asmgen.out("  stx  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                        else -> throw AssemblyError("non-word regs")
+                    }
+                } else if(targetDt.isLong) {
+                    when(regs) {
+                        RegisterOrPair.AX -> asmgen.out("  sta  $targetAsmVarName |  stx  $targetAsmVarName+1")
+                        RegisterOrPair.AY -> asmgen.out("  sta  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                        RegisterOrPair.XY -> asmgen.out("  stx  $targetAsmVarName |  sty  $targetAsmVarName+1")
+                        else -> throw AssemblyError("non-word regs")
+                    }
+                    asmgen.out("  lda  #0 |  sta  $targetAsmVarName+2 |  sta  $targetAsmVarName+3")
+                } else {
+                    throw AssemblyError("cannot assign pointer to $targetDt")
+                }
+            }
+            BaseDataType.LONG -> {
+                val startreg = regs.startregname()
+                if(targetDt.isByteOrBool) {
+                    asmgen.out("  lda  cx16.$startreg |  sta  $targetAsmVarName")
+                } else if(targetDt.isWord || targetDt.isPointer) {
+                    asmgen.out("  lda  cx16.$startreg |  sta  $targetAsmVarName |  lda  cx16.$startreg+1 |  sta  $targetAsmVarName+1")
+                } else if(targetDt.isFloat) {
+                    if(regs==RegisterOrPair.R0R1)
+                        throw AssemblyError("cannot assign long in R0R1 to float because r1 is used as a parameter $position")
+                    asmgen.out("""
+                        lda  #<cx16.$startreg
+                        ldy  #>cx16.$startreg
+                        sta  cx16.r1L
+                        sty  cx16.r1H
+                        lda  #<$targetAsmVarName
+                        ldy  #>$targetAsmVarName
+                        jsr  floats.internal_long_R1_to_float_AY""")
+                } else
+                    throw AssemblyError("weird type $targetDt")
+            }
+            else -> throw AssemblyError("weird type $sourceDt")
+        }
+    }
+
+    /**
+     * Compute 8-bit offset Y = idx*structSize + fieldOffset for a struct-array element.
+     * Assumes totalBytes <=256 so offset fits in Y. Uses A/Y and multiply_bytes when needed.
+     */
+
+}

codeGenCpu6502/test/TestAsmOptimizer.kt

@@ -5,30 +5,41 @@ import io.kotest.matchers.shouldBe
 import prog8.code.StMemVar
 import prog8.code.SymbolTable
 import prog8.code.ast.PtProgram
+import prog8.code.core.BaseDataType
 import prog8.code.core.DataType
 import prog8.code.target.C64Target
-import prog8.codegen.cpu6502.*
-import prog8tests.helpers.DummyMemsizer
-import prog8tests.helpers.DummyStringEncoder
+import prog8.code.target.VMTarget
+import prog8.codegen.cpu6502.hasInstr
+import prog8.codegen.cpu6502.haslabel
+import prog8.codegen.cpu6502.isBranch
+import prog8.codegen.cpu6502.isLoadReg
+import prog8.codegen.cpu6502.isStoreReg
+import prog8.codegen.cpu6502.isStoreRegOrZero
+import prog8.codegen.cpu6502.keeplabel
+import prog8.codegen.cpu6502.modifiesYRegister
+import prog8.codegen.cpu6502.optimizeAssembly
 
 
 class TestAsmOptimizer: FunSpec({
 
     val machine = C64Target()
-    val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+    val program = PtProgram("test", VMTarget())
     val symbolTable = SymbolTable(program)
 
     // Add test variables with known non-IO addresses so the optimizer can resolve them
     // Non-IO addresses in typical C64 RAM range
     symbolTable.add(StMemVar("myvar", DataType.UBYTE, 0x0300u, null, null))
     symbolTable.add(StMemVar("var", DataType.UBYTE, 0x0301u, null, null))
-    symbolTable.add(StMemVar("var1", DataType.UWORD, 0x0302u, null, null))
-    symbolTable.add(StMemVar("var2", DataType.UWORD, 0x0304u, null, null))
-    symbolTable.add(StMemVar("ptr", DataType.UWORD, 0x0306u, null, null))
+    symbolTable.add(StMemVar("array", DataType.arrayFor(BaseDataType.UBYTE, machine), 0x0302u, 16u, null))
+    symbolTable.add(StMemVar("wordarray", DataType.arrayFor(BaseDataType.UWORD, machine), 0x0310u, 32u, null))
+    symbolTable.add(StMemVar("nswordarray", DataType.arrayFor(BaseDataType.UWORD, machine), 0x0320u, 32u, null))
+    symbolTable.add(StMemVar("var1", DataType.UWORD, 0x0400u, null, null))
+    symbolTable.add(StMemVar("var2", DataType.UWORD, 0x0402u, null, null))
+    symbolTable.add(StMemVar("ptr", DataType.UWORD, 0x0404u, null, null))
     symbolTable.add(StMemVar("P8ZP_SCRATCH_PTR", DataType.UWORD, 0x02u, null, null))
     symbolTable.add(StMemVar("P8ZP_SCRATCH_W1", DataType.UWORD, 0x04u, null, null))
-    symbolTable.add(StMemVar("A1", DataType.UWORD, 0x0310u, null, null))
-    symbolTable.add(StMemVar("A2", DataType.UWORD, 0x0312u, null, null))
+    symbolTable.add(StMemVar("A1", DataType.UWORD, 0x0410u, null, null))
+    symbolTable.add(StMemVar("A2", DataType.UWORD, 0x0412u, null, null))
 
     fun optimize(lines: MutableList<String>) {
         optimizeAssembly(lines, machine, symbolTable)
@@ -161,13 +172,13 @@ class TestAsmOptimizer: FunSpec({
     }
 
     test("optimizeJsrRts: three consecutive jsr only tail jsr+rts becomes jmp") {
-        val lines = mutableListOf("  jsr  \$4000", "  jsr  \$5000", "  jsr  \$6000", "  rts", "  nop", "  nop")
+        val lines = mutableListOf($$"  jsr  $4000", $$"  jsr  $5000", $$"  jsr  $6000", "  rts", "  nop", "  nop")
         optimize(lines)
-        lines shouldBe listOf("  jsr  \$4000", "  jsr  \$5000", "  jmp  \$6000", "  nop", "  nop")
+        lines shouldBe listOf($$"  jsr  $4000", $$"  jsr  $5000", $$"  jmp  $6000", "  nop", "  nop")
     }
 
     test("optimizeJsrRts: single line with embedded newlines only tail converted defensively") {
-        val lines = mutableListOf("  jsr  \$4000\n  jsr  \$5000\n  jsr  \$6000", "  rts", "  nop", "  nop")
+        val lines = mutableListOf($$"  jsr  $4000\n  jsr  $5000\n  jsr  $6000", "  rts", "  nop", "  nop")
         optimize(lines)
         // Before fix this became jmp/jmp/jmp via replace("jsr","jmp"); after fix only one jmp via replaceMnemonic
         val jmpCount = lines.joinToString("\n").split("jmp").size - 1
@@ -259,6 +270,106 @@ class TestAsmOptimizer: FunSpec({
             "  nop", "  nop", "  nop", "  nop", "  nop", "  nop", "  nop", "  nop")
     }
 
+    // --- optimizeSamePointerIndexing: guards on the index reload ---
+
+    fun padded(vararg body: String): MutableList<String> =
+        (body.toList() + List(14 - body.size) { "  nop" }).toMutableList()
+
+    /** Same, but one line shorter, for cases where the optimizer is expected to drop a line. */
+    fun paddedAfterDrop(vararg body: String): MutableList<String> =
+        (body.toList() + List(13 - body.size) { "  nop" }).toMutableList()
+
+    test("samePointerIndexing: KEEPS reload when the index variable is written in between") {
+        val lines = padded("  ldy  i", "  lda  arr,y", "  inc  i", "  ldy  i", "  sta  arr,y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  i", "  lda  arr,y", "  inc  i", "  ldy  i", "  sta  arr,y", "  rts")
+    }
+
+    test("samePointerIndexing: KEEPS reload when the index variable is decremented in between") {
+        val lines = padded("  ldy  i", "  lda  arr,y", "  dec  i", "  ldy  i", "  sta  arr,y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  i", "  lda  arr,y", "  dec  i", "  ldy  i", "  sta  arr,y", "  rts")
+    }
+
+    test("samePointerIndexing: KEEPS reload across a jsr") {
+        val lines = padded("  ldy  #0", "  lda  (ptr),y", "  jsr  someSub", "  ldy  #0", "  sta  (ptr),y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  #0", "  lda  (ptr),y", "  jsr  someSub", "  ldy  #0", "  sta  (ptr),y", "  rts")
+    }
+
+    test("samePointerIndexing: KEEPS reload across a label") {
+        val lines = padded("  ldy  #0", "  lda  (ptr),y", "someLabel:", "  ldy  #0", "  sta  (ptr),y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  #0", "  lda  (ptr),y", "someLabel:", "  ldy  #0", "  sta  (ptr),y", "  rts")
+    }
+
+    test("samePointerIndexing: KEEPS reload when an indirect store may alias the index") {
+        val lines = padded("  ldy  i", "  lda  arr,y", "  sta  (vptr),y", "  ldy  i", "  lda  arr2,y")
+        optimize(lines)
+        lines shouldBe padded("  ldy  i", "  lda  arr,y", "  sta  (vptr),y", "  ldy  i", "  lda  arr2,y")
+    }
+
+    test("samePointerIndexing: KEEPS reload when an intervening instruction modifies Y") {
+        val lines = padded("  ldy  #0", "  lda  (ptr),y", "  tay", "  ldy  #0", "  sta  (ptr),y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  #0", "  lda  (ptr),y", "  tay", "  ldy  #0", "  sta  (ptr),y", "  rts")
+    }
+
+    test("samePointerIndexing: KEEPS reload for an IO index operand") {
+        val lines = padded($$"  ldy  c64.$d020", "  lda  arr,y", "  lda  arr2,y", $$"  ldy  c64.$d020", "  lda  arr3,y")
+        optimize(lines)
+        lines shouldBe padded($$"  ldy  c64.$d020", "  lda  arr,y", "  lda  arr2,y", $$"  ldy  c64.$d020", "  lda  arr3,y")
+    }
+
+    test("samePointerIndexing: still drops reload when a store to an unrelated var intervenes") {
+        val lines = padded("  ldy  i", "  lda  arr,y", "  sta  other", "  ldy  i", "  sta  arr,y", "  rts")
+        optimize(lines)
+        lines shouldBe paddedAfterDrop("  ldy  i", "  lda  arr,y", "  sta  other", "  sta  arr,y", "  rts")
+    }
+
+    test("samePointerIndexing: KEEPS reload when the two ldy load DIFFERENT immediates") {
+        // regression: seed[1] then seed[0] - the reload selects a different element
+        val lines = padded("  ldy  #1", "  lda  seed_msb,y", "  sta  x", "  ldy  #0", "  lda  seed_msb,y", "  sta  y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  #1", "  lda  seed_msb,y", "  sta  x", "  ldy  #0", "  lda  seed_msb,y", "  sta  y", "  rts")
+    }
+
+    test("samePointerIndexing: KEEPS reload when the two ldy load different variables") {
+        val lines = padded("  ldy  i", "  lda  arr,y", "  sta  x", "  ldy  j", "  lda  arr,y", "  sta  y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  i", "  lda  arr,y", "  sta  x", "  ldy  j", "  lda  arr,y", "  sta  y", "  rts")
+    }
+
+    test("samePointerIndexing: KEEPS reload when the two ldy differ by an offset") {
+        val lines = padded("  ldy  w+0", "  lda  arr,y", "  sta  x", "  ldy  w+1", "  lda  arr,y", "  sta  y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  w+0", "  lda  arr,y", "  sta  x", "  ldy  w+1", "  lda  arr,y", "  sta  y", "  rts")
+    }
+
+    test("samePointerIndexing: matched load/store arm also requires the same index") {
+        val lines = padded("  ldy  i", "  lda  arr,y", "  clc", "  ldy  j", "  sta  arr,y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  i", "  lda  arr,y", "  clc", "  ldy  j", "  sta  arr,y", "  rts")
+    }
+
+    test("samePointerIndexing: relaxed arm drops reload for two indexed loads") {
+        val lines = padded("  ldy  sp", "  lda  cx,y", "  sta  cxv", "  ldy  sp", "  lda  cy,y", "  sta  cyv", "  rts")
+        optimize(lines)
+        lines shouldBe paddedAfterDrop("  ldy  sp", "  lda  cx,y", "  sta  cxv", "  lda  cy,y", "  sta  cyv", "  rts")
+    }
+
+    test("samePointerIndexing: relaxed arm does not fire when Y is modified") {
+        val lines = padded("  ldy  sp", "  lda  cx,y", "  dey", "  ldy  sp", "  lda  cy,y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  sp", "  lda  cx,y", "  dey", "  ldy  sp", "  lda  cy,y", "  rts")
+    }
+
+    test("samePointerIndexing: relaxed arm keeps reload when index is written in between") {
+        val lines = padded("  ldy  i", "  lda  arr,y", "  inc  i", "  ldy  i", "  lda  arr2,y", "  rts")
+        optimize(lines)
+        lines shouldBe padded("  ldy  i", "  lda  arr,y", "  inc  i", "  ldy  i", "  lda  arr2,y", "  rts")
+    }
+
     // --- optimizeAddWordToSameVariableOrExtraRegisterLoadInWordStore ---
 
     test("optimizeAddWordToSameVariable: optimizes P8ZP_SCRATCH_PTR += AY") {
@@ -284,6 +395,179 @@ class TestAsmOptimizer: FunSpec({
             "  nop", "  nop", "  nop", "  nop", "  nop", "  nop", "  nop", "  nop")
     }
 
+    // --- optimizeFoldIndexOffset (byte arrays) ---
+
+    test("optimizeFoldIndexOffset: folds byte array lda+clc+adc+tay+lda to ldy+lda") {
+        val lines = mutableListOf(
+            "  lda  myvar", "  clc", "  adc  #3", "  tay", "  lda  array,y",
+            "  rts", "  nop", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines[0] shouldBe "  ldy  myvar"
+        lines[1] shouldBe "  lda  array+3,y"
+        lines.size shouldBe 7
+    }
+
+    test("optimizeFoldIndexOffset: folds with x register") {
+        val lines = mutableListOf(
+            "  lda  var", "  clc", "  adc  #5", "  tax", "  lda  array,x",
+            "  rts", "  nop", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines[0] shouldBe "  ldx  var"
+        lines[1] shouldBe "  lda  array+5,x"
+        lines.size shouldBe 7
+    }
+
+    test("optimizeFoldIndexOffset: works with sta instead of lda") {
+        val lines = mutableListOf(
+            "  lda  myvar", "  clc", "  adc  #1", "  tay", "  sta  array,y",
+            "  rts", "  nop", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines[0] shouldBe "  ldy  myvar"
+        lines[1] shouldBe "  sta  array+1,y"
+        lines.size shouldBe 7
+    }
+
+    test("optimizeFoldIndexOffset: does not fold indirect indexed (ptr),y") {
+        val lines = mutableListOf(
+            "  lda  myvar", "  clc", "  adc  #3", "  tay", "  lda  (ptr),y",
+            "  rts", "  nop", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines.size shouldBe 10
+    }
+
+    test("optimizeFoldIndexOffset: preserves hex constant format") {
+        val lines = mutableListOf(
+            "  lda  myvar", "  clc", "  adc  #$0f", "  tay", "  lda  array,y",
+            "  rts", "  nop", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines[0] shouldBe "  ldy  myvar"
+        lines[1] shouldBe "  lda  array+$0f,y"
+        lines.size shouldBe 7
+    }
+
+    test("optimizeFoldIndexOffset: does not fold when 5th op is not absolute indexed") {
+        val lines = mutableListOf(
+            "  lda  myvar", "  clc", "  adc  #3", "  tay", "  nop",
+            "  rts", "  nop", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines.size shouldBe 10
+    }
+
+    test("optimizeFoldIndexOffset: works with cmp instruction") {
+        val lines = mutableListOf(
+            "  lda  myvar", "  clc", "  adc  #7", "  tay", "  cmp  array,y",
+            "  rts", "  nop", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines[0] shouldBe "  ldy  myvar"
+        lines[1] shouldBe "  cmp  array+7,y"
+        lines.size shouldBe 7
+    }
+
+    test("optimizeFoldIndexOffset: preserves label on first instruction") {
+        val lines = mutableListOf(
+            "mylabel: lda  myvar", "  clc", "  adc  #3", "  tay", "  lda  array,y",
+            "  rts", "  nop", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines[0] shouldBe "mylabel: ldy  myvar"
+        lines[1] shouldBe "  lda  array+3,y"
+        lines.size shouldBe 7
+    }
+
+    // --- optimizeFoldIndexOffsetSplitWord (split word arrays) ---
+
+    test("optimizeFoldIndexOffsetSplitWord: folds split word array access with y") {
+        val lines = mutableListOf(
+            "  lda  var", "  clc", "  adc  #3", "  tay",
+            "  lda  wordarray_lsb,y", "  ldx  wordarray_msb,y",
+            "  rts", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines[0] shouldBe "  ldy  var"
+        lines[1] shouldBe "  lda  wordarray_lsb+3,y"
+        lines[2] shouldBe "  ldx  wordarray_msb+3,y"
+        lines.size shouldBe 7
+    }
+
+    test("optimizeFoldIndexOffsetSplitWord: folds split word array access with x") {
+        val lines = mutableListOf(
+            "  lda  var", "  clc", "  adc  #3", "  tax",
+            "  lda  wordarray_lsb,x", "  ldx  wordarray_msb,x",
+            "  rts", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines[0] shouldBe "  ldx  var"
+        lines[1] shouldBe "  lda  wordarray_lsb+3,x"
+        lines[2] shouldBe "  ldx  wordarray_msb+3,x"
+        lines.size shouldBe 7
+    }
+
+    test("optimizeFoldIndexOffsetSplitWord: does not fold mismatched registers") {
+        val lines = mutableListOf(
+            "  lda  var", "  clc", "  adc  #3", "  tay",
+            "  lda  wordarray_lsb,y", "  ldx  wordarray_msb,x",
+            "  rts", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines.size shouldBe 10
+    }
+
+    test("optimizeFoldIndexOffsetSplitWord: does not fold non-lsb-msb pair") {
+        val lines = mutableListOf(
+            "  lda  var", "  clc", "  adc  #3", "  tay",
+            "  lda  wordarray_lsb,y", "  ldx  other,y",
+            "  rts", "  nop", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines.size shouldBe 10
+    }
+
+    // --- optimizeFoldIndexOffsetNonSplitWord (non-split word arrays) ---
+
+    test("optimizeFoldIndexOffsetNonSplitWord: folds non-split word array access") {
+        val lines = mutableListOf(
+            "  lda  var", "  clc", "  adc  #3", "  asl  a", "  tay",
+            "  lda  nswordarray,y", "  ldx  nswordarray+1,y",
+            "  rts", "  nop"
+        )
+        optimize(lines)
+        lines[0] shouldBe "  lda  var"
+        lines[1] shouldBe "  asl  a"
+        lines[2] shouldBe "  tay"
+        lines[3] shouldBe "  lda  nswordarray+6,y"
+        lines[4] shouldBe "  ldx  nswordarray+1+6,y"
+        lines.size shouldBe 7
+    }
+
+    test("optimizeFoldIndexOffsetNonSplitWord: scales hex constants") {
+        val lines = mutableListOf(
+            "  lda  var", "  clc", "  adc  #$02", "  asl  a", "  tay",
+            "  lda  nswordarray,y", "  ldx  nswordarray+1,y",
+            "  rts", "  nop"
+        )
+        optimize(lines)
+        lines[3] shouldBe "  lda  nswordarray+$04,y"
+        lines[4] shouldBe "  ldx  nswordarray+1+$04,y"
+        lines.size shouldBe 7
+    }
+
+    test("optimizeFoldIndexOffsetNonSplitWord: does not fold without asl a") {
+        val lines = mutableListOf(
+            "  lda  var", "  clc", "  adc  #3", "  tay",
+            "  lda  nswordarray,y", "  ldx  nswordarray+1,y",
+            "  rts", "  nop", "  nop"
+        )
+        optimize(lines)
+        lines.size shouldBe 9
+    }
+
     // --- Edge cases ---
 
     test("handles empty input") {

codeGenCpu6502/test/TestCodegen.kt

@@ -12,10 +12,9 @@ import prog8.code.ast.*
 import prog8.code.core.*
 import prog8.code.source.SourceCode
 import prog8.code.target.C64Target
+import prog8.code.target.VMTarget
 import prog8.codegen.cpu6502.AsmGen6502
 import prog8.codegen.cpu6502.VariableAllocator
-import prog8tests.helpers.DummyMemsizer
-import prog8tests.helpers.DummyStringEncoder
 import prog8tests.helpers.ErrorReporterForTests
 import java.nio.file.Files
 import kotlin.io.path.Path
@@ -46,13 +45,15 @@ class TestCodegen: FunSpec({
 //        xx += cx16.r0
 //    }
 //}
-        val codegen = AsmGen6502(prefixSymbols = false, 0)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val compTarget = C64Target()
+        val codegen = AsmGen6502(0)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main",false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         sub.add(PtVariable(
             "pi",
             DataType.UBYTE,
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -62,7 +63,8 @@ class TestCodegen: FunSpec({
         ))
         sub.add(PtVariable(
             "particleX",
-            DataType.arrayFor(BaseDataType.UBYTE),
+            DataType.arrayFor(BaseDataType.UBYTE, compTarget),
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -72,7 +74,8 @@ class TestCodegen: FunSpec({
         ))
         sub.add(PtVariable(
             "particleDX",
-            DataType.arrayFor(BaseDataType.UBYTE),
+            DataType.arrayFor(BaseDataType.UBYTE, compTarget),
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -83,6 +86,7 @@ class TestCodegen: FunSpec({
         sub.add(PtVariable(
             "xx",
             DataType.WORD,
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -93,14 +97,14 @@ class TestCodegen: FunSpec({
 
         val assign = PtAugmentedAssign("+=", Position.DUMMY)
         val target = PtAssignTarget(false, Position.DUMMY).also {
-            val targetIdx = PtArrayIndexer(DataType.UBYTE, Position.DUMMY).also { idx ->
-                idx.add(PtIdentifier("main.start.particleX", DataType.arrayFor(BaseDataType.UBYTE), Position.DUMMY))
+            val targetIdx = PtArrayIndexer(DataType.UBYTE, false, Position.DUMMY).also { idx ->
+                idx.add(PtIdentifier("main.start.particleX", DataType.arrayFor(BaseDataType.UBYTE, compTarget), Position.DUMMY))
                 idx.add(PtNumber(BaseDataType.UBYTE, 2.0, Position.DUMMY))
             }
             it.add(targetIdx)
         }
-        val value = PtArrayIndexer(DataType.UBYTE, Position.DUMMY)
-        value.add(PtIdentifier("main.start.particleDX", DataType.arrayFor(BaseDataType.UBYTE), Position.DUMMY))
+        val value = PtArrayIndexer(DataType.UBYTE, false, Position.DUMMY)
+        value.add(PtIdentifier("main.start.particleDX", DataType.arrayFor(BaseDataType.UBYTE, compTarget), Position.DUMMY))
         value.add(PtNumber(BaseDataType.UBYTE, 2.0, Position.DUMMY))
         assign.add(target)
         assign.add(value)
@@ -166,7 +170,7 @@ class TestCodegen: FunSpec({
     }
 
     test("memory mapped zp var is correctly considered to be zp var") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val st = SymbolTable(program)
         st.add(StMemVar("zpmemvar", DataType.WORD, 0x20u, null, null))
         st.add(StMemVar("normalmemvar", DataType.WORD, 0x9000u, null, null))

codeGenIntermediate/src/prog8/codegen/intermediate/BuiltinFuncGen.kt

@@ -81,7 +81,7 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val result = mutableListOf<IRCodeChunkBase>()
         val tr = exprGen.translateExpression(call.args[0])
         addToResult(result, tr, tr.resultReg, -1)
-        addInstr(result, IRInstruction(Opcode.PUSH, IRDataType.BYTE, reg1=tr.resultReg), null)
+        addInstr(result, IRInstructions.push(IRDataType.BYTE, tr.resultReg), null)
         return ExpressionCodeResult(result, IRDataType.BYTE, -1, -1)
     }
 
@@ -89,7 +89,7 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val result = mutableListOf<IRCodeChunkBase>()
         val tr = exprGen.translateExpression(call.args[0])
         addToResult(result, tr, tr.resultReg, -1)
-        addInstr(result, IRInstruction(Opcode.PUSH, IRDataType.WORD, reg1=tr.resultReg), null)
+        addInstr(result, IRInstructions.push(IRDataType.WORD, tr.resultReg), null)
         return ExpressionCodeResult(result, IRDataType.BYTE, -1, -1)
     }
 
@@ -97,7 +97,7 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val result = mutableListOf<IRCodeChunkBase>()
         val tr = exprGen.translateExpression(call.args[0])
         addToResult(result, tr, tr.resultReg, -1)
-        addInstr(result, IRInstruction(Opcode.PUSH, IRDataType.LONG, reg1=tr.resultReg), null)
+        addInstr(result, IRInstructions.push(IRDataType.LONG, tr.resultReg), null)
         return ExpressionCodeResult(result, IRDataType.BYTE, -1, -1)
     }
 
@@ -105,36 +105,36 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val result = mutableListOf<IRCodeChunkBase>()
         val tr = exprGen.translateExpression(call.args[0])
         addToResult(result, tr, -1, tr.resultFpReg)
-        addInstr(result, IRInstruction(Opcode.PUSH, IRDataType.FLOAT, fpReg1 = RegisterNum(tr.resultFpReg)), null)
+        addInstr(result, IRInstructions.push(IRDataType.FLOAT, tr.resultFpReg), null)
         return ExpressionCodeResult(result, IRDataType.BYTE, -1, -1)
     }
 
     private fun funcPop(): ExpressionCodeResult {
         val result = mutableListOf<IRCodeChunkBase>()
         val resultRegister = codeGen.registers.next(IRDataType.BYTE)
-        addInstr(result, IRInstruction(Opcode.POP, IRDataType.BYTE, reg1=resultRegister), null)
+        addInstr(result, IRInstructions.pop(IRDataType.BYTE, resultRegister), null)
         return ExpressionCodeResult(result, IRDataType.BYTE, resultRegister, -1)
     }
 
     private fun funcPopW(): ExpressionCodeResult {
         val result = mutableListOf<IRCodeChunkBase>()
         val resultRegister = codeGen.registers.next(IRDataType.WORD)
-        addInstr(result, IRInstruction(Opcode.POP, IRDataType.WORD, reg1=resultRegister), null)
-        return ExpressionCodeResult(result, IRDataType.BYTE, resultRegister, -1)
+        addInstr(result, IRInstructions.pop(IRDataType.WORD, resultRegister), null)
+        return ExpressionCodeResult(result, IRDataType.WORD, resultRegister, -1)
     }
 
     private fun funcPopL(): ExpressionCodeResult {
         val result = mutableListOf<IRCodeChunkBase>()
         val resultRegister = codeGen.registers.next(IRDataType.LONG)
-        addInstr(result, IRInstruction(Opcode.POP, IRDataType.LONG, reg1=resultRegister), null)
-        return ExpressionCodeResult(result, IRDataType.BYTE, resultRegister, -1)
+        addInstr(result, IRInstructions.pop(IRDataType.LONG, resultRegister), null)
+        return ExpressionCodeResult(result, IRDataType.LONG, resultRegister, -1)
     }
 
     private fun funcPopF(): ExpressionCodeResult {
         val result = mutableListOf<IRCodeChunkBase>()
         val resultRegister = codeGen.registers.next(IRDataType.FLOAT)
-        addInstr(result, IRInstruction(Opcode.POP, IRDataType.FLOAT, fpReg1 = RegisterNum(resultRegister)), null)
-        return ExpressionCodeResult(result, IRDataType.BYTE, -1, resultRegister)
+        addInstr(result, IRInstructions.pop(IRDataType.FLOAT, resultRegister), null)
+        return ExpressionCodeResult(result, IRDataType.FLOAT, -1, resultRegister)
     }
 
     private fun funcCopyFromPointer1ToPointer2(call: PtFunctionCall, type: IRDataType): ExpressionCodeResult {
@@ -147,15 +147,15 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
             IRDataType.LONG -> {
                 val intermediateReg = codeGen.registers.next(IRDataType.LONG)
                 result += IRCodeChunk(null, null).also {
-                    it += IRInstruction(Opcode.LOADI, type, reg1=intermediateReg, reg2=trSourceAddr.resultReg, immediate = 0)
-                    it += IRInstruction(Opcode.STOREI, type, reg1=intermediateReg, reg2=trTargetAddr.resultReg, immediate = 0)
+                    it += IRInstructions.loadMemory(Opcode.LOADI, type, intermediateReg, IRMemory.indirect(trSourceAddr.resultReg))
+                    it += IRInstructions.storeMemory(Opcode.STOREI, type, intermediateReg, IRMemory.indirect(trTargetAddr.resultReg))
                 }
             }
             IRDataType.FLOAT -> {
                 val intermediateReg = codeGen.registers.next(IRDataType.FLOAT)
                 result += IRCodeChunk(null, null).also {
-                    it += IRInstruction(Opcode.LOADI, type, fpReg1=RegisterNum(intermediateReg), reg1=trSourceAddr.resultReg, immediate = 0)
-                    it += IRInstruction(Opcode.STOREI, type, fpReg1=RegisterNum(intermediateReg), reg1=trTargetAddr.resultReg, immediate = 0)
+                    it += IRInstructions.loadMemory(Opcode.LOADI, type, intermediateReg, IRMemory.indirect(trSourceAddr.resultReg))
+                    it += IRInstructions.storeMemory(Opcode.STOREI, type, intermediateReg, IRMemory.indirect(trTargetAddr.resultReg))
                 }
             }
             else -> throw AssemblyError("invalid type $type")
@@ -169,12 +169,12 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         addToResult(result, valueTr, valueTr.resultReg, valueTr.resultFpReg)
         return if(resultType==IRDataType.FLOAT) {
             val resultFpReg = codeGen.registers.next(IRDataType.FLOAT)
-            addInstr(result, IRInstruction(Opcode.SQUARE, resultType, fpReg1 = RegisterNum(resultFpReg), fpReg2 = RegisterNum(valueTr.resultFpReg)), null)
+            addInstr(result, IRInstructions.binary(Opcode.SQUARE, resultType, resultFpReg, valueTr.resultFpReg), null)
             ExpressionCodeResult(result, resultType, -1, resultFpReg)
         }
         else {
             val resultReg = codeGen.registers.next(resultType)
-            addInstr(result, IRInstruction(Opcode.SQUARE, resultType, reg1 = resultReg, reg2 = valueTr.resultReg), null)
+            addInstr(result, IRInstructions.binary(Opcode.SQUARE, resultType, resultReg, valueTr.resultReg), null)
             ExpressionCodeResult(result, resultType, resultReg, -1)
         }
     }
@@ -183,12 +183,13 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val result = mutableListOf<IRCodeChunkBase>()
         val addressTr = exprGen.translateExpression(call.args[0])
         addToResult(result, addressTr, addressTr.resultReg, -1)
-        addInstr(result, IRInstruction(Opcode.CALLI, reg1 = addressTr.resultReg), null)
+        addInstr(result, IRInstructions.call(Opcode.CALLI, CallSite(CallTarget.Direct(codeIndirect(addressTr.resultReg)))), null)
         return if(call.void)
             ExpressionCodeResult(result, IRDataType.BYTE, -1, -1)
         else {
             val resultvalueReg = codeGen.registers.next(IRDataType.WORD)
-            addInstr(result, IRInstruction(Opcode.LOADHR, IRDataType.WORD, reg1=resultvalueReg, immediate=4), null)
+            val slot = if(codeGen.options.compTarget.cpu.usesM68kConvention) 10 else 4
+            addInstr(result, IRInstructions.hardwareLoad(IRDataType.WORD, resultvalueReg, CallingConventionSlot(slot)), null)
             ExpressionCodeResult(result, IRDataType.WORD, resultvalueReg, -1)
         }
     }
@@ -201,8 +202,9 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         addToResult(result, bankTr, bankTr.resultReg, -1)
         addToResult(result, addressTr, addressTr.resultReg, -1)
         addToResult(result, argumentwordTr, argumentwordTr.resultReg, -1)
-        result += codeGen.makeSyscall(IMSyscall.CALLFAR, listOf(IRDataType.BYTE to bankTr.resultReg, IRDataType.WORD to addressTr.resultReg, IRDataType.WORD to argumentwordTr.resultReg), IRDataType.WORD to argumentwordTr.resultReg)
-        return ExpressionCodeResult(result, IRDataType.WORD, argumentwordTr.resultReg, -1)
+        val resultReg = codeGen.registers.next(IRDataType.WORD)
+        result += codeGen.makeSyscall(IMSyscall.CALLFAR, listOf(IRDataType.BYTE to bankTr.resultReg, codeGen.addressDt to addressTr.resultReg, IRDataType.WORD to argumentwordTr.resultReg), IRDataType.WORD to resultReg)
+        return ExpressionCodeResult(result, IRDataType.WORD, resultReg, -1)
     }
 
     private fun funcCallfar2(call: PtFunctionCall): ExpressionCodeResult {
@@ -219,15 +221,20 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         addToResult(result, argumentX, argumentX.resultReg, -1)
         addToResult(result, argumentY, argumentY.resultReg, -1)
         addToResult(result, argumentCarry, argumentCarry.resultReg, -1)
-        result += codeGen.makeSyscall(IMSyscall.CALLFAR2, listOf(IRDataType.BYTE to bankTr.resultReg, IRDataType.WORD to addressTr.resultReg,
+        val resultReg = codeGen.registers.next(IRDataType.WORD)
+        result += codeGen.makeSyscall(IMSyscall.CALLFAR2, listOf(IRDataType.BYTE to bankTr.resultReg, codeGen.addressDt to addressTr.resultReg,
             IRDataType.BYTE to argumentA.resultReg,
             IRDataType.BYTE to argumentX.resultReg,
             IRDataType.BYTE to argumentY.resultReg,
-            IRDataType.BYTE to argumentCarry.resultReg), IRDataType.WORD to addressTr.resultReg)
-        return ExpressionCodeResult(result, IRDataType.WORD, addressTr.resultReg, -1)
+            IRDataType.BYTE to argumentCarry.resultReg), IRDataType.WORD to resultReg)
+        return ExpressionCodeResult(result, IRDataType.WORD, resultReg, -1)
     }
 
     private fun funcDivmod(call: PtFunctionCall, type: IRDataType, signed: Boolean): ExpressionCodeResult {
+        if(signed && codeGen.options.compTarget.cpu.is6502) {
+            codeGen.errors.err("expected all ubyte or all uword arguments (no signed divmod support on 6502 yet)", call.position)
+            return ExpressionCodeResult(mutableListOf(), type, -1, -1)
+        }
         val result = mutableListOf<IRCodeChunkBase>()
         val number = call.args[0]
         val divident = call.args[1]
@@ -238,33 +245,56 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         if(divident is PtNumber) {
             val tr = exprGen.translateExpression(number)
             addToResult(result, tr, tr.resultReg, -1)
-            addInstr(result, IRInstruction(divmodOpcode, type, reg1 = tr.resultReg, immediate = divident.number.toInt()), null)
-            divisionReg = tr.resultReg
             remainderReg = codeGen.registers.next(type)
+            addInstr(result, IRInstructions.divmodImmediate(divmodOpcode, type, tr.resultReg, remainderReg, divident.number.toInt()), null)
+            divisionReg = tr.resultReg
         } else {
             val numTr = exprGen.translateExpression(number)
             addToResult(result, numTr, numTr.resultReg, -1)
             val dividentTr = exprGen.translateExpression(divident)
             addToResult(result, dividentTr, dividentTr.resultReg, -1)
-            addInstr(result, IRInstruction(divmodrOpcode, type, reg1 = numTr.resultReg, reg2=dividentTr.resultReg), null)
-            divisionReg = numTr.resultReg
             remainderReg = dividentTr.resultReg
+            addInstr(result, IRInstructions.divmodRegister(divmodrOpcode, type, numTr.resultReg, dividentTr.resultReg), null)
+            divisionReg = numTr.resultReg
         }
-        // DIVMOD instruction result convention: on value stack, division and remainder on top.
-        addInstr(result, IRInstruction(Opcode.POP, type, reg1=remainderReg), null)
-        addInstr(result, IRInstruction(Opcode.POP, type, reg1=divisionReg), null)
-        // return the 2 result values: division and remainder  division in AY, remainder in R15
-        if(type==IRDataType.BYTE) {
-            result += IRCodeChunk(null, null).also {
-                it += IRInstruction(Opcode.STOREHR, IRDataType.BYTE, reg1=divisionReg, immediate=0)
-                it += IRInstruction(Opcode.STOREM, IRDataType.BYTE, reg1=remainderReg, labelSymbol = "cx16.r15")
+        // DIVMOD result convention: quotient in reg1, remainder in reg2.
+        // Route to the return value locations expected by the caller (via returnsWhatWhere).
+        if(codeGen.options.compTarget.cpu.usesM68kConvention) {
+            // m68k: first return value in D0 (slot 10), second in D1 (slot 11)
+            when (type) {
+                IRDataType.BYTE -> {
+                    result += IRCodeChunk(null, null).also {
+                        it += IRInstructions.hardwareStore(IRDataType.BYTE, remainderReg, CallingConventionSlot(11))
+                        it += IRInstructions.hardwareStore(IRDataType.BYTE, divisionReg, CallingConventionSlot(10))
+                    }
+                }
+                IRDataType.WORD -> {
+                    result += IRCodeChunk(null, null).also {
+                        it += IRInstructions.hardwareStore(IRDataType.WORD, remainderReg, CallingConventionSlot(11))
+                        it += IRInstructions.hardwareStore(IRDataType.WORD, divisionReg, CallingConventionSlot(10))
+                    }
+                }
+                else -> throw AssemblyError("invalid type for DIVMOD")
             }
-        } else if(type==IRDataType.WORD) {
-            result += IRCodeChunk(null, null).also {
-                it += IRInstruction(Opcode.STOREHR, IRDataType.WORD, reg1=divisionReg, immediate=4)
-                it += IRInstruction(Opcode.STOREM, IRDataType.WORD, reg1=remainderReg, labelSymbol = "cx16.r15")
+        } else {
+            // 6502/cx16: quotient to A (slot 0) or AY (slot 4), remainder to cx16.r15
+            // Note: STOREM must come before STOREHR, otherwise STOREM clobbers A before LOADHR reads it.
+            when (type) {
+                IRDataType.BYTE -> {
+                    result += IRCodeChunk(null, null).also {
+                        it += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, remainderReg, IRMemory.direct("cx16.r15"))
+                        it += IRInstructions.hardwareStore(IRDataType.BYTE, divisionReg, CallingConventionSlot(0))
+                    }
+                }
+                IRDataType.WORD -> {
+                    result += IRCodeChunk(null, null).also {
+                        it += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.WORD, remainderReg, IRMemory.direct("cx16.r15"))
+                        it += IRInstructions.hardwareStore(IRDataType.WORD, divisionReg, CallingConventionSlot(4))
+                    }
+                }
+                else -> throw AssemblyError("invalid type for DIVMOD")
             }
-        } else throw AssemblyError("invalid type for DIVMOD")
+        }
         return ExpressionCodeResult(result, type, -1, -1)
     }
 
@@ -275,7 +305,8 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         addToResult(result, left, left.resultReg, -1)
         addToResult(result, right, right.resultReg, -1)
         val resultReg = codeGen.registers.next(IRDataType.BYTE)
-        result += codeGen.makeSyscall(IMSyscall.COMPARE_STRINGS, listOf(IRDataType.WORD to left.resultReg, IRDataType.WORD to right.resultReg), IRDataType.BYTE to resultReg)
+        val addressDt = codeGen.addressDt
+        result += codeGen.makeSyscall(IMSyscall.COMPARE_STRINGS, listOf(addressDt to left.resultReg, addressDt to right.resultReg), IRDataType.BYTE to resultReg)
         return ExpressionCodeResult(result, IRDataType.BYTE, resultReg, -1)
     }
 
@@ -285,9 +316,9 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         addToResult(result, leftTr, leftTr.resultReg, -1)
         val rightTr = exprGen.translateExpression(call.args[1])
         addToResult(result, rightTr, rightTr.resultReg, -1)
-        val dt = irType(call.args[0].type)
+        val dt = codeGen.irType(call.args[0].type)
         result += IRCodeChunk(null, null).also {
-            it += IRInstruction(Opcode.CMP, dt, reg1=leftTr.resultReg, reg2=rightTr.resultReg)
+            it += IRInstructions.compare(dt, leftTr.resultReg, rightTr.resultReg)
         }
         return ExpressionCodeResult(result, dt, leftTr.resultReg, -1)
     }
@@ -306,11 +337,11 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 val notNegativeLabel = codeGen.createLabelName()
                 val compareReg = codeGen.registers.next(IRDataType.BYTE)
                 result += IRCodeChunk(null, null).also {
-                    it += IRInstruction(Opcode.LOADR, IRDataType.BYTE, reg1=compareReg, reg2=tr.resultReg)
+                    it += IRInstructions.move(IRDataType.BYTE, compareReg, tr.resultReg)
                     if(needsExplicitCmpi)
-                        it += IRInstruction(Opcode.CMPI, IRDataType.BYTE, reg1=compareReg, immediate = 0)
-                    it += IRInstruction(Opcode.BSTPOS, labelSymbol = notNegativeLabel)
-                    it += IRInstruction(Opcode.NEG, IRDataType.BYTE, reg1=tr.resultReg)
+                        it += IRInstructions.compareImmediate(IRDataType.BYTE, compareReg, 0)
+                    it += IRInstructions.branch(Opcode.BSTPOS, codeLabel(notNegativeLabel))
+                    it += IRInstructions.unary(Opcode.NEG, IRDataType.BYTE, tr.resultReg)
                 }
                 result += IRCodeChunk(notNegativeLabel, null)
                 return ExpressionCodeResult(result, IRDataType.BYTE, tr.resultReg, -1)
@@ -319,11 +350,11 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 val notNegativeLabel = codeGen.createLabelName()
                 val compareReg = codeGen.registers.next(IRDataType.WORD)
                 result += IRCodeChunk(null, null).also {
-                    it += IRInstruction(Opcode.LOADR, IRDataType.WORD, reg1=compareReg, reg2=tr.resultReg)
+                    it += IRInstructions.move(IRDataType.WORD, compareReg, tr.resultReg)
                     if(needsExplicitCmpi)
-                        it += IRInstruction(Opcode.CMPI, IRDataType.WORD, reg1=compareReg, immediate = 0)
-                    it += IRInstruction(Opcode.BSTPOS, labelSymbol = notNegativeLabel)
-                    it += IRInstruction(Opcode.NEG, IRDataType.WORD, reg1=tr.resultReg)
+                        it += IRInstructions.compareImmediate(IRDataType.WORD, compareReg, 0)
+                    it += IRInstructions.branch(Opcode.BSTPOS, codeLabel(notNegativeLabel))
+                    it += IRInstructions.unary(Opcode.NEG, IRDataType.WORD, tr.resultReg)
                 }
                 result += IRCodeChunk(notNegativeLabel, null)
                 return ExpressionCodeResult(result, IRDataType.WORD, tr.resultReg, -1)
@@ -332,20 +363,24 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 val notNegativeLabel = codeGen.createLabelName()
                 val compareReg = codeGen.registers.next(IRDataType.LONG)
                 result += IRCodeChunk(null, null).also {
-                    it += IRInstruction(Opcode.LOADR, IRDataType.LONG, reg1=compareReg, reg2=tr.resultReg)
+                    it += IRInstructions.move(IRDataType.LONG, compareReg, tr.resultReg)
                     if(needsExplicitCmpi)
-                        it += IRInstruction(Opcode.CMPI, IRDataType.LONG, reg1=compareReg, immediate = 0)
-                    it += IRInstruction(Opcode.BSTPOS, labelSymbol = notNegativeLabel)
-                    it += IRInstruction(Opcode.NEG, IRDataType.LONG, reg1=tr.resultReg)
+                        it += IRInstructions.compareImmediate(IRDataType.LONG, compareReg, 0)
+                    it += IRInstructions.branch(Opcode.BSTPOS, codeLabel(notNegativeLabel))
+                    it += IRInstructions.unary(Opcode.NEG, IRDataType.LONG, tr.resultReg)
                 }
                 result += IRCodeChunk(notNegativeLabel, null)
                 return ExpressionCodeResult(result, IRDataType.LONG, tr.resultReg, -1)
             }
             BaseDataType.FLOAT -> {
                 val resultFpReg = codeGen.registers.next(IRDataType.FLOAT)
-                addInstr(result, IRInstruction(Opcode.FABS, IRDataType.FLOAT, fpReg1 = RegisterNum(resultFpReg), fpReg2 = RegisterNum(tr.resultFpReg)), null)
+                addInstr(result, IRInstructions.binary(Opcode.FABS, IRDataType.FLOAT, resultFpReg, tr.resultFpReg), null)
                 return ExpressionCodeResult(result, IRDataType.FLOAT, -1, resultFpReg)
             }
+            BaseDataType.POINTER -> {
+                // pointer is unsigned, abs has no effect
+                return ExpressionCodeResult(result, IRDataType.POINTER, tr.resultReg, -1)
+            }
             else -> throw AssemblyError("weird dt")
         }
     }
@@ -358,12 +393,12 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         if(tr.dt==IRDataType.FLOAT) {
             addToResult(result, tr, -1, tr.resultFpReg)
             result += IRCodeChunk(null, null).also {
-                it += IRInstruction(Opcode.SGN, tr.dt, reg1 = resultReg, fpReg1 = RegisterNum(tr.resultFpReg))
+                it += IRInstructions.binary(Opcode.SGN, tr.dt, resultReg, tr.resultFpReg)
             }
         } else {
             addToResult(result, tr, tr.resultReg, -1)
             result += IRCodeChunk(null, null).also {
-                it += IRInstruction(Opcode.SGN, tr.dt, reg1 = resultReg, reg2 = tr.resultReg)
+                it += IRInstructions.binary(Opcode.SGN, tr.dt, resultReg, tr.resultReg)
             }
         }
         return ExpressionCodeResult(result, IRDataType.BYTE, resultReg, -1)
@@ -378,7 +413,7 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 addToResult(result, tr, tr.resultReg, -1)
                 val resultReg = codeGen.registers.next(IRDataType.BYTE)
                 result += IRCodeChunk(null, null).also {
-                    it += IRInstruction(Opcode.SQRT, IRDataType.BYTE, reg1=resultReg, reg2=tr.resultReg)
+                    it += IRInstructions.binary(Opcode.SQRT, IRDataType.BYTE, resultReg, tr.resultReg)
                 }
                 return ExpressionCodeResult(result, IRDataType.BYTE, resultReg, -1)
             }
@@ -386,7 +421,7 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 addToResult(result, tr, tr.resultReg, -1)
                 val resultReg = codeGen.registers.next(IRDataType.BYTE)     // sqrt of a word still produces just a byte result
                 result += IRCodeChunk(null, null).also {
-                    it += IRInstruction(Opcode.SQRT, IRDataType.WORD, reg1=resultReg, reg2=tr.resultReg)
+                    it += IRInstructions.binary(Opcode.SQRT, IRDataType.WORD, resultReg, tr.resultReg)
                 }
                 return ExpressionCodeResult(result, IRDataType.BYTE, resultReg, -1)
             }
@@ -394,7 +429,7 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 addToResult(result, tr, tr.resultReg, -1)
                 val resultReg = codeGen.registers.next(IRDataType.WORD)     // sqrt of a long still produces just a word result
                 result += IRCodeChunk(null, null).also {
-                    it += IRInstruction(Opcode.SQRT, IRDataType.LONG, reg1=resultReg, reg2=tr.resultReg)
+                    it += IRInstructions.binary(Opcode.SQRT, IRDataType.LONG, resultReg, tr.resultReg)
                 }
                 return ExpressionCodeResult(result, IRDataType.WORD, resultReg, -1)
             }
@@ -402,7 +437,7 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 addToResult(result, tr, -1, tr.resultFpReg)
                 val resultFpReg = codeGen.registers.next(IRDataType.FLOAT)
                 result += IRCodeChunk(null, null).also {
-                    it += IRInstruction(Opcode.SQRT, IRDataType.FLOAT, fpReg1 = RegisterNum(resultFpReg), fpReg2 = RegisterNum(tr.resultFpReg))
+                    it += IRInstructions.binary(Opcode.SQRT, IRDataType.FLOAT, resultFpReg, tr.resultFpReg)
                 }
                 return ExpressionCodeResult(result, IRDataType.FLOAT, -1, resultFpReg)
             }
@@ -417,13 +452,13 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
             // msb is 0, use EXT
             val lsbTr = exprGen.translateExpression(call.args[1])
             addToResult(result, lsbTr, lsbTr.resultReg, -1)
-            addInstr(result, IRInstruction(Opcode.EXT, IRDataType.BYTE, reg1=resultReg, reg2 = lsbTr.resultReg), null)
+            addInstr(result, IRInstructions.binary(Opcode.EXT, IRDataType.BYTE, resultReg, lsbTr.resultReg), null)
         } else {
             val msbTr = exprGen.translateExpression(call.args[0])
             addToResult(result, msbTr, msbTr.resultReg, -1)
             val lsbTr = exprGen.translateExpression(call.args[1])
             addToResult(result, lsbTr, lsbTr.resultReg, -1)
-            addInstr(result, IRInstruction(Opcode.CONCAT, IRDataType.BYTE, reg1=resultReg, reg2 = msbTr.resultReg, reg3 = lsbTr.resultReg), null)
+            addInstr(result, IRInstructions.concat(IRDataType.BYTE, resultReg, msbTr.resultReg, lsbTr.resultReg), null)
         }
         return ExpressionCodeResult(result, IRDataType.WORD, resultReg, -1)
     }
@@ -437,23 +472,21 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 // msw is 0, use EXT
                 val lswTr = exprGen.translateExpression(call.args[1])
                 addToResult(result, lswTr, lswTr.resultReg, -1)
-                addInstr(result, IRInstruction(Opcode.EXT, IRDataType.WORD, reg1=resultReg, reg2 = lswTr.resultReg), null)
+                addInstr(result, IRInstructions.binary(Opcode.EXT, IRDataType.WORD, resultReg, lswTr.resultReg), null)
             } else {
                 val mswTr = exprGen.translateExpression(call.args[0])
                 addToResult(result, mswTr, mswTr.resultReg, -1)
                 val lswTr = exprGen.translateExpression(call.args[1])
                 addToResult(result, lswTr, lswTr.resultReg, -1)
-                addInstr(result, IRInstruction(Opcode.CONCAT, IRDataType.WORD, reg1=resultReg, reg2 = mswTr.resultReg, reg3 = lswTr.resultReg), null)
+                addInstr(result, IRInstructions.concat(IRDataType.WORD, resultReg, mswTr.resultReg, lswTr.resultReg), null)
             }
         } else {
             // mklong(msb, b3, b2, lsb)
             if((call.args[0] as? PtNumber)?.number == 0.0 && (call.args[1] as? PtNumber)?.number == 0.0 && (call.args[2] as? PtNumber)?.number == 0.0) {
-                // use EXT.b + EXT.w
+                // use EXTL.b: zero-extend byte directly to long
                 val lsbTr = exprGen.translateExpression(call.args[3])
                 addToResult(result, lsbTr, lsbTr.resultReg, -1)
-                val wordReg = codeGen.registers.next(IRDataType.WORD)
-                addInstr(result, IRInstruction(Opcode.EXT, IRDataType.BYTE, reg1=wordReg, reg2 = lsbTr.resultReg), null)
-                addInstr(result, IRInstruction(Opcode.EXT, IRDataType.WORD, reg1=resultReg, reg2 = wordReg), null)
+                addInstr(result, IRInstructions.binary(Opcode.EXTL, IRDataType.BYTE, resultReg, lsbTr.resultReg), null)
             } else {
                 val msbTr = exprGen.translateExpression(call.args[0])
                 val b2Tr = exprGen.translateExpression(call.args[1])
@@ -465,9 +498,9 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 addToResult(result, lsbTr, lsbTr.resultReg, -1)
                 val lswReg = codeGen.registers.next(IRDataType.WORD)
                 val mswReg = codeGen.registers.next(IRDataType.WORD)
-                addInstr(result, IRInstruction(Opcode.CONCAT, IRDataType.BYTE, reg1=mswReg, reg2 = msbTr.resultReg, reg3 = b2Tr.resultReg), null)
-                addInstr(result, IRInstruction(Opcode.CONCAT, IRDataType.BYTE, reg1=lswReg, reg2 = b1Tr.resultReg, reg3 = lsbTr.resultReg), null)
-                addInstr(result, IRInstruction(Opcode.CONCAT, IRDataType.WORD, reg1=resultReg, reg2 = mswReg, reg3 = lswReg), null)
+                addInstr(result, IRInstructions.concat(IRDataType.BYTE, mswReg, msbTr.resultReg, b2Tr.resultReg), null)
+                addInstr(result, IRInstructions.concat(IRDataType.BYTE, lswReg, b1Tr.resultReg, lsbTr.resultReg), null)
+                addInstr(result, IRInstructions.concat(IRDataType.WORD, resultReg, mswReg, lswReg), null)
             }
 
         }
@@ -475,7 +508,7 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
     }
 
     private fun funcClamp(call: PtFunctionCall): ExpressionCodeResult {
-        val type = irType(call.type)
+        val type = codeGen.irType(call.type)
         require(type != IRDataType.FLOAT)
         val result = mutableListOf<IRCodeChunkBase>()
         val valueTr = exprGen.translateExpression(call.args[0])
@@ -492,17 +525,18 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
             BaseDataType.LONG -> IMSyscall.CLAMP_LONG
             else -> throw AssemblyError("invalid dt")
         }
+        val resultReg = codeGen.registers.next(type)
         result += codeGen.makeSyscall(syscall, listOf(
                 valueTr.dt to valueTr.resultReg,
                 minimumTr.dt to minimumTr.resultReg,
                 maximumTr.dt to maximumTr.resultReg,
-            ), type to valueTr.resultReg
+            ), type to resultReg
         )
-        return ExpressionCodeResult(result, type, valueTr.resultReg, -1)
+        return ExpressionCodeResult(result, type, resultReg, -1)
     }
 
     private fun funcMin(call: PtFunctionCall): ExpressionCodeResult {
-        val type = irType(call.type)
+        val type = codeGen.irType(call.type)
         val result = mutableListOf<IRCodeChunkBase>()
         val leftTr = exprGen.translateExpression(call.args[0])
         addToResult(result, leftTr, leftTr.resultReg, -1)
@@ -511,17 +545,17 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val comparisonOpcode = if(call.type.isSigned) Opcode.BGTSR else Opcode.BGTR
         val after = codeGen.createLabelName()
         result += IRCodeChunk(null, null).also {
-            it += IRInstruction(comparisonOpcode, type, reg1 = rightTr.resultReg, reg2 = leftTr.resultReg, labelSymbol = after)
+            it += IRInstructions.branchRegister(comparisonOpcode, type, rightTr.resultReg, leftTr.resultReg, codeLabel(after))
             // right <= left, take right
-            it += IRInstruction(Opcode.LOADR, type, reg1=leftTr.resultReg, reg2=rightTr.resultReg)
-            it += IRInstruction(Opcode.JUMP, labelSymbol = after)
+            it += IRInstructions.move(type, leftTr.resultReg, rightTr.resultReg)
+            it += IRInstructions.jump(codeLabel(after))
         }
         result += IRCodeChunk(after, null)
         return ExpressionCodeResult(result, type, leftTr.resultReg, -1)
     }
 
     private fun funcMax(call: PtFunctionCall): ExpressionCodeResult {
-        val type = irType(call.type)
+        val type = codeGen.irType(call.type)
         val result = mutableListOf<IRCodeChunkBase>()
         val leftTr = exprGen.translateExpression(call.args[0])
         addToResult(result, leftTr, leftTr.resultReg, -1)
@@ -530,33 +564,35 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val comparisonOpcode = if(call.type.isSigned) Opcode.BGTSR else Opcode.BGTR
         val after = codeGen.createLabelName()
         result += IRCodeChunk(null, null).also {
-            it += IRInstruction(comparisonOpcode, type, reg1 = leftTr.resultReg, reg2 = rightTr.resultReg, labelSymbol = after)
+            it += IRInstructions.branchRegister(comparisonOpcode, type, leftTr.resultReg, rightTr.resultReg, codeLabel(after))
             // right >= left, take right
-            it += IRInstruction(Opcode.LOADR, type, reg1=leftTr.resultReg, reg2=rightTr.resultReg)
-            it += IRInstruction(Opcode.JUMP, labelSymbol = after)
+            it += IRInstructions.move(type, leftTr.resultReg, rightTr.resultReg)
+            it += IRInstructions.jump(codeLabel(after))
         }
         result += IRCodeChunk(after, null)
         return ExpressionCodeResult(result, type, leftTr.resultReg, -1)
     }
 
     private fun funcPoke(call: PtFunctionCall, dt: IRDataType): ExpressionCodeResult {
+        // Try struct-array field folding: pokew(&arr[idx]+fieldOff, value) -> STOREX arr+fieldOff,S=structSize
+        tryFoldStructArrayPoke(call, dt)?.let { return it }
         val result = mutableListOf<IRCodeChunkBase>()
         if(codeGen.isZero(call.args[1])) {
             if (call.args[0] is PtNumber) {
                 val address = (call.args[0] as PtNumber).number.toUInt()
                 result += IRCodeChunk(null, null).also {
-                    it += IRInstruction(Opcode.STOREZM, dt, address = address.toAddress())
+                    it += IRInstructions.storeZero(Opcode.STOREZM, dt, IRMemory.direct(address.toAddress()))
                 }
             } else {
                 val (address, offset) = exprGen.getAddressAndOffset(call.args[0])
                 if(address!=null) {
                     val tr = exprGen.translateExpression(address)
                     addToResult(result, tr, tr.resultReg, -1)
-                    addInstr(result, IRInstruction(Opcode.STOREZI, dt, reg1 = tr.resultReg, immediate = offset!!), null)
+                    addInstr(result, IRInstructions.storeZero(Opcode.STOREZI, dt, IRMemory.indirect(tr.resultReg, offset!!)), null)
                 } else {
                     val tr = exprGen.translateExpression(call.args[0])
                     addToResult(result, tr, tr.resultReg, -1)
-                    addInstr(result, IRInstruction(Opcode.STOREZI, dt, reg1 = tr.resultReg, immediate = 0), null)
+                    addInstr(result, IRInstructions.storeZero(Opcode.STOREZI, dt, IRMemory.indirect(tr.resultReg)), null)
                 }
             }
         } else {
@@ -565,9 +601,9 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 val address = (call.args[0] as PtNumber).number.toUInt()
                 addToResult(result, valueTr, valueTr.resultReg, valueTr.resultFpReg)
                 if(dt==IRDataType.FLOAT) {
-                    addInstr(result, IRInstruction(Opcode.STOREM, dt, fpReg1 = RegisterNum(valueTr.resultFpReg), address = address.toAddress()), null)
+                    addInstr(result, IRInstructions.storeMemory(Opcode.STOREM, dt, valueTr.resultFpReg, IRMemory.direct(address.toAddress())), null)
                 } else {
-                    addInstr(result, IRInstruction(Opcode.STOREM, dt, reg1 = valueTr.resultReg, address = address.toAddress()), null)
+                    addInstr(result, IRInstructions.storeMemory(Opcode.STOREM, dt, valueTr.resultReg, IRMemory.direct(address.toAddress())), null)
                 }
             } else {
                 val (address, offset) = exprGen.getAddressAndOffset(call.args[0])
@@ -576,18 +612,18 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                     addToResult(result, valueTr, valueTr.resultReg, valueTr.resultFpReg)
                     addToResult(result, addressTr, addressTr.resultReg, -1)
                     if(dt==IRDataType.FLOAT) {
-                        addInstr(result, IRInstruction(Opcode.STOREI, IRDataType.FLOAT, reg1 = addressTr.resultReg, fpReg1 = RegisterNum(valueTr.resultFpReg), immediate = offset!!), null)
+                        addInstr(result, IRInstructions.storeMemory(Opcode.STOREI, IRDataType.FLOAT, valueTr.resultFpReg, IRMemory.indirect(addressTr.resultReg, offset!!)), null)
                     } else {
-                        addInstr(result, IRInstruction(Opcode.STOREI, dt, reg1 = valueTr.resultReg, reg2 = addressTr.resultReg, immediate = offset!!), null)
+                        addInstr(result, IRInstructions.storeMemory(Opcode.STOREI, dt, valueTr.resultReg, IRMemory.indirect(addressTr.resultReg, offset!!)), null)
                     }
                 } else {
                     val addressTr = exprGen.translateExpression(call.args[0])
                     addToResult(result, valueTr, valueTr.resultReg, valueTr.resultFpReg)
                     addToResult(result, addressTr, addressTr.resultReg, -1)
                     if(dt==IRDataType.FLOAT) {
-                        addInstr(result, IRInstruction(Opcode.STOREI, IRDataType.FLOAT, reg1 = addressTr.resultReg, fpReg1 = RegisterNum(valueTr.resultFpReg), immediate = 0), null)
+                        addInstr(result, IRInstructions.storeMemory(Opcode.STOREI, IRDataType.FLOAT, valueTr.resultFpReg, IRMemory.indirect(addressTr.resultReg)), null)
                     } else {
-                        addInstr(result, IRInstruction(Opcode.STOREI, dt, reg1 = valueTr.resultReg, reg2 = addressTr.resultReg, immediate = 0), null)
+                        addInstr(result, IRInstructions.storeMemory(Opcode.STOREI, dt, valueTr.resultReg, IRMemory.indirect(addressTr.resultReg)), null)
                     }
                 }
             }
@@ -601,26 +637,26 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
             if(call.args[0] is PtNumber) {
                 val resultFpRegister = codeGen.registers.next(IRDataType.FLOAT)
                 val address = (call.args[0] as PtNumber).number.toUInt()
-                addInstr(result, IRInstruction(Opcode.LOADM, IRDataType.FLOAT, fpReg1 = RegisterNum(resultFpRegister), address = address.toAddress()), null)
+                addInstr(result, IRInstructions.loadMemory(Opcode.LOADM, IRDataType.FLOAT, resultFpRegister, IRMemory.direct(address.toAddress())), null)
                 ExpressionCodeResult(result, IRDataType.FLOAT, -1, resultFpRegister)
             } else {
                 val tr = exprGen.translateExpression(call.args.single())
                 addToResult(result, tr, tr.resultReg, -1)
                 val resultFpReg = codeGen.registers.next(IRDataType.FLOAT)
-                addInstr(result, IRInstruction(Opcode.LOADI, IRDataType.FLOAT, reg1 = tr.resultReg, fpReg1 = RegisterNum(resultFpReg), immediate = 0), null)
+                addInstr(result, IRInstructions.loadMemory(Opcode.LOADI, IRDataType.FLOAT, resultFpReg, IRMemory.indirect(tr.resultReg)), null)
                 ExpressionCodeResult(result, IRDataType.FLOAT, -1, resultFpReg)
             }
         } else {
             if (call.args[0] is PtNumber) {
                 val resultRegister = codeGen.registers.next(dt)
                 val address = (call.args[0] as PtNumber).number.toUInt()
-                addInstr(result, IRInstruction(Opcode.LOADM, dt, reg1 = resultRegister, address = address.toAddress()), null)
+                addInstr(result, IRInstructions.loadMemory(Opcode.LOADM, dt, resultRegister, IRMemory.direct(address.toAddress())), null)
                 ExpressionCodeResult(result, dt, resultRegister, -1)
             } else {
                 val tr = exprGen.translateExpression(call.args.single())
                 addToResult(result, tr, tr.resultReg, -1)
                 val resultReg = codeGen.registers.next(dt)
-                addInstr(result, IRInstruction(Opcode.LOADI, dt, reg1 = resultReg, reg2 = tr.resultReg, immediate = 0), null)
+                addInstr(result, IRInstructions.loadMemory(Opcode.LOADI, dt, resultReg, IRMemory.indirect(tr.resultReg)), null)
                 ExpressionCodeResult(result, dt, resultReg, -1)
             }
         }
@@ -637,11 +673,11 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
             val addressNum = address.number.toUInt()
             val resultRegister = codeGen.registers.next(IRDataType.BYTE)
             if (newValueTr != null) addToResult(result, newValueTr, newValueTr.resultReg, -1)
-            addInstr(result, IRInstruction(Opcode.LOADM, IRDataType.BYTE, reg1 = resultRegister, address = addressNum.toAddress()), null)
+            addInstr(result, IRInstructions.loadMemory(Opcode.LOADM, IRDataType.BYTE, resultRegister, IRMemory.direct(addressNum.toAddress())), null)
             if (newValueTr == null) {
-                addInstr(result, IRInstruction(Opcode.STOREZM, IRDataType.BYTE, address = addressNum.toAddress()), null)
+                addInstr(result, IRInstructions.storeZero(Opcode.STOREZM, IRDataType.BYTE, IRMemory.direct(addressNum.toAddress())), null)
             } else {
-                addInstr(result, IRInstruction(Opcode.STOREM, IRDataType.BYTE, reg1 = newValueTr.resultReg, address = addressNum.toAddress()), null)
+                addInstr(result, IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, newValueTr.resultReg, IRMemory.direct(addressNum.toAddress())), null)
             }
             ExpressionCodeResult(result, IRDataType.BYTE, resultRegister, -1)
         } else {
@@ -654,11 +690,11 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
             if (newValueTr != null) addToResult(result, newValueTr, newValueTr.resultReg, -1)
             addToResult(result, addressTr, addressTr.resultReg, -1)
 
-            addInstr(result, IRInstruction(Opcode.LOADI, IRDataType.BYTE, reg1 = resultReg, reg2 = addressTr.resultReg, immediate = finalOffset), null)
+            addInstr(result, IRInstructions.loadMemory(Opcode.LOADI, IRDataType.BYTE, resultReg, IRMemory.indirect(addressTr.resultReg, finalOffset)), null)
             if (newValueTr == null) {
-                addInstr(result, IRInstruction(Opcode.STOREZI, IRDataType.BYTE, reg1 = addressTr.resultReg, immediate = finalOffset), null)
+                addInstr(result, IRInstructions.storeZero(Opcode.STOREZI, IRDataType.BYTE, IRMemory.indirect(addressTr.resultReg, finalOffset)), null)
             } else {
-                addInstr(result, IRInstruction(Opcode.STOREI, IRDataType.BYTE, reg1 = newValueTr.resultReg, reg2 = addressTr.resultReg, immediate = finalOffset), null)
+                addInstr(result, IRInstructions.storeMemory(Opcode.STOREI, IRDataType.BYTE, newValueTr.resultReg, IRMemory.indirect(addressTr.resultReg, finalOffset)), null)
             }
             ExpressionCodeResult(result, IRDataType.BYTE, resultReg, -1)
         }
@@ -667,10 +703,10 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
 
     private fun funcStructAlloc(call: PtFunctionCall): ExpressionCodeResult {
         val code = IRCodeChunk(null, null)
-        val resultReg = codeGen.registers.next(IRDataType.WORD)
+        val resultReg = codeGen.registers.next(IRDataType.POINTER)
         val labelname = SymbolTable.labelnameForStructInstance(call)
-        code += IRInstruction(Opcode.LOAD, IRDataType.WORD, reg1=resultReg, labelSymbol = "${StStructInstanceBlockName}.$labelname")
-        return ExpressionCodeResult(code, IRDataType.WORD, resultReg, -1)
+        code += IRInstructions.loadAddress(IRDataType.POINTER, resultReg, "${StStructInstanceBlockName}.$labelname")
+        return ExpressionCodeResult(code, IRDataType.POINTER, resultReg, -1)
     }
 
     private fun funcLsb(call: PtFunctionCall, fromLong: Boolean): ExpressionCodeResult {
@@ -679,9 +715,9 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         addToResult(result, tr, tr.resultReg, -1)
         val resultReg = codeGen.registers.next(IRDataType.BYTE)
         if(fromLong)
-            addInstr(result, IRInstruction(Opcode.LSIGB, IRDataType.LONG, reg1 = resultReg, reg2 = tr.resultReg), null)
+            addInstr(result, IRInstructions.binary(Opcode.LSIGB, IRDataType.LONG, resultReg, tr.resultReg), null)
         else
-            addInstr(result, IRInstruction(Opcode.LSIGB, IRDataType.WORD, reg1 = resultReg, reg2 = tr.resultReg), null)
+            addInstr(result, IRInstructions.binary(Opcode.LSIGB, IRDataType.WORD, resultReg, tr.resultReg), null)
         // note: if a word result is needed, the upper byte is cleared by the typecast that follows. No need to do it here.
         return ExpressionCodeResult(result, IRDataType.BYTE, resultReg, -1)
     }
@@ -691,14 +727,25 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         // TODO this can be more optimal if the argument is a variable or memory address
         val tr = exprGen.translateExpression(call.args.single())
         addToResult(result, tr, tr.resultReg, -1)
-        // low byte returned in A, mid in R15, high (bank) in R14
         val byteReg = codeGen.registers.next(IRDataType.BYTE)
-        addInstr(result, IRInstruction(Opcode.BSIGB, IRDataType.LONG, reg1 = byteReg, reg2 = tr.resultReg), null)
-        addInstr(result, IRInstruction(Opcode.STOREM, IRDataType.BYTE, reg1 = byteReg, labelSymbol = "cx16.r14"), null)
-        addInstr(result, IRInstruction(Opcode.MIDB, IRDataType.LONG, reg1 = byteReg, reg2 = tr.resultReg), null)
-        addInstr(result, IRInstruction(Opcode.STOREM, IRDataType.BYTE, reg1 = byteReg, labelSymbol = "cx16.r15"), null)
-        addInstr(result, IRInstruction(Opcode.LSIGB, IRDataType.LONG, reg1 = byteReg, reg2 = tr.resultReg), null)
-        addInstr(result, IRInstruction(Opcode.STOREHR, IRDataType.BYTE, reg1 = byteReg, immediate = 0), null)
+        if(codeGen.options.compTarget.cpu.usesM68kConvention) {
+            // m68k: 3 return values in D0 (slot 10), D1 (slot 11), D2 (slot 12)
+            // store high byte first, then mid, then low (to avoid clobbering)
+            addInstr(result, IRInstructions.binary(Opcode.BSIGB, IRDataType.LONG, byteReg, tr.resultReg), null)
+            addInstr(result, IRInstructions.hardwareStore(IRDataType.BYTE, byteReg, CallingConventionSlot(12)), null)
+            addInstr(result, IRInstructions.binary(Opcode.MIDB, IRDataType.LONG, byteReg, tr.resultReg), null)
+            addInstr(result, IRInstructions.hardwareStore(IRDataType.BYTE, byteReg, CallingConventionSlot(11)), null)
+            addInstr(result, IRInstructions.binary(Opcode.LSIGB, IRDataType.LONG, byteReg, tr.resultReg), null)
+            addInstr(result, IRInstructions.hardwareStore(IRDataType.BYTE, byteReg, CallingConventionSlot(10)), null)
+        } else {
+            // 6502/cx16: low byte in A (slot 0), mid in cx16.r15, high (bank) in cx16.r14
+            addInstr(result, IRInstructions.binary(Opcode.BSIGB, IRDataType.LONG, byteReg, tr.resultReg), null)
+            addInstr(result, IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, byteReg, IRMemory.direct("cx16.r14")), null)
+            addInstr(result, IRInstructions.binary(Opcode.MIDB, IRDataType.LONG, byteReg, tr.resultReg), null)
+            addInstr(result, IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, byteReg, IRMemory.direct("cx16.r15")), null)
+            addInstr(result, IRInstructions.binary(Opcode.LSIGB, IRDataType.LONG, byteReg, tr.resultReg), null)
+            addInstr(result, IRInstructions.hardwareStore(IRDataType.BYTE, byteReg, CallingConventionSlot(0)), null)
+        }
         return ExpressionCodeResult(result, IRDataType.BYTE, -1, -1)
     }
 
@@ -707,7 +754,7 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val tr = exprGen.translateExpression(call.args.single())
         addToResult(result, tr, tr.resultReg, -1)
         val resultReg = codeGen.registers.next(IRDataType.WORD)
-        addInstr(result, IRInstruction(Opcode.LSIGW, IRDataType.LONG, reg1 = resultReg, reg2 = tr.resultReg), null)
+        addInstr(result, IRInstructions.binary(Opcode.LSIGW, IRDataType.LONG, resultReg, tr.resultReg), null)
         return ExpressionCodeResult(result, IRDataType.WORD, resultReg, -1)
     }
 
@@ -717,9 +764,9 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         addToResult(result, tr, tr.resultReg, -1)
         val resultReg = codeGen.registers.next(IRDataType.BYTE)
         if(fromLong)
-            addInstr(result, IRInstruction(Opcode.MSIGB, IRDataType.LONG, reg1 = resultReg, reg2 = tr.resultReg), null)
+            addInstr(result, IRInstructions.binary(Opcode.MSIGB, IRDataType.LONG, resultReg, tr.resultReg), null)
         else
-            addInstr(result, IRInstruction(Opcode.MSIGB, IRDataType.WORD, reg1 = resultReg, reg2 = tr.resultReg), null)
+            addInstr(result, IRInstructions.binary(Opcode.MSIGB, IRDataType.WORD, resultReg, tr.resultReg), null)
         // note: if a word result is needed, the upper byte is cleared by the typecast that follows. No need to do it here.
         return ExpressionCodeResult(result, IRDataType.BYTE, resultReg, -1)
     }
@@ -729,14 +776,14 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val tr = exprGen.translateExpression(call.args.single())
         addToResult(result, tr, tr.resultReg, -1)
         val resultReg = codeGen.registers.next(IRDataType.WORD)
-        addInstr(result, IRInstruction(Opcode.MSIGW, IRDataType.LONG, reg1 = resultReg, reg2 = tr.resultReg), null)
+        addInstr(result, IRInstructions.binary(Opcode.MSIGW, IRDataType.LONG, resultReg, tr.resultReg), null)
         return ExpressionCodeResult(result, IRDataType.WORD, resultReg, -1)
     }
 
     private fun funcRolRor(call: PtFunctionCall): ExpressionCodeResult {
         val result = mutableListOf<IRCodeChunkBase>()
         val arg = call.args[0]
-        val vmDt = irType(arg.type)
+        val vmDt = codeGen.irType(arg.type)
         val opcodeMemAndReg = when(call.name) {
             "rol" -> Opcode.ROXLM to Opcode.ROXL
             "ror" -> Opcode.ROXRM to Opcode.ROXR
@@ -747,13 +794,13 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
 
         val ident = arg as? PtIdentifier
         if(ident!=null) {
-            addInstr(result, IRInstruction(opcodeMemAndReg.first, vmDt, labelSymbol = ident.name), null)
+            addInstr(result, IRInstructions.memoryOp(opcodeMemAndReg.first, vmDt, IRMemory.direct(ident.name)), null)
             return ExpressionCodeResult(result, vmDt, -1, -1)
         }
 
         val memAddr: UInt? = (arg as? PtMemoryByte)?.address?.asConstInteger()?.toUInt()
         if(memAddr!=null) {
-            addInstr(result, IRInstruction(opcodeMemAndReg.first, vmDt, address = memAddr.toAddress()), null)
+            addInstr(result, IRInstructions.memoryOp(opcodeMemAndReg.first, vmDt, IRMemory.direct(memAddr.toAddress())), null)
             return ExpressionCodeResult(result, vmDt, -1, -1)
         }
 
@@ -765,19 +812,19 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                 result += IRCodeChunk(null, null).also {
                     when(opcodeMemAndReg.first) {
                         Opcode.ROXRM, Opcode.RORM -> {
-                            it += IRInstruction(opcodeMemAndReg.first, IRDataType.BYTE, labelSymbol = "${variable}_msb", symbolOffset = index)
-                            it += IRInstruction(opcodeMemAndReg.first, IRDataType.BYTE, labelSymbol = "${variable}_lsb", symbolOffset = index)
+                            it += IRInstructions.memoryOp(opcodeMemAndReg.first, IRDataType.BYTE, IRMemory.direct("${variable}_msb", index))
+                            it += IRInstructions.memoryOp(opcodeMemAndReg.first, IRDataType.BYTE, IRMemory.direct("${variable}_lsb", index))
                         }
                         Opcode.ROXLM, Opcode.ROLM -> {
-                            it += IRInstruction(opcodeMemAndReg.first, IRDataType.BYTE, labelSymbol = "${variable}_lsb", symbolOffset = index)
-                            it += IRInstruction(opcodeMemAndReg.first, IRDataType.BYTE, labelSymbol = "${variable}_msb", symbolOffset = index)
+                            it += IRInstructions.memoryOp(opcodeMemAndReg.first, IRDataType.BYTE, IRMemory.direct("${variable}_lsb", index))
+                            it += IRInstructions.memoryOp(opcodeMemAndReg.first, IRDataType.BYTE, IRMemory.direct("${variable}_msb", index))
                         }
                         else -> throw AssemblyError("wrong rol/ror opcode")
                     }
                 }
             } else {
                 val offset = codeGen.program.memsizer.memorySize(arr.type, index)
-                addInstr(result, IRInstruction(opcodeMemAndReg.first, vmDt, labelSymbol = variable, symbolOffset = offset), null)
+                addInstr(result, IRInstructions.memoryOp(opcodeMemAndReg.first, vmDt, IRMemory.direct(variable, offset)), null)
             }
             return ExpressionCodeResult(result, vmDt, -1, -1)
         }
@@ -785,17 +832,17 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val opcode = opcodeMemAndReg.second
         val saveCarry = opcode in OpcodesThatDependOnCarry && !arg.isSimple()
         if(saveCarry)
-            addInstr(result, IRInstruction(Opcode.PUSHST), null)    // save Carry
+            addInstr(result, IRInstructions.simple(Opcode.PUSHST), null)    // save Carry
         val tr = exprGen.translateExpression(arg)
         addToResult(result, tr, tr.resultReg, -1)
         if(saveCarry)
-            addInstr(result, IRInstruction(Opcode.POPST), null)
-        addInstr(result, IRInstruction(opcode, vmDt, reg1 = tr.resultReg), null)
+            addInstr(result, IRInstructions.simple(Opcode.POPST), null)
+        addInstr(result, IRInstructions.unary(opcode, vmDt, tr.resultReg), null)
         if(saveCarry)
-            addInstr(result, IRInstruction(Opcode.PUSHST), null)    // save Carry
+            addInstr(result, IRInstructions.simple(Opcode.PUSHST), null)    // save Carry
         result += codeGen.assignRegisterTo(arg, tr.resultReg)
         if(saveCarry)
-            addInstr(result, IRInstruction(Opcode.POPST), null)
+            addInstr(result, IRInstructions.simple(Opcode.POPST), null)
         return ExpressionCodeResult(result, vmDt, -1, -1)
     }
 
@@ -803,22 +850,24 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         val result = mutableListOf<IRCodeChunkBase>()
         val target = call.args[0]
         val isConstZeroValue = call.args[1].asConstInteger()==0
+        val isBigEndian = codeGen.options.compTarget.cpu.isBigEndian
+        fun byteOffset(elementSize: Int) = if(msb != isBigEndian) elementSize - 1 else 0
         when(target) {
             is PtIdentifier -> {
-                val offset = if (msb) { if(target.type.isLong) 3 else 1 } else 0
+                val offset = byteOffset(if(target.type.isLong) 4 else 2)
                 if(isConstZeroValue) {
                     result += IRCodeChunk(null, null).also {
-                        val pointerReg = codeGen.registers.next(IRDataType.WORD)
-                        it += IRInstruction(Opcode.LOAD, IRDataType.WORD, reg1 = pointerReg, labelSymbol = target.name)
-                        it += IRInstruction(Opcode.STOREZI, IRDataType.BYTE, reg1 = pointerReg, immediate = offset)
+                        val pointerReg = codeGen.registers.next(IRDataType.POINTER)
+                        it += IRInstructions.loadAddress(IRDataType.POINTER, pointerReg, target.name)
+                        it += IRInstructions.storeZero(Opcode.STOREZI, IRDataType.BYTE, IRMemory.indirect(pointerReg, offset))
                     }
                 } else {
                     val valueTr = exprGen.translateExpression(call.args[1])
                     addToResult(result, valueTr, valueTr.resultReg, -1)
                     result += IRCodeChunk(null, null).also {
-                        val pointerReg = codeGen.registers.next(IRDataType.WORD)
-                        it += IRInstruction(Opcode.LOAD, IRDataType.WORD, reg1 = pointerReg, labelSymbol = target.name)
-                        it += IRInstruction(Opcode.STOREI, IRDataType.BYTE, reg1 = valueTr.resultReg, reg2 = pointerReg, immediate = offset)
+                        val pointerReg = codeGen.registers.next(IRDataType.POINTER)
+                        it += IRInstructions.loadAddress(IRDataType.POINTER, pointerReg, target.name)
+                        it += IRInstructions.storeMemory(Opcode.STOREI, IRDataType.BYTE, valueTr.resultReg, IRMemory.indirect(pointerReg, offset))
                     }
                 }
             }
@@ -829,23 +878,26 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                     val pointerTr = exprGen.translateExpression(target.pointerderef!!)
                     addToResult(result, pointerTr, pointerTr.resultReg, -1)
                     val constIndex = target.index.asConstInteger()
+                    val elementByteOffset = byteOffset(eltSize)
                     if(constIndex != null) {
-                        val offset = eltSize * constIndex + if(msb) eltSize - 1 else 0
+                        val offset = eltSize * constIndex + elementByteOffset
                         if(offset > 0)
-                            addInstr(result, IRInstruction(Opcode.ADD, IRDataType.WORD, reg1 = pointerTr.resultReg, immediate = offset), null)
+                            addInstr(result, IRInstructions.binaryImmediate(Opcode.ADD, IRDataType.POINTER, pointerTr.resultReg, offset), null)
                     } else {
-                        val (code, indexWordReg) = codeGen.loadIndexReg(target.index, eltSize, true, false)
+                        val (code, indexWordReg) = codeGen.loadIndexReg(target.index, true)
                         result += code
-                        addInstr(result, IRInstruction(Opcode.ADDR, IRDataType.WORD, reg1 = pointerTr.resultReg, reg2 = indexWordReg), null)
-                        if(msb)
-                            addInstr(result, IRInstruction(Opcode.ADD, IRDataType.WORD, reg1 = pointerTr.resultReg, immediate = eltSize - 1), null)
+                        if(eltSize!=1)
+                            result += codeGen.multiplyByConst(DataType.UWORD, indexWordReg, eltSize)
+                        addInstr(result, IRInstructions.binary(Opcode.ADDR, IRDataType.POINTER, pointerTr.resultReg, indexWordReg), null)
+                        if(elementByteOffset > 0)
+                            addInstr(result, IRInstructions.binaryImmediate(Opcode.ADD, IRDataType.POINTER, pointerTr.resultReg, elementByteOffset), null)
                     }
                     if(isConstZeroValue) {
-                        addInstr(result, IRInstruction(Opcode.STOREZI, IRDataType.BYTE, reg1 = pointerTr.resultReg, immediate = 0), null)
+                        addInstr(result, IRInstructions.storeZero(Opcode.STOREZI, IRDataType.BYTE, IRMemory.indirect(pointerTr.resultReg)), null)
                     } else {
                         val valueTr = exprGen.translateExpression(call.args[1])
                         addToResult(result, valueTr, valueTr.resultReg, -1)
-                        addInstr(result, IRInstruction(Opcode.STOREI, IRDataType.BYTE, reg1 = valueTr.resultReg, reg2 = pointerTr.resultReg, immediate = 0), null)
+                        addInstr(result, IRInstructions.storeMemory(Opcode.STOREI, IRDataType.BYTE, valueTr.resultReg, IRMemory.indirect(pointerTr.resultReg)), null)
                     }
                 }
                 else if(target.splitWords) {
@@ -854,32 +906,36 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                     val varName = target.variable!!.name + if(msb) "_msb" else "_lsb"
                     if(isConstZeroValue) {
                         if(constIndex!=null) {
-                            val offsetReg = codeGen.registers.next(IRDataType.BYTE)
+                            val idxType = codeGen.options.compTarget.indexRegType
+                            val offsetReg = codeGen.registers.next(idxType)
                             result += IRCodeChunk(null, null).also {
-                                it += IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=offsetReg, immediate = constIndex)
-                                it += IRInstruction(Opcode.STOREZX, IRDataType.BYTE, reg1=offsetReg, labelSymbol = varName)
+                                it += IRInstructions.load(idxType, offsetReg, constIndex)
+                                it += IRInstructions.storeZero(Opcode.STOREZX, IRDataType.BYTE, IRMemory.indexed(varName, offsetReg, codeGen.options.compTarget.indexRegType))
                             }
                         } else {
                             val indexTr = exprGen.translateExpression(target.index)
                             addToResult(result, indexTr, indexTr.resultReg, -1)
+                            val indexReg = codeGen.canonicalizeIndexReg(result, indexTr)
                             result += IRCodeChunk(null, null).also {
-                                it += IRInstruction(Opcode.STOREZX, IRDataType.BYTE, reg1=indexTr.resultReg, labelSymbol = varName)
+                                it += IRInstructions.storeZero(Opcode.STOREZX, IRDataType.BYTE, IRMemory.indexed(varName, indexReg, codeGen.options.compTarget.indexRegType))
                             }
                         }
                     } else {
                         val valueTr = exprGen.translateExpression(call.args[1])
                         addToResult(result, valueTr, valueTr.resultReg, -1)
                         if(constIndex!=null) {
-                            val offsetReg = codeGen.registers.next(IRDataType.BYTE)
+                            val idxType = codeGen.options.compTarget.indexRegType
+                            val offsetReg = codeGen.registers.next(idxType)
                             result += IRCodeChunk(null, null).also {
-                                it += IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=offsetReg, immediate = constIndex)
-                                it += IRInstruction(Opcode.STOREX, IRDataType.BYTE, reg1=valueTr.resultReg, reg2=offsetReg, labelSymbol = varName)
+                                it += IRInstructions.load(idxType, offsetReg, constIndex)
+                                it += IRInstructions.storeMemory(Opcode.STOREX, IRDataType.BYTE, valueTr.resultReg, IRMemory.indexed(varName, offsetReg, codeGen.options.compTarget.indexRegType))
                             }
                         } else {
                             val indexTr = exprGen.translateExpression(target.index)
                             addToResult(result, indexTr, indexTr.resultReg, -1)
+                            val indexReg = codeGen.canonicalizeIndexReg(result, indexTr)
                             result += IRCodeChunk(null, null).also {
-                                it += IRInstruction(Opcode.STOREX, IRDataType.BYTE, reg1=valueTr.resultReg, reg2=indexTr.resultReg, labelSymbol = varName)
+                                it += IRInstructions.storeMemory(Opcode.STOREX, IRDataType.BYTE, valueTr.resultReg, IRMemory.indexed(varName, indexReg, codeGen.options.compTarget.indexRegType))
                             }
                         }
                     }
@@ -891,50 +947,36 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
                     val constIndex = target.index.asConstInteger()
                     if(isConstZeroValue) {
                         if(constIndex!=null) {
-                            val offsetReg = codeGen.registers.next(IRDataType.BYTE)
-                            val offset = eltSize*constIndex + if(msb) eltSize-1 else 0
+                            val offsetReg = codeGen.registers.next(if(codeGen.wordArrayIndex) IRDataType.WORD else IRDataType.BYTE)
+                            val offset = eltSize*constIndex + byteOffset(eltSize)
                             result += IRCodeChunk(null, null).also {
-                                it += IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=offsetReg, immediate = offset)
-                                it += IRInstruction(Opcode.STOREZX, IRDataType.BYTE, reg1=offsetReg, labelSymbol = targetVariable.name)
+                                it += IRInstructions.load(if(codeGen.wordArrayIndex) IRDataType.WORD else IRDataType.BYTE, offsetReg, offset)
+                                it += IRInstructions.storeZero(Opcode.STOREZX, IRDataType.BYTE, IRMemory.indexed(targetVariable.name, offsetReg, codeGen.options.compTarget.indexRegType))
                             }
                         } else {
-                            val indexTr = exprGen.translateExpression(target.index)
-                            addToResult(result, indexTr, indexTr.resultReg, -1)
+                            val (code, indexReg) = codeGen.loadIndexReg(target.index, codeGen.wordArrayIndex)
+                            result += code
                             result += IRCodeChunk(null, null).also {
-                                if(eltSize>1)
-                                    it += codeGen.multiplyByConst(DataType.UBYTE, indexTr.resultReg, eltSize)
-                                if(msb) {
-                                    it += if(target.type.isLong)
-                                        IRInstruction(Opcode.ADD, IRDataType.BYTE, reg1 = indexTr.resultReg, immediate = 3)
-                                    else
-                                        IRInstruction(Opcode.INC, IRDataType.BYTE, reg1 = indexTr.resultReg)
-                                }
-                                it += IRInstruction(Opcode.STOREZX, IRDataType.BYTE, reg1=indexTr.resultReg, labelSymbol = targetVariable.name)
+                                val offset = byteOffset(eltSize)
+                                it += IRInstructions.storeZero(Opcode.STOREZX, IRDataType.BYTE, IRMemory.indexed(targetVariable.name, indexReg, codeGen.options.compTarget.indexRegType, scale = eltSize, displacement = offset))
                             }
                         }
                     } else {
                         val valueTr = exprGen.translateExpression(call.args[1])
                         addToResult(result, valueTr, valueTr.resultReg, -1)
                         if(constIndex!=null) {
-                            val offsetReg = codeGen.registers.next(IRDataType.BYTE)
-                            val offset = eltSize*constIndex + if(msb) eltSize-1 else 0
+                            val offsetReg = codeGen.registers.next(if(codeGen.wordArrayIndex) IRDataType.WORD else IRDataType.BYTE)
+                            val offset = eltSize*constIndex + byteOffset(eltSize)
                             result += IRCodeChunk(null, null).also {
-                                it += IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=offsetReg, immediate = offset)
-                                it += IRInstruction(Opcode.STOREX, IRDataType.BYTE, reg1=valueTr.resultReg, reg2=offsetReg, labelSymbol = targetVariable.name)
+                                it += IRInstructions.load(if(codeGen.wordArrayIndex) IRDataType.WORD else IRDataType.BYTE, offsetReg, offset)
+                                it += IRInstructions.storeMemory(Opcode.STOREX, IRDataType.BYTE, valueTr.resultReg, IRMemory.indexed(targetVariable.name, offsetReg, codeGen.options.compTarget.indexRegType))
                             }
                         } else {
-                            val indexTr = exprGen.translateExpression(target.index)
-                            addToResult(result, indexTr, indexTr.resultReg, -1)
+                            val (code, indexReg) = codeGen.loadIndexReg(target.index, codeGen.wordArrayIndex)
+                            result += code
                             result += IRCodeChunk(null, null).also {
-                                if(eltSize>1)
-                                    it += codeGen.multiplyByConst(DataType.UBYTE, indexTr.resultReg, eltSize)
-                                if(msb) {
-                                    it += if(target.type.isLong)
-                                        IRInstruction(Opcode.ADD, IRDataType.BYTE, reg1 = indexTr.resultReg, immediate = 3)
-                                    else
-                                        IRInstruction(Opcode.INC, IRDataType.BYTE, reg1 = indexTr.resultReg)
-                                }
-                                it += IRInstruction(Opcode.STOREX, IRDataType.BYTE, reg1=valueTr.resultReg, reg2=indexTr.resultReg, labelSymbol = targetVariable.name)
+                                val offset = byteOffset(eltSize)
+                                it += IRInstructions.storeMemory(Opcode.STOREX, IRDataType.BYTE, valueTr.resultReg, IRMemory.indexed(targetVariable.name, indexReg, codeGen.options.compTarget.indexRegType, scale = eltSize, displacement = offset))
                             }
                         }
                     }
@@ -944,4 +986,31 @@ internal class BuiltinFuncGen(private val codeGen: IRCodeGen, private val exprGe
         }
         return ExpressionCodeResult(result, IRDataType.WORD, -1, -1)
     }
+
+    private fun tryFoldStructArrayPoke(call: PtFunctionCall, dt: IRDataType): ExpressionCodeResult? {
+        // Detect pokew(&arr[idx]+fieldOff, value) where arr is struct array with variable idx
+        // and fold to single STOREX/STOREZX with scale=structSize.
+        val info = extractStructArrayIndexInfo(call.args[0], codeGen) ?: return null
+        val isZero = codeGen.isZero(call.args[1])
+        val result = mutableListOf<IRCodeChunkBase>()
+        val (idxCode, indexReg) = codeGen.loadIndexReg(info.idxExpr, codeGen.wordArrayIndex)
+        result += idxCode
+        return if(isZero) {
+            result += IRCodeChunk(null, null).also {
+                it += IRInstructions.storeZero(Opcode.STOREZX, dt, IRMemory.indexed(info.arrayName, indexReg, codeGen.options.compTarget.indexRegType, scale=info.structSize, displacement=info.fieldOffset))
+            }
+            ExpressionCodeResult(result, IRDataType.BYTE, -1, -1)
+        } else {
+            val valueTr = exprGen.translateExpression(call.args[1])
+            addToResult(result, valueTr, valueTr.resultReg, valueTr.resultFpReg)
+            result += IRCodeChunk(null, null).also {
+                it += if(dt==IRDataType.FLOAT) {
+                    IRInstructions.storeMemory(Opcode.STOREX, dt, valueTr.resultFpReg, IRMemory.indexed(info.arrayName, indexReg, codeGen.options.compTarget.indexRegType, scale=info.structSize, displacement=info.fieldOffset))
+                } else {
+                    IRInstructions.storeMemory(Opcode.STOREX, dt, valueTr.resultReg, IRMemory.indexed(info.arrayName, indexReg, codeGen.options.compTarget.indexRegType, scale=info.structSize, displacement=info.fieldOffset))
+                }
+            }
+            ExpressionCodeResult(result, IRDataType.BYTE, -1, -1)
+        }
+    }
 }

codeGenIntermediate/src/prog8/codegen/intermediate/IRFactoryHelpers.kt

@@ -0,0 +1,36 @@
+package prog8.codegen.intermediate
+
+import prog8.intermediate.*
+
+/*
+Construction helpers for the few IR opcodes that have no dedicated IRInstructions factory,
+plus a small operand-inspection helper used by the optimizers.
+
+The structured IRInstructions factory covers almost every opcode, but three opcodes have an
+operand shape that no named factory produces (a lone source register, or a source register plus
+an immediate without a destination). They are built here from schema-derived operands, so this
+module never constructs IRInstruction directly.
+*/
+
+/** STOREHFACZERO / STOREHFACONE: a single source float register (no dedicated factory exists). */
+internal fun storeHwFac(opcode: Opcode, fpRegister: Int): IRInstruction =
+    IRInstructions.push(IRDataType.FLOAT, fpRegister).copy(opcode = opcode)
+
+/** BITTST: a source register plus a byte immediate bit number, with no destination register. */
+internal fun bitTest(type: IRDataType, register: Int, bit: Int): IRInstruction =
+    IRInstructions.simple(Opcode.NOP).copy(
+        opcode = Opcode.BITTST,
+        type = type,
+        srcA = IRInstructions.operandFor(Opcode.BITTST, type, InstructionSlot.SRC_A, register),
+        immediate = IRInstructions.immediateFor(Opcode.BITTST, type, bit)
+    )
+
+/**
+ * The name of the single symbol this instruction refers to, if any.
+ * That is either the symbol a memory reference is based on, the symbol whose address is taken
+ * in an immediate operand, or the label a branch/jump/call transfers control to.
+ */
+internal fun IRInstruction.referencedSymbol(): String? =
+    memory?.symbolName
+        ?: (immediate as? ImmediateOperand.SymbolAddress)?.symbol
+        ?: labelTarget

codeGenIntermediate/src/prog8/codegen/intermediate/IRUnusedCodeRemover.kt

@@ -8,12 +8,28 @@ class IRUnusedCodeRemover(
     private val irprog: IRProgram,
     private val errors: IErrorReporter
 ) {
+    /**
+     * The symbols that struct instance initializers take the address of. These are plain data
+     * references: nothing ever calls them, so without this the code they point at is considered
+     * unused and removed, leaving a dangling reference in the instance initializer.
+     * The same reasoning applies to the array initializers of variables, which are handled
+     * separately at each of the use sites below.
+     */
+    private fun structInstanceInitRefs(): Sequence<IRStSymbolicReference.Symbol> =
+        irprog.st.allStructInstances()
+            .flatMap { instance -> instance.values.asSequence().map { it.value } }
+            .filterIsInstance<IRStSymbolicReference.Symbol>()
+
     fun optimize(): Int {
         var numRemoved = removeUnusedSubroutines() + removeUnusedAsmSubroutines()
 
-        // remove empty blocks
+        // remove empty blocks (but keep blocks with %option force_output,
+        // and also blocks that contain labels -- labels are addressable symbols,
+        // and blocks that still have variables in the symbol table -- a block with variables is not truly empty)
         irprog.blocks.reversed().forEach { block ->
-            if(block.isEmpty()) {
+            val hasLabels = block.children.any { it.label != null }
+            val hasVariables = irprog.st.allVariables().any { it.name.startsWith(block.label + ".") }
+            if(!hasLabels && !hasVariables && block.isEmpty() && !block.options.forceOutput) {
                 irprog.blocks.remove(block)
                 pruneSymboltable(block.label)
                 numRemoved++
@@ -33,7 +49,7 @@ class IRUnusedCodeRemover(
         blockVars.forEach { stVar ->
             irprog.allSubs().flatMap { it.chunks }.forEach { chunk ->
                 chunk.instructions.forEach { ins ->
-                    if(ins.labelSymbol == stVar.name) {
+                    if(ins.referencedSymbol() == stVar.name) {
                         return  // symbol occurs in an instruction
                     }
                 }
@@ -48,6 +64,14 @@ class IRUnusedCodeRemover(
                         return   // symbol occurs in an initializer value (address-of this symbol)_
                 }
             }
+
+            irprog.st.allStructInstances().forEach { instance ->
+                if(instance.values.any { fieldValue ->
+                        val ref = fieldValue.value
+                        ref is IRStSymbolicReference.Symbol && ref.name.startsWith(blockLabel)
+                    })
+                    return   // symbol occurs in a struct instance initializer value
+            }
         }
 
         val blockStructs = irprog.st.allStructDefs().filter { it.name.startsWith(prefix) }
@@ -70,41 +94,26 @@ class IRUnusedCodeRemover(
     private fun removeBlockInits(code: IRProgram, blockLabel: String) {
         val instructions = code.globalInits.instructions
         instructions.toTypedArray().forEach {ins ->
-            if(ins.labelSymbol?.startsWith(blockLabel)==true) {
+            if(ins.referencedSymbol()?.startsWith(blockLabel)==true) {
                 instructions.remove(ins)
             }
         }
 
         // remove stray loads
-        val readRegs = mutableSetOf<Int>()
-        val readFpRegs = mutableSetOf<Int>()
+        val readRegs = mutableSetOf<VirtualRegister>()
         instructions.forEach { ins ->
-            val readRegsCounts = mutableMapOf<RegisterNum, Int>()
-            val readFpRegsCounts = mutableMapOf<RegisterNum, Int>()
-            val writeRegsCounts = mutableMapOf<RegisterNum, Int>()
-            val writeFpRegsCounts = mutableMapOf<RegisterNum, Int>()
-            val regsTypes = mutableMapOf<RegisterNum, IRDataType>()
-            ins.addUsedRegistersCounts(readRegsCounts, writeRegsCounts, readFpRegsCounts, writeFpRegsCounts, regsTypes, null)
-            readRegs.addAll(readRegsCounts.keys.map { it.value })
-            readFpRegs.addAll(readFpRegsCounts.keys.map { it.value })
+            val readRegsCounts = mutableMapOf<VirtualRegister, Int>()
+            val writeRegsCounts = mutableMapOf<VirtualRegister, Int>()
+            val regsTypes = mutableMapOf<VirtualRegister, IRDataType>()
+            ins.addUsedRegistersCounts(readRegsCounts, writeRegsCounts, regsTypes, null)
+            readRegs.addAll(readRegsCounts.keys)
         }
         instructions.toTypedArray().forEach { ins ->
             if(ins.opcode in arrayOf(Opcode.LOAD, Opcode.LOADR, Opcode.LOADM)) {
-                val reg1 = ins.reg1
-                val fpReg1 = ins.fpReg1
-                if(reg1!=null) {
-                    if(reg1 !in readRegs) {
-                        if(ins.labelSymbol!=null)
-                            code.st.removeIfExists(ins.labelSymbol!!)
-                        instructions.remove(ins)
-                    }
-                }
-                else if(fpReg1!=null) {
-                    if(fpReg1.value !in readFpRegs) {
-                        if(ins.labelSymbol!=null)
-                            code.st.removeIfExists(ins.labelSymbol!!)
-                        instructions.remove(ins)
-                    }
+                val dest = ins.dest
+                if(dest != null && dest.register !in readRegs) {
+                    ins.referencedSymbol()?.let { code.st.removeIfExists(it) }
+                    instructions.remove(ins)
                 }
             }
         }
@@ -187,8 +196,8 @@ class IRUnusedCodeRemover(
         // check if asmsub is linked or called from another regular subroutine
         irprog.foreachCodeChunk { chunk ->
             chunk.instructions.forEach {
-                it.labelSymbol?.let { label -> allSubs[label]?.let { cc -> linkedAsmSubs += cc } }
-                // note: branchTarget can't yet point to another IRAsmSubroutine, so do nothing when it's set
+                it.referencedSymbol()?.let { label -> allSubs[label]?.let { cc -> linkedAsmSubs += cc } }
+                // note: codeTarget can't yet point to another IRAsmSubroutine, so do nothing when it's set
             }
         }
 
@@ -232,21 +241,35 @@ class IRUnusedCodeRemover(
                 }
             }
 
+        // ... and likewise the ones referenced in struct instance initializers:
+        structInstanceInitRefs()
+            .filter { irprog.st.lookup(it.name) == null }
+            .forEach { ref -> irprog.getChunkWithLabel(ref.name).let { reachable += it } }
+
         fun grow() {
             val new = mutableSetOf<IRCodeChunkBase>()
             reachable.forEach {
                 it.next?.let { next -> new += next }
+                if (it is IRLoopChunk) {
+                    // A reachable counted loop makes every body chunk reachable too.
+                    // The body is no longer represented by next pointers in the outer list.
+                    new += it.body
+                }
                 it.instructions.forEach { instr ->
-                    if (instr.branchTarget == null)
-                        instr.labelSymbol?.let { label ->
+                    val codeTarget = instr.codeTarget
+                    if (codeTarget == null) {
+                        instr.referencedSymbol()?.let { label ->
                             val chunk = allLabeledChunks[label] ?: allLabeledChunks[label.substringBeforeLast('.')]
                             if(chunk!=null)
                                 new+=chunk
-                            else
-                                allLabeledChunks[label]?.let { c -> new += c }
                         }
-                    else
-                        new += instr.branchTarget!!
+                    }
+                    else {
+                        val chunk = irprog.resolveCodeTarget(codeTarget)
+                            ?: codeTarget.labelName?.let { label -> allLabeledChunks[label] }
+                        if(chunk!=null)
+                            new += chunk
+                    }
                 }
             }
             reachable += new
@@ -279,13 +302,22 @@ class IRUnusedCodeRemover(
                 }
             }
 
+        // ... and likewise the chunks referenced in struct instance initializers:
+        structInstanceInitRefs()
+            .filter { irprog.st.lookup(it.name) == null }
+            .forEach { ref -> allLabeledChunks[ref.name]?.let { linkedChunks += it } }
+
         irprog.foreachCodeChunk { chunk ->
             chunk.next?.let { next -> linkedChunks += next }
             chunk.instructions.forEach {
-                if(it.branchTarget==null) {
-                    it.labelSymbol?.let { label -> allLabeledChunks[label]?.let { cc -> linkedChunks += cc } }
+                val codeTarget = it.codeTarget
+                if(codeTarget==null) {
+                    it.referencedSymbol()?.let { label -> allLabeledChunks[label]?.let { cc -> linkedChunks += cc } }
                 } else {
-                    linkedChunks += it.branchTarget!!
+                    val target = irprog.resolveCodeTarget(codeTarget)
+                        ?: codeTarget.labelName?.let { label -> allLabeledChunks[label] }
+                    if(target!=null)
+                        linkedChunks += target
                 }
             }
             if (chunk.label in entrypointNames)
@@ -295,8 +327,8 @@ class IRUnusedCodeRemover(
         // make sure that chunks that are only used as a prefix of a label, are also marked as linked
         linkedChunks.toList().forEach { chunk ->
             chunk.instructions.forEach {
-                if(it.labelSymbol!=null) {
-                    val chunkName = it.labelSymbol!!.substringBeforeLast('.')
+                it.referencedSymbol()?.let { symbol ->
+                    val chunkName = symbol.substringBeforeLast('.')
                     allLabeledChunks[chunkName]?.let { c -> linkedChunks += c }
                 }
             }
@@ -312,7 +344,7 @@ class IRUnusedCodeRemover(
             // don't remove stuff in library modules or with %forceoutput
             // TODO this is needed to keep the subroutine body of sqrt_long from being wiped !??!   Sounds like the bad solution to the problem: why is the *body* being cleared , making an empty sqrt_long subrotuine end up in the p8ir output (and eventually in the .asm, which crashes the resulting program) 
             // TODO this causes other stuff to end up in the output IR that is not used......
-            // note: this wasn't a problem on the virtual target, but it is a problem when using the expericodegen target (because on the virtual target, the VM simply executes the SQRT instruction directly without needing a library routine)
+            // note: this wasn't a problem on the virtual target, but it is a problem when using the newcodegen targets (because on the virtual target, the VM simply executes the SQRT instruction directly without needing a library routine)
             val block = irprog.blocks.first { b -> b.children.any { it === sub } }
             if (block.library || block.options.forceOutput)
                 return@foreachSub
@@ -328,6 +360,9 @@ class IRUnusedCodeRemover(
                                     numRemoved++
                                 }
                             }
+                            is IRLoopChunk -> {
+                                // keep loop at subroutine entry; don't remove
+                            }
                             is IRInlineAsmChunk, is IRInlineBinaryChunk -> {
                                 sub.chunks[index] = IRCodeChunk(chunk.label, chunk.next)
                                 numRemoved++

codeGenIntermediate/src/prog8/codegen/intermediate/RegisterPacker.kt

@@ -0,0 +1,465 @@
+package prog8.codegen.intermediate
+
+import prog8.intermediate.*
+
+/*
+ * ============================================================================
+ * PROTOTYPE - NOT FOR PRODUCTION USE
+ * ============================================================================
+ *
+ * This is a PROTOTYPE memory-slot packer and the intended reusable foundation
+ * for a future hardware-register allocator targeting the m68k backend. It is
+ * NOT a complete or production-ready register allocator in its current form.
+ * It is DISABLED in production: its only call site in IRCodeGen.generate() is
+ * commented out, and it currently runs only under TestRegisterPacker.
+ *
+ * What this prototype demonstrates:
+ *   - A working graph-coloring register allocator skeleton: CFG construction,
+ *     intraprocedural liveness analysis (gen/kill fixed-point), live interval
+ *     derivation, conflict-graph construction, greedy coloring, and IR rewrite.
+ *   - Coalescing of virtual registers into shared memory slots in the flat
+ *     `p8_regfile` BSS block (vregs with non-overlapping live ranges share a
+ *     slot).
+ *
+ * What this prototype DOES NOT do (and why it is not production-ready):
+ *   - No hardware-register allocation: slots are memory addresses in a flat
+ *     BSS regfile, not physical D/A/FP registers.
+ *   - No CALL clobbering: a callee packed into the same slot as a caller's
+ *     live value silently overwrites it (see bug 3 below).
+ *   - No register classes: no distinction between DATA, ADDRESS, and FPU
+ *     registers (required by the m68k backend).
+ *   - No spilling, prologue/epilogue emission, or CALL-aware interference.
+ *   - Known-unsound liveness for complex control flow (nested loops,
+ *     conditionals, early returns) - see bug 2 below.
+ *
+ * Intended reuse path (per ideas/m68k-register-allocation.md):
+ *   - Stage 1: Make the m68k backend's instruction selection location-agnostic
+ *     (operand() indirection returning "d3" if allocated, "p8_regfile+N" if
+ *     spilled) - prerequisite bulk work, no behavior change.
+ *   - Stage 2: Replace this packer's interference model with a correct one
+ *     (CALL-aware, class-constrained) and point the coloring step at real
+ *     D2-D6 / A2-A4 / FP2-FP7 registers. Six open design decisions remain
+ *     in Stage 2 (see design doc section 7.1).
+ *   - Stage 3: Delete subsumed peephole optimizations, update test assertions.
+ */
+
+/*
+ * Known bugs / unsoundness (carried over from the removed register-packing.md):
+ *
+ * 1. Disjoint-interval value clobbering (observed on cx16 Fibonacci).
+ *    A register with multiple DISJOINT live ranges (e.g. a loop counter written
+ *    at loop entry and read at loop exit, with no uses in between) was split
+ *    into separate intervals; the packer allowed another register to share the
+ *    same slot during the gap, overwriting the value before its next read.
+ *    Fix applied (see packSubroutine interval merging): always merge ALL
+ *    intervals of the same register into one contiguous range, regardless of
+ *    overlap, so the value persists across gaps.
+ *
+ * 2. Complex control flow (observed on m68k TextElite).
+ *    The iterative gen/kill fixed-point liveness (computeLiveness) produced
+ *    incorrect live ranges for subroutines with complex CFGs (nested
+ *    conditionals, loops, early returns). Two registers with genuinely
+ *    overlapping live ranges were assigned the same slot, so one clobbered the
+ *    other at runtime (symptom: infinite loop printing spaces — galaxy map
+ *    data read from the wrong memory location). Root cause: the dataflow does
+ *    not converge to the true meet-over-all-paths solution for all loop /
+ *    conditional structures. The liveness MUST be validated against nested
+ *    loops, conditionals, early returns, and switch-like dispatch before reuse.
+ *
+ * 3. Cross-subroutine slot collision (fundamental design issue; see header
+ *    comment above). Per-subroutine packing into ONE flat shared regfile is
+ *    unsound: a callee packed into the same slot as a caller's live value
+ *    clobbers it. The packer does not model CALL clobbering. Soundness would
+ *    require the full-program call tree (call-graph-aware allocation, depth
+ *    ranges, or save/restore around calls) — which is exactly what the m68k
+ *    allocator with a calling convention avoids by allocating per-subroutine.
+ *
+ * 4. Interval-merge type loss / silent skips. When merging a register's
+ *    intervals, the merged interval keeps only the FIRST interval's type; if a
+ *    later interval has an incompatible type, the register is dropped via
+ *    skipRegs (and its original number is preserved via the startSlot =
+ *    maxReg + 1 invariant). When reusing this code, merged intervals must carry
+ *    the WIDEST type, not just the first, to avoid under-sized slots.
+ */
+
+object RegisterPacker {
+
+    data class Interval(val register: VirtualRegister, val start: Int, val end: Int, val type: IRDataType)
+
+    fun pack(irProg: IRProgram) {
+        val allRegTypes = mutableMapOf<VirtualRegister, IRDataType>()
+        irProg.forEachInstruction { instr ->
+            for (access in instr.registerAccesses)
+                allRegTypes.putIfAbsent(access.register, access.type)
+        }
+        val beforeCount = allRegTypes.size
+
+        // Start packing slots after the highest original register number to avoid
+        // collisions between packed slot numbers and original (non-packed) register numbers.
+        val maxReg = allRegTypes.keys.maxOfOrNull { it.num } ?: 0
+        val startSlot = maxReg + 1
+        val globalSlotTypes = mutableMapOf<VirtualRegister, IRDataType>()
+        irProg.foreachSub { sub -> packSubroutine(irProg, sub, allRegTypes, globalSlotTypes, startSlot) }
+
+        val afterTypes = rebuildTypeMap(irProg)
+        val afterCount = afterTypes.size
+        if (!irProg.options.quiet)
+            println("Register packing: $beforeCount -> $afterCount registers")
+    }
+
+    // Rebuild the register type map after packing, using the same type determination as the packer.
+    // This avoids the strict per-instruction validation in usedRegisters() that POINTER↔LONG↔WORD
+    // cross-type packing can trigger.
+    fun rebuildTypeMap(irProg: IRProgram): Map<VirtualRegister, IRDataType> {
+        val newTypes = mutableMapOf<VirtualRegister, IRDataType>()
+        irProg.forEachInstruction { instr ->
+            for (access in instr.registerAccesses)
+                newTypes.putIfAbsent(access.register, access.type)
+        }
+        return newTypes
+    }
+
+    private fun packSubroutine(irProg: IRProgram, sub: IRSubroutine, allRegTypes: Map<VirtualRegister, IRDataType>, globalSlotTypes: MutableMap<VirtualRegister, IRDataType>, startSlot: Int) {
+        if (sub.chunks.isEmpty())
+            return
+
+        val successors = buildCFG(sub, irProg)
+        val (_, liveOut) = computeLiveness(sub, successors)
+
+        val codeChunks = sub.chunks.filterIsInstance<IRCodeChunk>()
+        if (codeChunks.isEmpty())
+            return
+
+        val chunkRanges = mutableMapOf<IRCodeChunk, IntRange>()
+        var globalIdx = 0
+        for (chunk in codeChunks) {
+            val start = globalIdx
+            val end = globalIdx + chunk.instructions.size - 1
+            chunkRanges[chunk] = start..end
+            globalIdx += chunk.instructions.size
+        }
+
+        val registerIntervals = mutableMapOf<VirtualRegister, MutableList<Interval>>()
+        val registerTypes = mutableMapOf<VirtualRegister, IRDataType>()
+
+        for (chunk in codeChunks) {
+            val range = chunkRanges[chunk]!!
+            val chunkStart = range.first
+            if (chunk.instructions.isEmpty())
+                continue
+
+            val liveSet = liveOut[chunk]?.toMutableSet() ?: mutableSetOf()
+            val lastUse = mutableMapOf<VirtualRegister, Int>()
+
+            // Scan backward through instructions
+            for (i in chunk.instructions.indices.reversed()) {
+                val instr = chunk.instructions[i]
+                val globalI = chunkStart + i
+                val accesses = instr.registerAccesses
+
+                // Process READ (USE / USE_DEF) first so types are set before WRITE processing
+                for (access in accesses) {
+                    if (access.direction != OperandDirection.DEF) {
+                        val r = access.register
+                        registerTypes.putIfAbsent(r, access.type)
+                        if (r !in liveSet) {
+                            lastUse[r] = globalI
+                            liveSet.add(r)
+                        }
+                    }
+                }
+
+                // Process WRITTEN (DEF / USE_DEF) registers: end their live range (start of interval going forward)
+                for (access in accesses) {
+                    if (access.direction != OperandDirection.USE) {
+                        val r = access.register
+                        if (r in liveSet) {
+                            val end = lastUse.getOrElse(r) { globalI }
+                            registerIntervals.getOrPut(r) { mutableListOf() }
+                                .add(Interval(r, globalI, end, access.type))
+                            liveSet.remove(r)
+                            lastUse.remove(r)
+                        }
+                    }
+                }
+            }
+
+            // Registers still live at start of chunk are live-in
+            for (r in liveSet) {
+                val end = lastUse.getOrElse(r) { chunkStart }
+                registerIntervals.getOrPut(r) { mutableListOf() }
+                    .add(Interval(r, chunkStart, end, registerTypes.getOrElse(r) { IRDataType.BYTE }))
+            }
+        }
+
+        if (registerIntervals.isEmpty())
+            return
+
+        // Merge adjacent/overlapping intervals for each register.
+        // Also merge non-overlapping intervals of the same register into a single
+        // contiguous range, because the value in the slot must persist across the
+        // gap to the next use (otherwise another register packed to the same slot
+        // would clobber it between intervals).
+        val mergedIntervals = mutableListOf<Interval>()
+        val skipRegs = mutableSetOf<VirtualRegister>()
+        for ((reg, intervals) in registerIntervals) {
+            val sorted = intervals.sortedBy { it.start }
+            var current = sorted.first()
+            for (next in sorted.drop(1)) {
+                if (current.type != next.type && !typesCompatible(current.type, next.type)) {
+                    skipRegs.add(reg)
+                }
+                // Always merge: same register's intervals are merged into a single
+                // contiguous range to preserve the value across gaps.
+                current = Interval(reg, current.start, maxOf(current.end, next.end), current.type)
+            }
+            mergedIntervals.add(current)
+        }
+        if (skipRegs.isNotEmpty()) {
+            mergedIntervals.removeAll { it.register in skipRegs }
+        }
+
+        val conflictGraph = buildConflictGraph(sub)
+        val packing = greedyColor(mergedIntervals, conflictGraph, allRegTypes, globalSlotTypes, startSlot)
+
+        if (packing.isNotEmpty())
+            rewrite(sub, packing)
+    }
+
+    private fun buildConflictGraph(sub: IRSubroutine): Map<VirtualRegister, Set<VirtualRegister>> {
+        val conflicts = mutableMapOf<VirtualRegister, MutableSet<VirtualRegister>>()
+        for (chunk in sub.chunks.filterIsInstance<IRCodeChunk>()) {
+            val typeRegs = mutableMapOf<IRDataType, MutableSet<VirtualRegister>>()
+            for (instr in chunk.instructions) {
+                val accesses = instr.registerAccesses
+                for (access in accesses)
+                    typeRegs.getOrPut(access.type) { mutableSetOf() }.add(access.register)
+
+                // Registers accessed together within a single instruction must not share a slot
+                // (generalizes the old reg1/reg2-only conflict rule to every register access of
+                // the instruction, now that they are all enumerable via registerAccesses).
+                val regsInInstr = accesses.map { it.register }.distinct()
+                for (i in regsInInstr.indices) {
+                    for (j in i + 1 until regsInInstr.size) {
+                        val r1 = regsInInstr[i]
+                        val r2 = regsInInstr[j]
+                        conflicts.getOrPut(r1) { mutableSetOf() }.add(r2)
+                        conflicts.getOrPut(r2) { mutableSetOf() }.add(r1)
+                    }
+                }
+            }
+            // Add conflicts between registers with incompatible types in the same chunk
+            val typeList = typeRegs.entries.toList()
+            for (i in typeList.indices) {
+                for (j in i + 1 until typeList.size) {
+                    val (t1, regs1) = typeList[i]
+                    val (t2, regs2) = typeList[j]
+                    if (t1 != t2 && !typesCompatible(t1, t2)) {
+                        for (r1 in regs1) {
+                            for (r2 in regs2) {
+                                if (r1 != r2) {
+                                    conflicts.getOrPut(r1) { mutableSetOf() }.add(r2)
+                                    conflicts.getOrPut(r2) { mutableSetOf() }.add(r1)
+                                }
+                            }
+                        }
+                    }
+                }
+            }
+        }
+        return conflicts
+    }
+
+    private fun buildCFG(sub: IRSubroutine, irProg: IRProgram): Map<IRCodeChunkBase, List<IRCodeChunkBase>> {
+        val successors = mutableMapOf<IRCodeChunkBase, MutableList<IRCodeChunkBase>>()
+
+        val conditionals = setOf(
+            Opcode.BSTCC, Opcode.BSTCS, Opcode.BSTEQ, Opcode.BSTNE,
+            Opcode.BSTNEG, Opcode.BSTPOS, Opcode.BSTVC, Opcode.BSTVS,
+            Opcode.BGTR, Opcode.BGT, Opcode.BLT,
+            Opcode.BGTSR, Opcode.BGTS, Opcode.BLTS,
+            Opcode.BGER, Opcode.BGE, Opcode.BLE,
+            Opcode.BGESR, Opcode.BGES, Opcode.BLES
+        )
+
+        for (chunk in sub.chunks) {
+            val succ = mutableListOf<IRCodeChunkBase>()
+
+            if (chunk is IRCodeChunk) {
+                val lastInstr = chunk.instructions.lastOrNull()
+                val endsWithUnconditional = lastInstr != null && lastInstr.opcode in OpcodesThatBranchUnconditionally
+                val endsWithConditional = lastInstr != null && lastInstr.opcode in conditionals
+
+                if (!endsWithUnconditional) {
+                    chunk.next?.let { succ.add(it) }
+                }
+                if (endsWithConditional) {
+                    lastInstr.codeTarget?.let { irProg.resolveCodeTarget(it) }?.let { target ->
+                        if (!succ.contains(target))
+                            succ.add(target)
+                    }
+                } else if (lastInstr?.opcode == Opcode.JUMP) {
+                    succ.clear()
+                    lastInstr.codeTarget?.let { irProg.resolveCodeTarget(it) }?.let { succ.add(it) }
+                }
+            } else {
+                chunk.next?.let { succ.add(it) }
+            }
+
+            successors[chunk] = succ
+        }
+
+        return successors
+    }
+
+    private fun computeLiveness(
+        sub: IRSubroutine,
+        successors: Map<IRCodeChunkBase, List<IRCodeChunkBase>>
+    ): Pair<Map<IRCodeChunk, Set<VirtualRegister>>, Map<IRCodeChunk, Set<VirtualRegister>>> {
+
+        val liveIn = mutableMapOf<IRCodeChunk, MutableSet<VirtualRegister>>()
+        val liveOut = mutableMapOf<IRCodeChunk, MutableSet<VirtualRegister>>()
+        val gen = mutableMapOf<IRCodeChunk, MutableSet<VirtualRegister>>()
+        val kill = mutableMapOf<IRCodeChunk, MutableSet<VirtualRegister>>()
+
+        for (chunk in sub.chunks) {
+            if (chunk !is IRCodeChunk)
+                continue
+
+            val genSet = mutableSetOf<VirtualRegister>()
+            val killSet = mutableSetOf<VirtualRegister>()
+
+            for (instr in chunk.instructions) {
+                val written = instr.definitions
+                val read = instr.uses
+                for (r in read) {
+                    if (r !in killSet)
+                        genSet.add(r)
+                }
+                for (r in written) {
+                    killSet.add(r)
+                }
+            }
+
+            gen[chunk] = genSet
+            kill[chunk] = killSet
+            liveIn[chunk] = mutableSetOf()
+            liveOut[chunk] = mutableSetOf()
+        }
+
+        var changed = true
+
+        while (changed) {
+            changed = false
+
+            for (chunk in sub.chunks.reversed()) {
+                if (chunk !is IRCodeChunk)
+                    continue
+
+                val newLiveOut = mutableSetOf<VirtualRegister>()
+                for (succ in successors[chunk].orEmpty()) {
+                    if (succ is IRCodeChunk)
+                        newLiveOut.addAll(liveIn[succ].orEmpty())
+                }
+
+                if (newLiveOut != liveOut[chunk]) {
+                    liveOut[chunk] = newLiveOut
+                    changed = true
+                }
+
+                val newLiveIn = mutableSetOf<VirtualRegister>()
+                newLiveIn.addAll(gen[chunk].orEmpty())
+                newLiveIn.addAll(liveOut[chunk].orEmpty() - kill[chunk].orEmpty())
+
+                if (newLiveIn != liveIn[chunk]) {
+                    liveIn[chunk] = newLiveIn
+                    changed = true
+                }
+            }
+        }
+
+        return liveIn to liveOut
+    }
+
+    private fun greedyColor(
+        intervals: List<Interval>,
+        conflictGraph: Map<VirtualRegister, Set<VirtualRegister>>,
+        allRegTypes: Map<VirtualRegister, IRDataType>,
+        slotTypes: MutableMap<VirtualRegister, IRDataType>,
+        startSlot: Int
+    ): Map<VirtualRegister, VirtualRegister> {
+        if (intervals.isEmpty())
+            return emptyMap()
+
+        val packing = mutableMapOf<VirtualRegister, VirtualRegister>()
+        // Reserve slots for non-packed registers
+        val packedRegs = intervals.map { it.register }.toSet()
+        for ((reg, type) in allRegTypes) {
+            if (reg !in packedRegs)
+                slotTypes.putIfAbsent(reg, type)
+        }
+
+        val sorted = intervals.sortedWith(compareBy<Interval> { it.type.ordinal }.thenBy { it.start })
+        val activeSlots = mutableMapOf<VirtualRegister, Pair<VirtualRegister, Int>>()
+
+        for (interval in sorted) {
+            activeSlots.entries.removeAll { it.value.second < interval.start }
+
+            var slotNum = startSlot
+            var slot: VirtualRegister
+            while (true) {
+                // a packed slot stays in the same register file (int/float) as the register it replaces
+                slot = if (interval.register is VirtualRegister.FloatReg) VirtualRegister.float(slotNum) else VirtualRegister.int(slotNum)
+                if (slot !in activeSlots) {
+                    val existingType = slotTypes[slot]
+                    if (existingType == null || existingType == interval.type || typesCompatible(existingType, interval.type)) {
+                        if (!conflictsWithRegister(slot, interval.register, conflictGraph, packing))
+                            break
+                    }
+                }
+                slotNum++
+            }
+
+            packing[interval.register] = slot
+            slotTypes.putIfAbsent(slot, interval.type)  // keep the first-assigned type to prevent type narrowing via typesCompatible chain
+            activeSlots[slot] = Pair(interval.register, interval.end)
+        }
+        return packing
+    }
+
+    private fun typesCompatible(t1: IRDataType, t2: IRDataType): Boolean {
+        return t1 == t2 || (t1 == IRDataType.POINTER && t2 in setOf(IRDataType.WORD, IRDataType.LONG)) ||
+            (t2 == IRDataType.POINTER && t1 in setOf(IRDataType.WORD, IRDataType.LONG))
+    }
+
+    private fun conflictsWithRegister(
+        slot: VirtualRegister,
+        register: VirtualRegister,
+        conflictGraph: Map<VirtualRegister, Set<VirtualRegister>>,
+        packing: Map<VirtualRegister, VirtualRegister>
+    ): Boolean {
+        val conflictingRegs = conflictGraph[register] ?: return false
+        for (cr in conflictingRegs) {
+            val crSlot = packing[cr]
+            if (crSlot != null && crSlot == slot)
+                return true
+            if (cr !in packing && cr == slot)
+                return true
+        }
+        return false
+    }
+
+    private fun rewrite(sub: IRSubroutine, packing: Map<VirtualRegister, VirtualRegister>) {
+        for (chunk in sub.chunks) {
+            if (chunk !is IRCodeChunk)
+                continue
+
+            for (i in chunk.instructions.indices) {
+                val instr = chunk.instructions[i]
+                val newInstr = instr.mapRegisters { vr -> packing[vr] ?: vr }
+                if (newInstr != instr)
+                    chunk.instructions[i] = newInstr
+            }
+        }
+    }
+}

codeGenIntermediate/src/prog8/codegen/intermediate/RegisterPool.kt

@@ -14,6 +14,11 @@ internal class RegisterPool {
     private val registerTypes: MutableMap<RegisterNum, IRDataType> = mutableMapOf()
 
     fun getTypes(): Map<RegisterNum, IRDataType> = registerTypes
+    fun resetTypes(types: Map<RegisterNum, IRDataType>) {
+        // called after register packing
+        registerTypes.clear()
+        registerTypes.putAll(types)
+    }
 
     init {
         for(i in 99000..99099)

codeGenIntermediate/src/prog8/codegen/intermediate/StConvert.kt

@@ -2,8 +2,8 @@ package prog8.codegen.intermediate
 
 import prog8.code.*
 import prog8.code.ast.PtStructField
+import prog8.code.core.Position
 import prog8.intermediate.*
-import kotlin.collections.flatMap
 
 /**
  * Converter object for transforming SymbolTable nodes to IR symbol table nodes.
@@ -11,7 +11,7 @@ import kotlin.collections.flatMap
  */
 private class StToIrConverter(val romable: Boolean) {
     fun convert(struct: StStruct): IRStStructDef =
-        IRStStructDef(struct.scopedNameString, makeIrStructFieldsFrom(struct.fields), struct.size)
+        IRStStructDef(struct.scopedNameString, makeIrStructFieldsFrom(struct.fields), struct.size, struct.isUnion)
 
     private fun makeIrStructFieldsFrom(fields: List<PtStructField>): List<IRStStructField> =
         fields.map { IRStStructField(it.type, it.name, it.arraySize) }
@@ -77,9 +77,9 @@ private class StToIrConverter(val romable: Boolean) {
 
         val name = if('.' in variable.name) variable.name else variable.scopedNameString
         val hasInit = initValue != null
-        val inBss = !hasInit || (romable && hasInit)
+        val inBss = !hasInit || romable
         val readonly = romable && hasInit
-        return IRStStaticVariable(name, variable.dt, initValue, variable.length, variable.zpwish, variable.align, variable.dirty, inBss, readonly)
+        return IRStStaticVariable(name, variable.dt, initValue, variable.length, variable.zpwish, variable.align, variable.dirty, inBss, readonly, variable.shared, variable.astNode?.position ?: Position.DUMMY)
     }
 
     fun convert(variable: StMemVar): IRStMemVar {
@@ -112,13 +112,14 @@ private class StToIrConverter(val romable: Boolean) {
                 constant.name
             }
         }
+        val noPrefix = constant.astNode?.definingBlock()?.options?.noSymbolPrefixing == true
         if(constant.value != null)
-            return IRStConstant(scopedName, constant.dt, constant.value)
+            return IRStConstant(scopedName, constant.dt, constant.value, noPrefix = noPrefix)
         else if(constant.memorySlab != null) {
             // For memory() constants, store reference to slab name
             // The actual address will be resolved during code generation
             val slabName = constant.memorySlab!!.name
-            return IRStConstant(scopedName, constant.dt, null, slabName)
+            return IRStConstant(scopedName, constant.dt, null, slabName, noPrefix)
         }
         else
             TODO("constant without value or memory slab: $constant")

codeGenIntermediate/src/prog8/codegen/intermediate/StructArrayFolding.kt

@@ -0,0 +1,54 @@
+package prog8.codegen.intermediate
+
+import prog8.code.ast.*
+
+internal data class StructArrayIndexInfo(
+    val arrayName: String,
+    val idxExpr: PtExpression,
+    val structSize: Int,
+    val fieldOffset: Int
+)
+
+internal fun extractStructArrayIndexInfo(
+    addressExpr: PtExpression,
+    codeGen: IRCodeGen
+): StructArrayIndexInfo? {
+    fun unwrapCast(expr: PtExpression): PtExpression {
+        var cur = expr
+        while(cur is PtTypeCast) cur = cur.value
+        return cur
+    }
+    fun asAddressOf(expr: PtExpression): PtAddressOf? {
+        return unwrapCast(expr) as? PtAddressOf
+    }
+    val baseAddrOf: PtAddressOf
+    val fieldOffset: Int
+    val unwrapped = unwrapCast(addressExpr)
+    when (unwrapped) {
+        is PtAddressOf -> {
+            baseAddrOf = unwrapped
+            fieldOffset = 0
+        }
+
+        is PtBinaryExpression if unwrapped.operator == "+" -> {
+            val leftAddr = asAddressOf(unwrapped.left) ?: return null
+            val rightNum = (unwrapCast(unwrapped.right) as? PtNumber)?.number?.toInt() ?: return null
+            baseAddrOf = leftAddr
+            fieldOffset = rightNum
+        }
+
+        else -> return null
+    }
+    if(!baseAddrOf.isFromArrayElement) return null
+    val arrayIdent = baseAddrOf.identifier ?: return null
+    val idxExpr = baseAddrOf.arrayIndexExpr ?: return null
+    if(idxExpr is PtNumber) return null
+    val arrayType = arrayIdent.type
+    if(!arrayType.isArray) return null
+    val elemType = arrayType.elementType()
+    if(!elemType.isStructInstance) return null
+    val structSize = codeGen.program.memsizer.memorySize(elemType, null)
+    if(structSize !in 1..65535) return null
+    if(fieldOffset !in 0..65535) return null
+    return StructArrayIndexInfo(arrayIdent.name, idxExpr, structSize, fieldOffset)
+}

codeGenIntermediate/src/prog8/codegen/intermediate/SymbolPrefixer.kt

@@ -2,7 +2,10 @@ package prog8.codegen.intermediate
 
 import prog8.code.*
 import prog8.code.ast.*
-import prog8.code.core.*
+import prog8.code.core.AssemblyError
+import prog8.code.core.CompilationOptions
+import prog8.code.core.DataType
+import prog8.code.core.ISubType
 
 
 /**
@@ -76,7 +79,7 @@ fun prefixSymbols(program: PtProgram, options: CompilationOptions, st: SymbolTab
                     }
                 }
                 var lookupName = node.name
-                if(node.type.isSplitWordArray && (lookupName.endsWith("_lsb") || lookupName.endsWith("_msb"))) {
+                if(node.type.isSplitWordArray(program.memsizer) && (lookupName.endsWith("_lsb") || lookupName.endsWith("_msb"))) {
                     lookupName = lookupName.dropLast(4)
                 }
                 val stNode = st.lookup(lookupName) ?:
@@ -120,7 +123,7 @@ fun prefixSymbols(program: PtProgram, options: CompilationOptions, st: SymbolTab
         }
         else if (node is PtIdentifier) {
             var lookupName = node.name
-            if(node.type.isSplitWordArray && (lookupName.endsWith("_lsb") || lookupName.endsWith("_msb"))) {
+            if(node.type.isSplitWordArray(program.memsizer) && (lookupName.endsWith("_lsb") || lookupName.endsWith("_msb"))) {
                 lookupName = lookupName.dropLast(4)
             }
             val stNode = st.lookup(lookupName) ?: throw AssemblyError("unknown identifier $node")
@@ -193,7 +196,7 @@ fun prefixSymbols(program: PtProgram, options: CompilationOptions, st: SymbolTab
 }
 
 
-internal fun prefixScopedName(name: String, type: Char): String {
+internal fun prefixScopedName(name: String, type: Char, st: SymbolTable? = null): String {
     if('.' !in name) {
         if(name.startsWith(GENERATED_LABEL_PREFIX))
             return name
@@ -203,7 +206,11 @@ internal fun prefixScopedName(name: String, type: Char): String {
     val firstPrefixed = "p8b_${parts[0]}"
     val lastPart = parts.last()
     val lastPrefixed = if(lastPart.startsWith(GENERATED_LABEL_PREFIX)) lastPart else  "p8${type}_$lastPart"
-    val inbetweenPrefixed = parts.drop(1).dropLast(1).map{ "p8s_$it" }
+    val inbetweenPrefixed = parts.drop(1).dropLast(1).mapIndexed { index, part ->
+        val scopedName = parts.take(index + 2).joinToString(".")
+        val middleType = st?.lookup(scopedName)?.type ?: StNodeType.SUBROUTINE
+        "p8${typePrefixChar(middleType)}_$part"
+    }
     val prefixed = listOf(firstPrefixed) + inbetweenPrefixed + listOf(lastPrefixed)
     return prefixed.joinToString(".")
 }
@@ -226,7 +233,7 @@ internal fun typePrefixChar(targetNt: StNodeType): Char {
 
 
 private fun PtVariable.prefixVariableOrThis(parent: PtNode, st: SymbolTable): PtVariable {
-    name = prefixScopedName(name, 'v')
+    name = prefixScopedName(name, 'v', st)
     if(value==null)
         return this
 
@@ -253,13 +260,25 @@ private fun PtVariable.prefixVariableOrThis(parent: PtNode, st: SymbolTable): Pt
                 is PtFunctionCall if elt.builtin -> {
                     if (elt.name != "prog8_lib_structalloc")
                         throw AssemblyError("weird array value element $elt")
-                    else
+                    else {
+                        // The struct initializer's arguments are not reachable via children, so
+                        // collectRefs never visits them; prefix the address-of ones here.
+                        elt.children.forEachIndexed { idx, arg ->
+                            if (arg is PtAddressOf && arg.definingBlock()?.options?.noSymbolPrefixing != true) {
+                                val newAddr = PtAddressOf(arg.type, false, arg.position)
+                                newAddr.add(arg.identifier!!.prefixIdentifierOrThis(newAddr, st))
+                                if (arg.arrayIndexExpr != null)
+                                    newAddr.add(arg.arrayIndexExpr!!)
+                                elt.setChild(idx, newAddr)
+                            }
+                        }
                         newValue.add(elt)
+                    }
                 }
                 else -> throw AssemblyError("weird array value element $elt")
             }
         }
-        val result = PtVariable(name, type, zeropage, align, dirty, newValue, arraySize, position)
+        val result = PtVariable(name, type, isSplitWordArray, zeropage, align, dirty, newValue, arraySize, position, shared)
         result.parent = parent
         result
     }
@@ -287,7 +306,7 @@ private fun PtIdentifier.prefixIdentifierOrThis(parent: PtNode, st: SymbolTable)
     }
 
     val prefixType = typePrefixChar(targetNt)
-    val newName = prefixScopedName(name, prefixType)
+    val newName = prefixScopedName(name, prefixType, st)
     val node = PtIdentifier(newName, type, position)
     node.parent = parent
     return node

codeGenIntermediate/src/prog8/codegen/vm/VmCodeGen.kt

@@ -1,14 +1,14 @@
 package prog8.codegen.vm
 
-import prog8.code.IAssemblyProgram
 import prog8.code.ICodeGeneratorBackend
 import prog8.code.SymbolTable
+import prog8.code.assembly.IAssemblyProgram
 import prog8.code.ast.PtProgram
 import prog8.code.core.CompilationOptions
 import prog8.code.core.IErrorReporter
+import prog8.code.core.OutputFiles
 import prog8.codegen.intermediate.IRCodeGen
-import prog8.intermediate.IRFileWriter
-import prog8.intermediate.IRProgram
+import prog8.intermediate.*
 
 class VmCodeGen(val retainSSA: Boolean,
                 private val preassignedCallSiteIds: Map<String, UByte> = emptyMap()
@@ -22,7 +22,14 @@ class VmCodeGen(val retainSSA: Boolean,
         val irCodeGen = IRCodeGen(program, symbolTable, options, errors, retainSSA, preassignedCallSiteIds)
         val irProgram = irCodeGen.generate()
 
-        irProgram.verifyRegisterTypes(irCodeGen.registerTypes())
+        val virtualRegisterTypes = irCodeGen.registerTypes().entries.associate { (num, type) ->
+            val register: VirtualRegister = if (type == IRDataType.FLOAT) VirtualRegister.float(num.value) else VirtualRegister.int(num.value)
+            register to type
+        }
+        irProgram.verifyRegisterTypes(virtualRegisterTypes)
+
+        if (options.dumpVariables)
+            dumpVariables(irProgram)
 
         return VmAssemblyProgram(irProgram.name, irProgram)
     }
@@ -44,8 +51,9 @@ internal class VmAssemblyProgram(
         get() = irProgram.countUsedRegisters()
 
     override fun assemble(options: CompilationOptions, errors: IErrorReporter): Boolean {
-        // the VM reads the IR file from disk.
-        IRFileWriter(irProgram, null).write()
+        // the VM reads the IR file from disk; for the virtual target the IR file is the program artifact.
+        val programFile = OutputFiles.of(options, irProgram.name).programFile
+        IRFileWriter(irProgram, programFile).write()
         return true
     }
 }

codeGenIntermediate/test/TestIRPeepholeOpt.kt

@@ -10,6 +10,8 @@ import prog8.intermediate.*
 import prog8tests.helpers.ErrorReporterForTests
 
 class TestIRPeepholeOpt: FunSpec({
+    val indexRegType = VMTarget().indexRegType
+
     fun makeIRProgram(chunks: List<IRCodeChunkBase>): IRProgram {
         require(chunks.first().label=="p8b_main.p8s_start")
         val block = IRBlock("p8b_main", false, IRBlock.Options(), Position.DUMMY)
@@ -40,9 +42,9 @@ class TestIRPeepholeOpt: FunSpec({
 
     test("remove nops") {
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=1, immediate=42),
-            IRInstruction(Opcode.NOP),
-            IRInstruction(Opcode.NOP)
+            IRInstructions.load(IRDataType.BYTE, 1, 42),
+            IRInstructions.simple(Opcode.NOP),
+            IRInstructions.simple(Opcode.NOP)
         ))
         irProg.chunks().single().instructions.size shouldBe 3
         val opt = IRPeepholeOptimizer(irProg, false)
@@ -52,13 +54,13 @@ class TestIRPeepholeOpt: FunSpec({
 
     test("remove jmp to label below but keep labels") {
         val c1 = IRCodeChunk("p8b_main.p8s_start", null)
-        c1 += IRInstruction(Opcode.JUMP, labelSymbol = "label")
+        c1 += IRInstructions.jump(codeLabel("label"))
         val c2 = IRCodeChunk("label", null)
-        c2 += IRInstruction(Opcode.JUMP, labelSymbol = "label2")
-        c2 += IRInstruction(Opcode.NOP)  // removed
+        c2 += IRInstructions.jump(codeLabel("label2"))
+        c2 += IRInstructions.simple(Opcode.NOP)  // removed
         val c3 = IRCodeChunk("label2", null)
-        c3 += IRInstruction(Opcode.JUMP, labelSymbol = "label3")
-        c3 += IRInstruction(Opcode.INC, IRDataType.BYTE, reg1=1)
+        c3 += IRInstructions.jump(codeLabel("label3"))
+        c3 += IRInstructions.unary(Opcode.INC, IRDataType.BYTE, 1)
         val c4 = IRCodeChunk("label3", null)
         val irProg = makeIRProgram(listOf(c1, c2, c3, c4))
 
@@ -84,18 +86,18 @@ class TestIRPeepholeOpt: FunSpec({
 
     test("remove double sec/clc/sei/cli") {
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.SEC),
-            IRInstruction(Opcode.SEC),
-            IRInstruction(Opcode.SEC),
-            IRInstruction(Opcode.CLC),
-            IRInstruction(Opcode.CLC),
-            IRInstruction(Opcode.CLC),
-            IRInstruction(Opcode.SEI),
-            IRInstruction(Opcode.SEI),
-            IRInstruction(Opcode.SEI),
-            IRInstruction(Opcode.CLI),
-            IRInstruction(Opcode.CLI),
-            IRInstruction(Opcode.CLI),
+            IRInstructions.simple(Opcode.SEC),
+            IRInstructions.simple(Opcode.SEC),
+            IRInstructions.simple(Opcode.SEC),
+            IRInstructions.simple(Opcode.CLC),
+            IRInstructions.simple(Opcode.CLC),
+            IRInstructions.simple(Opcode.CLC),
+            IRInstructions.simple(Opcode.SEI),
+            IRInstructions.simple(Opcode.SEI),
+            IRInstructions.simple(Opcode.SEI),
+            IRInstructions.simple(Opcode.CLI),
+            IRInstructions.simple(Opcode.CLI),
+            IRInstructions.simple(Opcode.CLI),
         ))
         irProg.chunks().single().instructions.size shouldBe 12
         val opt = IRPeepholeOptimizer(irProg, false)
@@ -108,18 +110,18 @@ class TestIRPeepholeOpt: FunSpec({
 
     test("remove double sec/clc/sei/cli reversed") {
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.CLC),
-            IRInstruction(Opcode.CLC),
-            IRInstruction(Opcode.CLC),
-            IRInstruction(Opcode.SEC),
-            IRInstruction(Opcode.SEC),
-            IRInstruction(Opcode.SEC),
-            IRInstruction(Opcode.CLI),
-            IRInstruction(Opcode.CLI),
-            IRInstruction(Opcode.CLI),
-            IRInstruction(Opcode.SEI),
-            IRInstruction(Opcode.SEI),
-            IRInstruction(Opcode.SEI),
+            IRInstructions.simple(Opcode.CLC),
+            IRInstructions.simple(Opcode.CLC),
+            IRInstructions.simple(Opcode.CLC),
+            IRInstructions.simple(Opcode.SEC),
+            IRInstructions.simple(Opcode.SEC),
+            IRInstructions.simple(Opcode.SEC),
+            IRInstructions.simple(Opcode.CLI),
+            IRInstructions.simple(Opcode.CLI),
+            IRInstructions.simple(Opcode.CLI),
+            IRInstructions.simple(Opcode.SEI),
+            IRInstructions.simple(Opcode.SEI),
+            IRInstructions.simple(Opcode.SEI),
         ))
         irProg.chunks().single().instructions.size shouldBe 12
         val opt = IRPeepholeOptimizer(irProg, false)
@@ -132,10 +134,10 @@ class TestIRPeepholeOpt: FunSpec({
 
     test("push followed by pop") {
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.PUSH, IRDataType.BYTE, reg1=42),
-            IRInstruction(Opcode.POP, IRDataType.BYTE, reg1=42),
-            IRInstruction(Opcode.PUSH, IRDataType.BYTE, reg1=99),
-            IRInstruction(Opcode.POP, IRDataType.BYTE, reg1=222)
+            IRInstructions.push(IRDataType.BYTE, 42),
+            IRInstructions.pop(IRDataType.BYTE, 42),
+            IRInstructions.push(IRDataType.BYTE, 99),
+            IRInstructions.pop(IRDataType.BYTE, 222)
         ))
         irProg.chunks().single().instructions.size shouldBe 4
         val opt = IRPeepholeOptimizer(irProg, false)
@@ -143,23 +145,23 @@ class TestIRPeepholeOpt: FunSpec({
         val instr = irProg.chunks().single().instructions
         instr.size shouldBe 1
         instr[0].opcode shouldBe Opcode.LOADR
-        instr[0].reg1 shouldBe 222
-        instr[0].reg2 shouldBe 99
+        instr[0].requireDest().register shouldBe VirtualRegister.int(222)
+        instr[0].requireSrcA().register shouldBe VirtualRegister.int(99)
     }
 
     test("remove useless div/mul, add/sub") {
         // Use different registers for each test case to avoid dead store elimination removing them
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.DIV, IRDataType.BYTE, reg1=1, immediate = 1),
-            IRInstruction(Opcode.DIVS, IRDataType.BYTE, reg1=2, immediate = 1),
-            IRInstruction(Opcode.MUL, IRDataType.BYTE, reg1=3, immediate = 1),
-            IRInstruction(Opcode.MOD, IRDataType.BYTE, reg1=4, immediate = 1),
-            IRInstruction(Opcode.DIV, IRDataType.BYTE, reg1=5, immediate = 2),
-            IRInstruction(Opcode.DIVS, IRDataType.BYTE, reg1=6, immediate = 2),
-            IRInstruction(Opcode.MUL, IRDataType.BYTE, reg1=7, immediate = 2),
-            IRInstruction(Opcode.MOD, IRDataType.BYTE, reg1=8, immediate = 2),
-            IRInstruction(Opcode.ADD, IRDataType.BYTE, reg1=9, immediate = 0),
-            IRInstruction(Opcode.SUB, IRDataType.BYTE, reg1=10, immediate = 0)
+            IRInstructions.binaryImmediate(Opcode.DIV, IRDataType.BYTE, 1, 1),
+            IRInstructions.binaryImmediate(Opcode.DIVS, IRDataType.BYTE, 2, 1),
+            IRInstructions.binaryImmediate(Opcode.MUL, IRDataType.BYTE, 3, 1),
+            IRInstructions.binaryImmediate(Opcode.MOD, IRDataType.BYTE, 4, 1),
+            IRInstructions.binaryImmediate(Opcode.DIV, IRDataType.BYTE, 5, 2),
+            IRInstructions.binaryImmediate(Opcode.DIVS, IRDataType.BYTE, 6, 2),
+            IRInstructions.binaryImmediate(Opcode.MUL, IRDataType.BYTE, 7, 2),
+            IRInstructions.binaryImmediate(Opcode.MOD, IRDataType.BYTE, 8, 2),
+            IRInstructions.binaryImmediate(Opcode.ADD, IRDataType.BYTE, 9, 0),
+            IRInstructions.binaryImmediate(Opcode.SUB, IRDataType.BYTE, 10, 0)
         ))
         irProg.chunks().single().instructions.size shouldBe 10
         val opt = IRPeepholeOptimizer(irProg, false)
@@ -168,11 +170,66 @@ class TestIRPeepholeOpt: FunSpec({
         irProg.chunks().single().instructions.size shouldBe 4
     }
 
+    test("replace integer multiply by zero with load") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.binaryImmediate(Opcode.MUL, IRDataType.BYTE, 1, 0),
+            IRInstructions.binaryImmediate(Opcode.MULS, IRDataType.WORD, 2, 0),
+            IRInstructions.binaryImmediate(Opcode.MULS, IRDataType.LONG, 3, 0),
+            IRInstructions.binaryImmediateFloat(Opcode.MUL, 4, 0.0),
+            IRInstructions.returnVoid()
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.count { it.opcode == Opcode.LOAD } shouldBe 3
+        instr.count { it.opcode == Opcode.MUL } shouldBe 1
+    }
+
+    test("fold adjacent integer immediate multiplications") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.binaryImmediate(Opcode.MUL, IRDataType.BYTE, 1, 7),
+            IRInstructions.binaryImmediate(Opcode.MUL, IRDataType.BYTE, 1, 9),
+            IRInstructions.binaryImmediate(Opcode.MULS, IRDataType.WORD, 2, -3),
+            IRInstructions.binaryImmediate(Opcode.MULS, IRDataType.WORD, 2, 4),
+            IRInstructions.binaryImmediate(Opcode.MULS, IRDataType.LONG, 3, 100000),
+            IRInstructions.binaryImmediate(Opcode.MULS, IRDataType.LONG, 3, 30000),
+            IRInstructions.returnVoid()
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.size shouldBe 4
+        instr[0].requireImmediateInt() shouldBe 63
+        instr[1].requireImmediateInt() shouldBe 65524
+        instr[2].requireImmediateInt() shouldBe 3000000000L.toInt()
+    }
+
+    test("collapse adjacent integer and float loadr chains") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.move(IRDataType.BYTE, 2, 1),
+            IRInstructions.move(IRDataType.BYTE, 3, 2),
+            IRInstructions.move(IRDataType.FLOAT, 5, 4),
+            IRInstructions.move(IRDataType.FLOAT, 6, 5),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 3, IRMemory.direct(MemoryAddress(100u))),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.FLOAT, 6, IRMemory.direct(MemoryAddress(200u)))
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.map { it.opcode } shouldBe listOf(Opcode.LOADR, Opcode.LOADR, Opcode.STOREM, Opcode.STOREM)
+        instr[0].requireDest().register shouldBe VirtualRegister.int(3)
+        instr[0].requireSrcA().register shouldBe VirtualRegister.int(1)
+        instr[1].requireDest().register shouldBe VirtualRegister.float(6)
+        instr[1].requireSrcA().register shouldBe VirtualRegister.float(4)
+        instr[2].requireSrcA().register shouldBe VirtualRegister.int(3)
+        instr[3].requireSrcA().register shouldBe VirtualRegister.float(6)
+    }
+
     test("replace add/sub 1 by inc/dec") {
         // Use different registers for each test case to avoid dead store elimination removing them
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.ADD, IRDataType.BYTE, reg1=1, immediate = 1),
-            IRInstruction(Opcode.SUB, IRDataType.BYTE, reg1=2, immediate = 1)
+            IRInstructions.binaryImmediate(Opcode.ADD, IRDataType.BYTE, 1, 1),
+            IRInstructions.binaryImmediate(Opcode.SUB, IRDataType.BYTE, 2, 1)
         ))
         irProg.chunks().single().instructions.size shouldBe 2
         val opt = IRPeepholeOptimizer(irProg, false)
@@ -186,23 +243,23 @@ class TestIRPeepholeOpt: FunSpec({
     test("remove useless and/or/xor") {
         // Use different registers for each test case to avoid dead store elimination removing them
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.AND, IRDataType.BYTE, reg1=1, immediate = 0),
-            IRInstruction(Opcode.AND, IRDataType.WORD, reg1=2, immediate = 0),
-            IRInstruction(Opcode.AND, IRDataType.LONG, reg1=3, immediate = 0),
-            IRInstruction(Opcode.AND, IRDataType.BYTE, reg1=4, immediate = 255),
-            IRInstruction(Opcode.AND, IRDataType.WORD, reg1=5, immediate = 65535),
-            IRInstruction(Opcode.AND, IRDataType.LONG, reg1=6, immediate = 2147483647),
-            IRInstruction(Opcode.AND, IRDataType.LONG, reg1=7, immediate = -1),
-            IRInstruction(Opcode.OR, IRDataType.BYTE, reg1=8, immediate = 0),
-            IRInstruction(Opcode.OR, IRDataType.BYTE, reg1=9, immediate = 255),
-            IRInstruction(Opcode.OR, IRDataType.WORD, reg1=10, immediate = 65535),
-            IRInstruction(Opcode.OR, IRDataType.LONG, reg1=11, immediate = 2147483647),
-            IRInstruction(Opcode.OR, IRDataType.LONG, reg1=12, immediate = -1),
-            IRInstruction(Opcode.XOR, IRDataType.BYTE, reg1=13, immediate = 0),
-            IRInstruction(Opcode.AND, IRDataType.BYTE, reg1=14, immediate = 200),
-            IRInstruction(Opcode.AND, IRDataType.WORD, reg1=15, immediate = 60000),
-            IRInstruction(Opcode.OR, IRDataType.BYTE, reg1=16, immediate = 1),
-            IRInstruction(Opcode.XOR, IRDataType.BYTE, reg1=17, immediate = 1)
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.BYTE, 1, 0),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.WORD, 2, 0),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.LONG, 3, 0),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.BYTE, 4, 255),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.WORD, 5, 65535),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.LONG, 6, 2147483647),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.LONG, 7, -1),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.BYTE, 8, 0),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.BYTE, 9, 255),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.WORD, 10, 65535),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.LONG, 11, 2147483647),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.LONG, 12, -1),
+            IRInstructions.binaryImmediate(Opcode.XOR, IRDataType.BYTE, 13, 0),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.BYTE, 14, 200),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.WORD, 15, 60000),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.BYTE, 16, 1),
+            IRInstructions.binaryImmediate(Opcode.XOR, IRDataType.BYTE, 17, 1)
         ))
         irProg.chunks().single().instructions.size shouldBe 17
         val opt = IRPeepholeOptimizer(irProg, false)
@@ -223,40 +280,86 @@ class TestIRPeepholeOpt: FunSpec({
     test("replace and/or/xor by constant number") {
         // Use different registers for each test case to avoid dead store elimination removing them
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.AND, IRDataType.BYTE, reg1=1, immediate = 0),
-            IRInstruction(Opcode.AND, IRDataType.WORD, reg1=2, immediate = 0),
-            IRInstruction(Opcode.AND, IRDataType.LONG, reg1=3, immediate = 0),
-            IRInstruction(Opcode.OR, IRDataType.BYTE, reg1=4, immediate = 255),
-            IRInstruction(Opcode.OR, IRDataType.WORD, reg1=5, immediate = 65535),
-            IRInstruction(Opcode.OR, IRDataType.LONG, reg1=6, immediate = -1)
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.BYTE, 1, 0),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.WORD, 2, 0),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.LONG, 3, 0),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.BYTE, 4, 255),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.WORD, 5, 65535),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.LONG, 6, -1)
         ))
         irProg.chunks().single().instructions.size shouldBe 6
         val opt = IRPeepholeOptimizer(irProg, false)
         opt.optimize(true, ErrorReporterForTests())
         val instr = irProg.chunks().single().instructions
         instr.size shouldBe 6
+        instr.map { it.opcode } shouldBe List(6) { Opcode.LOAD }
+        instr.map { it.requireImmediateInt() } shouldBe listOf(0, 0, 0, 255, 65535, -1)
+    }
+
+    test("combine adjacent immediate and/or operations") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.BYTE, 1, 0xf0),
+            IRInstructions.binaryImmediate(Opcode.AND, IRDataType.BYTE, 1, 0xcc),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.WORD, 2, 0x0010),
+            IRInstructions.binaryImmediate(Opcode.OR, IRDataType.WORD, 2, 0x0003),
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.size shouldBe 2
+        instr[0].opcode shouldBe Opcode.AND
+        instr[0].requireImmediateInt() shouldBe 0xc0
+        instr[1].opcode shouldBe Opcode.OR
+        instr[1].requireImmediateInt() shouldBe 0x0013
+    }
+
+    test("cancel identical adjacent immediate xor operations") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.binaryImmediate(Opcode.XOR, IRDataType.BYTE, 1, 3),
+            IRInstructions.binaryImmediate(Opcode.XOR, IRDataType.BYTE, 1, 3),
+            IRInstructions.load(IRDataType.BYTE, 2, 42)
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.size shouldBe 1
         instr[0].opcode shouldBe Opcode.LOAD
-        instr[1].opcode shouldBe Opcode.LOAD
-        instr[2].opcode shouldBe Opcode.LOAD
-        instr[3].opcode shouldBe Opcode.LOAD
-        instr[4].opcode shouldBe Opcode.LOAD
-        instr[5].opcode shouldBe Opcode.LOAD
-        instr[0].immediate shouldBe 0
-        instr[1].immediate shouldBe 0
-        instr[2].immediate shouldBe 0
-        instr[3].immediate shouldBe 255
-        instr[4].immediate shouldBe 65535
-        instr[5].immediate shouldBe -1
+        instr[0].requireImmediateInt() shouldBe 42
+    }
+
+    test("fold adjacent immediate arithmetic with inc/dec") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.binaryImmediate(Opcode.ADD, IRDataType.BYTE, 1, 4),
+            IRInstructions.unary(Opcode.INC, IRDataType.BYTE, 1),
+            IRInstructions.unary(Opcode.INC, IRDataType.BYTE, 2),
+            IRInstructions.binaryImmediate(Opcode.ADD, IRDataType.BYTE, 2, 4),
+            IRInstructions.binaryImmediate(Opcode.SUB, IRDataType.BYTE, 3, 4),
+            IRInstructions.unary(Opcode.DEC, IRDataType.BYTE, 3),
+            IRInstructions.unary(Opcode.DEC, IRDataType.BYTE, 4),
+            IRInstructions.binaryImmediate(Opcode.SUB, IRDataType.BYTE, 4, 4)
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.size shouldBe 4
+        instr[0].opcode shouldBe Opcode.ADD
+        instr[0].requireImmediateInt() shouldBe 5
+        instr[1].opcode shouldBe Opcode.ADD
+        instr[1].requireImmediateInt() shouldBe 5
+        instr[2].opcode shouldBe Opcode.SUB
+        instr[2].requireImmediateInt() shouldBe 5
+        instr[3].opcode shouldBe Opcode.SUB
+        instr[3].requireImmediateInt() shouldBe 5
     }
 
     test("replace register arithmetic by immediate operations") {
         // Test the new optimization: LOAD + SUBR/DIVR/MODR → immediate operation
         // Pattern: LOAD r1, #const  followed by  SUBR r2, r1  →  SUB r2, #const
-        
+
         // SUBR optimization
         val subrProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=1, immediate = 5),
-            IRInstruction(Opcode.SUBR, IRDataType.BYTE, reg1=2, reg2=1)
+            IRInstructions.load(IRDataType.BYTE, 1, 5),
+            IRInstructions.binary(Opcode.SUBR, IRDataType.BYTE, 2, 1)
         ))
         subrProg.chunks().single().instructions.size shouldBe 2
         val opt1 = IRPeepholeOptimizer(subrProg, false)
@@ -264,12 +367,12 @@ class TestIRPeepholeOpt: FunSpec({
         val subInstr = subrProg.chunks().single().instructions
         subInstr.size shouldBe 1
         subInstr[0].opcode shouldBe Opcode.SUB
-        subInstr[0].immediate shouldBe 5
-        
+        subInstr[0].requireImmediateInt() shouldBe 5
+
         // DIVR optimization
         val divrProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.LOAD, IRDataType.WORD, reg1=10, immediate = 100),
-            IRInstruction(Opcode.DIVR, IRDataType.WORD, reg1=20, reg2=10)
+            IRInstructions.load(IRDataType.WORD, 10, 100),
+            IRInstructions.binary(Opcode.DIVR, IRDataType.WORD, 20, 10)
         ))
         divrProg.chunks().single().instructions.size shouldBe 2
         val opt2 = IRPeepholeOptimizer(divrProg, false)
@@ -277,12 +380,12 @@ class TestIRPeepholeOpt: FunSpec({
         val divInstr = divrProg.chunks().single().instructions
         divInstr.size shouldBe 1
         divInstr[0].opcode shouldBe Opcode.DIV
-        divInstr[0].immediate shouldBe 100
-        
+        divInstr[0].requireImmediateInt() shouldBe 100
+
         // MODR optimization
         val modrProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=30, immediate = 7),
-            IRInstruction(Opcode.MODR, IRDataType.BYTE, reg1=40, reg2=30)
+            IRInstructions.load(IRDataType.BYTE, 30, 7),
+            IRInstructions.binary(Opcode.MODR, IRDataType.BYTE, 40, 30)
         ))
         modrProg.chunks().single().instructions.size shouldBe 2
         val opt3 = IRPeepholeOptimizer(modrProg, false)
@@ -290,34 +393,42 @@ class TestIRPeepholeOpt: FunSpec({
         val modInstr = modrProg.chunks().single().instructions
         modInstr.size shouldBe 1
         modInstr[0].opcode shouldBe Opcode.MOD
-        modInstr[0].immediate shouldBe 7
+        modInstr[0].requireImmediateInt() shouldBe 7
     }
 
     test("remove self identity operations") {
-        // LOADR with same fp register is allowed (constructor only blocks integer reg1==reg2)
+        // The IR is not SSA: a written register may also be a read-only source register.
+        // Self-identity LOADR/ANDR/ORR become NOP, self-identity XORR becomes load 0.
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.LOADR, IRDataType.FLOAT, fpReg1=RegisterNum(10), fpReg2=RegisterNum(10)),
-            IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=99, immediate=42)       // keep chunk non-empty
+            IRInstructions.move(IRDataType.FLOAT, 10, 10),
+            IRInstructions.move(IRDataType.BYTE, 11, 11),
+            IRInstructions.binary(Opcode.ANDR, IRDataType.BYTE, 12, 12),
+            IRInstructions.binary(Opcode.ORR, IRDataType.BYTE, 13, 13),
+            IRInstructions.binary(Opcode.XORR, IRDataType.BYTE, 14, 14),
+            IRInstructions.load(IRDataType.BYTE, 99, 42)       // keep chunk non-empty
         ))
-        irProg.chunks().single().instructions.size shouldBe 2
+        irProg.chunks().single().instructions.size shouldBe 6
         val opt = IRPeepholeOptimizer(irProg, false)
         opt.optimize(true, ErrorReporterForTests())
-        val instr = irProg.chunks().single().instructions
-        instr.size shouldBe 1
+        val instr = irProg.chunks().single().instructions.filter { it.opcode != Opcode.NOP }
+        instr.size shouldBe 2
         instr[0].opcode shouldBe Opcode.LOAD
-        instr[0].reg1 shouldBe 99
-        instr[0].immediate shouldBe 42
+        instr[0].requireDest().register shouldBe VirtualRegister.int(14)
+        instr[0].requireImmediateInt() shouldBe 0
+        instr[1].opcode shouldBe Opcode.LOAD
+        instr[1].requireDest().register shouldBe VirtualRegister.int(99)
+        instr[1].requireImmediateInt() shouldBe 42
     }
 
     test("remove shift by zero") {
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=10, immediate=0),
-            IRInstruction(Opcode.LSLN, IRDataType.BYTE, reg1=1, reg2=10),
-            IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=11, immediate=0),
-            IRInstruction(Opcode.LSRN, IRDataType.BYTE, reg1=2, reg2=11),
-            IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=12, immediate=0),
-            IRInstruction(Opcode.ASRN, IRDataType.BYTE, reg1=3, reg2=12),
-            IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=99, immediate=42)       // keep chunk non-empty
+            IRInstructions.load(IRDataType.BYTE, 10, 0),
+            IRInstructions.binary(Opcode.LSLN, IRDataType.BYTE, 1, 10),
+            IRInstructions.load(IRDataType.BYTE, 11, 0),
+            IRInstructions.binary(Opcode.LSRN, IRDataType.BYTE, 2, 11),
+            IRInstructions.load(IRDataType.BYTE, 12, 0),
+            IRInstructions.binary(Opcode.ASRN, IRDataType.BYTE, 3, 12),
+            IRInstructions.load(IRDataType.BYTE, 99, 42)       // keep chunk non-empty
         ))
         irProg.chunks().single().instructions.size shouldBe 7
         val opt = IRPeepholeOptimizer(irProg, false)
@@ -325,25 +436,68 @@ class TestIRPeepholeOpt: FunSpec({
         val instr = irProg.chunks().single().instructions
         instr.size shouldBe 1
         instr[0].opcode shouldBe Opcode.LOAD
-        instr[0].reg1 shouldBe 99
-        instr[0].immediate shouldBe 42
+        instr[0].requireDest().register shouldBe VirtualRegister.int(99)
+        instr[0].requireImmediateInt() shouldBe 42
+    }
+
+    test("replace concat with ext after harmless instruction") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.load(IRDataType.BYTE, 10, 0),
+            IRInstructions.unary(Opcode.INC, IRDataType.BYTE, 20),
+            IRInstructions.concat(IRDataType.BYTE, 1, 10, 11),
+            IRInstructions.load(IRDataType.BYTE, 99, 42)
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.any {
+            it.opcode == Opcode.EXT &&
+                    it.requireDest().register == VirtualRegister.int(1) &&
+                    it.requireSrcA().register == VirtualRegister.int(11)
+        } shouldBe true
+        instr.any { it.opcode == Opcode.CONCAT } shouldBe false
+    }
+
+    test("do not replace concat after intervening write") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.load(IRDataType.BYTE, 10, 0),
+            IRInstructions.unary(Opcode.INC, IRDataType.BYTE, 10),
+            IRInstructions.concat(IRDataType.BYTE, 1, 10, 11),
+            IRInstructions.load(IRDataType.BYTE, 99, 42)
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        irProg.chunks().single().instructions.any { it.opcode == Opcode.CONCAT } shouldBe true
+    }
+
+    test("do not replace concat across control flow") {
+        val c1 = IRCodeChunk("p8b_main.p8s_start", null)
+        c1 += IRInstructions.load(IRDataType.BYTE, 10, 0)
+        c1 += IRInstructions.jump(codeLabel("after"))
+        c1 += IRInstructions.concat(IRDataType.BYTE, 1, 10, 11)
+        c1 += IRInstructions.load(IRDataType.BYTE, 99, 42)
+        val c2 = IRCodeChunk("after", null)
+        val irProg = makeIRProgram(listOf(c1, c2))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        irProg.chunks().flatMap { it.instructions }.any { it.opcode == Opcode.CONCAT } shouldBe true
     }
 
     test("cancel adjacent ops") {
         val irProg = makeIRProgram(listOf(
-            IRInstruction(Opcode.INV, IRDataType.BYTE, reg1=1),
-            IRInstruction(Opcode.INV, IRDataType.BYTE, reg1=1),
-            IRInstruction(Opcode.NEG, IRDataType.BYTE, reg1=2),
-            IRInstruction(Opcode.NEG, IRDataType.BYTE, reg1=2),
-            IRInstruction(Opcode.EXT, IRDataType.BYTE, reg1=3, reg2=33),
-            IRInstruction(Opcode.EXT, IRDataType.BYTE, reg1=3, reg2=33),
-            IRInstruction(Opcode.EXTS, IRDataType.WORD, reg1=4, reg2=44),
-            IRInstruction(Opcode.EXTS, IRDataType.WORD, reg1=4, reg2=44),
-            IRInstruction(Opcode.INC, IRDataType.BYTE, reg1=5),
-            IRInstruction(Opcode.DEC, IRDataType.BYTE, reg1=5),
-            IRInstruction(Opcode.DEC, IRDataType.BYTE, reg1=6),
-            IRInstruction(Opcode.INC, IRDataType.BYTE, reg1=6),
-            IRInstruction(Opcode.LOAD, IRDataType.BYTE, reg1=99, immediate=42)       // keep chunk non-empty
+            IRInstructions.unary(Opcode.INV, IRDataType.BYTE, 1),
+            IRInstructions.unary(Opcode.INV, IRDataType.BYTE, 1),
+            IRInstructions.unary(Opcode.NEG, IRDataType.BYTE, 2),
+            IRInstructions.unary(Opcode.NEG, IRDataType.BYTE, 2),
+            IRInstructions.binary(Opcode.EXT, IRDataType.BYTE, 3, 33),
+            IRInstructions.binary(Opcode.EXT, IRDataType.BYTE, 3, 33),
+            IRInstructions.binary(Opcode.EXTS, IRDataType.WORD, 4, 44),
+            IRInstructions.binary(Opcode.EXTS, IRDataType.WORD, 4, 44),
+            IRInstructions.unary(Opcode.INC, IRDataType.BYTE, 5),
+            IRInstructions.unary(Opcode.DEC, IRDataType.BYTE, 5),
+            IRInstructions.unary(Opcode.DEC, IRDataType.BYTE, 6),
+            IRInstructions.unary(Opcode.INC, IRDataType.BYTE, 6),
+            IRInstructions.load(IRDataType.BYTE, 99, 42)       // keep chunk non-empty
         ))
         irProg.chunks().single().instructions.size shouldBe 13
         val opt = IRPeepholeOptimizer(irProg, false)
@@ -351,12 +505,210 @@ class TestIRPeepholeOpt: FunSpec({
         val instr = irProg.chunks().single().instructions
         instr.size shouldBe 3
         instr[0].opcode shouldBe Opcode.EXT
-        instr[0].reg1 shouldBe 3
-        instr[0].reg2 shouldBe 33
+        instr[0].requireDest().register shouldBe VirtualRegister.int(3)
+        instr[0].requireSrcA().register shouldBe VirtualRegister.int(33)
         instr[1].opcode shouldBe Opcode.EXTS
-        instr[1].reg1 shouldBe 4
-        instr[1].reg2 shouldBe 44
+        instr[1].requireDest().register shouldBe VirtualRegister.int(4)
+        instr[1].requireSrcA().register shouldBe VirtualRegister.int(44)
         instr[2].opcode shouldBe Opcode.LOAD
-        instr[2].immediate shouldBe 42
+        instr[2].requireImmediateInt() shouldBe 42
+    }
+
+    test("collapse lsigb.w + ext.b into loadr.w + and.w") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.loadMemory(Opcode.LOADM, IRDataType.WORD, 7, IRMemory.direct("src")),
+            IRInstructions.binary(Opcode.LSIGB, IRDataType.WORD, 8, 7),
+            IRInstructions.binary(Opcode.EXT, IRDataType.BYTE, 9, 8),
+            IRInstructions.binary(Opcode.ADDR, IRDataType.WORD, 2, 9),
+            IRInstructions.returnVoid()
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.any { it.opcode == Opcode.LSIGB } shouldBe false
+        instr.any { it.opcode == Opcode.EXT && it.requireDest().register == VirtualRegister.int(9) } shouldBe false
+        instr.any {
+            it.opcode == Opcode.LOADR &&
+                    it.type == IRDataType.WORD &&
+                    it.requireDest().register == VirtualRegister.int(9) &&
+                    it.requireSrcA().register == VirtualRegister.int(7)
+        } shouldBe true
+        instr.any {
+            it.opcode == Opcode.AND &&
+                    it.type == IRDataType.WORD &&
+                    it.requireDest().register == VirtualRegister.int(9) &&
+                    it.requireImmediateInt() == 0xff
+        } shouldBe true
+    }
+
+    test("do not collapse lsigb.l + ext.b when source and dest types differ") {
+        // lsigb.l produces a byte from a long; ext.b widens it to word. Rewriting this as
+        // loadr + and would require accessing the long source as word, which conflicts with
+        // the register type system. The optimizer must leave the original sequence intact.
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.loadMemory(Opcode.LOADM, IRDataType.LONG, 7, IRMemory.direct("src")),
+            IRInstructions.binary(Opcode.LSIGB, IRDataType.LONG, 8, 7),
+            IRInstructions.binary(Opcode.EXT, IRDataType.BYTE, 9, 8),
+            IRInstructions.binary(Opcode.ADDR, IRDataType.WORD, 2, 9),
+            IRInstructions.returnVoid()
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.any { it.opcode == Opcode.LSIGB } shouldBe true
+        instr.any { it.opcode == Opcode.EXT && it.requireDest().register == VirtualRegister.int(9) } shouldBe true
+    }
+
+    test("coalesce redundant LOADX/STOREX to same index is removed") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.loadMemory(Opcode.LOADX, IRDataType.BYTE, 1, IRMemory.indexed("myArray", 10, indexRegType)),
+            IRInstructions.unary(Opcode.INC, IRDataType.BYTE, 99),
+            IRInstructions.storeMemory(Opcode.STOREX, IRDataType.BYTE, 1, IRMemory.indexed("myArray", 10, indexRegType)),
+            IRInstructions.returnVoid()
+        ))
+        irProg.chunks().single().instructions.count { it.opcode==Opcode.STOREX } shouldBe 1
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.count { it.opcode==Opcode.STOREX } shouldBe 0
+        instr.count { it.opcode==Opcode.LOADX } shouldBe 1
+        instr.size shouldBe 3
+    }
+
+    test("do not coalesce LOADX/STOREX when index differs") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.loadMemory(Opcode.LOADX, IRDataType.BYTE, 1, IRMemory.indexed("myArray", 10, indexRegType)),
+            IRInstructions.storeMemory(Opcode.STOREX, IRDataType.BYTE, 1, IRMemory.indexed("myArray", 11, indexRegType)),
+            IRInstructions.returnVoid()
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.count { it.opcode==Opcode.LOADX } shouldBe 1
+        instr.count { it.opcode==Opcode.STOREX } shouldBe 1
+    }
+
+    test("do not coalesce LOADX/STOREX with intervening read of value") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.loadMemory(Opcode.LOADX, IRDataType.BYTE, 1, IRMemory.indexed("myArray", 10, indexRegType)),
+            IRInstructions.binary(Opcode.ADDR, IRDataType.BYTE, 1, 2),
+            IRInstructions.storeMemory(Opcode.STOREX, IRDataType.BYTE, 1, IRMemory.indexed("myArray", 10, indexRegType)),
+            IRInstructions.returnVoid()
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.count { it.opcode==Opcode.STOREX } shouldBe 1
+    }
+
+    test("do not coalesce LOADX/STOREX across chunks single-chunk") {
+        val c1 = IRCodeChunk("p8b_main.p8s_start", null)
+        c1 += IRInstructions.loadMemory(Opcode.LOADX, IRDataType.BYTE, 1, IRMemory.indexed("myArray", 10, indexRegType))
+        val c2 = IRCodeChunk("other", null)
+        c2 += IRInstructions.storeMemory(Opcode.STOREX, IRDataType.BYTE, 1, IRMemory.indexed("myArray", 10, indexRegType))
+        c2 += IRInstructions.returnVoid()
+        val irProg = makeIRProgram(listOf(c1, c2))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val all = irProg.chunks().flatMap { it.instructions }
+        all.count { it.opcode==Opcode.LOADX } shouldBe 1
+        all.count { it.opcode==Opcode.STOREX } shouldBe 1
+    }
+
+    test("do not fold storem+loadm same symbol different offsets") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 5, IRMemory.direct("myVar", 0)),
+            IRInstructions.loadMemory(Opcode.LOADM, IRDataType.BYTE, 6, IRMemory.direct("myVar", 1)),
+            IRInstructions.returnVoid()
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.count { it.opcode==Opcode.STOREM } shouldBe 1
+        instr.count { it.opcode==Opcode.LOADM } shouldBe 1
+        instr.count { it.opcode==Opcode.LOADR } shouldBe 0
+    }
+
+    test("do not fold storem+loadm across chunks single-chunk") {
+        val c1 = IRCodeChunk("p8b_main.p8s_start", null)
+        c1 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 5, IRMemory.direct("myVar"))
+        val c2 = IRCodeChunk("other", null)
+        c2 += IRInstructions.loadMemory(Opcode.LOADM, IRDataType.BYTE, 6, IRMemory.direct("myVar"))
+        c2 += IRInstructions.returnVoid()
+        val irProg = makeIRProgram(listOf(c1, c2))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val all = irProg.chunks().flatMap { it.instructions }
+        all.count { it.opcode==Opcode.STOREM } shouldBe 1
+        all.count { it.opcode==Opcode.LOADM } shouldBe 1
+        all.count { it.opcode==Opcode.LOADR } shouldBe 0
+    }
+
+    test("do not fold storem+loadm different types") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 5, IRMemory.direct("myVar")),
+            IRInstructions.loadMemory(Opcode.LOADM, IRDataType.WORD, 6, IRMemory.direct("myVar")),
+            IRInstructions.returnVoid()
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.count { it.opcode==Opcode.STOREM } shouldBe 1
+        instr.count { it.opcode==Opcode.LOADM } shouldBe 1
+        instr.count { it.opcode==Opcode.LOADR } shouldBe 0
+    }
+
+    test("dead store analysis sees reads via secondary operand slots") {
+        // r10 is only read as the INDEX register inside the LOADX memory reference. The dead
+        // store detector must see that read, otherwise the first LOAD r10 is wrongly removed
+        // and the array would be indexed with an uninitialized register.
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.load(indexRegType, 10, 3),
+            IRInstructions.loadMemory(Opcode.LOADX, IRDataType.BYTE, 1, IRMemory.indexed("myArray", 10, indexRegType)),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 1, IRMemory.direct("sym")),
+            IRInstructions.load(indexRegType, 10, 7),
+            IRInstructions.binaryImmediate(Opcode.ADD, indexRegType, 10, 1),
+            IRInstructions.storeMemory(Opcode.STOREM, indexRegType, 10, IRMemory.direct("sym2")),
+            IRInstructions.returnVoid()
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.count { it.opcode == Opcode.LOADX } shouldBe 1
+        instr.any { it.opcode == Opcode.LOAD && it.requireImmediateInt() == 3 } shouldBe true
+    }
+
+    test("dead store to the same register is removed") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.load(IRDataType.BYTE, 1, 5),
+            IRInstructions.load(IRDataType.BYTE, 1, 10),
+            IRInstructions.unary(Opcode.INC, IRDataType.BYTE, 1),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 1, IRMemory.direct("sym")),
+            IRInstructions.returnVoid()
+        ))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val instr = irProg.chunks().single().instructions
+        instr.any { it.opcode == Opcode.LOAD && it.requireImmediateInt() == 5 } shouldBe false
+        instr.any { it.opcode == Opcode.LOAD && it.requireImmediateInt() == 10 } shouldBe true
+    }
+
+    test("loadr forwarding does not delete live-out source") {
+        // r1 is live-out to next chunk, so LOAD r1 should be kept even though r1 has single use in first chunk.
+        // The forwarder is conservative: it keeps the source live when it may be needed in a successor chunk,
+        // preventing the previous bug where the source LOAD was deleted.
+        val c1 = IRCodeChunk("p8b_main.p8s_start", null)
+        c1 += IRInstructions.load(IRDataType.BYTE, 1, 42)
+        c1 += IRInstructions.move(IRDataType.BYTE, 2, 1)
+        c1 += IRInstructions.jump(codeLabel("next"))
+        val c2 = IRCodeChunk("next", null)
+        c2 += IRInstructions.move(IRDataType.BYTE, 3, 1)
+        c2 += IRInstructions.returnVoid()
+        val irProg = makeIRProgram(listOf(c1, c2))
+        val opt = IRPeepholeOptimizer(irProg, false)
+        opt.optimize(true, ErrorReporterForTests())
+        val all = irProg.chunks().flatMap { it.instructions }
+        // LOAD r1 must be kept because r1 is live-out; whether LOADR is forwarded to LOAD r2 is secondary.
+        all.count { it.opcode == Opcode.LOAD && it.dest?.register == VirtualRegister.int(1) } shouldBe 1
     }
 })

codeGenIntermediate/test/TestRegisterPacker.kt

@@ -0,0 +1,537 @@
+import io.kotest.assertions.throwables.shouldThrow
+import io.kotest.core.spec.style.FunSpec
+import io.kotest.matchers.shouldBe
+import io.kotest.matchers.shouldNotBe
+import io.kotest.matchers.string.shouldContain
+import prog8.code.core.CompilationOptions
+import prog8.code.core.OutputType
+import prog8.code.core.Position
+import prog8.code.core.ZeropageType
+import prog8.code.target.VMTarget
+import prog8.codegen.intermediate.RegisterPacker
+import prog8.intermediate.*
+
+class TestRegisterPacker: FunSpec({
+
+    fun makeIRProgram(chunks: List<IRCodeChunkBase>): IRProgram {
+        require(chunks.first().label=="p8b_main.p8s_start")
+        val block = IRBlock("p8b_main", false, IRBlock.Options(), Position.DUMMY)
+        val sub = IRSubroutine("p8b_main.p8s_start", emptyList(), emptyList(), Position.DUMMY)
+        chunks.forEach { sub += it }
+        block += sub
+        val target = VMTarget()
+        val options = CompilationOptions.builder(target)
+            .output(OutputType.RAW)
+            .zeropage(ZeropageType.DONTUSE)
+            .noSysInit(true)
+            .compilerVersion("99.99")
+            .build()
+        val prog = IRProgram("test", IRSymbolTable(), options, target)
+        prog.addBlock(block)
+        prog.linkChunks()
+        prog.validate()
+        return prog
+    }
+
+    fun makeIRProgram(instructions: List<IRInstruction>): IRProgram {
+        val chunk = IRCodeChunk("p8b_main.p8s_start", null)
+        instructions.forEach { chunk += it }
+        return makeIRProgram(listOf(chunk as IRCodeChunkBase))
+    }
+
+    fun IRProgram.chunks(): List<IRCodeChunkBase> = this.allSubs().flatMap { it.chunks }.toList()
+
+    fun IRProgram.instructions(): List<IRInstruction> = chunks().flatMap { it.instructions }
+
+    fun IRProgram.registerAccesses(): List<RegisterOperand> = instructions().flatMap { it.registerAccesses }
+
+    test("simple coalescing") {
+        // r1=[0,1], r2=[2,3]  non-overlapping -> same packed slot
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.load(IRDataType.BYTE, 1, 10),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 1, IRMemory.direct("sym")),
+            IRInstructions.load(IRDataType.BYTE, 2, 20),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 2, IRMemory.direct("sym2"))
+        ))
+        RegisterPacker.pack(irProg)
+        val regs = irProg.registerAccesses().map { it.register }
+        // After packing, all register accesses should refer to the same register (both r1 and r2 mapped to 1 slot)
+        regs.toSet().size shouldBe 1
+        val used = irProg.registersUsed()
+        (used.readRegs.keys + used.writeRegs.keys).size shouldBe 1
+    }
+
+    test("no coalescing with overlapping intervals") {
+        // r1=[0,3], r2=[1,4] overlap -> 2 distinct slots needed
+        // LOADR uses both at same time -> destination must differ from source
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.load(IRDataType.BYTE, 1, 10),
+            IRInstructions.load(IRDataType.BYTE, 2, 20),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 1, IRMemory.direct("s")),
+            IRInstructions.move(IRDataType.BYTE, 1, 2)
+        ))
+        RegisterPacker.pack(irProg)
+        val instrs = irProg.instructions()
+        val loadr = instrs.first { it.opcode == Opcode.LOADR }
+        loadr.requireDest().register shouldNotBe loadr.requireSrcA().register
+    }
+
+    test("cross-chunk liveness") {
+        // Chunk A: def r1, def r2, use r1
+        //   r1=[0,2] (lives only in A), r2=defined but not used in A,
+        //   r2 is live-out from A because it's used in B
+        // Chunk B: def r3, use r2, use r3
+        //   r2=[3,4] (first read in B), r3=[3,5] (lives only in B)
+        // Non-overlapping: r1=[0,2] and r2=[3,4] -> can share
+        // Overlapping:     r2=[3,4] and r3=[3,5] -> must differ
+        val c1 = IRCodeChunk("p8b_main.p8s_start", null)
+        c1 += IRInstructions.load(IRDataType.BYTE, 1, 10)
+        c1 += IRInstructions.load(IRDataType.BYTE, 2, 20)
+        c1 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 1, IRMemory.direct("sym"))
+
+        val c2 = IRCodeChunk(null, null)
+        c2 += IRInstructions.load(IRDataType.BYTE, 3, 30)
+        c2 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 2, IRMemory.direct("sym2"))
+        c2 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 3, IRMemory.direct("sym3"))
+
+        val irProg = makeIRProgram(listOf(c1, c2))
+        RegisterPacker.pack(irProg)
+        val distinctRegs = irProg.registerAccesses().map { it.register }.toSet()
+
+        // r1 and r2 can share a slot (non-overlapping), r3 must differ from r2 (overlapping)
+        // Results in exactly 2 distinct slots used
+        distinctRegs.size shouldBe 2
+    }
+
+    test("different types must not share the same slot number") {
+        // BYTE R5=[0,1] and POINTER R1=[2,3] in same subroutine.
+        // Non-overlapping but different incompatible types.
+        // Packer must NOT assign both to slot 1 (would cause type conflict).
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.load(IRDataType.BYTE, 5, 10),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 5, IRMemory.direct("s")),
+            IRInstructions.load(IRDataType.POINTER, 1, 0x1000),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.POINTER, 1, IRMemory.direct("s2"))
+        ))
+        RegisterPacker.pack(irProg)
+        // After packing, usedRegisters() must not throw (internal type consistency)
+        // and all registers must have exactly one type each
+        val chunkTypes = irProg.chunks().first().usedRegisters().regsTypes
+        chunkTypes.size shouldBe 2  // exactly 2 distinct registers
+        // Each register should appear with exactly one type (no conflicts)
+        chunkTypes.values.toSet().size shouldBe 2  // BYTE and POINTER, each on different regs
+    }
+
+    test("unconditional branch only one successor") {
+        // Chunk A -> unconditional jump -> chunk C
+        // Chunk B is unreachable from A (but still a predecessor of C via fall-through)
+        val c1 = IRCodeChunk("p8b_main.p8s_start", null)
+        c1 += IRInstructions.load(IRDataType.BYTE, 1, 10)
+        c1 += IRInstructions.jump(codeLabel("label_c"))
+
+        val c2 = IRCodeChunk("label_b", null)
+        c2 += IRInstructions.load(IRDataType.BYTE, 2, 20)
+
+        val c3 = IRCodeChunk("label_c", null)
+        c3 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 1, IRMemory.direct("sym"))
+        c3 += IRInstructions.load(IRDataType.BYTE, 2, 99)
+        c3 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 2, IRMemory.direct("sym2"))
+
+        val irProg = makeIRProgram(listOf(c1, c2, c3))
+        RegisterPacker.pack(irProg)
+        // r1 is live-out of A into C via JUMP; r2 is NOT live-in to C from A via the branch
+        // but IS live-in to C from B's fall-through -> conservative: r2 is live-in to C
+        // r1 and r3 both defined across non-overlapping chunks
+        // Just verify no crash and packing happened
+        val instrs = irProg.instructions()
+        instrs shouldNotBe emptyList<IRInstruction>()
+    }
+
+    test("conditional branch two successors") {
+        // Chunk A -> conditional branch to C, fall-through to B
+        // Liveness flows from A to both B and C
+        val c1 = IRCodeChunk("p8b_main.p8s_start", null)
+        c1 += IRInstructions.compareImmediate(IRDataType.BYTE, 1, 0)
+        c1 += IRInstructions.branch(Opcode.BSTNE, codeLabel("label_c"))
+
+        val c2 = IRCodeChunk(null, null)  // fall-through (B)
+        c2 += IRInstructions.load(IRDataType.BYTE, 2, 42)
+        c2 += IRInstructions.jump(codeLabel("label_end"))
+
+        val c3 = IRCodeChunk("label_c", null)
+        c3 += IRInstructions.load(IRDataType.BYTE, 2, 99)
+
+        val c4 = IRCodeChunk("label_end", null)
+        c4 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 2, IRMemory.direct("sym"))
+
+        val irProg = makeIRProgram(listOf(c1, c2, c3, c4))
+        RegisterPacker.pack(irProg)
+        val instrs = irProg.instructions()
+        instrs shouldNotBe emptyList<IRInstruction>()
+    }
+
+    test("multi-chunk register reduction") {
+        // 3 chunks with 8 registers, non-overlapping patterns
+        val c1 = IRCodeChunk("p8b_main.p8s_start", null)
+        c1 += IRInstructions.load(IRDataType.BYTE, 1, 10)
+        c1 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 1, IRMemory.direct("sym"))
+        c1 += IRInstructions.load(IRDataType.BYTE, 2, 20)
+        c1 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 2, IRMemory.direct("sym2"))
+        c1 += IRInstructions.jump(codeLabel("cont"))
+
+        val c2 = IRCodeChunk("cont", null)
+        c2 += IRInstructions.load(IRDataType.BYTE, 3, 30)
+        c2 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 3, IRMemory.direct("sym3"))
+        c2 += IRInstructions.load(IRDataType.BYTE, 4, 40)
+        c2 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 4, IRMemory.direct("sym4"))
+        c2 += IRInstructions.jump(codeLabel("cont2"))
+
+        val c3 = IRCodeChunk("cont2", null)
+        c3 += IRInstructions.load(IRDataType.BYTE, 5, 50)
+        c3 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 5, IRMemory.direct("sym5"))
+        c3 += IRInstructions.load(IRDataType.BYTE, 6, 60)
+        c3 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 6, IRMemory.direct("sym6"))
+
+        val irProg = makeIRProgram(listOf(c1, c2, c3))
+        RegisterPacker.pack(irProg)
+        val distinctRegs = irProg.registerAccesses().map { it.register }.toSet()
+        // Each chunk defines and uses 2 regs. Within each chunk the regs are non-overlapping
+        // (def rX at i, use at i+1, then def rY at i+2). Across chunks they don't overlap either
+        // (sequential, jumps between chunks). So all 6 regs can share 1 slot.
+        distinctRegs.size shouldBe 1
+    }
+
+    test("floating point registers packed independently") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.loadFloat(1, 1.0),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.FLOAT, 1, IRMemory.direct("sym")),
+            IRInstructions.loadFloat(2, 2.0),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.FLOAT, 2, IRMemory.direct("sym2"))
+        ))
+        RegisterPacker.pack(irProg)
+        // Both fp regs have non-overlapping intervals -> same packed slot
+        val fpRegs = irProg.registerAccesses().filter { it.isFloat }.map { it.register }
+        fpRegs.distinct().size shouldBe 1
+    }
+
+    test("function call args are remapped") {
+        val callSite = CallSite(
+            CallTarget.SystemCall(0),
+            arguments = listOf(
+                Calls.argument(1, IRDataType.BYTE),
+                Calls.argument(2, IRDataType.BYTE)
+            ),
+            results = listOf(Calls.result(3, IRDataType.BYTE))
+        )
+
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.load(IRDataType.BYTE, 1, 10),
+            IRInstructions.load(IRDataType.BYTE, 2, 20),
+            // Non-overlapping: r1=[0,0], r2=[1,1], r3=live only as return from SYSCALL
+            IRInstructions.call(Opcode.SYSCALL, callSite)
+        ))
+        RegisterPacker.pack(irProg)
+        val instrs = irProg.instructions()
+        val syscall = instrs.first { it.opcode == Opcode.SYSCALL }
+        val remapped = syscall.requireCallSite()
+        // Should have same packed register numbers
+        remapped.arguments.map { it.source.register }.toSet().size shouldBe 2
+        remapped.results.single().destination!!.registerNumber shouldBe 3
+    }
+
+    test("empty subroutine no crash") {
+        val c1 = IRCodeChunk("p8b_main.p8s_start", null)
+        val irProg = makeIRProgram(listOf(c1))
+        // Should not throw
+        RegisterPacker.pack(irProg)
+    }
+
+    test("no registers used") {
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.simple(Opcode.NOP),
+            IRInstructions.simple(Opcode.NOP)
+        ))
+        RegisterPacker.pack(irProg)
+        irProg.instructions().size shouldBe 2
+    }
+
+    test("call arguments with mixed types must not share same slot") {
+        // Regression: call site registers with incompatible types (WORD vs LONG)
+        // in the same chunk must be assigned different packed slots.
+        // The conflict graph must track call site register types.
+        val callSite = CallSite(
+            CallTarget.SystemCall(0),
+            arguments = listOf(
+                Calls.argument(1, IRDataType.WORD),
+                Calls.argument(2, IRDataType.LONG)
+            )
+        )
+
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.load(IRDataType.WORD, 1, 100),
+            IRInstructions.call(Opcode.SYSCALL, callSite)
+        ))
+        RegisterPacker.pack(irProg)
+        // usedRegisters() must not throw (type consistency within chunk)
+        val used = irProg.registersUsed()
+        val regs = (used.readRegs.keys + used.writeRegs.keys).map { it.num }.toSet()
+        // Must have at least 2 distinct packed slots (WORD and LONG cannot share)
+        regs.size shouldBe 2
+    }
+
+    test("call results with mixed types must not share same slot") {
+        // Same as above but for return registers
+        val callSite = CallSite(
+            CallTarget.SystemCall(0),
+            results = listOf(
+                Calls.result(1, IRDataType.WORD),
+                Calls.result(2, IRDataType.LONG)
+            )
+        )
+
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.call(Opcode.SYSCALL, callSite),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.WORD, 1, IRMemory.direct("s")),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.LONG, 2, IRMemory.direct("s2"))
+        ))
+        RegisterPacker.pack(irProg)
+        // usedRegisters() must not throw
+        val used = irProg.registersUsed()
+        val regs = (used.readRegs.keys + used.writeRegs.keys).map { it.num }.toSet()
+        regs.size shouldBe 2
+    }
+
+    test("WORD and LONG non-overlapping in same chunk must not share slot") {
+        // Regression: two registers with incompatible types (WORD vs LONG) but
+        // non-overlapping live ranges in the same chunk must NOT be assigned
+        // the same packed slot, because usedRegisters() enforces a single type
+        // per register number within a chunk.
+        val irProg = makeIRProgram(listOf(
+            IRInstructions.load(IRDataType.WORD, 1, 100),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.WORD, 1, IRMemory.direct("s")),
+            IRInstructions.load(IRDataType.LONG, 2, 0x10000),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.LONG, 2, IRMemory.direct("s2"))
+        ))
+        RegisterPacker.pack(irProg)
+        // usedRegisters() must not throw
+        val used = irProg.registersUsed()
+        val types = used.regsTypes
+        // Exactly 2 registers, one WORD and one LONG (different slots)
+        types.size shouldBe 2
+        types.values.toSet() shouldBe setOf(IRDataType.WORD, IRDataType.LONG)
+    }
+
+    test("cross-subroutine slot type is preserved via putIfAbsent") {
+        // Regression: when a slot is first assigned a type (e.g. POINTER),
+        // subsequent compatible assignments (WORD via typesCompatible) must not
+        // overwrite the slot's type. Otherwise a later LONG interval would be
+        // blocked (WORD vs LONG incompatible), forcing it to a different slot,
+        // and registersUsed() would see the same slot with both POINTER and
+        // WORD types across subroutines → should be compatible (POINTER bridges both).
+        fun makeMultiSubProgram(sub1Chunks: List<IRCodeChunkBase>, sub2Chunks: List<IRCodeChunkBase>): IRProgram {
+            val target = VMTarget()
+            val options = CompilationOptions.builder(target)
+                .output(OutputType.RAW)
+                .zeropage(ZeropageType.DONTUSE)
+                .noSysInit(true)
+                .compilerVersion("99.99")
+                .build()
+            val prog = IRProgram("test", IRSymbolTable(), options, target)
+            val block = IRBlock("p8b_main", false, IRBlock.Options(), Position.DUMMY)
+            val sub1 = IRSubroutine("p8b_main.sub1", emptyList(), emptyList(), Position.DUMMY)
+            sub1Chunks.forEach { sub1 += it }
+            val sub2 = IRSubroutine("p8b_main.sub2", emptyList(), emptyList(), Position.DUMMY)
+            sub2Chunks.forEach { sub2 += it }
+            block += sub1
+            block += sub2
+            prog.addBlock(block)
+            prog.linkChunks()
+            prog.validate()
+            return prog
+        }
+
+        // Sub 1: POINTER interval (gets slot ≥ startSlot), and then the same slot again with WORD
+        val c1 = IRCodeChunk("p8b_main.sub1", null)
+        c1 += IRInstructions.load(IRDataType.POINTER, 100, 0x1000)
+        c1 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.POINTER, 100, IRMemory.direct("sp"))
+        c1 += IRInstructions.load(IRDataType.WORD, 101, 200)
+        c1 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.WORD, 101, IRMemory.direct("sw"))
+
+        // Sub 2: LONG interval that should be able to share the same slot as sub 1's POINTER
+        val c2 = IRCodeChunk("p8b_main.sub2", null)
+        c2 += IRInstructions.load(IRDataType.LONG, 200, 0x100000)
+        c2 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.LONG, 200, IRMemory.direct("sl"))
+
+        val irProg = makeMultiSubProgram(listOf(c1), listOf(c2))
+        RegisterPacker.pack(irProg)
+        // registersUsed() must not throw: POINTER and WORD compatible,
+        // POINTER and LONG compatible, and putIfAbsent keeps the slot type
+        // as POINTER so WORD and LONG intervals can share the same slot.
+        val used = irProg.registersUsed()
+        used.regsTypes.size shouldBe 2  // POINTER (or POINTER+WORD+LONG sharing) + WORD
+    }
+
+    test("original register numbers must not collide with packed slots") {
+        // Regression: pre-existing low register numbers (e.g. r5) in the program
+        // must not collide with packed slot numbers. Packed slots start at
+        // maxReg+1 to avoid this.
+        val irProg = makeIRProgram(listOf(
+            // Register 5 is used directly (not packed? actually it HAS an interval and gets packed)
+            IRInstructions.load(IRDataType.BYTE, 5, 99),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 5, IRMemory.direct("s")),
+            // Register 1 and 2 can be packed
+            IRInstructions.load(IRDataType.WORD, 1, 100),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.WORD, 1, IRMemory.direct("s2")),
+            IRInstructions.load(IRDataType.WORD, 2, 200),
+            IRInstructions.storeMemory(Opcode.STOREM, IRDataType.WORD, 2, IRMemory.direct("s3"))
+        ))
+        RegisterPacker.pack(irProg)
+        // after packing:
+        // - all registers should have been packed to slots > 5 (since maxReg=5, startSlot=6)
+        // - no packed slot should equal an original register number
+        val used = irProg.registersUsed()
+        val allRegNums = (used.readRegs.keys + used.writeRegs.keys).map { it.num }.toSet()
+        // packed slots start at 6, so no register should be 5 anymore
+        (5 in allRegNums) shouldBe false
+        // all registers should be >= 6
+        for (rn in allRegNums)
+            rn shouldNotBe 5
+    }
+
+    test("strict registersUsed throws on incompatible types across subroutines") {
+        // Regression: registersUsed() in strict mode (no packing applied) must detect
+        // the same register number used with incompatible types in different subroutines.
+        val target = VMTarget()
+        val options = CompilationOptions.builder(target)
+            .output(OutputType.RAW)
+            .zeropage(ZeropageType.DONTUSE)
+            .noSysInit(true)
+            .compilerVersion("99.99")
+            .build()
+        val prog = IRProgram("test", IRSymbolTable(), options, target)
+        val block = IRBlock("p8b_main", false, IRBlock.Options(), Position.DUMMY)
+
+        // Sub 1: uses r37 as POINTER
+        val sub1 = IRSubroutine("p8b_main.sub1", emptyList(), emptyList(), Position.DUMMY)
+        val c1 = IRCodeChunk("p8b_main.sub1", null)
+        c1 += IRInstructions.load(IRDataType.POINTER, 37, 0x1000)
+        c1 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.POINTER, 37, IRMemory.direct("sp"))
+        sub1 += c1
+
+        // Sub 2: uses same r37 as BYTE (incompatible)
+        val sub2 = IRSubroutine("p8b_main.sub2", emptyList(), emptyList(), Position.DUMMY)
+        val c2 = IRCodeChunk("p8b_main.sub2", null)
+        c2 += IRInstructions.load(IRDataType.BYTE, 37, 42)
+        c2 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 37, IRMemory.direct("sb"))
+        sub2 += c2
+
+        block += sub1
+        block += sub2
+        prog.addBlock(block)
+        prog.linkChunks()
+        prog.validate()
+
+        // No packing applied - strict mode should detect the incompatibility
+        val ex = shouldThrow<IllegalArgumentException> {
+            prog.registersUsed()
+        }
+        ex.message shouldContain "register r37 given multiple types"
+        ex.message shouldContain "POINTER"
+        ex.message shouldContain "BYTE"
+    }
+
+    test("intra-subroutine cross-type register is skipped by packer") {
+        // When a register has non-overlapping intervals with incompatible types (e.g.
+        // BYTE and LONG in different chunks), the packer skips it entirely because its
+        // single-entry packing map can't handle a register with multiple types.
+        // After packing, the register retains its original number and is not rewritten.
+        // registersUsed() in permissive mode uses putIfAbsent and should not throw.
+        val target = VMTarget()
+        val options = CompilationOptions.builder(target)
+            .output(OutputType.RAW)
+            .zeropage(ZeropageType.DONTUSE)
+            .noSysInit(true)
+            .compilerVersion("99.99")
+            .build()
+        val prog = IRProgram("test", IRSymbolTable(), options, target)
+        val block = IRBlock("p8b_main", false, IRBlock.Options(), Position.DUMMY)
+
+        // Single subroutine with two chunks using same register with incompatible types
+        val sub = IRSubroutine("p8b_main.sub1", emptyList(), emptyList(), Position.DUMMY)
+
+        val c1 = IRCodeChunk("p8b_main.sub1", null)
+        c1 += IRInstructions.load(IRDataType.POINTER, 37, 0x1000)
+        c1 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.POINTER, 37, IRMemory.direct("sp"))
+        sub += c1
+
+        val c2 = IRCodeChunk("p8b_main.sub1.c2", c1)
+        c2 += IRInstructions.load(IRDataType.BYTE, 37, 42)
+        c2 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 37, IRMemory.direct("sb"))
+        sub += c2
+
+        block += sub
+        prog.addBlock(block)
+        prog.linkChunks()
+        prog.validate()
+
+        RegisterPacker.pack(prog)
+        prog.wasPackingApplied = true
+
+        // registersUsed() in permissive mode: putIfAbsent, should not throw
+        val used = prog.registersUsed()
+        // r37 should still appear (was skipped by packer)
+        used.regsTypes.containsKey(VirtualRegister.int(37)) shouldBe true
+    }
+
+    test("disjoint intervals of same register must be merged into one contiguous range") {
+        // Regression: if register r1 has disjoint intervals (written in chunk 1,
+        // read in chunk 3 with a gap in chunk 2), the packer must merge them into
+        // a single interval spanning the gap. Otherwise another register (r2) with
+        // an interval in chunk 2 could share the same slot and clobber r1's value.
+        val target = VMTarget()
+        val options = CompilationOptions.builder(target)
+            .output(OutputType.RAW)
+            .zeropage(ZeropageType.DONTUSE)
+            .noSysInit(true)
+            .compilerVersion("99.99")
+            .build()
+        val prog = IRProgram("test", IRSymbolTable(), options, target)
+        val block = IRBlock("p8b_main", false, IRBlock.Options(), Position.DUMMY)
+        val sub = IRSubroutine("p8b_main.sub1", emptyList(), emptyList(), Position.DUMMY)
+
+        // Chunk 1: write r1 (value should persist to chunk 3)
+        val c1 = IRCodeChunk("p8b_main.sub1", null)
+        c1 += IRInstructions.load(IRDataType.BYTE, 1, 42)
+        c1 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 1, IRMemory.direct("s1"))
+        sub += c1
+
+        // Chunk 2: write and read r2 (this interval falls in the gap between r1's uses)
+        val c2 = IRCodeChunk("p8b_main.sub1.c2", c1)
+        c2 += IRInstructions.load(IRDataType.BYTE, 2, 99)
+        c2 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 2, IRMemory.direct("s2"))
+        sub += c2
+
+        // Chunk 3: read r1 (uses the value from chunk 1)
+        val c3 = IRCodeChunk("p8b_main.sub1.c3", c2)
+        c3 += IRInstructions.move(IRDataType.BYTE, 3, 1)
+        c3 += IRInstructions.storeMemory(Opcode.STOREM, IRDataType.BYTE, 3, IRMemory.direct("s3"))
+        sub += c3
+
+        block += sub
+        prog.addBlock(block)
+        prog.linkChunks()
+        prog.validate()
+
+        RegisterPacker.pack(prog)
+
+        // After packing, r1 and r2 must be in DIFFERENT slots because r1's
+        // merged interval spans [0..?] covering the gap, preventing r2
+        // from sharing the same slot.
+        val instructions = prog.instructions()
+        // Find the register operands used for the STOREM to s1 (was r1) and s2 (was r2)
+        val s1Reg = instructions.first { it.memory?.symbolName=="s1" }.requireSrcA().register
+        val s2Reg = instructions.first { it.memory?.symbolName=="s2" }.requireSrcA().register
+        // They must be different (can't share a slot)
+        s1Reg shouldNotBe s2Reg
+    }
+})

codeGenIntermediate/test/TestSymbolPrefixer.kt

@@ -2,15 +2,15 @@ import io.kotest.assertions.throwables.shouldThrow
 import io.kotest.core.spec.style.FunSpec
 import io.kotest.matchers.shouldBe
 import io.kotest.matchers.shouldNotBe
-import prog8.code.*
+import prog8.code.StNodeType
+import prog8.code.SymbolTableMaker
 import prog8.code.ast.*
 import prog8.code.core.*
 import prog8.code.source.SourceCode
+import prog8.code.target.VMTarget
 import prog8.codegen.intermediate.prefixScopedName
 import prog8.codegen.intermediate.prefixSymbols
 import prog8.codegen.intermediate.typePrefixChar
-import prog8tests.helpers.DummyMemsizer
-import prog8tests.helpers.DummyStringEncoder
 
 
 class TestSymbolPrefixer : FunSpec({
@@ -34,6 +34,17 @@ class TestSymbolPrefixer : FunSpec({
         prefixScopedName("lib.module.func", 's') shouldBe "p8b_lib.p8s_module.p8s_func"
     }
 
+    test("prefixScopedName uses the symbol type for middle components") {
+        val program = PtProgram("test", VMTarget())
+        val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
+        val struct = PtStructDecl("Point", listOf(PtStructField(DataType.UBYTE, "x", null)), Position.DUMMY)
+        block.add(struct)
+        program.add(block)
+        val st = SymbolTableMaker(program, basicTestOptions()).make()
+
+        prefixScopedName("main.Point.field", 'v', st) shouldBe "p8b_main.p8t_Point.p8v_field"
+    }
+
     test("prefixScopedName generated label is not prefixed") {
         val genLabel = "p8_label_gen_42"
         prefixScopedName(genLabel, 's') shouldBe genLabel
@@ -59,10 +70,10 @@ class TestSymbolPrefixer : FunSpec({
     // --- full program prefixing tests ---
 
     test("prefixSymbols prefixes a simple program") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
-        val varX = PtVariable("x", DataType.UBYTE, ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY)
+        val varX = PtVariable("x", DataType.UBYTE, false, ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY)
         sub.add(varX)
         block.add(sub)
         program.add(block)
@@ -84,7 +95,7 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("prefixSymbols handles subroutine parameters") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         val param = PtSubroutineParameter("arg", DataType.UBYTE, null, Position.DUMMY)
@@ -100,7 +111,7 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("prefixSymbols handles constants") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val constant = PtConstant("SIZE", DataType.UWORD, 100.0, null, Position.DUMMY)
         block.add(constant)
@@ -114,7 +125,7 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("prefixSymbols handles labels") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         val label = PtLabel("mylabel", Position.DUMMY)
@@ -130,7 +141,7 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("prefixSymbols handles generated labels without prefixing them") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         val genLabel = PtLabel("p8_label_gen_42", Position.DUMMY)
@@ -146,7 +157,7 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("prefixSymbols handles struct declarations") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val struct = PtStructDecl("Point", listOf(
             PtStructField(DataType.UBYTE, "x", null),
@@ -163,7 +174,7 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("prefixSymbols handles asmsub") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val asmsub = PtAsmSub("routine", PtAsmSub.Address(null, null, 0x5000u), setOf(CpuRegister.Y), emptyList(), emptyList(), false, Position.DUMMY)
         block.add(asmsub)
@@ -177,7 +188,7 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("prefixSymbols handles memmapped variables") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val memvar = PtMemMapped("reg", DataType.UBYTE, 0xd020u, null, Position.DUMMY)
         block.add(memvar)
@@ -193,7 +204,7 @@ class TestSymbolPrefixer : FunSpec({
     // --- noSymbolPrefixing tests ---
 
     test("noSymbolPrefixing blocks keep their names intact") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val libOptions = PtBlock.Options(noSymbolPrefixing = true)
         val libBlock = PtBlock("lib", false, SourceCode.Generated("test"), libOptions, Position.DUMMY)
         val libSub = PtSub("func", emptyList(), emptyList(), Position.DUMMY)
@@ -209,7 +220,7 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("references from user block to noSymbolPrefixing block are not prefixed") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
 
         // Library block with no symbol prefixing
         val libOptions = PtBlock.Options(noSymbolPrefixing = true)
@@ -240,7 +251,7 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("references from noSymbolPrefixing block to regular block get prefixed") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
 
         // Regular user block
         val userBlock = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
@@ -267,10 +278,10 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("identifier references are prefixed according to their target type") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
-        val varX = PtVariable("x", DataType.UBYTE, ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY)
+        val varX = PtVariable("x", DataType.UBYTE, false, ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY)
         sub.add(varX)
 
         // Add an identifier reference to the variable
@@ -290,7 +301,7 @@ class TestSymbolPrefixer : FunSpec({
     }
 
     test("prefixSymbols exception for unknown identifier") {
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         val idRef = PtIdentifier("nonexistent", DataType.UBYTE, Position.DUMMY)
@@ -308,7 +319,7 @@ class TestSymbolPrefixer : FunSpec({
 })
 
 private fun basicTestOptions(): CompilationOptions {
-    val target = prog8.code.target.VMTarget()
+    val target = VMTarget()
     return CompilationOptions.builder(target)
         .output(OutputType.RAW)
         .zeropage(ZeropageType.DONTUSE)

codeGenIntermediate/test/TestVmCodeGen.kt

@@ -5,14 +5,16 @@ import prog8.code.SymbolTableMaker
 import prog8.code.ast.*
 import prog8.code.core.*
 import prog8.code.source.SourceCode
+import prog8.code.target.Cx16Target
 import prog8.code.target.VMTarget
 import prog8.codegen.vm.VmAssemblyProgram
 import prog8.codegen.vm.VmCodeGen
 import prog8.intermediate.IRSubroutine
 import prog8.intermediate.Opcode
+import prog8.intermediate.codeAddress
+import prog8.intermediate.requireCallSite
 import prog8.intermediate.toAddress
-import prog8tests.helpers.DummyMemsizer
-import prog8tests.helpers.DummyStringEncoder
+
 import prog8tests.helpers.ErrorReporterForTests
 
 class TestVmCodeGen: FunSpec({
@@ -40,13 +42,15 @@ class TestVmCodeGen: FunSpec({
 //        xx += cx16.r0
 //    }
 //}
+        val compTarget = VMTarget()
         val codegen = VmCodeGen(false)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         sub.add(PtVariable(
             "pi",
             DataType.UBYTE,
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -56,7 +60,8 @@ class TestVmCodeGen: FunSpec({
         ))
         sub.add(PtVariable(
             "particleX",
-            DataType.arrayFor(BaseDataType.UBYTE),
+            DataType.arrayFor(BaseDataType.UBYTE, compTarget),
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -66,7 +71,8 @@ class TestVmCodeGen: FunSpec({
         ))
         sub.add(PtVariable(
             "particleDX",
-            DataType.arrayFor(BaseDataType.UBYTE),
+            DataType.arrayFor(BaseDataType.UBYTE, compTarget),
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -77,6 +83,7 @@ class TestVmCodeGen: FunSpec({
         sub.add(PtVariable(
             "xx",
             DataType.WORD,
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -87,17 +94,17 @@ class TestVmCodeGen: FunSpec({
 
         val assign = PtAugmentedAssign("+=", Position.DUMMY)
         val target = PtAssignTarget(false, Position.DUMMY).also {
-            val targetIdx = PtArrayIndexer(DataType.UBYTE, Position.DUMMY).also { idx ->
+            val targetIdx = PtArrayIndexer(DataType.UBYTE, false, Position.DUMMY).also { idx ->
                 idx.add(PtIdentifier("main.start.particleX",
-                    DataType.arrayFor(BaseDataType.UBYTE),
+                    DataType.arrayFor(BaseDataType.UBYTE, compTarget),
                     Position.DUMMY))
                 idx.add(PtNumber(BaseDataType.UBYTE, 2.0, Position.DUMMY))
             }
             it.add(targetIdx)
         }
-        val value = PtArrayIndexer(DataType.UBYTE, Position.DUMMY)
+        val value = PtArrayIndexer(DataType.UBYTE, false, Position.DUMMY)
         value.add(PtIdentifier("main.start.particleDX",
-            DataType.arrayFor(BaseDataType.UBYTE),
+            DataType.arrayFor(BaseDataType.UBYTE, compTarget),
             Position.DUMMY))
         value.add(PtNumber(BaseDataType.UBYTE, 2.0, Position.DUMMY))
         assign.add(target)
@@ -152,11 +159,11 @@ class TestVmCodeGen: FunSpec({
 // Pattern: @(ptr+offset) = @(ptr+offset) + val   i.e.  ptr[offset] += val
 // Non-zero offsets are used to avoid potential optimization of the address expression.
         val codegen = VmCodeGen(false)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
-        sub.add(PtVariable("ptr", DataType.pointer(BaseDataType.UBYTE), ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY))
-        sub.add(PtVariable("val", DataType.UBYTE, ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY))
+        sub.add(PtVariable("ptr", DataType.pointer(BaseDataType.UBYTE), false, ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY))
+        sub.add(PtVariable("val", DataType.UBYTE, false, ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY))
 
         fun makeMemByteAssign(op: String, offset: Int): PtAugmentedAssign {
             // Use non-zero offset to avoid optimization
@@ -203,12 +210,12 @@ class TestVmCodeGen: FunSpec({
         andrCount shouldBe 1
         xorrCount shouldBe 1
 
-        // Verify that the immediate-only variants are NOT used with reg2
-        val wrongAdd = instructions.any { it.opcode == Opcode.ADD && it.reg2 != null }
-        val wrongSub = instructions.any { it.opcode == Opcode.SUB && it.reg2 != null }
-        val wrongOr = instructions.any { it.opcode == Opcode.OR && it.reg2 != null }
-        val wrongAnd = instructions.any { it.opcode == Opcode.AND && it.reg2 != null }
-        val wrongXor = instructions.any { it.opcode == Opcode.XOR && it.reg2 != null }
+        // Verify that the immediate-only variants are NOT used with a register source
+        val wrongAdd = instructions.any { it.opcode == Opcode.ADD && it.srcA != null }
+        val wrongSub = instructions.any { it.opcode == Opcode.SUB && it.srcA != null }
+        val wrongOr = instructions.any { it.opcode == Opcode.OR && it.srcA != null }
+        val wrongAnd = instructions.any { it.opcode == Opcode.AND && it.srcA != null }
+        val wrongXor = instructions.any { it.opcode == Opcode.XOR && it.srcA != null }
 
         (wrongAdd || wrongSub || wrongOr || wrongAnd || wrongXor) shouldBe false
     }
@@ -229,12 +236,13 @@ class TestVmCodeGen: FunSpec({
 //    }
 //}
         val codegen = VmCodeGen(false)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         sub.add(PtVariable(
             "f1",
             DataType.FLOAT,
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -301,12 +309,13 @@ class TestVmCodeGen: FunSpec({
 //    }
 //}
         val codegen = VmCodeGen(false)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         sub.add(PtVariable(
             "f1",
             DataType.FLOAT,
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -369,12 +378,13 @@ class TestVmCodeGen: FunSpec({
 //    }
 //}
         val codegen = VmCodeGen(false)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         sub.add(PtVariable(
             "f1",
             DataType.FLOAT,
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -425,12 +435,13 @@ class TestVmCodeGen: FunSpec({
 //    }
 //}
         val codegen = VmCodeGen(false)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         sub.add(PtVariable(
             "sb1",
             DataType.BYTE,
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -497,12 +508,13 @@ class TestVmCodeGen: FunSpec({
 //    }
 //}
         val codegen = VmCodeGen(false)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         sub.add(PtVariable(
             "sb1",
             DataType.BYTE,
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -565,12 +577,13 @@ class TestVmCodeGen: FunSpec({
 //    }
 //}
         val codegen = VmCodeGen(false)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
         sub.add(PtVariable(
             "ub1",
             DataType.BYTE,
+            false,
             ZeropageWish.DONTCARE,
             0u,
             false,
@@ -614,7 +627,7 @@ class TestVmCodeGen: FunSpec({
 //    }
 //}
         val codegen = VmCodeGen(false)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val extsub = PtAsmSub("routine", PtAsmSub.Address(null, null, 0x5000u), setOf(CpuRegister.Y), emptyList(), emptyList(), false, Position.DUMMY)
         block.add(extsub)
@@ -634,15 +647,15 @@ class TestVmCodeGen: FunSpec({
         irChunks[0].instructions.size shouldBe 1
         val callInstr = irChunks[0].instructions[0]
         callInstr.opcode shouldBe Opcode.CALL
-        callInstr.address shouldBe 0x5000u.toAddress()
+        callInstr.requireCallSite().codeReference shouldBe codeAddress(0x5000u.toAddress())
     }
 
     test("ir codegen for target 'virtual' produces no prefixed names") {
         val codegen = VmCodeGen(false)
-        val program = PtProgram("test", DummyMemsizer, DummyStringEncoder)
+        val program = PtProgram("test", VMTarget())
         val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
         val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
-        sub.add(PtVariable("x", DataType.UBYTE, ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY))
+        sub.add(PtVariable("x", DataType.UBYTE, false, ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY))
         block.add(sub)
         program.add(block)
 
@@ -664,4 +677,39 @@ class TestVmCodeGen: FunSpec({
         val irVar = irProg.st.allVariables().single()
         irVar.name shouldBe "main.start.x"
     }
+
+    test("ir codegen for target 'cx16' via newcodegen produces prefixed names") {
+        val codegen = VmCodeGen(false)
+        val program = PtProgram("test", VMTarget())
+        val block = PtBlock("main", false, SourceCode.Generated("test"), PtBlock.Options(), Position.DUMMY)
+        val sub = PtSub("start", emptyList(), emptyList(), Position.DUMMY)
+        sub.add(PtVariable("x", DataType.UBYTE, false, ZeropageWish.DONTCARE, 0u, false, null, null, Position.DUMMY))
+        block.add(sub)
+        program.add(block)
+
+        val target = Cx16Target()
+        val options = CompilationOptions.builder(target)
+            .output(OutputType.RAW)
+            .zeropage(ZeropageType.DONTUSE)
+            .floats(true)
+            .compilerVersion("99.99")
+            .newCodegen(true)
+            .build()
+        val st = SymbolTableMaker(program, options).make()
+        val errors = ErrorReporterForTests()
+        val result = codegen.generate(program, st, options, errors) as VmAssemblyProgram
+        val irProg = result.irProgram
+
+        // Block label should be prefixed
+        val irBlock = irProg.blocks.first()
+        irBlock.label shouldBe "p8b_main"
+
+        // Subroutine label should be fully scoped and prefixed
+        val irSub = irBlock.children.single() as IRSubroutine
+        irSub.label shouldBe "p8b_main.p8s_start"
+
+        // Variable in IR symbol table should be fully scoped and prefixed
+        val irVar = irProg.st.allVariables().single()
+        irVar.name shouldBe "p8b_main.p8s_start.p8v_x"
+    }    
 })

codeGenM68k/build.gradle.kts

@@ -0,0 +1,97 @@
+plugins {
+    kotlin("jvm")
+    id("application")
+    id("com.peterabeles.gversion") version "1.11.0"
+}
+
+dependencies {
+    implementation(project(":codeCore"))
+    implementation(project(":simpleAst"))
+    implementation(project(":intermediate"))
+    implementation(project(":codeGenIntermediate"))
+    implementation("com.michael-bull.kotlin-result:kotlin-result-jvm:2.3.1")
+
+    testImplementation(testFixtures(project(":codeCore")))
+    testImplementation("io.kotest:kotest-runner-junit5")
+}
+
+
+val debugPort = 8000
+val debugArgs = "-agentlib:jdwp=transport=dt_socket,server=y,address=8000,suspend=n,quiet=y"
+
+val serverMainClassName = "prog8.codegen.m68k.MainKt"
+val applicationName = "prog8-m68kgen"
+
+application {
+    mainClass.set(serverMainClassName)
+    description = "New codegen based on IR files"
+    applicationDistribution.into("bin") {
+        filePermissions {
+            user {
+                read=true
+                execute=true
+                write=true
+            }
+            other.execute = true
+            group.execute = true
+        }
+    }
+}
+
+repositories {
+    mavenCentral()
+}
+
+dependencies {
+    implementation(project(":codeCore"))
+    implementation(project(":intermediate"))
+    implementation("com.github.ajalt.clikt:clikt:5.1.0")
+    implementation("com.github.ajalt.mordant:mordant:3.1.0")
+
+    // Test dependencies - Kotest BOM is provided by root build.gradle.kts
+    testImplementation("io.kotest:kotest-runner-junit5")
+}
+
+configurations.forEach { config ->
+    config.resolutionStrategy {
+        preferProjectModules()
+    }
+}
+
+sourceSets {
+    main {
+        java.srcDir("src")
+        resources.srcDir("resources")
+    }
+}
+
+tasks.startScripts {
+    applicationName = "prog8-m68kgen"
+}
+
+tasks.register<Exec>("fixFilePermissions") {
+    description = "When running on macOS or Linux the start script needs executable permissions to run, this sets them"
+
+    onlyIf { !System.getProperty("os.name").lowercase().contains("windows") }
+    commandLine("chmod", "+x", "${tasks.installDist.get().destinationDir}/bin/prog8-m68kgen")
+}
+
+tasks.register<JavaExec>("debugRun") {
+    description = "run with remote debugging"
+    mainClass.set(serverMainClassName)
+    classpath(sourceSets.main.get().runtimeClasspath)
+    standardInput = System.`in`
+
+    jvmArgs(debugArgs)
+    doLast {
+        println("Using debug port $debugPort")
+    }
+}
+
+tasks.installDist {
+    finalizedBy("fixFilePermissions")
+}
+
+tasks.build {
+    finalizedBy("installDist")
+}

codeGenM68k/docs/kernel-example.asm

@@ -0,0 +1,411 @@
+; A working m68k kernel for use in QEMU `-M virt` m68k machine.
+; It parses the bootinfo chunks, prints stuff to the uart output, and reads input as well.
+
+
+        section .text,code
+        xref _end_of_code
+
+        global _start
+_start: move.l #$fffc,sp
+        bsr bootinfo_parse
+        move.l ram_top,d0
+        beq 1$
+        sub.l #256,d0
+        move.l d0,sp
+1$:
+
+        movea.l #str_welcome,a0
+        bsr puts
+
+2$:     bsr getchar
+        move.b d0,d1
+        move.b d1,(a0)
+        bsr putchar
+        cmpi.b #13,d1
+        bne 3$
+        move.b #10,d0
+        bsr putchar
+3$:     cmpi.b #$1b,d1
+        bne 2$
+        bsr do_halt
+        bra 2$
+
+; ============================================================
+; bootinfo_parse -- iterate bootinfo records and print them
+; ============================================================
+bootinfo_parse:
+        move.l d2,-(sp)
+        move.l d4,-(sp)
+        moveq #0,d4
+        move.l #_end_of_code,d0
+        addq.l #1,d0
+        lsr.l #1,d0
+        lsl.l #1,d0
+        movea.l d0,a2
+bi_loop:
+        clr.l d0
+        move.w (a2)+,d0
+        beq bi_done
+        move.l d0,d2
+        cmpi.w #2,d2
+        bne 1$
+        moveq #1,d4
+1$:
+        clr.l d1
+        move.w (a2)+,d1
+        move.l a2,a3
+        lea (-4,a2,d1.l),a2
+        subq.l #4,d1
+
+        movea.l #str_indent,a0
+        bsr puts
+        move.l a3,d0
+        subq.l #4,d0
+        bsr puthex8
+        movea.l #str_colon_space,a0
+        bsr puts
+
+        move.l d2,d0
+        bsr bi_print_tag
+
+        bsr bi_newline
+
+        movea.l #str_pad,a0
+        bsr puts
+        move.l a3,-(sp)
+        move.l d1,-(sp)
+        move.l d2,-(sp)
+        bsr bi_print_data
+        add.w #12,sp
+
+        bsr bi_newline
+        bra bi_loop
+
+bi_done:
+        tst.l d4
+        bne 1$
+        movea.l #str_indent,a0
+        bsr puts
+        movea.l #str_no_cputype,a0
+        bsr puts
+        bsr bi_newline
+1$:     move.l (sp)+,d4
+        move.l (sp)+,d2
+        rts
+
+bi_newline:
+        movea.l #str_newline,a0
+        bra puts
+
+; ============================================================
+; bi_print_tag -- print tag name for value in d0
+; ============================================================
+bi_print_tag:
+        cmpi.w #1,d0
+        beq bi_tag_machtype
+        cmpi.w #2,d0
+        beq bi_tag_cputype
+        cmpi.w #3,d0
+        beq bi_tag_fputype
+        cmpi.w #4,d0
+        beq bi_tag_mmutype
+        cmpi.w #5,d0
+        beq bi_tag_memchunk
+        cmpi.w #6,d0
+        beq bi_tag_ramdisk
+        cmpi.w #7,d0
+        beq bi_tag_cmdline
+        cmpi.w #8,d0
+        beq bi_tag_rngseed
+        cmpi.w #$8000,d0
+        beq bi_tag_version
+        cmpi.w #$8001,d0
+        beq bi_tag_pic
+        cmpi.w #$8002,d0
+        beq bi_tag_rtc
+        cmpi.w #$8003,d0
+        beq bi_tag_tty
+        cmpi.w #$8004,d0
+        beq bi_tag_virtio
+        cmpi.w #$8005,d0
+        beq bi_tag_ctrl
+
+        movea.l #str_hex_prefix,a0
+        bsr puts
+        bsr puthex8
+        rts
+
+bi_tag_machtype: movea.l #str_machtype,a0
+                bra puts
+bi_tag_cputype:  movea.l #str_cputype,a0
+                bra puts
+bi_tag_fputype:  movea.l #str_fputype,a0
+                bra puts
+bi_tag_mmutype:  movea.l #str_mmutype,a0
+                bra puts
+bi_tag_memchunk: movea.l #str_memchunk,a0
+                bra puts
+bi_tag_ramdisk:  movea.l #str_ramdisk,a0
+                bra puts
+bi_tag_cmdline:  movea.l #str_cmdline,a0
+                bra puts
+bi_tag_rngseed:  movea.l #str_rngseed,a0
+                bra puts
+bi_tag_version:  movea.l #str_version,a0
+                bra puts
+bi_tag_pic:      movea.l #str_pic,a0
+                bra puts
+bi_tag_rtc:      movea.l #str_rtc,a0
+                bra puts
+bi_tag_tty:      movea.l #str_tty,a0
+                bra puts
+bi_tag_ctrl:     movea.l #str_ctrl,a0
+                bra puts
+bi_tag_virtio:   movea.l #str_virtio,a0
+                bra puts
+
+; ============================================================
+; bi_print_data -- print data based on tag type
+; stack args (low to high): tag, size, data_ptr
+; ============================================================
+bi_print_data:
+        move.l 4(sp),d0
+        move.l 8(sp),d1
+        movea.l 12(sp),a3
+
+        cmpi.w #1,d0
+        beq bi_data_machtype
+        cmpi.w #2,d0
+        beq bi_data_cputype
+        cmpi.w #3,d0
+        beq bi_data_u32
+        cmpi.w #4,d0
+        beq bi_data_u32
+        cmpi.w #$8000,d0
+        beq bi_data_u32
+        cmpi.w #5,d0
+        beq bi_data_memchunk
+        cmpi.w #6,d0
+        beq bi_data_base_size
+        cmpi.w #7,d0
+        beq bi_data_str
+        cmpi.w #8,d0
+        beq bi_data_hex
+        cmpi.w #$8001,d0
+        beq bi_data_base_count
+        cmpi.w #$8002,d0
+        beq bi_data_base_count
+        cmpi.w #$8003,d0
+        beq bi_data_base_count
+        cmpi.w #$8004,d0
+        beq bi_data_base_count
+        cmpi.w #$8005,d0
+        beq bi_data_base_count
+
+bi_data_hex:
+        move.l d1,d3
+1$:     move.b (a3)+,d0
+        bsr puthex2
+        subq.l #1,d3
+        beq 2$
+        movea.l #str_space,a0
+        bsr puts
+        bra 1$
+2$:     rts
+
+bi_data_u32:
+        move.l (a3),d0
+        bra puthex8
+
+bi_data_memchunk:
+        move.l (a3),-(sp)
+        movea.l #str_mem_at,a0
+        bsr puts
+        move.l (sp)+,d0
+        bsr puthex8
+        movea.l #str_to,a0
+        bsr puts
+        move.l 4(a3),d0
+        add.l (a3),d0
+        move.l d0,ram_top
+        bsr puthex8
+        movea.l #str_size,a0
+        bsr puts
+        move.l 4(a3),d0
+        bra puthex8
+
+bi_data_base_size:
+        movea.l #str_base,a0
+        bsr puts
+        move.l (a3),d0
+        bsr puthex8
+        movea.l #str_size,a0
+        bsr puts
+        move.l 4(a3),d0
+        bra puthex8
+
+bi_data_base_count:
+        movea.l #str_base,a0
+        bsr puts
+        move.l (a3),d0
+        bsr puthex8
+        movea.l #str_count,a0
+        bsr puts
+        move.l 4(a3),d0
+        bra puthex8
+
+bi_data_str:
+        movea.l a3,a0
+        bra puts
+
+bi_data_machtype:
+        move.l (a3),d0
+        cmpi.l #14,d0
+        beq 1$
+        bra puthex8
+1$:     movea.l #str_mach_virt,a0
+        bra puts
+
+bi_data_cputype:
+        move.l (a3),d0
+        beq 6$
+        btst #0,d0
+        bne 1$
+        btst #1,d0
+        bne 2$
+        btst #2,d0
+        bne 3$
+        btst #3,d0
+        bne 4$
+        btst #4,d0
+        bne 5$
+        bra puthex8
+1$:     movea.l #str_cpu68020,a0
+        bra puts
+2$:     movea.l #str_cpu68030,a0
+        bra puts
+3$:     movea.l #str_cpu68040,a0
+        bra puts
+4$:     movea.l #str_cpu68060,a0
+        bra puts
+5$:     movea.l #str_coldfire,a0
+        bra puts
+6$:     movea.l #str_cpu68000,a0
+        bra puts
+
+; ============================================================
+puts:   move.b (a0)+,d0
+        beq puts_rts
+        bsr putchar
+        bra puts
+puts_rts:
+        rts
+
+; ============================================================
+puthex8:
+        move.l d2,-(sp)
+        move.l d3,-(sp)
+        move.l d0,d2
+        moveq #28,d3
+1$:     move.l d2,d0
+        lsr.l d3,d0
+        and.b #$f,d0
+        add.b #'0',d0
+        cmp.b #'9',d0
+        bls 2$
+        add.b #7,d0
+2$:     bsr putchar
+        subq.l #4,d3
+        bpl 1$
+        move.l (sp)+,d3
+        move.l (sp)+,d2
+        rts
+
+; ============================================================
+puthex2:
+        move.l d2,-(sp)
+        move.l d3,-(sp)
+        move.l d0,d2
+        moveq #4,d3
+1$:     move.l d2,d0
+        lsr.l d3,d0
+        and.b #$f,d0
+        add.b #'0',d0
+        cmp.b #'9',d0
+        bls 2$
+        add.b #7,d0
+2$:     bsr putchar
+        subq.l #4,d3
+        bpl 1$
+        move.l (sp)+,d3
+        move.l (sp)+,d2
+        rts
+
+; ============================================================
+putchar:
+        movea.l #$ff008000,a1
+        move.l d0,(a1)
+        rts
+
+; ============================================================
+getchar:
+        suba.l #16,sp
+        movea.l #$ff008004,a1
+1$:     move.l (a1),d0
+        tst.l d0
+        beq 1$
+        movea.l #$ff008010,a1
+        move.l sp,(a1)
+        movea.l #$ff008014,a1
+        move.l #1,(a1)
+        movea.l #$ff008008,a1
+        move.l #3,(a1)
+        move.b (sp),d0
+        adda.l #16,sp
+        rts
+
+; ============================================================
+do_halt:
+        movea.l #$ff009004,a1
+        move.l #2,(a1)
+        rts
+
+; ============================================================
+; Strings
+; ============================================================
+str_indent:      dc.b "  ",0
+str_colon_space: dc.b ": ",0
+str_pad:         dc.b "         ",0
+str_newline:     dc.b 10,0
+str_base:        dc.b "base=0x",0
+str_size:        dc.b "  size=0x",0
+str_count:       dc.b "  count=0x",0
+str_mem_at:      dc.b "mem at 0x",0
+str_to:          dc.b " - 0x",0
+str_space:       dc.b " ",0
+str_hex_prefix: dc.b "TAG_0x",0
+str_mach_virt:  dc.b "MACH_VIRT",0
+str_cpu68000:   dc.b "CPU_68000",0
+str_cpu68020:   dc.b "CPU_68020",0
+str_cpu68030:   dc.b "CPU_68030",0
+str_cpu68040:   dc.b "CPU_68040",0
+str_cpu68060:   dc.b "CPU_68060",0
+str_coldfire:   dc.b "ColdFire",0
+str_no_cputype: dc.b "(no BI_CPUTYPE - assuming CPU_68000)",0
+str_machtype:   dc.b "BI_MACHTYPE",0
+str_cputype:    dc.b "BI_CPUTYPE",0
+str_fputype:    dc.b "BI_FPUTYPE",0
+str_mmutype:    dc.b "BI_MMUTYPE",0
+str_memchunk:   dc.b "BI_MEMCHUNK",0
+str_ramdisk:    dc.b "BI_RAMDISK",0
+str_cmdline:    dc.b "BI_COMMAND_LINE",0
+str_rngseed:    dc.b "BI_RNG_SEED",0
+str_version:    dc.b "BI_VIRT_QEMU_VER",0
+str_pic:        dc.b "BI_VIRT_GF_PIC",0
+str_rtc:        dc.b "BI_VIRT_GF_RTC",0
+str_tty:        dc.b "BI_VIRT_GF_TTY",0
+str_ctrl:       dc.b "BI_VIRT_CTRL",0
+str_virtio:     dc.b "BI_VIRT_VIRTIO",0
+str_welcome:    dc.b "Hello from m68k! (ESC to halt)",10,0
+        align 4
+ram_top:        dc.l 0

codeGenM68k/docs/linking.md

@@ -0,0 +1,94 @@
+# Linking the kernel
+
+The kernel is linked using **vlink**, producing an ELF32 binary (`kernel.elf`) that qemu loads directly via its `-kernel` option.
+
+## Toolchain
+
+| Tool | Role |
+|---|---|
+| `vasmm68k_mot` | Assembles `kernel.S` → `kernel.o` |
+| `vlink` | Links `kernel.o` → `kernel.elf` using `link.ld` |
+
+The `Makefile` invocation:
+
+```makefile
+vlink -b elf32m68k -n -T link.ld -o kernel.elf kernel.o
+```
+
+## Command-line flags
+
+### `-b elf32m68k` — output format
+
+Tells vlink to produce a 32-bit big-endian m68k ELF binary.
+
+vlink supports many output formats; without this flag it would default to something else. Note that vlink ignores the `OUTPUT_FORMAT()` directive inside linker scripts, so the format **must** be specified on the command line.
+
+### `-n` — NMAGIC / no page alignment
+
+Disables page-boundary alignment of ELF segments.
+
+Without `-n`, vlink aligns the LOAD segment to a full page (0x2000 = 8 KB). The resulting ELF has a large gap of padding between the ELF header and the actual code:
+
+```
+Type   Offset   VirtAddr   PhysAddr   Align
+LOAD   0x002000 0x00010000 0x00010000 0x2000
+```
+
+qemu's ELF loader gets confused by this layout and either fails to load the kernel or jumps to the wrong address.
+
+With `-n`, segment alignment is reduced to the minimum (0x10) and the code starts right after the ELF header:
+
+```
+Type   Offset   VirtAddr   PhysAddr   Align
+LOAD   0x000080 0x00010000 0x00010000 0x10
+```
+
+qemu loads this cleanly — the code is at file offset 0x80, maps to virtual address 0x10000, and execution starts at `_start` as expected.
+
+`-n` is the vlink equivalent of GNU ld's `-N` / `--omagic` option.
+
+### `-T link.ld` — linker script
+
+Passes the linker script (see below).
+
+## The linker script (`link.ld`)
+
+The compiler resolves the linker script for a target in this order:
+
+1. `link.ld` in the target's library directory (the `library` setting). This is how a custom
+   target, defined outside of the compiler, supplies its own memory layout.
+2. `/prog8lib/<targetname>/link.ld`, for targets built into the compiler.
+3. `/prog8lib/qemu68k/link.ld` as the fallback, which is the script shown below.
+
+Only the `ELF` output type links at all. The `AMIGAHUNK` output type (used by the built-in
+Amiga targets) assembles straight to a hunk executable and ignores the linker script.
+
+```
+ENTRY(_start)
+SECTIONS
+{
+  . = 0x10000;
+  .text : { *(.text) }
+  .data : { *(.data) }
+  .bss  : { *(.bss)  *(COMMON) }
+  _end_of_code = .;
+}
+```
+
+### Line by line
+
+| Line | What it does |
+|---|---|
+| `ENTRY(_start)` | Sets the ELF entry point to `_start` — the first instruction in `kernel.S`. qemu uses this to know where to jump after loading the kernel. |
+| `SECTIONS { … }` | Opens the section layout block — the core of any linker script. Everything inside describes how to arrange the output binary. |
+| `. = 0x10000;` | Sets the **load address** to `0x10000` (64 KB). This is where qemu places the kernel in memory and begins execution. The `.` is the *location counter* — it advances automatically as sections are placed. |
+| `.text : { *(.text) }` | Collects all `.text` (code) sections from every input object file into one output `.text` section. |
+| `.data : { *(.data) }` | Same for initialised data. |
+| `.bss  : { *(.bss) *(COMMON) }` | Same for uninitialised (zero-filled) data. `COMMON` catches any unresolved global variables that the assembler emits as common symbols. |
+| `_end_of_code = .;` | Defines the symbol `_end_of_code` at the address immediately after the last section. `kernel.S` references this symbol to find where the bootinfo records begin in memory (qemu appends them just past the kernel image). |
+
+### Why a linker script is required
+
+vlink has no command-line flag equivalent to `. = 0x10000` (setting the load address). Attempting `-Ttext=0x10000` causes a fatal error because vlink treats the `=` as part of a filename. The correct syntax `-Ttext 0x10000` (space-separated) does set the address, but then `_end_of_code` is undefined because there is no linker script to define it.
+
+The linker script is therefore the minimum required: it sets the load address and defines `_end_of_code`.

codeGenM68k/docs/qemu-io.md

@@ -0,0 +1,456 @@
+# QEMU m68k "virt" Machine -- IO Interfaces Reference
+
+All MMIO addresses and register layouts for the QEMU `-M virt` m68k machine.
+
+## Memory Map
+
+| Device | Base Address | Size | CPU IRQ |
+|---|---|---|---|
+| Goldfish PIC (x6) | `0xFF000000` | 6 x 0x1000 | #1-#6 |
+| Goldfish RTC (x2) | `0xFF006000` | 2 x 0x1000 | #6 (PIC#6, IRQ#1) |
+| Goldfish TTY | `0xFF008000` | 0x1000 | #1 (PIC#1, IRQ#32) |
+| Virt Ctrl | `0xFF009000` | 0x100 | #1 (PIC#1, IRQ#1) |
+| Virtio-MMIO (x128) | `0xFF010000` | 128 x 0x200 | #2-#5 (PIC#2-#5) |
+
+### IRQ Hierarchy
+
+```
+CPU IRQ #1  -> PIC #1   (goldfish_pic.0)  -> IRQ#1 = virt-ctrl, IRQ#32 = tty
+CPU IRQ #2  -> PIC #2   (goldfish_pic.1)  -> IRQ#1..#32 = virtio-mmio #1..#32
+CPU IRQ #3  -> PIC #3   (goldfish_pic.2)  -> IRQ#1..#32 = virtio-mmio #33..#64
+CPU IRQ #4  -> PIC #4   (goldfish_pic.3)  -> IRQ#1..#32 = virtio-mmio #65..#96
+CPU IRQ #5  -> PIC #5   (goldfish_pic.4)  -> IRQ#1..#32 = virtio-mmio #97..#128
+CPU IRQ #6  -> PIC #6   (goldfish_pic.5)  -> IRQ#1 = RTC, IRQ#2..#32 = unused
+CPU IRQ #7  -> NMI
+```
+
+All registers are **32 bits wide**. Accesses must be 4-byte aligned.
+
+---
+
+## Goldfish TTY -- `0xFF008000`
+
+Primary console. All access via 32-bit loads/stores.
+
+### Registers
+
+| Offset | Name | Access | Description |
+|---|---|---|---|
+| `0x00` | `REG_PUT_CHAR` | WO | Write one character to serial output |
+| `0x04` | `REG_BYTES_READY` | RO | Number of bytes available in RX FIFO |
+| `0x08` | `REG_CMD` | WO | Issue command |
+| `0x10` | `REG_DATA_PTR` | WO | DMA buffer address (low 32 bits) |
+| `0x14` | `REG_DATA_LEN` | WO | DMA transfer byte count |
+| `0x18` | `REG_DATA_PTR_HIGH` | WO | DMA buffer address (high 32 bits) |
+| `0x20` | `REG_VERSION` | RO | Returns 1 |
+
+### Commands (write to REG_CMD)
+
+| Value | Name | Description |
+|---|---|---|
+| `0` | `CMD_INT_DISABLE` | Disable interrupts |
+| `1` | `CMD_INT_ENABLE` | Enable interrupts |
+| `2` | `CMD_WRITE_BUFFER` | Write N bytes from guest memory to chardev |
+| `3` | `CMD_READ_BUFFER` | Read N bytes from chardev into guest memory |
+
+### Usage: putchar (blocking, non-DMA)
+
+Write the character directly to `REG_PUT_CHAR`. QEMU sends it immediately.
+
+```asm
+putchar:
+    movea.l #0xff008000, %a1
+    move.l  %d0, (%a1)
+    rts
+```
+
+### Usage: getchar (blocking, polled, DMA)
+
+1. Poll `REG_BYTES_READY` until non-zero
+2. Set `REG_DATA_PTR` to buffer address
+3. Set `REG_DATA_LEN` to number of bytes to read (e.g. 1)
+4. Write `3` (`CMD_READ_BUFFER`) to `REG_CMD`
+5. Read character(s) from buffer
+
+```asm
+getchar:
+    suba.l  #16, %sp
+    movea.l #0xff008004, %a1     @ REG_BYTES_READY
+1:  move.l  (%a1), %d0
+    tst.l   %d0
+    beq     1b                   @ wait until data available
+    movea.l #0xff008010, %a1     @ REG_DATA_PTR
+    move.l  %sp, (%a1)
+    movea.l #0xff008014, %a1     @ REG_DATA_LEN
+    move.l  #1, (%a1)
+    movea.l #0xff008008, %a1     @ REG_CMD
+    move.l  #3, (%a1)            @ CMD_READ_BUFFER
+    move.b  (%sp), %d0
+    adda.l  #16, %sp
+    rts
+```
+
+To use interrupts instead of polling:
+1. Write `1` (`CMD_INT_ENABLE`) to `REG_CMD` (offset `0x08`)
+2. Enable IRQ in PIC (PIC#1 at `0xFF000000`, IRQ#32)
+3. On interrupt, read `REG_BYTES_READY` then DMA data from buffer
+
+---
+
+## Virt Controller -- `0xFF009000`
+
+Controls machine power state.
+
+### Registers
+
+| Offset | Name | Access | Description |
+|---|---|---|---|
+| `0x00` | `REG_FEATURES` | RO | Feature bitmask |
+| `0x04` | `REG_CMD` | WO | Issue power/reset command |
+
+### Feature Bits (REG_FEATURES)
+
+| Bit | Name | Description |
+|---|---|---|
+| `0` | `FEAT_POWER_CTRL` | Power control is available |
+
+### Commands (write to REG_CMD)
+
+| Value | Name | Description |
+|---|---|---|
+| `0` | `CMD_NOOP` | No operation |
+| `1` | `CMD_RESET` | Reset the VM |
+| `2` | `CMD_HALT` | Shut down the VM |
+| `3` | `CMD_PANIC` | Shut down with guest panic |
+
+### Usage: halt
+
+```asm
+halt:
+    movea.l #0xff009004, %a1     @ REG_CMD
+    move.l  #2, (%a1)            @ CMD_HALT
+    rts
+```
+
+### Usage: reset
+
+```asm
+reset:
+    movea.l #0xff009004, %a1     @ REG_CMD
+    move.l  #1, (%a1)            @ CMD_RESET
+    rts
+```
+
+---
+
+## Goldfish PIC -- `0xFF000000` (+ idx * 0x1000)
+
+Programmable Interrupt Controller. Six PICs in the virt machine:
+
+| Base Address | PIC # | CPU IRQ | Devices |
+|---|---|---|---|
+| `0xFF000000` | 1 | #1 | virt-ctrl (IRQ#1), tty (IRQ#32) |
+| `0xFF001000` | 2 | #2 | virtio-mmio #1..#32 |
+| `0xFF002000` | 3 | #3 | virtio-mmio #33..#64 |
+| `0xFF003000` | 4 | #4 | virtio-mmio #65..#96 |
+| `0xFF004000` | 5 | #5 | virtio-mmio #97..#128 |
+| `0xFF005000` | 6 | #6 | RTC (IRQ#1) |
+
+### Registers
+
+| Offset | Name | Access | Description |
+|---|---|---|---|
+| `0x00` | `REG_STATUS` | RO | Count of pending+enabled IRQs (popcount) |
+| `0x04` | `REG_IRQ_PENDING` | RO | Bitmask of pending AND enabled IRQs |
+| `0x08` | `REG_IRQ_DISABLE_ALL` | WO | Disable ALL interrupts (clears pending too) |
+| `0x0c` | `REG_DISABLE` | WO | Disable specific IRQs (bitmask) |
+| `0x10` | `REG_ENABLE` | WO | Enable specific IRQs (bitmask) |
+
+Interrupts are level-sensitive. Each PIC manages 32 IRQ inputs (IRQ#1 = bit 0, IRQ#32 = bit 31).
+
+### Usage: enable IRQ in PIC
+
+```asm
+    @ Enable IRQ#32 (tty) in PIC#1 (0xFF000000)
+    movea.l #0xff000010, %a1     @ REG_ENABLE of PIC#1
+    move.l  #1 << 31, (%a1)      @ bit 31 = IRQ#32
+```
+
+### Usage: read pending IRQs
+
+```asm
+    movea.l #0xff000004, %a1     @ REG_IRQ_PENDING
+    move.l  (%a1), %d0           @ d0 = bitmask of pending IRQs
+    btst    #31, %d0             @ check if IRQ#32 (tty) is firing
+```
+
+### Usage: disable all interrupts (mask)
+
+```asm
+    movea.l #0xff000008, %a1     @ REG_IRQ_DISABLE_ALL
+    move.l  #0, (%a1)            @ any value disables all
+```
+
+---
+
+## Goldfish RTC -- `0xFF006000`
+
+Real-time clock with a 64-bit nanosecond counter. **Big-endian** on m68k (unlike other goldfish devices which are little-endian).
+
+### Registers
+
+| Offset | Name | Access | Description |
+|---|---|---|---|
+| `0x00` | `RTC_TIME_LOW` | RW | Time value, lower 32 bits |
+| `0x04` | `RTC_TIME_HIGH` | RW | Time value, upper 32 bits |
+| `0x08` | `RTC_ALARM_LOW` | RW | Alarm value, lower 32 bits |
+| `0x0c` | `RTC_ALARM_HIGH` | RW | Alarm value, upper 32 bits |
+| `0x10` | `RTC_IRQ_ENABLED` | RW | IRQ enable (bit 0) |
+| `0x14` | `RTC_CLEAR_ALARM` | WO | Clear the alarm |
+| `0x18` | `RTC_ALARM_STATUS` | RO | 1 if alarm is set |
+| `0x1c` | `RTC_CLEAR_INTERRUPT` | WO | Clear pending interrupt |
+
+Reading `RTC_TIME_LOW` latches the high 32 bits; the subsequent read of `RTC_TIME_HIGH` returns that latched value. This ensures atomic 64-bit time reads.
+
+Writing either `RTC_TIME_LOW` or `RTC_TIME_HIGH` adjusts the tick offset to set the time.
+
+The RTC interrupt goes to PIC#6 (CPU IRQ #6), IRQ #1.
+
+### Usage: read time
+
+```asm
+    movea.l #0xff006000, %a1     @ RTC_TIME_LOW
+    move.l  (%a1), %d0           @ d0 = low 32 bits (latches high)
+    movea.l #0xff006004, %a1     @ RTC_TIME_HIGH
+    move.l  (%a1), %d1           @ d1 = high 32 bits
+    @ now d1:d0 = 64-bit nanosecond counter
+```
+
+---
+
+## Virtio-MMIO -- `0xFF010000` (+ idx * 0x200)
+
+128 virtio-mmio transport devices. Each occupies 0x200 bytes.
+
+| Start | End | Slot Range |
+|---|---|---|
+| `0xFF010000` | `0xFF01FFFF` | 128 slots (0x200 each) |
+
+### Identification Registers (RO)
+
+| Offset | Name | Value | Description |
+|---|---|---|---|
+| `0x000` | `MAGIC_VALUE` | `0x74726976` ("virt") |
+| `0x004` | `VERSION` | 2 (modern) | 1 = legacy |
+| `0x008` | `DEVICE_ID` | varies | 0 = no device |
+| `0x00c` | `VENDOR_ID` | `0x554D4551` ("QEMU") |
+
+### Key Registers (32-bit)
+
+| Offset | Name | Access | Description |
+|---|---|---|---|
+| `0x010` | `DEVICE_FEATURES` | RO | Features offered by host (low 32) |
+| `0x014` | `DEVICE_FEATURES_SEL` | WO | Select feature bits (high/low) |
+| `0x020` | `DRIVER_FEATURES` | WO | Features accepted by driver (low 32) |
+| `0x024` | `DRIVER_FEATURES_SEL` | WO | Select driver feature bits |
+| `0x030` | `QUEUE_SEL` | WO | Select virtqueue |
+| `0x034` | `QUEUE_NUM_MAX` | RO | Max queue size |
+| `0x038` | `QUEUE_NUM` | WO | Set queue size |
+| `0x044` | `QUEUE_READY` | RW | Queue ready bit |
+| `0x050` | `QUEUE_NOTIFY` | WO | Kick the virtqueue |
+| `0x060` | `INTERRUPT_STATUS` | RO | Interrupt reason bits |
+| `0x064` | `INTERRUPT_ACK` | WO | Acknowledge interrupt |
+| `0x070` | `STATUS` | RW | Device status (driver negotiation) |
+| `0x080` | `QUEUE_DESC_LOW` | WO | Descriptor table address (low) |
+| `0x084` | `QUEUE_DESC_HIGH` | WO | Descriptor table address (high) |
+| `0x090` | `QUEUE_AVAIL_LOW` | WO | Available ring address (low) |
+| `0x094` | `QUEUE_AVAIL_HIGH` | WO | Available ring address (high) |
+| `0x0a0` | `QUEUE_USED_LOW` | WO | Used ring address (low) |
+| `0x0a4` | `QUEUE_USED_HIGH` | WO | Used ring address (high) |
+| `0x0fc` | `CONFIG_GENERATION` | RO | Config space version |
+| `0x100` | `CONFIG` | RW | Device-specific config space |
+
+### Interrupt Status Bits
+
+| Bit | Name | Description |
+|---|---|---|
+| `0` | `VIRTIO_MMIO_INT_VRING` | Used buffer notification |
+| `1` | `VIRTIO_MMIO_INT_CONFIG` | Config change notification |
+
+### Usage: detect virtio device
+
+```asm
+    movea.l #0xff010000, %a1     @ first virtio slot
+    move.l  0x000(%a1), %d0      @ MAGIC_VALUE
+    cmp.l   #0x74726976, %d0     @ "virt" ?
+    bne     no_device
+    move.l  0x008(%a1), %d0      @ DEVICE_ID
+    tst.l   %d0
+    beq     no_device
+    @ valid virtio device found, d0 = device ID
+```
+
+---
+
+## m68k Bootinfo Protocol
+
+Kernels can query device base addresses, memory layout, and machine properties via the m68k bootinfo protocol. QEMU places a bootinfo blob immediately after the kernel's ELF LOAD segment in guest physical memory.
+
+### Record Format
+
+Every bootinfo record has a 4-byte header followed by variable-length data:
+
+```
+struct bi_record {
+    uint16 tag;        // tag ID (big-endian)
+    uint16 size;       // total record size in bytes, including 4-byte header (big-endian)
+    uint8  data[size - 4];  // tag-specific data
+};
+```
+
+Records are contiguous (no padding between them). The list is terminated by a sentinel record with `tag = 0` (`BI_LAST`).
+
+### Locating the Bootinfo Blob
+
+QEMU loads the kernel ELF at `0x10000` (the `LOAD_ADDR`). The `_end` linker symbol marks the end of the LOAD segment. QEMU places the bootinfo at the next word-aligned address:
+
+```c
+parameters_base = (_end + 1) & ~1;
+```
+
+When writing a kernel in assembly, use:
+
+```asm
+    move.l  #_end, d0
+    addq.l  #1, d0
+    lsr.l   #1, d0
+    lsl.l   #1, d0
+    movea.l d0, a2        ; a2 = bootinfo base
+```
+
+### Parsing Loop
+
+```asm
+    movea.l d0, a2        ; a2 = bootinfo base
+1$:
+    clr.l   d0
+    move.w  (a2)+, d0     ; tag (2 bytes)
+    beq     3$            ; BI_LAST sentinel
+    clr.l   d1
+    move.w  (a2)+, d1     ; size = total record size (2 bytes)
+    move.l  a2, a3        ; a3 -> data portion
+    lea     (-4, a2, d1.l), a2  ; advance to next record
+    subq.l  #4, d1        ; d1 = data size (total - header)
+    ; handle tag d0 with data at a3 (d1 bytes)
+    bra     1$
+3$: ; done
+```
+
+**Important**: The `size` field stores the *total* record size (including the 4-byte header), not just the data size. To advance `a2` to the next record after reading the header, compute `a2 + size - 4`. To get the data payload size, compute `size - 4`.
+
+### Tag Definitions
+
+#### Standard Tags (Linux/m68k)
+
+Defined in Linux `include/uapi/asm-generic/bootinfo.h`:
+
+| Tag  | Name                | Data Size | Description                                |
+|------|---------------------|-----------|--------------------------------------------|
+| 0x0000 | `BI_LAST`          | 0         | End-of-list sentinel                       |
+| 0x0001 | `BI_MACHTYPE`      | 4         | Machine type (enum, see below)             |
+| 0x0002 | `BI_CPUTYPE`       | 4         | CPU type bitmask (see below)               |
+| 0x0003 | `BI_FPUTYPE`       | 4         | FPU type bitmask                           |
+| 0x0004 | `BI_MMUTYPE`       | 4         | MMU type bitmask                           |
+| 0x0005 | `BI_MEMCHUNK`      | 8         | Memory chunk: `{uint32 base, uint32 size}` |
+| 0x0006 | `BI_RAMDISK`       | 8         | Ramdisk: `{uint32 addr, uint32 size}`      |
+| 0x0007 | `BI_COMMAND_LINE`  | variable  | Null-terminated kernel command line string |
+| 0x0008 | `BI_RNG_SEED`      | 32        | Random seed for RNG initialization         |
+
+**BI_MACHTYPE values:**
+
+| Value | Machine    |
+|-------|------------|
+| 1     | Amiga      |
+| 2     | Atari      |
+| 3     | Mac        |
+| 4     | Apollo     |
+| 5     | Sun3       |
+| 6     | MVME147    |
+| 7     | MVME16x    |
+| 8     | BVME6000   |
+| 9     | HP300      |
+| 10    | Q40        |
+| 11    | Sun3x      |
+| 12    | M54xx      |
+| 13    | M5441x     |
+| 14    | **VIRT**   |
+
+**BI_CPUTYPE (bitmask):**
+
+| Bit | CPU       |
+|-----|-----------|
+| 0   | 68020     |
+| 1   | 68030     |
+| 2   | 68040     |
+| 3   | 68060     |
+| 4   | ColdFire  |
+
+#### Virt-Specific Tags (BI_VIRT_*)
+
+Defined in `bootinfo-virt.h`:
+
+| Tag    | Name                    | Data Size | Description                                      |
+|--------|-------------------------|-----------|--------------------------------------------------|
+| 0x8000 | `BI_VIRT_QEMU_VERSION` | 4         | QEMU version: `(major << 16) \| minor`           |
+| 0x8001 | `BI_VIRT_GF_PIC_BASE`  | 8         | Goldfish PIC base address + IRQ count            |
+| 0x8002 | `BI_VIRT_GF_RTC_BASE`  | 8         | Goldfish RTC base address + count                |
+| 0x8003 | `BI_VIRT_GF_TTY_BASE`  | 8         | Goldfish TTY base address + count                |
+| 0x8004 | `BI_VIRT_VIRTIO_BASE`  | 8         | Virtio-MMIO base address + mmio region size      |
+| 0x8005 | `BI_VIRT_CTRL_BASE`    | 4         | Virt controller base address                     |
+
+For tags 0x8001-0x8004, the data is two `uint32` values: `{base address, count_or_size}`.
+
+### Example: QEMU 11.0.0 on `m68020`
+
+Bootinfo blob (captured via QEMU monitor `xp/32x <addr>`):
+
+```
+00010490: 0x00010008 0x0000000e 0x00020008 0x00000001
+000104a0: 0x0005000c 0x00000000 0x08000000 0x80000008
+000104b0: 0x0b000000 0x8001000c 0xff000000 0x00000001
+000104c0: 0x8002000c 0xff006000 0x000000a8 0x8003000c
+000104d0: 0xff008000 0x00000027 0x8005000c 0xff009000
+000104e0: 0x00000008 0x8004000c 0xff010000 0x00000028
+000104f0: 0x00070008 0x00000000 0x00080028 0x0020ac0e
+00010500: 0xd72bb388 0x3ec5e0e0 0x5a1534b3 0xe295bd9a
+```
+
+Decoded (from `xp/40x 0x10490` with QEMU 11.0.0, m68020, 128 MiB RAM):
+
+| Offset | Tag                | Total | Data                                     | Meaning                           |
+|--------|---------------------|-------|------------------------------------------|-----------------------------------|
+| 0x10490 | BI_MACHTYPE        | 8     | 0x0000000e (14)                          | MACH_VIRT                         |
+| 0x10498 | BI_CPUTYPE         | 8     | 0x00000001 (bit 0)                       | CPU_68020                         |
+| 0x104a0 | BI_MEMCHUNK        | 12    | base=0x00000000, size=0x08000000         | 128 MiB RAM at 0x0                |
+| 0x104ac | BI_VIRT_QEMU_VER   | 8     | 0x0B000000 (major=11, minor=0)           | QEMU 11.0                         |
+| 0x104b4 | BI_VIRT_GF_PIC     | 12    | base=0xFF000000, count=1                 | 1 PIC instance                    |
+| 0x104c0 | BI_VIRT_GF_RTC     | 12    | base=0xFF006000, count=168               | 168 RTC regs                      |
+| 0x104cc | BI_VIRT_GF_TTY     | 12    | base=0xFF008000, count=39                | 39 TTY channels                   |
+| 0x104d8 | BI_VIRT_CTRL       | 12    | base=0xFF009000, count=8                 | CTRL regs + feature bits           |
+| 0x104e4 | BI_VIRT_VIRTIO     | 12    | base=0xFF010000, count=40                | 40 VIRTIO slots                   |
+| 0x104f0 | BI_COMMAND_LINE    | 8     | (empty)                                  | No kernel command line            |
+| 0x104f8 | BI_RNG_SEED        | 40    | 36 bytes random seed                     | RNG initialization seed           |
+| 0x10520 | BI_LAST            | 4     | --                                       | End-of-list sentinel              |
+
+Key observations:
+
+- The `size` field is the **total** record size including the 4-byte header.
+- Most virt device records use 12 bytes total (8 bytes data: `{base, count}`).
+- Singletons (MACHTYPE, CPUTYPE, QEMU_VERSION) use 8 bytes total (4 bytes data).
+- BI_COMMAND_LINE is always present (4 bytes of zero when empty).
+- BI_RNG_SEED: total=40 (0x28), data=36 bytes — the last record before the sentinel.
+- BI_LAST has `size=4` (a 4-byte all-zero header with tag=0).
+
+### Fixed Addresses vs Bootinfo
+
+Although the bootinfo protocol provides device addresses, the QEMU `-M virt` machine uses fixed addresses for all MMIO devices (see Memory Map above). A standalone kernel that skips bootinfo can hardcode these addresses directly.

codeGenM68k/src/prog8/codegen/m68k/AsmOptimizer.kt

@@ -0,0 +1,783 @@
+package prog8.codegen.m68k
+
+// M68k assembly-level peephole optimizer.
+//
+// It operates on a MutableList<String> of assembly lines. Only instructions that a peephole
+// actually matches are modified or removed; every other line (including all comments) is left
+// completely untouched. This keeps the generated .asm readable and comments intact.
+//
+// To add a new peephole:
+//   1. Write a private function `optimizeXxx(window: Sequence<List<TrimmedLine>>): List<Modification>`
+//      that inspects a sliding window of consecutive instructions and returns the Modifications
+//      it wants to apply.
+//   2. Register it in optimizeAssembly() by adding a line:
+//          byN = runPass(optimizeXxx(byN), N, byN)
+// The optimizer re-runs all passes until a full pass makes no changes (so cascading optimizations
+// are picked up too).
+
+internal fun optimizeAssembly(lines: MutableList<String>) {
+    val pretrimmed = lines.map { it.trimStart() }.toMutableList()
+
+    while (true) {
+        var modified = false
+
+        /** Runs a single optimization pass: applies any modifications, then recomputes the line windows. */
+        fun runPass(
+            mods: List<Modification>,
+            windowSize: Int,
+            currentLines: Sequence<List<TrimmedLine>>
+        ): Sequence<List<TrimmedLine>> {
+            if (mods.isNotEmpty()) {
+                applyModifications(mods, lines, pretrimmed)
+                modified = true
+                return getLinesBy(pretrimmed, lines, windowSize)
+            }
+            return currentLines
+        }
+
+        var by2 = getLinesBy(pretrimmed, lines, 2)
+        by2 = runPass(optimizeRedundantReload(by2), 2, by2)  // REGFILE-DEPENDENT
+        by2 = runPass(optimizeBounceToGlobal(by2, pretrimmed), 2, by2)  // REGFILE-DEPENDENT
+        by2 = runPass(optimizeJmpToNextLabel(by2), 2, by2)
+        by2 = runPass(optimizeTailCall(by2), 2, by2)
+        by2 = runPass(optimizeRedundantTst(by2), 2, by2)
+        by2 = runPass(optimizeMsigbSpill(by2, pretrimmed), 2, by2)  // REGFILE-DEPENDENT
+
+        // DBRA peephole for repeat loops is now obsolete: counted loops are emitted as IRLoopChunk and directly translated to dbra.
+        // The old REGFILE-DEPENDENT pattern (move #N,p8_regfile+slot / label: body / subq #1,slot / bne label) no longer occurs.
+
+        if (!modified)
+            break
+    }
+    // M3 is a full-file scan (not a 2-window peephole) and is idempotent; run once after the fixed-point loop
+    // REGFILE-DEPENDENT
+    val clampMods = optimizeClampImmediate(pretrimmed, lines)
+    if (clampMods.isNotEmpty()) {
+        applyModifications(clampMods, lines, pretrimmed)
+    }
+    // M4: fuse LOADX+JUMPI - eliminate the register file store/reload between array load and indirect jump
+    // REGFILE-DEPENDENT
+    val fuseMods = optimizeFuseLoadxJumpi(pretrimmed, lines)
+    if (fuseMods.isNotEmpty()) {
+        applyModifications(fuseMods, lines, pretrimmed)
+    }
+    // M5: drop a `lea base,a0` when a0 already holds that same base.
+    // NOT REGFILE-DEPENDENT: a0 is a real CPU register, so this pass is independent of the
+    // p8_regfile layout and of the planned register allocator.
+    val leaMods = optimizeRedundantLea(pretrimmed)
+    if (leaMods.isNotEmpty()) {
+        applyModifications(leaMods, lines, pretrimmed)
+    }
+}
+
+/**
+ * M5: remove `lea base,a0` instructions that recompute a base a0 is already known to hold.
+ *
+ * Indexed array access is emitted as `lea base,a0` followed by `(a0,d0.w*2)`. The lea itself is
+ * required - absolute-indexed addressing (`base(a0,d0.w*2)`) does not exist on the m68k - but in
+ * a sequence of accesses to the same array the base is reloaded several times over.
+ *
+ * A tracked value is dropped (a0 treated as unknown) whenever the scan sees:
+ *  - a label: incoming control flow may not be the fallthrough path
+ *  - any control transfer: a branch may reach elsewhere, and a call clobbers a0 (it is a
+ *    scratch register in the m68k calling convention)
+ *  - any instruction that names a0 outside a parenthesised address, i.e. writes it
+ *
+ * Anything else - including pure reads such as `move.w (a0,d0.w),d1` - leaves it untouched.
+ */
+private fun optimizeRedundantLea(allPretrimmed: List<String>): List<Modification> {
+    // Unconditional control flow, calls and returns end the tracked value: control leaves via a
+    // path where a0's contents are unknown (and a call clobbers a0, a scratch register in the
+    // m68k calling convention). The whole DBcc family is included as loop control.
+    // Plain condition-code branches (bne, bge, ...) are deliberately NOT in this set: they do not
+    // modify a0, and their target is always a label, which resets the tracked value below. So the
+    // fallthrough keeps the tracked value and the branch target is re-entered conservatively.
+    val controlTransfers = setOf(
+        "bra", "bsr", "jmp", "jsr", "rts", "rtr", "rte", "trap",
+        "dbcc", "dbcs", "dbhi", "dbhs", "dblo", "dbls", "dbne", "dbnz", "dbeq", "dbz",
+        "dbmi", "dbpl", "dbge", "dblt", "dbgt", "dble", "dbvc", "dbvs", "dbtst", "dbset",
+        "dbclr", "dbchg", "dbf", "dbra"
+    )
+    // A lea whose destination is a0, and the value it leaves there.
+    fun leaBase(instr: String): String? {
+        if (!instr.startsWith("lea ") && !instr.startsWith("lea\t"))
+            return null
+        val operands = instr.substringAfter(' ').substringBefore(';')
+        val parts = operands.split(',', limit = 2)
+        if (parts.size != 2 || parts[1].trim() != "a0")
+            return null
+        return parts[0].trim()
+    }
+
+    val filtered = allPretrimmed.withIndex()
+        .filter { it.value.isNotBlank() && !it.value.trimStart().startsWith(';') }
+        .map { TrimmedLine(it.value.trimStart(), it.value.trimStart(), it.index) }
+
+    val mods = mutableListOf<Modification>()
+    var a0Base: String? = null
+    for (line in filtered) {
+        val instr = line.instruction
+        if (hasLabel(line.value)) {
+            // fallthrough into a label is fine, but the label may also be entered from
+            // elsewhere, so the tracked value is no longer known to hold
+            a0Base = null
+            continue
+        }
+        val mnemonic = instr.substringBefore(' ').substringBefore('\t').lowercase()
+        if (mnemonic in controlTransfers) {
+            a0Base = null
+            continue
+        }
+        val base = leaBase(instr.lowercase())
+        if (base != null) {
+            if (base == a0Base)
+                mods.add(Modification(line.index, remove = true, replacement = null))
+            else
+                a0Base = base
+            continue
+        }
+        // Any non-parenthesised mention of a0 means it is written
+        if (instr.replace(parenthesised, "").contains("a0", ignoreCase = true))
+            a0Base = null
+    }
+    return mods
+}
+
+/** Matches a parenthesised group, for stripping addressing modes. */
+private val parenthesised = Regex("""\([^)]*\)""")
+private val d0IndexOperand = Regex("""\(\s*a0\s*,\s*d0\.w(?:\*\d+)?\s*\)""")
+
+/** A single edit to apply: either remove a line (keeping its label if present) or replace it. */
+private class Modification(val lineIndex: Int, val remove: Boolean, val replacement: String?)
+
+/** Pre-trimmed assembly line that also knows its original text and source line index. */
+private class TrimmedLine(val value: String, val trimmed: String, val index: Int) {
+    val instruction: String = trimmed.instructionPart()
+}
+
+private fun getLinesBy(pretrimmed: MutableList<String>, originalLines: MutableList<String>, windowSize: Int): Sequence<List<TrimmedLine>> =
+    pretrimmed.asSequence()
+        .withIndex()
+        .filter { it.value.isNotBlank() && !it.value.startsWith(';') }
+        .map { TrimmedLine(originalLines[it.index], it.value, it.index) }
+        .windowed(windowSize, partialWindows = false)
+
+private fun applyModifications(modifications: List<Modification>, lines: MutableList<String>, pretrimmed: MutableList<String>) {
+    for (modification in modifications.sortedBy { it.lineIndex }.reversed()) {
+        val idx = modification.lineIndex
+        if (modification.remove) {
+            val pretrim = pretrimmed.getOrNull(idx)
+            // If the line carries a label we must keep it (other code may jump to it);
+            // otherwise the line is dropped entirely.
+            if (pretrim.isNullOrBlank() || pretrim.startsWith(';')) {
+                lines.removeAt(idx)
+            } else if (hasLabel(pretrim)) {
+                val label = keepLabel(pretrim)
+                if (label.isNotEmpty())
+                    lines[idx] = label
+                else
+                    lines.removeAt(idx)
+            } else {
+                lines.removeAt(idx)
+            }
+            pretrimmed.removeAt(idx)
+        } else {
+            lines[idx] = modification.replacement!!
+            pretrimmed[idx] = modification.replacement.trimStart()
+        }
+    }
+}
+
+// === helpers ===
+
+/** Strips a leading label (e.g. "mylabel:  move.l ..." -> "move.l ..."). */
+private fun String.instructionPart(): String {
+    val idx = indexOf(':')
+    return if (idx >= 0) substring(idx + 1).trimStart() else this
+}
+
+private fun hasLabel(line: String): Boolean =
+    line.length > 1 && !line.startsWith(';') && ':' in line
+
+/** Returns just the "label:" part of a line, or "" if there is no label. */
+private fun keepLabel(line: String): String {
+    val idx = line.indexOf(':')
+    return if (idx >= 0) line.substring(0, idx + 1) else ""
+}
+
+// Deliberately only recognizes move.b/w/l (data register moves), not movea or movem:
+// movea always writes to an address register (rejected by isSafeMemoryOperand), and
+// movem uses register-list operands (also rejected). So they can never participate in
+// the patterns we optimize, and excluding them keeps moveOperands() simple.
+private fun String.isMove(): Boolean =
+    startsWith("move.b") || startsWith("move.w") || startsWith("move.l")
+
+/** Size suffix of a move instruction: 'b', 'w' or 'l'. */
+private fun String.moveSize(): Char = substringAfter('.').firstOrNull() ?: ' '
+
+/** Splits a move instruction "move.S  src, dst" into (src, dst), or null if it isn't a move. */
+private fun String.moveOperands(): Pair<String, String>? {
+    val sp = indexOf(' ')
+    if (sp < 0) return null
+    val rest = substring(sp + 1)
+    val commentIdx = rest.indexOf(';')
+    val ops = (if (commentIdx >= 0) rest.substring(0, commentIdx) else rest).split(',', limit = 2)
+    if (ops.size != 2) return null
+    return ops[0].trim() to ops[1].trim()
+}
+
+private fun String.isRegister(): Boolean =
+    this.matches(Regex("^(d[0-7]|a[0-7]|fp[0-9]+|sp|pc)$"))
+
+/**
+ * A plain memory operand that can be safely read twice and assumed to hit the same location
+ * (e.g. "p8_regfile+268", "sys.DOSBase", "$1234", "symbol"). Excludes data/address registers
+ * and any indirect/indexed addressing ((a0), (a0)+, ($1234,a0), ...) whose effective address may
+ * change between the two accesses.
+ */
+private fun String.isSafeMemoryOperand(): Boolean {
+    if (isEmpty()) return false
+    if (isRegister()) return false
+    if ('(' in this || ')' in this) return false
+    return true
+}
+
+// === peephole passes ===
+
+private fun optimizeRedundantReload(linesBy: Sequence<List<TrimmedLine>>): List<Modification> {
+    // REGFILE-DEPENDENT: this pattern is produced by the current regfile-based
+    // codegen (store to p8_regfile then immediately reload to d0). If values are
+    // kept in registers instead of spilled to p8_regfile, this peephole will
+    // rarely fire and can be removed or generalized.
+    //  move.S  d0, MEM
+    //  move.S  MEM, d0        ->  remove the second line
+    //
+    // The reload is redundant: d0 already holds the value (it was just stored from d0 and is
+    // unchanged between the two instructions), so dropping the reload leaves d0 intact and the
+    // store to MEM stays for any later readers.
+    // MEM must be a plain memory operand (see isSafeMemoryOperand) so both accesses hit the
+    // exact same location.
+    val mods = mutableListOf<Modification>()
+    for (lines in linesBy) {
+        val first = lines[0].instruction
+        val second = lines[1].instruction
+        // Never remove an instruction that is a jump target: arriving via the label, d0 is
+        // not guaranteed to hold the stored value, so the reload would be needed there.
+        if (hasLabel(lines[1].trimmed))
+            continue
+        if (first.isMove() && second.isMove() && first.moveSize() == second.moveSize()) {
+            val (src0, dst0) = first.moveOperands() ?: continue
+            val (src1, dst1) = second.moveOperands() ?: continue
+            if (src0 == "d0" && dst1 == "d0" && dst0 == src1 && dst0.isSafeMemoryOperand()) {
+                mods.add(Modification(lines[1].index, true, null))
+            }
+        }
+    }
+    return mods
+}
+
+private fun optimizeBounceToGlobal(linesBy: Sequence<List<TrimmedLine>>, allPretrimmed: List<String>): List<Modification> {
+    // REGFILE-DEPENDENT: explicitly matches p8_regfile spill followed by copy
+    // to a global. Depends on the current strategy of spilling all intermediate
+    // values to p8_regfile. If regfile is refactored, this bounce disappears.
+    //  move.S  d0, p8_regfile+OFF     ; spill result to regfile temp
+    //  move.S  p8_regfile+OFF, GLOBAL ; copy temp -> global
+    //  ->
+    //  move.S  d0, GLOBAL             ; store directly, skip regfile bounce
+    //
+    // Only safe when p8_regfile+OFF is NOT read again in the remainder of the current
+    // subroutine. We scan forward from the global store to the next subroutine boundary
+    // to confirm this.
+    val mods = mutableListOf<Modification>()
+    for (lines in linesBy) {
+        val first = lines[0].instruction
+        val second = lines[1].instruction
+        if (hasLabel(lines[0].trimmed) || hasLabel(lines[1].trimmed))
+            continue
+        if (first.isMove() && second.isMove() && first.moveSize() == second.moveSize()) {
+            val (src0, dst0) = first.moveOperands() ?: continue
+            val (src1, dst1) = second.moveOperands() ?: continue
+            if (src0 == "d0" && dst1 != "d0" && dst0 == src1
+                && dst0.isRegfileSlot() && dst1.isSafeMemoryOperand()
+                && !isSlotReadAfter(dst0, lines[1].index + 1, allPretrimmed)
+            ) {
+                val size = first.moveSize()
+                val indent = lines[1].value.takeWhile { it.isWhitespace() }
+                mods.add(Modification(lines[0].index, true, null))   // remove regfile spill
+                mods.add(Modification(lines[1].index, false, "${indent}move.$size  d0, $dst1"))  // direct store
+            }
+        }
+    }
+    return mods
+}
+
+private fun optimizeJmpToNextLabel(linesBy: Sequence<List<TrimmedLine>>): List<Modification> {
+    //  bra  LABEL
+    //  LABEL:          ->  remove the branch (keep any label on the branch line itself)
+    //
+    // The branch is dead code because the target label immediately follows.
+    // (also handles the older jmp form)
+    val mods = mutableListOf<Modification>()
+    for (lines in linesBy) {
+        val firstInstr = lines[0].instruction
+        val secondLine = lines[1].trimmed
+        val branchMnemonic = when {
+            firstInstr.startsWith("bra ") -> "bra"
+            firstInstr.startsWith("jmp ") -> "jmp"
+            else -> null
+        } ?: continue
+        val target = firstInstr.substringAfter("$branchMnemonic ").trim()
+        // Skip indirect jumps like jmp ([p8_regfile+12]) or jmp (a0)
+        if (target.startsWith("("))
+            continue
+        if (hasLabel(secondLine)) {
+            val label = keepLabel(secondLine).removeSuffix(":")
+            if (target == label) {
+                if (hasLabel(lines[0].trimmed)) {
+                    // The branch line also carries a label - keep the label, drop the branch
+                    mods.add(Modification(lines[0].index, false, keepLabel(lines[0].trimmed)))
+                } else {
+                    mods.add(Modification(lines[0].index, true, null))
+                }
+            }
+        }
+    }
+    return mods
+}
+
+private fun optimizeTailCall(linesBy: Sequence<List<TrimmedLine>>): List<Modification> {
+    //  bsr  LABEL
+    //  rts              ->  bra  LABEL
+    //
+    // When a subroutine call is immediately followed by a return, replace the call with a branch.
+    // The called routine will return directly to the caller's caller (tail call optimization).
+    // Only handles direct label targets, not indirect (bsr (a0) etc.).
+    // (also handles the older jsr form, normalizing it to bra as well)
+    val mods = mutableListOf<Modification>()
+    for (lines in linesBy) {
+        val firstInstr = lines[0].instruction
+        val secondInstr = lines[1].instruction
+        // Don't optimize if the rts line has a label (it's a jump target)
+        if (hasLabel(lines[1].trimmed))
+            continue
+        val callMnemonic = when {
+            firstInstr.startsWith("bsr ") -> "bsr"
+            firstInstr.startsWith("jsr ") -> "jsr"
+            else -> null
+        } ?: continue
+        if (secondInstr == "rts") {
+            val target = firstInstr.substringAfter("$callMnemonic ").trim()
+            // Skip indirect/library calls like bsr (a0), jsr (a0) or jsr dos.VPrintf(a6)
+            if ('(' in target)
+                continue
+            val indent = lines[0].value.takeWhile { it.isWhitespace() }
+            if (hasLabel(lines[0].trimmed)) {
+                // Keep the label, replace call with branch and remove rts
+                val label = keepLabel(lines[0].trimmed)
+                mods.add(Modification(lines[0].index, false, "$label\n${indent}bra  $target"))
+                mods.add(Modification(lines[1].index, true, null))
+            } else {
+                // Replace call with branch and remove rts
+                mods.add(Modification(lines[0].index, false, "${indent}bra  $target"))
+                mods.add(Modification(lines[1].index, true, null))
+            }
+        }
+    }
+    return mods
+}
+
+private fun optimizeRedundantTst(linesBy: Sequence<List<TrimmedLine>>): List<Modification> {
+    //  move.S  SRC, DST
+    //  tst.S   DST              ->  remove the tst
+    //
+    //  andi.S  #imm, DST
+    //  tst.S   DST              ->  remove the tst
+    //
+    //  ori.S   #imm, DST
+    //  tst.S   DST              ->  remove the tst
+    //
+    // On M68k, move/andi/ori/eori set N, Z based on the result. So a tst
+    // immediately after to the same location is redundant.
+    val mods = mutableListOf<Modification>()
+    for (lines in linesBy) {
+        val first = lines[0].instruction
+        val second = lines[1].instruction
+        // Don't remove tst if it's a jump target
+        if (hasLabel(lines[1].trimmed))
+            continue
+        if (!second.startsWith("tst."))
+            continue
+        val dst0 = first.flagSettingDest() ?: continue
+        if (first.moveSize() != second.moveSize())
+            continue
+        val tstRaw = second.substringAfter('.').substringAfter(' ').trim()
+        val tstOperand = if (';' in tstRaw) tstRaw.substringBefore(';').trim() else tstRaw
+        if (dst0 == tstOperand && dst0.isSafeMemoryOperand()) {
+            mods.add(Modification(lines[1].index, true, null))
+        }
+    }
+    return mods
+}
+
+/**
+ * If this instruction sets N/Z flags and writes to a single destination, return that destination.
+ * Covers move, andi, ori, eori (the ones the codegen emits before tst).
+ * Note: add/sub also set condition codes, but the codegen always inserts a move before branching,
+ * so add/sub are never followed by a redundant tst.
+ */
+private fun String.flagSettingDest(): String? {
+    val stripped = if (';' in this) substring(0, indexOf(';')).trim() else this
+    return when {
+        isMove() -> moveOperands()?.second
+        startsWith("andi.") || startsWith("ori.") || startsWith("eori.") -> {
+            // andi.S #imm, dst
+            val sp = stripped.indexOf(' ')
+            if (sp < 0) return null
+            val ops = stripped.substring(sp + 1).split(',', limit = 2)
+            if (ops.size != 2) return null
+            ops[1].trim()
+        }
+        else -> null
+    }
+}
+
+private fun optimizeMsigbSpill(linesBy: Sequence<List<TrimmedLine>>, allPretrimmed: List<String>): List<Modification> {
+    // REGFILE-DEPENDENT: optimizes the spill of a word/long to p8_regfile
+    // followed by a byte read of its MSB. This spill is an artifact of the
+    // current regfile-based codegen and big-endian layout. If values stay in
+    // registers, the spill is never emitted.
+    //  move.w/l  SYMBOL, p8_regfile+N   ; spill word/long to slot
+    //  move.b    p8_regfile+N, d0       ; read MSB
+    //  ->  move.b  SYMBOL, d0           ; big-endian: MSB is at SYMBOL+0, no spill needed
+    // Keep the spill only if the slot is read again later in the subroutine.
+    val mods = mutableListOf<Modification>()
+    for (lines in linesBy) {
+        if (hasLabel(lines[0].trimmed) || hasLabel(lines[1].trimmed)) continue
+        val first = lines[0].instruction
+        val second = lines[1].instruction
+        if (!first.isMove() || !second.isMove()) continue
+        if (second.moveSize() != 'b') continue
+        val firstSize = first.moveSize()
+        if (firstSize != 'w' && firstSize != 'l') continue
+        val (src0, dst0) = first.moveOperands() ?: continue
+        val (src1, dst1) = second.moveOperands() ?: continue
+        if (dst1 != "d0") continue
+        if (src1 != dst0) continue
+        if (!dst0.isRegfileSlot()) continue
+        if (src0.isRegfileSlot() || src0.isRegister()) continue
+        if ('(' in src0 || ')' in src0) continue
+        // src0 must be a plain memory operand (symbol)
+        if (!src0.isSafeMemoryOperand()) continue
+        // If the slot is read again later, keep the spill; otherwise we can remove it
+        val indent2 = lines[1].value.takeWhile { it.isWhitespace() }
+        if (isSlotReadAfter(dst0, lines[1].index + 1, allPretrimmed)) {
+            // keep spill, just rewrite the byte load to read directly from the symbol
+            mods.add(Modification(lines[1].index, false, "${indent2}move.b  $src0,d0"))
+        } else {
+            mods.add(Modification(lines[0].index, true, null))
+            mods.add(Modification(lines[1].index, false, "${indent2}move.b  $src0,d0"))
+        }
+    }
+    return mods
+}
+
+private fun optimizeClampImmediate(allPretrimmed: List<String>, lines: MutableList<String>): List<Modification> {
+    // REGFILE-DEPENDENT: folds clamp bounds that were materialized as
+    // move #imm,p8_regfile+SLOT. Depends on the current codegen that spills
+    // clamp immediates to regfile slots before the compare block. If immediates
+    // are kept in registers or encoded as cmpi, this pattern disappears.
+    // M3: fold clamp bounds that were materialized as move #imm,SLOT.
+    // Only fold when we can identify the full clamp block, to avoid touching unrelated compares.
+    // Pattern (filtered lines, b/w/l size must match):
+    //   cmp.S  SLOT_MIN,d0
+    //   bge/bhs  .clamp_max
+    //   move.S SLOT_MIN,d0
+    //   bra  .clamp_done
+    //   .clamp_max:
+    //   cmp.S  SLOT_MAX,d0
+    //   ble/bls  .clamp_done
+    //   move.S SLOT_MAX,d0
+    //   .clamp_done:
+    // where SLOT_MIN/MAX were defined as move.S #imm,SLOT within a few lines before the block.
+    val filtered = allPretrimmed.withIndex()
+        .filter { it.value.isNotBlank() && !it.value.trimStart().startsWith(';') }
+        .map { TrimmedLine(lines[it.index], it.value.trimStart(), it.index) }
+
+    fun findDefine(slot: String, size: Char, useIdx: Int, limit: Int = 6): Pair<Int,String>? {
+        var count = 0
+        for (k in useIdx-1 downTo 0) {
+            if (count >= limit) break
+            count++
+            val cand = filtered[k]
+            val instr = cand.instruction
+            if (!instr.isMove()) continue
+            if (instr.moveSize() != size) continue
+            val (src, dst) = instr.moveOperands() ?: continue
+            if (dst != slot) continue
+            if (!src.startsWith("#")) continue
+            var overwritten = false
+            for (t in k+1 until useIdx) {
+                val mid = filtered[t]
+                if (mid.instruction.isMove()) {
+                    val ops = mid.instruction.moveOperands()
+                    if (ops != null && ops.second == slot) { overwritten = true; break }
+                }
+            }
+            if (!overwritten) return cand.index to src
+        }
+        return null
+    }
+
+    val mods = mutableListOf<Modification>()
+    // Scan for clamp blocks
+    for (idx in filtered.indices) {
+        val instr = filtered[idx].instruction
+        if (!instr.startsWith("cmp.")) continue
+        // need at least 8 more filtered lines for a full block
+        if (idx + 8 >= filtered.size) continue
+        val sizeMin = instr[4]
+        if (sizeMin != 'b' && sizeMin != 'w' && sizeMin != 'l') continue
+        val cmpMinParts = instr.substringAfter(' ').trim().split(',', limit = 2).map { it.trim().substringBefore(';').trim() }
+        if (cmpMinParts.size != 2 || cmpMinParts[1] != "d0" || !cmpMinParts[0].isRegfileSlot()) continue
+        val slotMin = cmpMinParts[0]
+
+        val bgeInstr = filtered[idx+1].instruction
+        if (!bgeInstr.startsWith("bge ") && !bgeInstr.startsWith("bhs ")) continue
+        // bge target should be the next label
+        val labelMax = bgeInstr.substringAfter(' ').trim().substringBefore(';').trim()
+
+        val moveMinInstr = filtered[idx+2].instruction
+        if (!moveMinInstr.startsWith("move.")) continue
+        if (moveMinInstr.moveSize() != sizeMin) continue
+        val moveMinOps = moveMinInstr.moveOperands() ?: continue
+        if (moveMinOps.first != slotMin || moveMinOps.second != "d0") continue
+
+        val braInstr = filtered[idx+3].instruction
+        if (!braInstr.startsWith("bra ")) continue
+        val labelDone = braInstr.substringAfter(' ').trim().substringBefore(';').trim()
+
+        val labelMaxLine = filtered[idx+4].trimmed
+        if (!labelMaxLine.startsWith("$labelMax:")) continue
+
+        val cmpMaxInstr = filtered[idx+5].instruction
+        if (!cmpMaxInstr.startsWith("cmp.")) continue
+        if (cmpMaxInstr[4] != sizeMin) continue
+        val cmpMaxParts = cmpMaxInstr.substringAfter(' ').trim().split(',', limit = 2).map { it.trim().substringBefore(';').trim() }
+        if (cmpMaxParts.size != 2 || cmpMaxParts[1] != "d0" || !cmpMaxParts[0].isRegfileSlot()) continue
+        val slotMax = cmpMaxParts[0]
+
+        val bleInstr = filtered[idx+6].instruction
+        if (!bleInstr.startsWith("ble ") && !bleInstr.startsWith("bls ")) continue
+        val bleTarget = bleInstr.substringAfter(' ').trim().substringBefore(';').trim()
+        if (bleTarget != labelDone) continue
+
+        val moveMaxInstr = filtered[idx+7].instruction
+        if (!moveMaxInstr.startsWith("move.")) continue
+        if (moveMaxInstr.moveSize() != sizeMin) continue
+        val moveMaxOps = moveMaxInstr.moveOperands() ?: continue
+        if (moveMaxOps.first != slotMax || moveMaxOps.second != "d0") continue
+
+        val labelDoneLine = filtered[idx+8].trimmed
+        if (!labelDoneLine.startsWith("$labelDone:")) continue
+
+        // Found a clamp block; try to fold each bound if its slot has an immediate define
+        val defineMin = findDefine(slotMin, sizeMin, idx)
+        if (defineMin != null) {
+            val immMin = defineMin.second
+            val indentCmp = filtered[idx].value.takeWhile { it.isWhitespace() }
+            val indentMove = filtered[idx+2].value.takeWhile { it.isWhitespace() }
+            mods.add(Modification(filtered[idx].index, false, "${indentCmp}cmpi.$sizeMin  $immMin,d0"))
+            mods.add(Modification(filtered[idx+2].index, false, "${indentMove}move.$sizeMin  $immMin,d0"))
+            // If the slot is not read after the clamp block, the defining move is dead
+            val defineIdx = defineMin.first
+            if (!isSlotReadAfter(slotMin, filtered[idx+8].index + 1, allPretrimmed)) {
+                mods.add(Modification(defineIdx, true, null))
+            }
+        }
+        val defineMax = findDefine(slotMax, sizeMin, idx+5)
+        if (defineMax != null) {
+            val immMax = defineMax.second
+            val indentCmp2 = filtered[idx+5].value.takeWhile { it.isWhitespace() }
+            val indentMove2 = filtered[idx+7].value.takeWhile { it.isWhitespace() }
+            mods.add(Modification(filtered[idx+5].index, false, "${indentCmp2}cmpi.$sizeMin  $immMax,d0"))
+            mods.add(Modification(filtered[idx+7].index, false, "${indentMove2}move.$sizeMin  $immMax,d0"))
+            val defineIdxMax = defineMax.first
+            if (!isSlotReadAfter(slotMax, filtered[idx+8].index + 1, allPretrimmed)) {
+                mods.add(Modification(defineIdxMax, true, null))
+            }
+        }
+    }
+    return mods
+}
+
+// REGFILE-DEPENDENT helper: identifies operands that refer to the current
+// p8_regfile spill area. If the regfile is refactored, this predicate and
+// all peepholes that use it must be updated.
+/** True if this operand is a prog8 register-file slot (e.g. "p8_regfile+268", "p8_fregfile+12"). */
+private fun String.isRegfileSlot(): Boolean =
+    startsWith("p8_regfile+") || startsWith("p8_fregfile+")
+
+// REGFILE-DEPENDENT helper: scans for future reads of a regfile slot.
+// Used by bounce/msigb/clamp/fuse peepholes to decide if a spill can be removed.
+/**
+ * Scans forward from [startIndex] to the next subroutine boundary (or end of file) and
+ * returns true if [slot] is referenced in any instruction line in that range.
+ * A "reference" means the slot string appears as a standalone operand (not as a prefix
+ * of a longer offset like p8_regfile+640 when checking p8_regfile+64).
+ */
+private fun isSlotReadAfter(slot: String, startIndex: Int, allPretrimmed: List<String>): Boolean {
+    for (i in startIndex until allPretrimmed.size) {
+        val line = allPretrimmed[i]
+        if (line.startsWith("; ---- Subroutine:") || line.startsWith("; End of subroutine:")
+            || line.startsWith("prog8_program_end"))
+            break
+        if (line.isBlank() || line.startsWith(';')) continue
+        if (isSlotReferencedInLine(slot, line)) return true
+    }
+    return false
+}
+
+// REGFILE-DEPENDENT helper
+private fun isSlotReferencedInLine(slot: String, line: String): Boolean {
+    var from = 0
+    while (true) {
+        val idx = line.indexOf(slot, from)
+        if (idx < 0) return false
+        val after = idx + slot.length
+        // Ensure it's not a prefix of a longer offset (e.g. p8_regfile+6 vs p8_regfile+64)
+        if (after < line.length && line[after].isDigit()) {
+            from = after
+            continue
+        }
+        return true
+    }
+}
+
+/**
+ * REGFILE-DEPENDENT: fuses the LOADX+JUMPI / LOADX+CALLI sequence for
+ * `on .. goto` / `on .. call` / `when` dispatch. The codegen emits an index
+ * load, table lookup into d0, spill to p8_regfile, reload to a0 and indirect
+ * jump/call. If the regfile is refactored so the table lookup can target a0
+ * (or an address register) directly, this spill/reload disappears and the
+ * peephole becomes dead code.
+ *
+ * Current codegen (byte index, 68000-style JUMPI/CALLI):
+ *   moveq   #0, d0
+ *   move.b  p8_regfile+N, d0     ; or move.w p8_regfile+N,d0 for word index
+ *   lea     table_label, a0
+ *   move.l  (a0,d0.w), d0        ; table lookup (unscaled, see scaled-indexing-IR.md)
+ *   move.l  d0, p8_regfile+M     ; spill (LOADX tail)
+ *   movea.l p8_regfile+M, a0     ; reload (JUMPI/CALLI head, 68000)
+ *   jmp/jsr (a0)
+ * or for 68020+ JUMPI/CALLI:
+ *   jmp/jsr ([p8_regfile+M])     ; memory-indirect
+ *
+ * The same sequence with the scale folded into the addressing mode instead of
+ * into d0 (68020+; this is what a pointer-array dispatch such as on..goto emits):
+ *   lea     table_label, a0
+ *   move.l  (a0,d0.w*4), d0
+ *   move.l  d0, p8_regfile+M
+ *   movea.l p8_regfile+M, a0
+ *   jmp/jsr (a0)
+ *
+ * Fused, keeping whichever index operand was matched so the scale is preserved:
+ *   moveq   #0, d0
+ *   move.b  p8_regfile+N, d0
+ *   lea     table_label, a0
+ *   move.l  (a0,d0.w), a0        ; or (a0,d0.w*4), a0 on 68020+
+ *   jmp/jsr (a0)
+ */
+private fun optimizeFuseLoadxJumpi(allPretrimmed: List<String>, lines: MutableList<String>): List<Modification> {
+    val filtered = allPretrimmed.withIndex()
+        .filter { it.value.isNotBlank() && !it.value.trimStart().startsWith(';') }
+        .map { TrimmedLine(lines[it.index], it.value.trimStart(), it.index) }
+
+    val mods = mutableListOf<Modification>()
+    val alreadyModified = mutableSetOf<Int>()
+
+    fun loadIndexedOperand(instr: String): String? {
+        // matches "move.l  (a0,d0.w), d0" (68000, index pre-scaled into d0) and the scaled
+        // "move.l  (a0,d0.w*4), d0" forms (68020+, scale folded into the addressing mode).
+        // The matched operand is returned so the fused instruction keeps the same scale.
+        if (!instr.startsWith("move.l")) return null
+        if (!(instr.trim().endsWith(", d0") || instr.trim().endsWith(",d0"))) return null
+        return d0IndexOperand.find(instr)?.value
+    }
+
+    fun isSpill(instr: String): String? {
+        // "move.l  d0, p8_regfile+M"
+        if (!instr.startsWith("move.l")) return null
+        val ops = instr.moveOperands() ?: return null
+        if (ops.first != "d0") return null
+        if (!ops.second.isRegfileSlot()) return null
+        return ops.second.removePrefix("p8_regfile+").substringBefore(',').substringBefore(' ').trim()
+    }
+
+    for (leaIdx in filtered.indices) {
+        if (filtered[leaIdx].index in alreadyModified) continue
+        val leaInstr = filtered[leaIdx].instruction
+        if (!leaInstr.startsWith("lea ")) continue
+        if (!leaInstr.endsWith(",a0")) continue
+
+        // The index is usually already in d0 at this point (scaled via lsl.w #2
+        // or zero-extended). Earlier peepholes may have removed the explicit
+        // reload of the scaled index from p8_regfile, so we do not require a
+        // preceding moveq/move.b or move.w pattern. We only verify the lea
+        // itself and the following LOADX+JUMPI spill/jump sequence.
+
+        // Need at least: lea, loadIndexed, spill, jump/reload
+        if (leaIdx + 2 >= filtered.size) continue
+        val loadIdx = leaIdx + 1
+        val spillIdx = leaIdx + 2
+        if (filtered[loadIdx].index in alreadyModified || filtered[spillIdx].index in alreadyModified) continue
+        val loadOperand = loadIndexedOperand(filtered[loadIdx].instruction) ?: continue
+        val slotM = isSpill(filtered[spillIdx].instruction) ?: continue
+
+        // Check jump/call variant: 68000 uses movea reload + jmp/jsr (a0),
+        // 68020 uses memory-indirect jmp/jsr ([slot])
+        val nextIdx = leaIdx + 3
+        if (nextIdx >= filtered.size) continue
+        val nextInstr = filtered[nextIdx].instruction
+        val is68000Reload = nextInstr.startsWith("movea.l") && slotM in nextInstr && nextInstr.trim().endsWith(",a0")
+        val is68020MemIndirectJmp = nextInstr.startsWith("jmp") && "([p8_regfile+$slotM" in nextInstr.replace(" ", "")
+        val is68020MemIndirectJsr = nextInstr.startsWith("jsr") && "([p8_regfile+$slotM" in nextInstr.replace(" ", "")
+
+        if (is68000Reload) {
+            // Need jmp/jsr (a0) after the reload
+            if (nextIdx + 1 >= filtered.size) continue
+            val jmpIdx = nextIdx + 1
+            if (filtered[jmpIdx].index in alreadyModified) continue
+            val jmpInstr = filtered[jmpIdx].instruction
+            val isJmp = jmpInstr == "jmp  (a0)" || jmpInstr.replace(" ", "") == "jmp(a0)"
+            val isJsr = jmpInstr.startsWith("jsr") && "(a0)" in jmpInstr.replace(" ", "")
+            if (!isJmp && !isJsr) continue
+            // No labels on lea, load, spill, reload, jmp/jsr
+            if ((0..1).any { hasLabel(filtered[loadIdx + it].trimmed) }) continue
+            if (hasLabel(filtered[leaIdx].trimmed)) continue
+            if (hasLabel(filtered[spillIdx].trimmed) || hasLabel(filtered[nextIdx].trimmed) || hasLabel(filtered[jmpIdx].trimmed)) continue
+
+            val loadIndent = lines[filtered[loadIdx].index].takeWhile { it.isWhitespace() }
+            mods.add(Modification(filtered[loadIdx].index, false, "${loadIndent}move.l  $loadOperand, a0"))
+            mods.add(Modification(filtered[spillIdx].index, true, null))
+            mods.add(Modification(filtered[nextIdx].index, true, null))
+            alreadyModified.add(filtered[loadIdx].index)
+            alreadyModified.add(filtered[spillIdx].index)
+            alreadyModified.add(filtered[nextIdx].index)
+            // jmp/jsr stays as is
+        } else if (is68020MemIndirectJmp || is68020MemIndirectJsr) {
+            if (hasLabel(filtered[spillIdx].trimmed) || hasLabel(filtered[nextIdx].trimmed)) continue
+            if (hasLabel(filtered[leaIdx].trimmed) || hasLabel(filtered[loadIdx].trimmed)) continue
+            val loadIndent = lines[filtered[loadIdx].index].takeWhile { it.isWhitespace() }
+            val jmpIndent = lines[filtered[nextIdx].index].takeWhile { it.isWhitespace() }
+            val mnemonic = if (is68020MemIndirectJsr) "jsr" else "jmp"
+            mods.add(Modification(filtered[loadIdx].index, false, "${loadIndent}move.l  $loadOperand, a0"))
+            mods.add(Modification(filtered[spillIdx].index, true, null))
+            mods.add(Modification(filtered[nextIdx].index, false, "${jmpIndent}$mnemonic  (a0)"))
+            alreadyModified.add(filtered[loadIdx].index)
+            alreadyModified.add(filtered[spillIdx].index)
+            alreadyModified.add(filtered[nextIdx].index)
+        }
+    }
+
+    return mods
+}

codeGenM68k/src/prog8/codegen/m68k/AssemblyProgramM68k.kt

@@ -0,0 +1,182 @@
+package prog8.codegen.m68k
+
+import prog8.code.assembly.IAssemblyProgram
+import prog8.code.core.*
+import prog8.code.target.Amiga1200Target
+import prog8.code.target.Amiga500Target
+import prog8.code.target.Qemu68kTarget
+import java.nio.file.Path
+import kotlin.io.path.deleteIfExists
+import kotlin.io.path.exists
+import kotlin.io.path.readText
+import kotlin.io.path.writeText
+
+class AssemblyProgramM68k(override val name: String) : IAssemblyProgram {
+
+    override val irInstructionCount: Int = 0
+    override val irChunkCount: Int = 0
+    override val irRegisterCount: Int = 0
+
+    private fun elfFile(outs: OutputFiles): Path = outs.programFile
+
+    /**
+     * A target supplies its own memory layout as a "link.ld" file.  A target defined by a
+     * config file (i.e. living outside of the compiler) keeps it in its library directory, so
+     * that is looked at first.  Targets built into the compiler are looked up as a
+     * /prog8lib/<targetname>/link.ld resource.  Anything else falls back to the qemu68k layout.
+     */
+    private fun resolveLinkerScript(target: ICompilationTarget): String {
+        target.libraryPath?.resolve("link.ld")?.takeIf { it.exists() }?.let { return it.readText() }
+        val targetResource = "/prog8lib/${target.name}/link.ld"
+        AssemblyProgramM68k::class.java.getResource(targetResource)?.let { return it.readText() }
+        return AssemblyProgramM68k::class.java.getResource("/prog8lib/${Qemu68kTarget.NAME}/link.ld")?.readText()
+            ?: error("cannot find $targetResource resource, and no qemu68k fallback either")
+    }
+
+    private fun runProcess(command: List<String>, quiet: Boolean, tool: String? = null): Boolean {
+        val proc = ProcessBuilder(command).redirectErrorStream(true)
+        if (quiet)
+            proc.redirectOutput(ProcessBuilder.Redirect.DISCARD)
+        val process = try {
+            proc.start()
+        } catch (e: Exception) {
+            when (tool) {
+                "vasm" -> {
+                    System.err.println("Cannot find '${command[0]}' (vasm assembler for m68k). Install it via your package manager if it's on there, or build it from source: http://sun.hasenbraten.de/vasm/")
+                }
+                "vlink" -> {
+                    System.err.println("Cannot find 'vlink' (linker). Install it via your package manager if it's on there, or build it from source: http://sun.hasenbraten.de/vlink/")
+                }
+                else -> {
+                    System.err.println("process failed to start: ${e.message}")
+                }
+            }
+            return false
+        }
+        if (!quiet) {
+            process.inputStream.bufferedReader().use { reader ->
+                reader.forEachLine { println(it) }
+            }
+        }
+        return process.waitFor() == 0
+    }
+
+    override fun assemble(options: CompilationOptions, errors: IErrorReporter): Boolean {
+        val cpu = when(options.compTarget.cpu) {
+            CpuType.M68000 -> "68000"
+            CpuType.M68020 -> "68020"
+            else -> error("invalid cpu type for m68k codegen ${options.compTarget.cpu}")
+        }
+        val assemblerCpu = if(options.compTarget.name == "amiga500" && options.floats) "68020" else cpu
+        val outs = OutputFiles.of(options, name)
+
+        when(options.output) {
+            OutputType.ELF -> {
+                // Step 1: assemble to ELF object file
+                val objFile = outs.obj()
+                val listFile = outs.list()
+                val assembleCmd = mutableListOf(
+                    "vasmm68k_mot",
+                    "-m$assemblerCpu",
+                    "-m68881",  // enable FPU
+                    "-Felf",
+                    "-opt-speed",
+                    "-warnunaligned",
+                    "-ldots",
+                    "-spaces",
+                    "-o", objFile.toString(),
+                    outs.asm().toString()
+                )
+                if (options.asmListfile)
+                    assembleCmd.addAll(listOf("-L", listFile.toString()))
+                if (options.asmQuiet)
+                    assembleCmd.add("-quiet")
+                if (!runProcess(assembleCmd, options.quiet, "vasm"))
+                    return false
+                // clean up any leftover ELF/obj files from previous builds
+                outs.binFile().deleteIfExists()
+
+                // Step 2: write linker script and link to ELF executable
+                val linkScript = outs.linkScript()
+                val elfFile = elfFile(outs)
+                linkScript.writeText(resolveLinkerScript(options.compTarget))
+
+                val linkCmd = listOf(
+                    "vlink",
+                    "-b", "elf32m68k",
+                    "-n",
+                    "-T", linkScript.toString(),
+                    "-o", elfFile.toString(),
+                    objFile.toString()
+                )
+                val linkOk = runProcess(linkCmd, options.quiet, "vlink")
+                linkScript.deleteIfExists()
+                if(linkOk && !options.quiet)
+                    println("Executable written to '$elfFile'")
+                return linkOk
+            }
+            OutputType.AMIGAHUNK -> {
+                // Step 1: assemble directly to AmigaHunk executable file
+                val exefile = outs.programFile
+                val listfile = outs.list()
+                val assembleCmd = when(options.compTarget) {
+                    is Amiga500Target -> {
+                        // amiga 500 with kickstart 1.3
+                        mutableListOf(
+                            "vasmm68k_mot",
+                            "-m$assemblerCpu",
+                            *if(options.floats) arrayOf("-m68881") else emptyArray(),
+                            "-Fhunkexe",
+                            "-kick1hunks",   // old hunk format compatible with AmigaDOS 1.3
+                            "-opt-speed",
+                            "-warnunaligned",
+                            "-ldots",
+                            "-spaces",
+                            "-nosym",       // no debug symbols
+                            "-o", exefile.toString(),
+                            outs.asm().toString()
+                        ).also { cmd ->
+                            if (options.asmListfile) {
+                                cmd.add("-L")
+                                cmd.add(listfile.toString())
+                                cmd.add("-Lns")
+                            }
+                            if (options.asmQuiet)
+                                cmd.add("-quiet")
+                        }
+                    }
+                    is Amiga1200Target -> {
+                        // amiga 1200 with 68020 and optional FPU
+                        mutableListOf(
+                            "vasmm68k_mot",
+                            "-m$assemblerCpu",
+                            "-m68881",  // enable FPU
+                            "-Fhunkexe",
+                            "-opt-speed",
+                            "-warnunaligned",
+                            "-ldots",
+                            "-spaces",
+                            "-nosym",       // no debug symbols
+                            "-o", exefile.toString(),
+                            outs.asm().toString()
+                        ).also { cmd ->
+                            if (options.asmListfile) {
+                                cmd.add("-L")
+                                cmd.add(listfile.toString())
+                            }
+                            if (options.asmQuiet)
+                                cmd.add("-quiet")
+                        }
+                    }
+                    else -> error("unsupported target for AMIGAHUNK output: ${options.compTarget.name}")
+                }
+                if (!runProcess(assembleCmd, options.quiet, "vasm"))
+                    return false
+                if(!options.quiet)
+                    println("Executable written to '$exefile'")
+                return true
+            }
+            else -> error("Unsupported output type: ${options.output}")
+        }
+    }
+}

codeGenM68k/src/prog8/codegen/m68k/InstrArithmetic.kt

@@ -0,0 +1,820 @@
+package prog8.codegen.m68k
+
+import prog8.code.core.CpuType
+import prog8.intermediate.*
+
+internal fun AsmGen.translateArithmetic(insn: IRInstruction) {
+    val type = insn.type ?: IRDataType.BYTE
+
+    if (type == IRDataType.FLOAT) {
+        translateFloatArithmetic(insn)
+        return
+    }
+
+    when (insn.opcode) {
+        Opcode.INC -> {
+            val reg = insn.requireIntDest().intNumber
+            emitLine("addq${dtSuffix(type)}  #1, ${regAddr(reg)}")
+            invalidateD0CacheForSlot(reg)
+        }
+
+        Opcode.INCM -> {
+            val target = resolveMemory(insn.requireMemory())
+            emitLine("addq${dtSuffix(type)}  #1, $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.DEC -> {
+            val reg = insn.requireIntDest().intNumber
+            emitLine("subq${dtSuffix(type)}  #1, ${regAddr(reg)}")
+            invalidateD0CacheForSlot(reg)
+        }
+
+        Opcode.DECM -> {
+            val target = resolveMemory(insn.requireMemory())
+            emitLine("subq${dtSuffix(type)}  #1, $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.NEG -> {
+            val reg = insn.requireIntDest().intNumber
+            emitLine("neg${dtSuffix(type)}  ${regAddr(reg)}")
+            invalidateD0CacheForSlot(reg)
+        }
+
+        Opcode.NEGM -> {
+            val target = resolveMemory(insn.requireMemory())
+            emitLine("neg${dtSuffix(type)}  $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.ADDR -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            loadRegOrZeroExtendToD0(srcReg, type)
+            emitLine("add${dtSuffix(type)}  d0, ${regAddr(dstReg)}")
+            invalidateD0CacheForSlot(dstReg)
+        }
+
+        Opcode.ADD -> {
+            val reg = insn.requireIntDest().intNumber
+            val value = insn.requireImmediateInt()
+            if(value in 1..8) {
+                emitLine("addq${dtSuffix(type)}  #$value, ${regAddr(reg)}")
+            } else {
+                emitLine("add${dtSuffix(type)}  #${immVal(value, type)}, ${regAddr(reg)}")
+            }
+            invalidateD0CacheForSlot(reg)
+        }
+
+        Opcode.ADDM -> {
+            val reg = insn.requireIntSourceA().intNumber
+            val target = resolveMemory(insn.requireMemory())
+            val sv = dtSuffix(type)
+            emitLoadD0(reg, type)
+            emitLine("add$sv  d0, $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.ADDIM -> {
+            val value = insn.requireImmediateInt()
+            val target = resolveMemory(insn.requireMemory())
+            val sv = dtSuffix(type)
+            if(value in 1..8) {
+                emitLine("addq$sv  #$value,$target")
+            } else {
+                when (type) {
+                    IRDataType.BYTE -> {
+                        require(value in 2..255)
+                        emitLine("add$sv  #$value,$target")
+                    }
+                    IRDataType.WORD -> {
+                        require(value in 2..65535)
+                        emitLine("add$sv  #$value,$target")
+                    }
+                    IRDataType.LONG, IRDataType.POINTER -> {
+                        require(value in 2..0x7fffffff)
+                        emitLine("add$sv  #$value,$target")
+                    }
+                }
+            }
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.SUBR -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            loadRegOrZeroExtendToD0(srcReg, type)
+            emitLine("sub${dtSuffix(type)}  d0, ${regAddr(dstReg)}")
+            invalidateD0CacheForSlot(dstReg)
+        }
+
+        Opcode.SUB -> {
+            val reg = insn.requireIntDest().intNumber
+            val value = insn.requireImmediateInt()
+            if(value in 1..8) {
+                emitLine("subq${dtSuffix(type)}  #$value, ${regAddr(reg)}")
+            } else {
+                emitLine("sub${dtSuffix(type)}  #${immVal(value, type)}, ${regAddr(reg)}")
+            }
+            invalidateD0CacheForSlot(reg)
+        }
+
+        Opcode.SUBM -> {
+            val reg = insn.requireIntSourceA().intNumber
+            val target = resolveMemory(insn.requireMemory())
+            val sv = dtSuffix(type)
+            emitLoadD0(reg, type)
+            emitLine("sub$sv  d0, $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.SUBIM -> {
+            val value = insn.requireImmediateInt()
+            val target = resolveMemory(insn.requireMemory())
+            val sv = dtSuffix(type)
+            if(value in 1..8) {
+                emitLine("subq$sv  #$value,$target")
+            } else {
+                when (type) {
+                    IRDataType.BYTE -> {
+                        require(value in 2..255)
+                        emitLine("sub$sv  #$value,$target")
+                    }
+                    IRDataType.WORD -> {
+                        require(value in 2..65535)
+                        emitLine("sub$sv  #$value,$target")
+                    }
+                    IRDataType.LONG, IRDataType.POINTER -> {
+                        require(value in 2..0x7fffffff)
+                        emitLine("sub$sv  #$value,$target")
+                    }
+                }
+            }
+            invalidateD0CacheForAddress(target)
+        }
+
+        // --- Multiply (unsigned) ---
+        // M68k has no mulu.b or muls.b; use .w with zero-extension for byte.
+
+        Opcode.MULR -> emitMulOp(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type, unsigned=true, imm=null, target=null)
+        Opcode.MUL -> emitMulOp(insn.requireIntDest().intNumber, null, type, unsigned=true, imm=insn.requireImmediateInt(), target=null)
+        Opcode.MULM -> emitMulOp(insn.requireIntSourceA().intNumber, null, type, unsigned=true, imm=null, target=resolveMemory(insn.requireMemory()))
+
+        // --- Multiply (signed) ---
+
+        Opcode.MULSR -> emitMulOp(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type, unsigned=false, imm=null, target=null)
+        Opcode.MULS -> emitMulOp(insn.requireIntDest().intNumber, null, type, unsigned=false, imm=insn.requireImmediateInt(), target=null)
+        Opcode.MULSM -> emitMulOp(insn.requireIntSourceA().intNumber, null, type, unsigned=false, imm=null, target=resolveMemory(insn.requireMemory()))
+
+        // --- Divide (unsigned) ---
+
+        Opcode.DIVR -> emitDivOp(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type, unsigned=true, imm=null, target=null)
+        Opcode.DIV -> emitDivOp(insn.requireIntDest().intNumber, null, type, unsigned=true, imm=insn.requireImmediateInt(), target=null)
+        Opcode.DIVM -> emitDivOp(insn.requireIntSourceA().intNumber, null, type, unsigned=true, imm=null, target=resolveMemory(insn.requireMemory()))
+
+        // --- Divide (signed) ---
+
+        Opcode.DIVSR -> emitDivOp(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type, unsigned=false, imm=null, target=null)
+        Opcode.DIVS -> emitDivOp(insn.requireIntDest().intNumber, null, type, unsigned=false, imm=insn.requireImmediateInt(), target=null)
+        Opcode.DIVSM -> emitDivOp(insn.requireIntSourceA().intNumber, null, type, unsigned=false, imm=null, target=resolveMemory(insn.requireMemory()))
+
+        // --- Modulus ---
+        // Use divu.w/divs.w remainder (upper 16 bits) for byte/word; divul/divsl for long (68020+)
+
+        Opcode.MODR -> emitModOp(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type, unsigned=true, imm=null)
+        Opcode.MOD -> emitModOp(insn.requireIntDest().intNumber, null, type, unsigned=true, imm=insn.requireImmediateInt())
+        Opcode.MODSR -> emitModOp(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type, unsigned=false, imm=null)
+        Opcode.MODS -> emitModOp(insn.requireIntDest().intNumber, null, type, unsigned=false, imm=insn.requireImmediateInt())
+
+        // --- DIVMOD (unsigned) ---
+        // 68020+ divul.l gives quotient + remainder in one instruction
+
+        Opcode.DIVMODR -> emitDivModOp(insn.requireIntDest().intNumber, insn.requireDestB().intNumber, type, unsigned=true, imm=null)
+        Opcode.DIVMOD -> emitDivModOp(insn.requireIntDest().intNumber, insn.requireDestB().intNumber, type, unsigned=true, imm=insn.requireImmediateInt())
+        Opcode.SDIVMODR -> emitDivModOp(insn.requireIntDest().intNumber, insn.requireDestB().intNumber, type, unsigned=false, imm=null)
+        Opcode.SDIVMOD -> emitDivModOp(insn.requireIntDest().intNumber, insn.requireDestB().intNumber, type, unsigned=false, imm=insn.requireImmediateInt())
+
+        // --- Integer sqrt/square ---
+
+        Opcode.SQRT -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            when (type) {
+                IRDataType.BYTE -> {
+                    emitLine("move.b  ${regAddr(srcReg)}, math._sqrt_ub.value", "sqrt byte")
+                    invalidateD0Cache()
+                    emitLine("bsr  math._sqrt_ub")
+                    emitStoreD0(dstReg, IRDataType.BYTE)
+                }
+                IRDataType.WORD -> {
+                    emitLine("move.w  ${regAddr(srcReg)}, math._sqrt_uw.value", "sqrt word")
+                    invalidateD0Cache()
+                    emitLine("bsr  math._sqrt_uw")
+                    emitStoreD0(dstReg, IRDataType.BYTE)
+                }
+                IRDataType.LONG, IRDataType.POINTER -> {
+                    emitLine("move.l  ${regAddr(srcReg)}, math._sqrt_l.value", "sqrt long")
+                    invalidateD0Cache()
+                    emitLine("bsr  math._sqrt_l")
+                    emitStoreD0(dstReg, IRDataType.WORD)
+                }
+            }
+        }
+        Opcode.SQUARE -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            when (type) {
+                IRDataType.BYTE -> {
+                    emitLoadD0(srcReg, IRDataType.BYTE)
+                    emitLine($$"and.l  #$ff, d0", "zero extend")
+                    invalidateD0Cache()
+                    emitLine("mulu.w  d0, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.BYTE)
+                }
+                IRDataType.WORD -> {
+                    emitLoadD0(srcReg, IRDataType.WORD)
+                    emitLine("mulu.w  d0, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.WORD)
+                }
+                IRDataType.LONG, IRDataType.POINTER -> {
+                    emitLoadD0(srcReg, IRDataType.LONG)
+                    emitLine("mulu.l  d0, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.LONG)
+                }
+            }
+        }
+
+        Opcode.CMP -> {
+            val leftReg = insn.requireIntSourceA().intNumber
+            val rightReg = insn.requireSrcB().intNumber
+            val s = dtSuffix(type)
+            emitLoadD0(leftReg, type)
+            emitLine("cmp$s  ${regAddr(rightReg)}, d0")
+        }
+
+        Opcode.CMPI -> {
+            val reg = insn.requireIntSourceA().intNumber
+            val masked = immVal(insn.requireImmediateInt(), type)
+            // cmpi.x #0, operand can be replaced by the faster tst.x operand
+            if(masked==0)
+                emitLine("tst${dtSuffix(type)}  ${regAddr(reg)}")
+            else
+                emitLine("cmpi${dtSuffix(type)}  #$masked, ${regAddr(reg)}")
+        }
+
+        else -> error("Unknown arithmetic opcode: ${insn.opcode}")
+    }
+}
+
+// === Multiply / Divide / Modulus / DIVMOD for all int sizes (68030-aware) ===
+
+// emit a call to one of the 68000 long math helper routines (utility.library fast path + software fallback)
+private fun AsmGen.emitLongMathCall(dstReg: Int, srcReg: Int?, imm: Int?, target: String?, routine: String, resultReg: String) {
+    when {
+        srcReg != null -> {
+            emitLoadD0(dstReg, IRDataType.LONG)
+            emitLine("move.l  ${regAddr(srcReg)}, d1")
+        }
+        imm != null -> {
+            emitLoadD0(dstReg, IRDataType.LONG)
+            emitLine("move.l  #$imm, d1")
+        }
+        target != null -> {
+            emitLoadD0FromAddress(target, IRDataType.LONG)
+            emitLine("move.l  ${regAddr(dstReg)}, d1")
+        }
+    }
+    invalidateD0Cache()
+    emitLine("bsr  $routine")
+    val storeTarget = target ?: regAddr(dstReg)
+    if (resultReg == "d0") {
+        emitStoreD0ToAddress(storeTarget, IRDataType.LONG)
+    } else {
+        invalidateD0CacheForAddress(storeTarget)
+        emitLine("move.l  $resultReg, $storeTarget")
+    }
+}
+
+private fun AsmGen.emitMulOp(dstReg: Int, srcReg: Int?, type: IRDataType, unsigned: Boolean, imm: Int?, target: String?) {
+    val op = if (unsigned) "mulu" else "muls"
+    when (type) {
+        IRDataType.BYTE -> {
+            // No .b multiply on M68k: zero-extend to word, mulu.w/muls.w, store low byte
+            when {
+                srcReg != null -> {
+                    emitLoadD0(dstReg, IRDataType.BYTE)
+                    if (unsigned) emitLine($$"and.l  #$ff, d0") else emitSignExtendByteToLong("d0")
+                    invalidateD0Cache()
+                    emitLine("move.w  d0, d2")
+                    emitLoadD0(srcReg, IRDataType.BYTE)
+                    if (unsigned) emitLine($$"and.l  #$ff, d0") else emitSignExtendByteToLong("d0")
+                    invalidateD0Cache()
+                    emitLine("$op.w  d2, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.BYTE)
+                }
+                imm != null -> {
+                    emitLoadD0(dstReg, IRDataType.BYTE)
+                    if (unsigned) emitLine($$"and.l  #$ff, d0") else emitSignExtendByteToLong("d0")
+                    invalidateD0Cache()
+                    emitLine("$op.w  #${imm.and(0xffff)}, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.BYTE)
+                }
+                target != null -> {
+                    emitLine("move.b  $target, d1")
+                    if (unsigned) emitLine($$"and.l  #$ff, d1") else emitSignExtendByteToLong("d1")
+                    emitLine("move.w  d1, d2")
+                    emitLoadD0(dstReg, IRDataType.BYTE)
+                    if (unsigned) emitLine($$"and.l  #$ff, d0") else emitSignExtendByteToLong("d0")
+                    invalidateD0Cache()
+                    emitLine("$op.w  d2, d0")
+                    invalidateD0Cache()
+                    emitStoreD0ToAddress(target, IRDataType.BYTE)
+                }
+            }
+        }
+
+        IRDataType.WORD -> {
+            // mulu.w/muls.w is native (16x16→32, lower 16 are result)
+            // destination must be a data register, not memory
+            when {
+                srcReg != null -> {
+                    emitLoadD0(dstReg, IRDataType.WORD)
+                    emitLine("$op.w  ${regAddr(srcReg)}, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.WORD)
+                }
+                imm != null -> {
+                    emitLoadD0(dstReg, IRDataType.WORD)
+                    emitLine("$op.w  #${imm.and(0xffff)}, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.WORD)
+                }
+                target != null -> {
+                    emitLoadD0(dstReg, IRDataType.WORD)
+                    emitLine("$op.w  $target, d0")
+                    invalidateD0Cache()
+                    emitStoreD0ToAddress(target, IRDataType.WORD)
+                }
+            }
+        }
+
+        IRDataType.LONG -> {
+            if(program.options.compTarget.cpu < CpuType.M68020) {
+                // 68000: no mulu.l/muls.l; use a helper routine (utility.library or software fallback)
+                val routine = if (unsigned) "p8_umult32" else "p8_smult32"
+                emitLongMathCall(dstReg, srcReg, imm, target, routine, "d0")
+            } else {
+                // 68020+ mulu.l/muls.l (32x32→64, lower 32 are result)
+                // destination must be a data register, not memory
+                when {
+                    srcReg != null -> {
+                        emitLoadD0(dstReg, IRDataType.LONG)
+                        emitLine("$op.l  ${regAddr(srcReg)}, d0")
+                        invalidateD0Cache()
+                        emitStoreD0(dstReg, IRDataType.LONG)
+                    }
+                    imm != null -> {
+                        emitLoadD0(dstReg, IRDataType.LONG)
+                        emitLine("$op.l  #${imm}, d0")
+                        invalidateD0Cache()
+                        emitStoreD0(dstReg, IRDataType.LONG)
+                    }
+                    target != null -> {
+                        emitLoadD0(dstReg, IRDataType.LONG)
+                        emitLine("$op.l  $target, d0")
+                        invalidateD0Cache()
+                        emitStoreD0ToAddress(target, IRDataType.LONG)
+                    }
+                }
+            }
+        }
+
+        else -> TODO("MUL for ${type.name}")
+    }
+}
+
+private fun AsmGen.emitDivOp(dstReg: Int, srcReg: Int?, type: IRDataType, unsigned: Boolean, imm: Int?, target: String?) {
+    val op = if (unsigned) "divu" else "divs"
+    when (type) {
+        IRDataType.BYTE -> {
+            // No .b divide on M68k: extend to word, divu.w/divs.w, take quotient from low byte
+            when {
+                srcReg != null -> {
+                    emitLoadD0(dstReg, IRDataType.BYTE)
+                    if (unsigned) emitLine($$"and.l  #$ff, d0") else emitSignExtendByteToLong("d0")
+                    invalidateD0Cache()
+                    emitLine("move.b  ${regAddr(srcReg)}, d1")
+                    if (unsigned) emitLine($$"and.l  #$ff, d1") else emitSignExtendByteToLong("d1")
+                    emitLine("$op.w  d1, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.BYTE)
+                }
+                imm != null -> {
+                    emitLoadD0(dstReg, IRDataType.BYTE)
+                    if (unsigned) emitLine($$"and.l  #$ff, d0") else emitSignExtendByteToLong("d0")
+                    invalidateD0Cache()
+                    emitLine("$op.w  #${imm.and(0xffff)}, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.BYTE)
+                }
+                target != null -> {
+                    emitLoadD0FromAddress(target, IRDataType.BYTE)
+                    if (unsigned) emitLine($$"and.l  #$ff, d0") else emitSignExtendByteToLong("d0")
+                    invalidateD0Cache()
+                    emitLine("move.b  ${regAddr(dstReg)}, d1")
+                    if (unsigned) emitLine($$"and.l  #$ff, d1") else emitSignExtendByteToLong("d1")
+                    emitLine("$op.w  d1, d0")
+                    invalidateD0Cache()
+                    emitStoreD0ToAddress(target, IRDataType.BYTE)
+                }
+            }
+        }
+
+        IRDataType.WORD -> {
+            // divu.w/divs.w divide a 32-bit dividend (Dd) by a 16-bit divisor.
+            // divs.w requires the 32-bit dividend to be SIGN-extended; for unsigned it
+            // must be ZERO-extended. A plain `move.w` does not extend the upper word, so
+            // extend explicitly here (this is what was wrong: signed dividends ended up
+            // zero-extended, e.g. -1000 became +64536).
+            fun loadDividend() {
+                emitLoadD0(dstReg, IRDataType.WORD)
+                if (unsigned) {
+                    emitLine($$"and.l  #$ffff, d0", "zero-extend upper word for divu.w")
+                    invalidateD0Cache()
+                } else {
+                    emitLine("ext.l  d0", "sign-extend upper word for divs.w (signed 32-bit dividend)")
+                    invalidateD0Cache()
+                }
+            }
+            when {
+                srcReg != null -> {
+                    loadDividend()
+                    emitLine("$op.w  ${regAddr(srcReg)}, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.WORD)
+                }
+                imm != null -> {
+                    loadDividend()
+                    emitLine("$op.w  #${imm.and(0xffff)}, d0")
+                    invalidateD0Cache()
+                    emitStoreD0(dstReg, IRDataType.WORD)
+                }
+                target != null -> {
+                    emitLoadD0FromAddress(target, IRDataType.WORD)
+                    if (unsigned) {
+                        emitLine($$"and.l  #$ffff, d0", "zero-extend upper word for divu.w")
+                        invalidateD0Cache()
+                    } else {
+                        emitLine("ext.l  d0", "sign-extend upper word for divs.w (signed 32-bit dividend)")
+                        invalidateD0Cache()
+                    }
+                    emitLine("$op.w  ${regAddr(dstReg)}, d0")
+                    invalidateD0Cache()
+                    emitStoreD0ToAddress(target, IRDataType.WORD)
+                }
+            }
+        }
+
+        IRDataType.LONG -> {
+            if(program.options.compTarget.cpu < CpuType.M68020) {
+                // 68000: no divu.l/divs.l; use a helper routine (utility.library or software fallback)
+                val routine = if (unsigned) "p8_udivmod32" else "p8_sdivmod32"
+                emitLongMathCall(dstReg, srcReg, imm, target, routine, "d0")
+            } else {
+                // 68020+ divu.l/divs.l (32/32→32)
+                when {
+                    srcReg != null -> {
+                        emitLoadD0(dstReg, IRDataType.LONG)
+                        emitLine("$op.l  ${regAddr(srcReg)}, d0")
+                        invalidateD0Cache()
+                        emitStoreD0(dstReg, IRDataType.LONG)
+                    }
+                    imm != null -> {
+                        emitLoadD0(dstReg, IRDataType.LONG)
+                        emitLine("$op.l  #${imm}, d0")
+                        invalidateD0Cache()
+                        emitStoreD0(dstReg, IRDataType.LONG)
+                    }
+                    target != null -> {
+                        emitLoadD0FromAddress(target, IRDataType.LONG)
+                        emitLine("$op.l  ${regAddr(dstReg)}, d0")
+                        invalidateD0Cache()
+                        emitStoreD0ToAddress(target, IRDataType.LONG)
+                    }
+                }
+            }
+        }
+
+        else -> TODO("DIV for ${type.name}")
+    }
+}
+
+private fun AsmGen.emitModOp(dstReg: Int, srcReg: Int?, type: IRDataType, unsigned: Boolean, imm: Int?) {
+    val opDiv = if (unsigned) "divu" else "divs"
+    when (type) {
+        IRDataType.BYTE -> {
+            // divu.w gives quotient low word, remainder high word. Swap to get remainder.
+            when {
+                srcReg != null -> {
+                    emitLoadD0(dstReg, IRDataType.BYTE)
+                    if (unsigned) emitLine($$"and.l  #$ff, d0") else emitSignExtendByteToLong("d0")
+                    invalidateD0Cache()
+                    emitLine("move.b  ${regAddr(srcReg)}, d1")
+                    if (unsigned) emitLine($$"and.l  #$ff, d1") else emitSignExtendByteToLong("d1")
+                    emitLine("$opDiv.w  d1, d0")
+                    invalidateD0Cache()
+                }
+                imm != null -> {
+                    emitLoadD0(dstReg, IRDataType.BYTE)
+                    if (unsigned) emitLine($$"and.l  #$ff, d0") else emitSignExtendByteToLong("d0")
+                    invalidateD0Cache()
+                    emitLine("$opDiv.w  #${imm.and(0xffff)}, d0")
+                    invalidateD0Cache()
+                }
+            }
+            // After divu.w/divs.w: quotient in lower word, remainder in upper word of d0.
+            emitLine("swap  d0", "remainder to low word")
+            invalidateD0Cache()
+            emitStoreD0(dstReg, IRDataType.BYTE)
+        }
+
+        IRDataType.WORD -> {
+            // divu.w/divs.w: remainder in upper 16 bits, swap to get it.
+            // divs.w requires the 32-bit dividend to be SIGN-extended; for unsigned it
+            // must be ZERO-extended. A plain `move.w` does not extend the upper word, so
+            // extend explicitly (signed dividends were wrongly zero-extended before).
+            fun loadDividend() {
+                emitLoadD0(dstReg, IRDataType.WORD)
+                if (unsigned) {
+                    emitLine($$"and.l  #$ffff, d0", "zero-extend upper word for divu.w")
+                    invalidateD0Cache()
+                } else {
+                    emitLine("ext.l  d0", "sign-extend upper word for divs.w (signed 32-bit dividend)")
+                    invalidateD0Cache()
+                }
+            }
+            when {
+                srcReg != null -> {
+                    loadDividend()
+                    emitLine("$opDiv.w  ${regAddr(srcReg)}, d0")
+                    invalidateD0Cache()
+                }
+                imm != null -> {
+                    loadDividend()
+                    emitLine("$opDiv.w  #${imm.and(0xffff)}, d0")
+                    invalidateD0Cache()
+                }
+            }
+            emitLine("swap  d0", "remainder to low word")
+            invalidateD0Cache()
+            emitStoreD0(dstReg, IRDataType.WORD)
+        }
+
+        IRDataType.LONG -> {
+            if(program.options.compTarget.cpu < CpuType.M68020) {
+                // 68000: no divul.l/divsl.l; use a helper routine (utility.library or software fallback)
+                val routine = if (unsigned) "p8_udivmod32" else "p8_sdivmod32"
+                emitLongMathCall(dstReg, srcReg, imm, null, routine, "d1")
+            } else {
+                // 68020+ divul.l <ea>, Dr, Dq: Dr=remainder, Dq=quotient
+                val opLong = if (unsigned) "divul.l" else "divsl.l"
+                when {
+                    srcReg != null -> {
+                        emitLoadD0(dstReg, IRDataType.LONG)
+                        emitLine("$opLong  ${regAddr(srcReg)}, d1:d0")
+                        invalidateD0Cache()
+                        invalidateD0CacheForSlot(dstReg)
+                        emitLine("move.l  d1, ${regAddr(dstReg)}", "remainder")
+                    }
+                    imm != null -> {
+                        emitLoadD0(dstReg, IRDataType.LONG)
+                        emitLine("move.l  #${imm}, d2", "divisor")
+                        emitLine("$opLong  d2, d1:d0")
+                        invalidateD0Cache()
+                        invalidateD0CacheForSlot(dstReg)
+                        emitLine("move.l  d1, ${regAddr(dstReg)}", "remainder")
+                    }
+                }
+            }
+        }
+
+        else -> TODO("MOD for ${type.name}")
+    }
+}
+
+private fun AsmGen.emitDivModOp(dstReg: Int, remainderReg: Int, type: IRDataType, unsigned: Boolean, imm: Int?) {
+    val opLong = if (unsigned) "divul.l" else "divsl.l"
+    when (type) {
+        IRDataType.BYTE -> {
+            val opDiv = if (unsigned) "divu.w" else "divs.w"
+            emitLoadD0(dstReg, IRDataType.BYTE)
+            if (unsigned) emitLine($$"and.l  #$ff, d0") else emitSignExtendByteToLong("d0")
+            invalidateD0Cache()
+            if(imm!=null) {
+                if (unsigned)
+                    emitLine("move.l  #${imm.and(0xff)}, d1")
+                else
+                    emitLine("moveq  #${imm.and(0xff)}, d1")
+                if (!unsigned) emitSignExtendByteToLong("d1")
+            } else {
+                emitLine("move.b  ${regAddr(remainderReg)}, d1")
+                if (unsigned) emitLine($$"and.l  #$ff, d1") else emitSignExtendByteToLong("d1")
+            }
+            emitLine("$opDiv  d1, d0")
+            invalidateD0Cache()
+            emitStoreD0(dstReg, IRDataType.BYTE)
+            emitLine("swap  d0", "remainder to low word")
+            invalidateD0Cache()
+            emitStoreD0(remainderReg, IRDataType.BYTE)
+        }
+
+        IRDataType.WORD -> {
+            val opDiv = if (unsigned) "divu.w" else "divs.w"
+            if (unsigned) {
+                emitLine("moveq  #0, d0", "clear upper word for divu.w")
+                invalidateD0Cache()
+            }
+            emitLoadD0(dstReg, IRDataType.WORD)
+            if (!unsigned) {
+                emitLine("ext.l  d0", "sign-extend for divs.w")
+                invalidateD0Cache()
+            }
+            if(imm!=null) {
+                emitLine("$opDiv  #${imm.and(0xffff)}, d0")
+            } else {
+                emitLine("$opDiv  ${regAddr(remainderReg)}, d0")
+            }
+            invalidateD0Cache()
+            emitStoreD0(dstReg, IRDataType.WORD)
+            emitLine("swap  d0", "remainder to low word")
+            invalidateD0Cache()
+            emitStoreD0(remainderReg, IRDataType.WORD)
+        }
+
+        IRDataType.LONG -> {
+            if(program.options.compTarget.cpu < CpuType.M68020) {
+                // 68000: no divul.l/divsl.l; use a helper routine (utility.library or software fallback)
+                val routine = if (unsigned) "p8_udivmod32" else "p8_sdivmod32"
+                emitLoadD0(dstReg, IRDataType.LONG)
+                if(imm!=null) {
+                    emitLine("move.l  #${imm}, d1", "divisor")
+                } else {
+                    emitLine("move.l  ${regAddr(remainderReg)}, d1", "divisor")
+                }
+                invalidateD0Cache()
+                emitLine("bsr  $routine")
+                emitStoreD0(dstReg, IRDataType.LONG)
+                invalidateD0CacheForSlot(remainderReg)
+                emitLine("move.l  d1, ${regAddr(remainderReg)}", "remainder")
+            } else {
+                emitLoadD0(dstReg, IRDataType.LONG)
+                if(imm!=null) {
+                    emitLine("move.l  #${imm}, d1", "divisor")
+                } else {
+                    emitLine("move.l  ${regAddr(remainderReg)}, d1", "divisor")
+                }
+                emitLine("$opLong  d1, d0")
+                invalidateD0Cache()
+                emitStoreD0(dstReg, IRDataType.LONG)
+                invalidateD0CacheForSlot(remainderReg)
+                emitLine("move.l  d1, ${regAddr(remainderReg)}", "remainder")
+            }
+        }
+
+        else -> TODO("DIVMOD for ${type.name}")
+    }
+}
+
+// === Float arithmetic via 68881 FPU ===
+// Float virtual registers live in the memory regfile. Physical fp0 (FP_ACC) and
+// fp1 (FP_SRC) are used as scratch during FPU operations.
+
+private fun AsmGen.emitFloatBinaryOp(op: String, dst: RegisterNum, src: RegisterNum) {
+    emitLine("fmove.s  ${floatRegFileAddr(dst)}, $FP_ACC")
+    emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_SRC")
+    emitLine("$op  $FP_SRC, $FP_ACC")
+    emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+}
+
+private fun AsmGen.emitFloatOpWithConstant(fpReg: RegisterNum, op: String, value: Double) {
+    emitLine("fmove.s  ${floatRegFileAddr(fpReg)}, $FP_ACC")
+    val native = nativeFloatConst(value)
+    if (native != null) {
+        emitLine("fmovecr  #$native, $FP_SRC")
+    } else {
+        val label = makeFloatConstLabel(value)
+        emitLine("lea  $label, a0")
+        emitLine("fmove.s  (a0), $FP_SRC")
+    }
+    emitLine("$op  $FP_SRC, $FP_ACC")
+    emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(fpReg)}")
+}
+
+private fun AsmGen.translateFloatArithmetic(insn: IRInstruction) {
+    when (insn.opcode) {
+        Opcode.INCM -> emitFloatMemUnary("fadd.s", resolveMemory(insn.requireMemory(), forFloat = true), immediateStr = "#1.0")
+        Opcode.DECM -> emitFloatMemUnary("fsub.s", resolveMemory(insn.requireMemory(), forFloat = true), immediateStr = "#1.0")
+        Opcode.NEGM -> emitFloatMemUnary("fneg", resolveMemory(insn.requireMemory(), forFloat = true))
+
+        Opcode.INC -> {
+            val dst = insn.requireFloatDest().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(dst)}, $FP_ACC")
+            emitLine("fadd.s  #1.0, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+        Opcode.DEC -> {
+            val dst = insn.requireFloatDest().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(dst)}, $FP_ACC")
+            emitLine("fsub.s  #1.0, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+        Opcode.NEG -> {
+            val dst = insn.requireFloatDest().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(dst)}, $FP_ACC")
+            emitLine("fneg  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.ADDR -> emitFloatBinaryOp("fadd", insn.requireFloatDest().floatNumber, insn.requireFloatSourceA().floatNumber)
+        Opcode.ADD -> emitFloatOpWithConstant(insn.requireFloatDest().floatNumber, "fadd", insn.requireImmediateFloat())
+
+        Opcode.SUBR -> emitFloatBinaryOp("fsub", insn.requireFloatDest().floatNumber, insn.requireFloatSourceA().floatNumber)
+        Opcode.SUB -> emitFloatOpWithConstant(insn.requireFloatDest().floatNumber, "fsub", insn.requireImmediateFloat())
+
+        Opcode.MULR, Opcode.MULSR -> emitFloatBinaryOp("fmul", insn.requireFloatDest().floatNumber, insn.requireFloatSourceA().floatNumber)
+        Opcode.MUL, Opcode.MULS -> emitFloatOpWithConstant(insn.requireFloatDest().floatNumber, "fmul", insn.requireImmediateFloat())
+
+        Opcode.DIVR, Opcode.DIVSR -> emitFloatBinaryOp("fdiv", insn.requireFloatDest().floatNumber, insn.requireFloatSourceA().floatNumber)
+        Opcode.DIV, Opcode.DIVS -> emitFloatOpWithConstant(insn.requireFloatDest().floatNumber, "fdiv", insn.requireImmediateFloat())
+
+        Opcode.SQRT -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fsqrt  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.SQUARE -> emitFloatBinaryOp("fmul", insn.requireFloatDest().floatNumber, insn.requireFloatSourceA().floatNumber)
+
+        Opcode.ADDM -> emitFloatMemBinary("fadd", insn.requireFloatSourceA().floatNumber, resolveMemory(insn.requireMemory(), forFloat = true))
+        Opcode.SUBM -> emitFloatMemBinary("fsub", insn.requireFloatSourceA().floatNumber, resolveMemory(insn.requireMemory(), forFloat = true))
+        Opcode.MULM, Opcode.MULSM -> emitFloatMemBinary("fmul", insn.requireFloatSourceA().floatNumber, resolveMemory(insn.requireMemory(), forFloat = true))
+        Opcode.DIVM, Opcode.DIVSM -> emitFloatMemBinary("fdiv", insn.requireFloatSourceA().floatNumber, resolveMemory(insn.requireMemory(), forFloat = true))
+
+        Opcode.ADDIM -> emitFloatMemBinaryImmediate("fadd", insn.requireImmediateFloat(), resolveMemory(insn.requireMemory(), forFloat = true))
+        Opcode.SUBIM -> emitFloatMemBinaryImmediate("fsub", insn.requireImmediateFloat(), resolveMemory(insn.requireMemory(), forFloat = true))
+
+        else -> TODO("FLOAT arithmetic: ${insn.opcode}")
+    }
+}
+
+private fun AsmGen.emitFloatMemBinaryImmediate(op: String, value: Double, target: String) {
+    emitLine("fmove.s $target,$FP_ACC")
+    when (value) {
+        0.0 -> { emitLine("fmove.s #0.0,$FP_SRC"); emitLine("$op $FP_SRC,$FP_ACC") }
+        1.0 -> {
+            if (op == "fadd" || op == "fsub") {
+                emitLine("f${op.substring(1)}.s #1.0,$FP_ACC")
+            } else {
+                return  // mul/div by 1.0 is no-op, just keep loaded value
+            }
+        }
+        else -> {
+            val lbl = makeFloatConstLabel(value)
+            emitLine("lea $lbl,a0")
+            emitLine("fmove.s (a0),$FP_SRC")
+            emitLine("$op $FP_SRC,$FP_ACC")
+        }
+    }
+    emitLine("fmove.s $FP_ACC,$target")
+}
+
+private fun AsmGen.emitFloatMemUnary(op: String, target: String, immediateStr: String? = null) {
+    emitLine("fmove.s  $target, fp0")
+    if (immediateStr != null)
+        emitLine("$op  $immediateStr, fp0")
+    else
+        emitLine("$op  fp0, fp0")
+    emitLine("fmove.s  fp0, $target")
+}
+
+private fun AsmGen.emitFloatMemBinary(op: String, fpReg1: RegisterNum, target: String) {
+    emitLine("fmove.s  $target, $FP_ACC")
+    emitLine("fmove.s  ${floatRegFileAddr(fpReg1)}, $FP_SRC")
+    emitLine("$op  $FP_SRC, $FP_ACC")
+    emitLine("fmove.s  $FP_ACC, $target")
+}
+
+private fun immVal(value: Int, type: IRDataType): Int = when(type) {
+    IRDataType.BYTE -> value and 0xff
+    IRDataType.WORD -> value and 0xffff
+    IRDataType.LONG, IRDataType.POINTER -> value
+    else -> value and 0xffff
+}

codeGenM68k/src/prog8/codegen/m68k/InstrBitwise.kt

@@ -0,0 +1,424 @@
+/*
+ * Bitwise and logical IR instruction translations for the M68k code generator.
+ *
+ * Handles: AND, OR, XOR (register/imm/memory), INV (bitwise not),
+ * shifts (LSL, LSR, ASR) by constant or variable count,
+ * rotates (ROL, ROR, ROXL, ROXR) and bit manipulation (BITTST, BITSET, BITCLR, BITTOG).
+ *
+ * M68k has native .B/.W/.L support for all logical operations,
+ * including immediate and register-count shifts.
+ * Rotates ROL/ROR are true rotates (bit wraps around, no carry involvement for entry).
+ * ROXL/ROXR rotate through the extend (X/C) bit.
+ */
+
+package prog8.codegen.m68k
+
+import prog8.code.target.Qemu68kTarget
+import prog8.intermediate.IRDataType
+import prog8.intermediate.IRInstruction
+import prog8.intermediate.Opcode
+import prog8.intermediate.intNumber
+import prog8.intermediate.requireImmediateInt
+import prog8.intermediate.requireIntDest
+import prog8.intermediate.requireIntSourceA
+import prog8.intermediate.requireMemory
+
+internal fun AsmGen.translateBitwise(insn: IRInstruction) {
+    val type = insn.type ?: IRDataType.BYTE
+
+    when (insn.opcode) {
+        Opcode.ANDR -> andRegisters(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type)
+        Opcode.AND -> andImmediate(insn.requireIntDest().intNumber, insn.requireImmediateInt(), type)
+        Opcode.ANDM -> andMemory(insn.requireIntSourceA().intNumber, resolveMemory(insn.requireMemory()), type)
+
+        Opcode.ORR -> orRegisters(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type)
+        Opcode.OR -> orImmediate(insn.requireIntDest().intNumber, insn.requireImmediateInt(), type)
+        Opcode.ORM -> orMemory(insn.requireIntSourceA().intNumber, resolveMemory(insn.requireMemory()), type)
+
+        Opcode.XORR -> xorRegisters(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type)
+        Opcode.XOR -> xorImmediate(insn.requireIntDest().intNumber, insn.requireImmediateInt(), type)
+        Opcode.XORM -> xorMemory(insn.requireIntSourceA().intNumber, resolveMemory(insn.requireMemory()), type)
+
+        Opcode.INV -> invertRegister(insn.requireIntDest().intNumber, type)
+        Opcode.INVM -> invertMemory(resolveMemory(insn.requireMemory()), type)
+
+        Opcode.ASRN -> arithmeticShiftRightVar(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type)
+        Opcode.LSRN -> logicalShiftRightVar(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type)
+        Opcode.LSLN -> logicalShiftLeftVar(insn.requireIntDest().intNumber, insn.requireIntSourceA().intNumber, type)
+
+        Opcode.ASRNM -> shiftMemoryVar(resolveMemory(insn.requireMemory()), insn.requireIntSourceA().intNumber, type, isArithmetic = true)
+        Opcode.LSRNM -> shiftMemoryVar(resolveMemory(insn.requireMemory()), insn.requireIntSourceA().intNumber, type, isArithmetic = false)
+        Opcode.LSLNM -> shiftMemoryLeftVar(resolveMemory(insn.requireMemory()), insn.requireIntSourceA().intNumber, type)
+
+        // Immediate-count shifts (LSLI/LSRI/ASRI): the count is a literal in the immediate operand.
+        Opcode.ASRI -> shiftRegister(insn.requireIntDest().intNumber, insn.requireImmediateInt(), type, isArithmetic = true, isLeft = false)
+        Opcode.LSRI -> shiftRegister(insn.requireIntDest().intNumber, insn.requireImmediateInt(), type, isArithmetic = false, isLeft = false)
+        Opcode.LSLI -> shiftRegister(insn.requireIntDest().intNumber, insn.requireImmediateInt(), type, isArithmetic = false, isLeft = true)
+
+        Opcode.ASR -> shiftRegister(insn.requireIntDest().intNumber, 1, type, isArithmetic = true, isLeft = false)
+        Opcode.ASRM -> memoryShiftRotate(resolveMemory(insn.requireMemory()), 1, type, isArithmetic = true, isLeft = false, isRotate = false, throughCarry = false)
+        Opcode.LSR -> shiftRegister(insn.requireIntDest().intNumber, 1, type, isArithmetic = false, isLeft = false)
+        Opcode.LSRM -> memoryShiftRotate(resolveMemory(insn.requireMemory()), 1, type, isArithmetic = false, isLeft = false, isRotate = false, throughCarry = false)
+        Opcode.LSL -> shiftRegister(insn.requireIntDest().intNumber, 1, type, isArithmetic = false, isLeft = true)
+        Opcode.LSLM -> memoryShiftRotate(resolveMemory(insn.requireMemory()), 1, type, isArithmetic = false, isLeft = true, isRotate = false, throughCarry = false)
+
+        Opcode.ROR -> rotateRight(insn.requireIntDest().intNumber, type)
+        Opcode.RORM -> memoryShiftRotate(resolveMemory(insn.requireMemory()), 1, type, isArithmetic = false, isLeft = false, isRotate = true, throughCarry = false)
+        Opcode.ROL -> rotateLeft(insn.requireIntDest().intNumber, type)
+        Opcode.ROLM -> memoryShiftRotate(resolveMemory(insn.requireMemory()), 1, type, isArithmetic = false, isLeft = true, isRotate = true, throughCarry = false)
+        Opcode.ROXR -> rotateRightThroughCarry(insn.requireIntDest().intNumber, type)
+        Opcode.ROXRM -> memoryShiftRotate(resolveMemory(insn.requireMemory()), 1, type, isArithmetic = false, isLeft = false, isRotate = true, throughCarry = true)
+        Opcode.ROXL -> rotateLeftThroughCarry(insn.requireIntDest().intNumber, type)
+        Opcode.ROXLM -> memoryShiftRotate(resolveMemory(insn.requireMemory()), 1, type, isArithmetic = false, isLeft = true, isRotate = true, throughCarry = true)
+
+        Opcode.BITTST -> bitTest(insn.requireIntSourceA().intNumber, insn.requireImmediateInt(), type)
+        Opcode.BITSET -> bitSet(insn.requireIntDest().intNumber, insn.requireImmediateInt(), type)
+        Opcode.BITCLR -> bitClear(insn.requireIntDest().intNumber, insn.requireImmediateInt(), type)
+        Opcode.BITTOG -> bitToggle(insn.requireIntDest().intNumber, insn.requireImmediateInt(), type)
+
+        else -> error("Unknown bitwise opcode: ${insn.opcode}")
+    }
+}
+
+// === AND ===
+
+private fun AsmGen.andRegisters(dstReg: Int, srcReg: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLoadD0(srcReg, type)
+    emitLine("and$s  d0, ${regAddr(dstReg)}")
+    invalidateD0CacheForSlot(dstReg)
+}
+
+private fun AsmGen.andImmediate(dstReg: Int, value: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    val mask = when (type) {
+        IRDataType.BYTE -> value and 0xff
+        IRDataType.WORD -> value and 0xffff
+        IRDataType.LONG -> value
+        else -> error("unsupported type for AND immediate")
+    }
+    emitLine("andi$s  #$mask, ${regAddr(dstReg)}")
+    invalidateD0CacheForSlot(dstReg)
+}
+
+private fun AsmGen.andMemory(dstReg: Int, sourceAddress: String, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLoadD0(dstReg, type)
+    emitLine("and$s  d0, $sourceAddress")
+    invalidateD0CacheForAddress(sourceAddress)
+}
+
+// === OR ===
+
+private fun AsmGen.orRegisters(dstReg: Int, srcReg: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLoadD0(srcReg, type)
+    emitLine("or$s  d0, ${regAddr(dstReg)}")
+    invalidateD0CacheForSlot(dstReg)
+}
+
+private fun AsmGen.orImmediate(dstReg: Int, value: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    val mask = when (type) {
+        IRDataType.BYTE -> value and 0xff
+        IRDataType.WORD -> value and 0xffff
+        IRDataType.LONG -> value
+        else -> error("unsupported type for OR immediate")
+    }
+    emitLine("ori$s  #$mask, ${regAddr(dstReg)}")
+    invalidateD0CacheForSlot(dstReg)
+}
+
+private fun AsmGen.orMemory(dstReg: Int, sourceAddress: String, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLoadD0(dstReg, type)
+    emitLine("or$s  d0, $sourceAddress")
+    invalidateD0CacheForAddress(sourceAddress)
+}
+
+// === XOR ===
+
+private fun AsmGen.xorRegisters(dstReg: Int, srcReg: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLoadD0(srcReg, type)
+    emitLine("eor$s  d0, ${regAddr(dstReg)}")
+    invalidateD0CacheForSlot(dstReg)
+}
+
+private fun AsmGen.xorImmediate(dstReg: Int, value: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    val mask = when (type) {
+        IRDataType.BYTE -> value and 0xff
+        IRDataType.WORD -> value and 0xffff
+        IRDataType.LONG -> value
+        else -> error("unsupported type for XOR immediate")
+    }
+    emitLine("eori$s  #$mask, ${regAddr(dstReg)}")
+    invalidateD0CacheForSlot(dstReg)
+}
+
+private fun AsmGen.xorMemory(dstReg: Int, sourceAddress: String, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLoadD0(dstReg, type)
+    emitLine("eor$s  d0, $sourceAddress")
+    invalidateD0CacheForAddress(sourceAddress)
+}
+
+// === Invert ===
+
+private fun AsmGen.invertRegister(reg: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLine("not$s  ${regAddr(reg)}")
+    invalidateD0CacheForSlot(reg)
+}
+
+private fun AsmGen.invertMemory(target: String, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLine("not$s  $target")
+    invalidateD0CacheForAddress(target)
+}
+
+// === Shift/rotate size helpers ===
+
+private fun AsmGen.shiftOpcode(isLeft: Boolean, isArithmetic: Boolean, isRotate: Boolean = false): String {
+    if (isRotate) {
+        // Always use roxl/roxr (never rol/ror). X is managed as rotate-carry.
+        return if (isLeft) "roxl" else "roxr"
+    }
+    return if (isLeft) {
+        "lsl"
+    } else {
+        if (isArithmetic) "asr" else "lsr"
+    }
+}
+
+// === Shifts by 1 (constant count 1) ===
+
+// Emit a .w count=1 memory-form shift/rotate. The absolute form (`op.w addr`)
+// is one instruction and optimal, but QEMU's 68020 emulation is broken for
+// word-sized memory *right* shifts in both addressing forms:
+//   `asr.w addr`  -> sign-extends the 16-bit operand instead of zero-extending
+//                    it, producing a logical shift result for negative values
+//   `lsr.w (a0)`  -> sign-extends, producing an arithmetic shift result
+// The register form (`move.w addr,d0` / `op.w #1,d0` / `move.w d0,addr`) is
+// correct in QEMU, and both the absolute and `(a0)` forms are correct on a real
+// 68000 (vamos). Left shifts and rotates are unaffected, so they keep the
+// compact memory form. d0 is scratch for the shift, matching the callers which
+// already invalidate their d0 cache.
+private fun AsmGen.emitMemoryWordShiftOrRotate(op: String, address: String) {
+    when (target.name) {
+        Qemu68kTarget.NAME if (op == "asr" || op == "lsr") -> {
+            emitLine("move.w  $address, d0")
+            emitLine("$op.w  #1, d0")
+            emitLine("move.w  d0, $address")
+        }
+        Qemu68kTarget.NAME -> {
+            emitLine("lea  $address, a0")
+            emitLine("$op.w  (a0)")
+        }
+        else -> {
+            emitLine("$op.w  $address")
+        }
+    }
+}
+
+private fun AsmGen.shiftRegister(reg: Int, count: Int, type: IRDataType, isArithmetic: Boolean, isLeft: Boolean) {
+    val s = dtSuffix(type)
+    val op = shiftOpcode(isLeft, isArithmetic)
+    if (isLeft && type == IRDataType.LONG && count == 16) {
+        // x << 16 for long: `swap; clr.w`. Left shift by 16 has the same
+        // result for signed and unsigned long: the sign bit lives in the
+        // top half, which is replaced by the bottom half; the bottom
+        // half is always cleared.
+        emitLoadD0(reg, IRDataType.LONG)
+        emitLine("swap  d0")
+        emitLine("clr.w  d0")
+        emitStoreD0(reg, IRDataType.LONG)
+        return
+    }
+    if (type == IRDataType.WORD && count == 1) {
+        // m68k memory shift/rotate is supported for .w count=1; collapse the
+        // d0 round-trip into a single memory form
+        emitMemoryWordShiftOrRotate(op, regAddr(reg))
+        invalidateD0CacheForSlot(reg)
+        return
+    }
+    if (count in 1..8) {
+        emitLoadD0(reg, type)
+        emitLine("$op$s  #$count, d0")
+        emitStoreD0(reg, type)
+    } else {
+        // m68k immediate-shift range is 1..8; for larger counts the count
+        // has to go into a data register first. `move.w` preserves the
+        // count up to 32767 (Prog8 shift counts are 0..255, so this is
+        // always safe).
+        emitLine("move.w  #$count, d1")
+        emitLoadD0(reg, type)
+        emitLine("$op$s  d1, d0")
+        emitStoreD0(reg, type)
+    }
+}
+
+private fun AsmGen.memoryShiftRotate(target: String, count: Int, type: IRDataType, isArithmetic: Boolean, isLeft: Boolean, isRotate: Boolean, throughCarry: Boolean) {
+    // For logical rotates (ROL/ROR), clear X before roxl/roxr
+    // so the injected bit is 0 (logical rotate, not through-carry).
+    if (isRotate && !throughCarry) {
+        emitLine($$"andi  #$ef, ccr")
+    }
+    // On 68k, shift/rotate instructions can operate directly on memory operands
+    // (only for .w size, and shift/rotate count must be 1)
+    if (type == IRDataType.WORD && count == 1) {
+        val op = shiftOpcode(isLeft, isArithmetic, isRotate)
+        emitMemoryWordShiftOrRotate(op, target)
+        invalidateD0CacheForAddress(target)
+        return
+    }
+    val s = dtSuffix(type)
+    val op = shiftOpcode(isLeft, isArithmetic, isRotate)
+    emitLoadD0FromAddress(target, type)
+    emitLine("$op$s  #$count, d0")
+    emitStoreD0ToAddress(target, type)
+}
+
+// === Variable-count shifts ===
+
+private fun AsmGen.logicalShiftLeftVar(reg: Int, countReg: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLine("move.b  ${regAddr(countReg)}, d1")
+    emitLoadD0(reg, type)
+    emitLine("lsl$s  d1, d0")
+    emitStoreD0(reg, type)
+}
+
+private fun AsmGen.logicalShiftRightVar(reg: Int, countReg: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLine("move.b  ${regAddr(countReg)}, d1")
+    emitLoadD0(reg, type)
+    emitLine("lsr$s  d1, d0")
+    emitStoreD0(reg, type)
+}
+
+private fun AsmGen.arithmeticShiftRightVar(reg: Int, countReg: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLine("move.b  ${regAddr(countReg)}, d1")
+    emitLoadD0(reg, type)
+    emitLine("asr$s  d1, d0")
+    emitStoreD0(reg, type)
+}
+
+// === Memory variable-count shifts ===
+
+private fun AsmGen.shiftMemoryVar(target: String, countReg: Int, type: IRDataType, isArithmetic: Boolean) {
+    val s = dtSuffix(type)
+    val op = if (isArithmetic) "asr" else "lsr"
+    emitLine("move.b  ${regAddr(countReg)}, d1")
+    emitLoadD0FromAddress(target, type)
+    emitLine("$op$s  d1, d0")
+    emitStoreD0ToAddress(target, type)
+}
+
+private fun AsmGen.shiftMemoryLeftVar(target: String, countReg: Int, type: IRDataType) {
+    val s = dtSuffix(type)
+    emitLine("move.b  ${regAddr(countReg)}, d1")
+    emitLoadD0FromAddress(target, type)
+    emitLine("lsl$s  d1, d0")
+    emitStoreD0ToAddress(target, type)
+}
+
+// === Rotates ===
+// On M68k, the X (extend/CCR bit 4) bit serves as the "carry for rotates",
+// while C (carry/CCR bit 0) is used for comparisons.
+// This maps naturally: ROL/ROR IR (logical rotate) = clear X + roxl/roxr,
+// ROXL/ROXR IR (rotate through carry) = roxl/roxr directly.
+// CLC/SEC manage both C+X to keep them in sync.
+
+private fun AsmGen.rotateLeft(reg: Int, type: IRDataType) {
+    // IR ROL: logical rotate left (inject 0 into LSB)
+    if (type == IRDataType.WORD) {
+        emitLine($$"andi  #$ef, ccr")       // clear X (and leave C alone)
+        emitMemoryWordShiftOrRotate("roxl", regAddr(reg))
+        invalidateD0CacheForSlot(reg)
+        return
+    }
+    val s = dtSuffix(type)
+    emitLoadD0(reg, type)
+    emitLine($$"andi  #$ef, ccr")       // clear X (and leave C alone)
+    emitLine("roxl$s  #1, d0")           // rotate through X: X=0 -> bit 0 gets 0
+    emitStoreD0(reg, type)
+}
+
+private fun AsmGen.rotateRight(reg: Int, type: IRDataType) {
+    // IR ROR: logical rotate right (inject 0 into MSB)
+    if (type == IRDataType.WORD) {
+        emitLine($$"andi  #$ef, ccr")       // clear X (and leave C alone)
+        emitMemoryWordShiftOrRotate("roxr", regAddr(reg))
+        invalidateD0CacheForSlot(reg)
+        return
+    }
+    val s = dtSuffix(type)
+    emitLoadD0(reg, type)
+    emitLine($$"andi  #$ef, ccr")       // clear X (and leave C alone)
+    emitLine("roxr$s  #1, d0")           // rotate through X: X=0 -> MSB gets 0
+    emitStoreD0(reg, type)
+}
+
+// === Rotates through carry (extend) ===
+
+private fun AsmGen.rotateLeftThroughCarry(reg: Int, type: IRDataType) {
+    // IR ROXL: rotate left through carry (on M68k, X serves as rotate-carry)
+    if (type == IRDataType.WORD) {
+        emitMemoryWordShiftOrRotate("roxl", regAddr(reg))
+        invalidateD0CacheForSlot(reg)
+        return
+    }
+    val s = dtSuffix(type)
+    emitLoadD0(reg, type)
+    emitLine("roxl$s  #1, d0")
+    emitStoreD0(reg, type)
+}
+
+private fun AsmGen.rotateRightThroughCarry(reg: Int, type: IRDataType) {
+    // IR ROXR: rotate right through carry (on M68k, X serves as rotate-carry)
+    if (type == IRDataType.WORD) {
+        emitMemoryWordShiftOrRotate("roxr", regAddr(reg))
+        invalidateD0CacheForSlot(reg)
+        return
+    }
+    val s = dtSuffix(type)
+    emitLoadD0(reg, type)
+    emitLine("roxr$s  #1, d0")
+    emitStoreD0(reg, type)
+}
+
+// === Bit manipulation ===
+
+private fun AsmGen.bitOpMem(op: String, reg: Int, bit: Int, type: IRDataType) {
+    // The register slot is big-endian in memory, so the target bit lives in
+    // byte (size-1-bit/8) at bit position (bit % 8). A byte-sized bit op with
+    // an immediate bit number operates on that byte directly, no d0 round-trip.
+    val size = when(type) {
+        IRDataType.BYTE -> 1
+        IRDataType.WORD -> 2
+        IRDataType.LONG -> 4
+        else -> error("bit op on unsupported type $type")
+    }
+    val byteOffset = size - 1 - bit / 8
+    val bitInByte = bit % 8
+    emitLine("$op  #$bitInByte, ${regAddrByte(reg, byteOffset)}")
+}
+
+private fun AsmGen.bitTest(reg: Int, bit: Int, type: IRDataType) {
+    bitOpMem("btst", reg, bit, type)
+}
+
+private fun AsmGen.bitSet(reg: Int, bit: Int, type: IRDataType) {
+    bitOpMem("bset", reg, bit, type)
+}
+
+private fun AsmGen.bitClear(reg: Int, bit: Int, type: IRDataType) {
+    bitOpMem("bclr", reg, bit, type)
+}
+
+private fun AsmGen.bitToggle(reg: Int, bit: Int, type: IRDataType) {
+    bitOpMem("bchg", reg, bit, type)
+}

codeGenM68k/src/prog8/codegen/m68k/InstrBranch.kt

@@ -0,0 +1,155 @@
+/*
+ * Branch/compare IR instruction translations for the M68k code generator.
+ *
+ * Status bit branches (BSTCC/BSTCS/etc.) map directly to M68k conditional
+ * branch instructions (BCC/BCS/BEQ/BNE/BMI/BPL/BVC/BVS).
+ *
+ * Comparison branches (BGT/BGE/BLT/BLE and signed variants) first emit
+ * a CMP or CMPI between the operands, then branch using the appropriate
+ * M68k condition code:
+ *
+ *   Unsigned: BHI (C=0 & Z=0), BHS/BCC (C=0), BLO/BCS (C=1), BLS (C=1 | Z=1)
+ *   Signed:   BGT (N=V & Z=0), BGE (N=V), BLT (N!=V), BLE (Z=1 | N!=V)
+ *
+ * M68k CMP/CMPI sets all flags correctly for .B, .W, and .L — no cascading
+ * or overflow correction needed (unlike 6502).
+ */
+
+package prog8.codegen.m68k
+
+import prog8.intermediate.CodeReference
+import prog8.intermediate.IRDataType
+import prog8.intermediate.IRInstruction
+import prog8.intermediate.Opcode
+import prog8.intermediate.intNumber
+import prog8.intermediate.requireImmediateInt
+import prog8.intermediate.requireSrcA
+import prog8.intermediate.requireSrcB
+import prog8.intermediate.requireTarget
+
+internal fun AsmGen.translateBranch(insn: IRInstruction) {
+    val target = insn.requireTarget()
+    val label: String = when (target) {
+        is CodeReference.Label -> fixNameSymbols(target.name)
+        is CodeReference.Absolute -> target.address.toHex()
+        is CodeReference.Indirect -> error("branch needs a static target")
+    }
+    val leavesSubroutine = when (target) {
+        is CodeReference.Label -> labelIsOutsideCurrentSub(target.name)
+        is CodeReference.Absolute -> true
+        is CodeReference.Indirect -> false
+    }
+
+    when (insn.opcode) {
+        Opcode.BSTCC -> emitBranch("bcc", label, leavesSubroutine)
+        Opcode.BSTCS -> emitBranch("bcs", label, leavesSubroutine)
+        Opcode.BSTEQ -> emitBranch("beq", label, leavesSubroutine)
+        Opcode.BSTNE -> emitBranch("bne", label, leavesSubroutine)
+        Opcode.BSTNEG -> emitBranch("bmi", label, leavesSubroutine)
+        Opcode.BSTPOS -> emitBranch("bpl", label, leavesSubroutine)
+        Opcode.BSTVC -> emitBranch("bvc", label, leavesSubroutine)
+        Opcode.BSTVS -> emitBranch("bvs", label, leavesSubroutine)
+
+        // Unsigned integer comparison branches
+        Opcode.BGT -> cmpBranchUnsignedImm(insn, label, "bhi", leavesSubroutine)
+        Opcode.BGE -> cmpBranchUnsignedImm(insn, label, "bhs", leavesSubroutine)
+        Opcode.BLT -> cmpBranchUnsignedImm(insn, label, "blo", leavesSubroutine)
+        Opcode.BLE -> cmpBranchUnsignedImm(insn, label, "bls", leavesSubroutine)
+
+        Opcode.BGTR -> cmpBranchUnsignedReg(insn, label, "bhi", leavesSubroutine)
+        Opcode.BGER -> cmpBranchUnsignedReg(insn, label, "bhs", leavesSubroutine)
+        // BLTR doesn't exist in IR — uses BGTR with swapped operands
+
+        // Signed integer comparison branches
+        Opcode.BGTS -> cmpBranchSignedImm(insn, label, "bgt", leavesSubroutine)
+        Opcode.BGES -> cmpBranchSignedImm(insn, label, "bge", leavesSubroutine)
+        Opcode.BLTS -> cmpBranchSignedImm(insn, label, "blt", leavesSubroutine)
+        Opcode.BLES -> cmpBranchSignedImm(insn, label, "ble", leavesSubroutine)
+
+        Opcode.BGTSR -> cmpBranchSignedReg(insn, label, "bgt", leavesSubroutine)
+        Opcode.BGESR -> cmpBranchSignedReg(insn, label, "bge", leavesSubroutine)
+        // BLTSR doesn't exist in IR — uses BGTSR with swapped operands
+
+        else -> error("Unknown branch opcode: ${insn.opcode}")
+    }
+}
+
+// === Unsigned comparisons: register vs immediate ===
+
+private fun AsmGen.cmpBranchUnsignedImm(insn: IRInstruction, label: String, branchOp: String, leavesSubroutine: Boolean) {
+    val type = insn.type ?: IRDataType.BYTE
+    val reg = insn.requireSrcA().intNumber
+    val imm = insn.requireImmediateInt()
+    val s = dtSuffix(type)
+    // cmpi.x #imm, <ea> and tst.x <ea> accept memory operands, so the
+    // register-file load into d0 is redundant
+    if(imm==0)
+        emitLine("tst$s  ${regAddr(reg)}")
+    else
+        emitLine("cmpi$s  #$imm, ${regAddr(reg)}")
+    emitBranch(branchOp, label, leavesSubroutine)
+}
+
+// === Unsigned comparisons: register vs register ===
+
+private fun AsmGen.cmpBranchUnsignedReg(insn: IRInstruction, label: String, branchOp: String, leavesSubroutine: Boolean) {
+    val type = insn.type ?: IRDataType.BYTE
+    val left = insn.requireSrcA().intNumber
+    val right = insn.requireSrcB().intNumber
+    val s = dtSuffix(type)
+    emitLoadD0(left, type)
+    emitLine("cmp$s  ${regAddr(right)}, d0")
+    emitBranch(branchOp, label, leavesSubroutine)
+}
+
+// === Signed comparisons: register vs immediate ===
+
+private fun AsmGen.cmpBranchSignedImm(insn: IRInstruction, label: String, branchOp: String, leavesSubroutine: Boolean) {
+    val type = insn.type ?: IRDataType.BYTE
+    val reg = insn.requireSrcA().intNumber
+    val imm = insn.requireImmediateInt()
+    val s = dtSuffix(type)
+    // cmpi.x #imm, <ea> and tst.x <ea> accept memory operands, so the
+    // register-file load into d0 is redundant
+    if(imm==0)
+        emitLine("tst$s  ${regAddr(reg)}")
+    else
+        emitLine("cmpi$s  #$imm, ${regAddr(reg)}")
+    emitBranch(branchOp, label, leavesSubroutine)
+}
+
+// === Signed comparisons: register vs register ===
+
+private fun AsmGen.cmpBranchSignedReg(insn: IRInstruction, label: String, branchOp: String, leavesSubroutine: Boolean) {
+    val type = insn.type ?: IRDataType.BYTE
+    val left = insn.requireSrcA().intNumber
+    val right = insn.requireSrcB().intNumber
+    val s = dtSuffix(type)
+    emitLoadD0(left, type)
+    emitLine("cmp$s  ${regAddr(right)}, d0")
+    emitBranch(branchOp, label, leavesSubroutine)
+}
+
+private val inverseBranchOps = mapOf(
+    "bcc" to "bcs", "bcs" to "bcc",
+    "beq" to "bne", "bne" to "beq",
+    "bmi" to "bpl", "bpl" to "bmi",
+    "bvc" to "bvs", "bvs" to "bvc",
+    "bhi" to "bls", "bls" to "bhi",
+    "bhs" to "blo", "blo" to "bhs",
+    "bgt" to "ble", "ble" to "bgt",
+    "bge" to "blt", "blt" to "bge")
+
+private fun AsmGen.emitBranch(branchOp: String, label: String, leavesSubroutine: Boolean) {
+    invalidateD0Cache()
+    if (!leavesSubroutine || !frameActive) {
+        emitLine("$branchOp  $label")
+        return
+    }
+
+    val skipLabel = makeLabel("branch_keep_frame")
+    emitLine("${inverseBranchOps.getValue(branchOp)}  $skipLabel")
+    emitFrameUnlk()
+    emitLine("jmp  $label")
+    emitLabel(skipLabel)
+}

codeGenM68k/src/prog8/codegen/m68k/InstrControl.kt

@@ -0,0 +1,937 @@
+/*
+ * Control flow IR instruction translations for the M68k code generator.
+ *
+ * Status flag returns (Pc/Pz/Pv/Pn) are handled by the IR's branch-based pattern
+ * (bsteq/bstneg/bstvs). The backend skips status flag returns in translateCall/translateReturnValue
+ * and trusts the IR to emit the appropriate branch instructions.
+ *
+ * PUSHST/POPST: save/restore CCR via the stack (Scc moves CCR to D0).
+ *
+ * Handles: JUMP, JUMPI, CALL, CALLI, CALLFAR, CALLFARVB, SYSCALL,
+ * RETURN, RETURNR, RETURNI, PUSH, POP, PUSHST, POPST,
+ * CLC, SEC, CLI, SEI, ALIGN,
+ * byte/word extraction (LSIGB, LSIGW, MSIGB, MSIGW, BSIGB, MIDB),
+ * sign extension (EXT, EXTS), and sign (SGN).
+ */
+
+package prog8.codegen.m68k
+
+import prog8.code.core.CpuType
+import prog8.code.core.toHex
+import prog8.code.target.Amiga500Target
+import prog8.intermediate.*
+
+internal fun AsmGen.translateControl(insn: IRInstruction, forwardedImmediateCall: ImmediateCallOptimization? = null) {
+    when (insn.opcode) {
+        Opcode.JUMP -> {
+            invalidateD0Cache()
+            when (val t = insn.requireTarget()) {
+                is CodeReference.Label -> {
+                    // a jump to a label in another subroutine ends this activation, so our frame has to be gone
+                    // before we leave: whatever runs next returns to our caller, not to us
+                    if (frameActive && labelIsOutsideCurrentSub(t.name))
+                        emitFrameUnlk()
+                    emitLine("bra  ${fixNameSymbols(t.name)}")     // PC-relative branch; vasm picks the optimal size and falls back to jmp if out of range
+                }
+                is CodeReference.Absolute -> {
+                    // a fixed address can't be resolved to a chunk of this subroutine, so assume it
+                    // leaves for good: whatever runs there returns to our caller, not to us
+                    emitFrameUnlk()
+                    emitLine("jmp  ${t.address.value.toHex()}")
+                }
+                is CodeReference.Indirect -> error("JUMP needs target")
+            }
+        }
+
+        Opcode.JUMPI -> {
+            val ref = insn.requireTarget() as? CodeReference.Indirect ?: error("JUMPI needs an indirect target register")
+            val reg = ref.pointer.intNumber
+            invalidateD0Cache()
+            // Keep the frame for a table whose targets all belong to this subroutine.
+            if (!jumpTableStaysInCurrentSub(ref.pointer.register))
+                emitFrameUnlk()
+            if(program.options.compTarget.cpu >= CpuType.M68020) {
+                // 68020+ supports memory-indirect addressing: fetch the target address from memory directly.
+                emitLine("jmp  ([${regAddr(reg)}])")
+            } else {
+                emitLine("movea.l  ${regAddr(reg).removeSuffix("+0")}, a0")
+                emitLine("jmp  (a0)")
+            }
+        }
+
+        Opcode.CALL -> {
+            val callSite = insn.requireCallSite()
+            val fnLabel = insn.labelTarget?.let { fixNameSymbols(it) }
+                ?: (insn.codeTarget as? CodeReference.Absolute)?.address?.value?.toHex()
+                ?: error("CALL needs label or address")
+            invalidateD0Cache()
+            translateCall(fnLabel, callSite, forwardedImmediateCall)
+        }
+
+        Opcode.CALLI -> {
+            val ref = insn.codeTarget as? CodeReference.Indirect ?: error("CALLI needs an indirect target register")
+            val reg = ref.pointer.intNumber
+            invalidateD0Cache()
+            if(program.options.compTarget.cpu >= CpuType.M68020) {
+                // 68020+ supports memory-indirect addressing: fetch the target address from memory directly.
+                emitLine("jsr  ([${regAddr(reg)}])")
+            } else {
+                emitLine("movea.l  ${regAddr(reg).removeSuffix("+0")}, a0")
+                emitLine("jsr  (a0)")
+            }
+        }
+
+        Opcode.CALLFAR -> {
+            // On the amiga target, CALLFAR is repurposed to encode automatic library LVO calls:
+            // library = library number (1=exec, 2=dos, 3=graphics, 4=intuition, ...)
+            // lvo = negative LVO offset (the offset into the library's jump table)
+            // name = symbolic function name, emitted as the displacement in jsr Symbol(a6)
+            if(program.options.compTarget.name.contains("amiga")) {
+                val callSite = insn.requireCallSite()
+                val amigaTarget = callSite.target as? CallTarget.AmigaLibrary ?: error("CALLFAR needs an amiga library target")
+                val offset = amigaTarget.lvo
+                require(offset<0) { "amiga libcall offsets should all be <0" }
+                val bank = amigaTarget.library
+                if(bank==1) {
+                    // exec.library is always at 4.w
+                    emitLine("move.l  4.w,a6")
+                } else {
+                    val library = Amiga500Target.LibraryNumbers.getValue(bank)
+                    val baseVar = "sys.${library}Base"
+                    emitLine("move.l  $baseVar,a6")
+                }
+
+                invalidateD0Cache()
+                for(arg in callSite.arguments)
+                    translateArgument(arg)
+                if(amigaTarget.name!=null) {
+                    emitLine("jsr  ${amigaTarget.name}(a6)")    // do the actual call (symbolic ref)
+                } else {
+                    emitLine("jsr  $offset(a6)")    // do the actual call
+                }
+                // Skip status flag returns: handled by IR branch pattern.
+                // In multi-assign context (returns.size > 1), also skip slot-based returns:
+                // the IR generates LOADHR for them.
+                // In single-return context, process slot returns normally.
+                // Multiple status flag returns in one multi-assign are protected by the IR:
+                // it wraps each flag's extraction in PUSHST/POPST so they don't clobber each other.
+                // On the 68000 cpu this fails, because it cannot save/restore the CCR with a
+                // nonprivileged instruction (compile for 68010 or higher to use this).
+                val isMultiReturn = callSite.results.size > 1
+                for(ret in callSite.results) {
+                    if (ret.location is CallLocation.StatusFlag)
+                        continue
+                    if (isMultiReturn)
+                        continue
+                    translateReturnValue(ret)
+                }
+            }
+            else  error("CALLFAR not applicable on this M68k")
+        }
+
+        Opcode.CALLFARVB -> {
+            error("CALLFARVB not applicable on M68k")
+        }
+
+        Opcode.SYSCALL -> {
+            val callSite = insn.requireCallSite()
+            val num = (callSite.target as CallTarget.SystemCall).number
+            invalidateD0Cache()
+            if (callSite.arguments.isNotEmpty())
+                translateSyscall(num, callSite)
+            else
+                emitLine("; syscall #$num   (no args)")
+        }
+
+        Opcode.RETURN -> {
+            invalidateD0Cache()
+            emitFrameUnlk()
+            emitLine("rts")
+        }
+
+        Opcode.RETURNR -> {
+            invalidateD0Cache()
+            val type = insn.type ?: IRDataType.BYTE
+            if (type == IRDataType.FLOAT) {
+                val fpReg = insn.requireFloatSourceA().floatNumber
+                emitLine("fmove.s  ${floatRegFileAddr(fpReg)}, $FP_ACC")
+            } else {
+                val reg = insn.requireIntSourceA().intNumber
+                emitLoadD0(reg, type)
+            }
+            emitFrameUnlk()
+            emitLine("rts")
+        }
+
+        Opcode.RETURNI -> {
+            invalidateD0Cache()
+            val value = insn.immediate?.integerValue ?: 0
+            val type = insn.type ?: IRDataType.BYTE
+            val s = dtSuffix(type)
+            if (value in -128..127)
+                emitLine("moveq  #$value, d0")
+            else
+                emitLine("move$s  #$value, d0")
+            emitFrameUnlk()
+            emitLine("rts")
+        }
+
+        // === Stack operations ===
+
+        Opcode.PUSH -> {
+            invalidateD0Cache()
+            val reg = insn.requireIntSourceA().intNumber
+            val type = insn.type ?: IRDataType.BYTE
+            val s = dtSuffix(type)
+            emitLine("move$s  ${regAddr(reg)}, -(sp)")
+        }
+
+        Opcode.POP -> {
+            invalidateD0Cache()
+            val reg = insn.requireIntDest().intNumber
+            val type = insn.type ?: IRDataType.BYTE
+            val s = dtSuffix(type)
+            emitLine("move$s  (sp)+, ${regAddr(reg)}")
+        }
+
+        // === Status flag stack ops via Scc ===
+
+        Opcode.PUSHST -> {
+            // Save CCR to the stack. The 68010+ memory form keeps the CCR save
+            // register-neutral (no data register is clobbered) and word-aligned.
+            if(program.options.compTarget.cpu == CpuType.M68000)
+                error("the 68000 cpu cannot save/restore the status bits using a nonprivileged instruction. This is required to implement the 'PUSHST/POPST' IR opcodes. Compile for 68010 or higher cpu or change the code such that PUSHST/POPST are no longer used (sometimes complicated rol/ror operations need it)")
+            emitLine("move  ccr, -(sp)")
+        }
+
+        Opcode.POPST -> {
+            // Restore CCR from the stack (register-neutral on 68010+).
+            if(program.options.compTarget.cpu == CpuType.M68000)
+                error("the 68000 cpu cannot save/restore the status bits using a nonprivileged instruction. This is required to implement the 'PUSHST/POPST' IR opcodes. Compile for 68010 or higher cpu or change the code such that PUSHST/POPST are no longer used (sometimes complicated rol/ror operations need it)")
+            emitLine("move  (sp)+, ccr")
+        }
+
+        // === Flag manipulation ===
+
+        // On M68k, X (extend/CCR bit 4) serves as the rotate-carry,
+        // while C (carry/CCR bit 0) is used for comparisons.
+        // CLC/SEC must manage both to keep them in sync.
+        Opcode.CLC -> emitLine($$"andi  #$ee, ccr")   // clear C and X
+        Opcode.SEC -> emitLine($$"ori  #$11, ccr")    // set C and X
+        Opcode.CLI -> emitLine($$"andi  #$fb, ccr")
+        Opcode.SEI -> emitLine($$"ori  #$04, ccr")
+
+        Opcode.ALIGN -> {
+            val alignment = insn.immediate?.integerValue ?: 2
+            emitLine("ALIGN  $alignment")
+        }
+
+        // === Byte/word extraction (no Scc needed) ===
+
+        Opcode.LSIGB -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            val type = insn.type ?: IRDataType.WORD
+            emitLoadD0(srcReg, type)
+            emitStoreD0(dstReg, IRDataType.BYTE)
+        }
+
+        Opcode.LSIGW -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            emitLoadD0(srcReg, IRDataType.LONG)
+            emitStoreD0(dstReg, IRDataType.WORD)
+        }
+
+        Opcode.MSIGB -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            // Big-endian: MSB is at byte offset 0 for both word and long.
+            emitLoadD0FromAddress(regAddrByte(srcReg, 0), IRDataType.BYTE)
+            emitStoreD0(dstReg, IRDataType.BYTE)
+        }
+
+        Opcode.MSIGW -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            // Big-endian: word 0 (offset+0) is most significant. Extract directly.
+            emitLoadD0FromAddress(regAddrByte(srcReg, 0), IRDataType.WORD)
+            emitStoreD0(dstReg, IRDataType.WORD)
+        }
+
+        Opcode.BSIGB -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            // Big-endian: bits 16-23 are at byte offset 1 of a long.
+            emitLoadD0FromAddress(regAddrByte(srcReg, 1), IRDataType.BYTE)
+            emitStoreD0(dstReg, IRDataType.BYTE)
+        }
+
+        Opcode.MIDB -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            // Big-endian: bits 8-15 are at byte offset 2 of a long.
+            emitLoadD0FromAddress(regAddrByte(srcReg, 2), IRDataType.BYTE)
+            emitStoreD0(dstReg, IRDataType.BYTE)
+        }
+
+        // === Sign/zero extension ===
+
+        Opcode.EXT -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            val type = insn.type ?: IRDataType.WORD
+            when (type) {
+                IRDataType.BYTE -> {
+                    // zero-extend byte to word
+                    invalidateD0Cache()
+                    emitLine("moveq  #0, d0")
+                    emitLoadD0(srcReg, IRDataType.BYTE)
+                    emitStoreD0(dstReg, IRDataType.WORD)
+                }
+                IRDataType.WORD -> {
+                    // zero-extend word to long
+                    invalidateD0Cache()
+                    emitLine("moveq  #0, d0")
+                    emitLoadD0(srcReg, IRDataType.WORD)
+                    emitStoreD0(dstReg, IRDataType.LONG)
+                }
+                IRDataType.LONG -> {
+                    // no extension needed
+                    emitLoadD0(srcReg, IRDataType.LONG)
+                    emitStoreD0(dstReg, IRDataType.LONG)
+                }
+                else -> TODO("EXT for ${type.name}")
+            }
+        }
+
+        Opcode.EXTS -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            val type = insn.type ?: IRDataType.WORD
+            when (type) {
+                IRDataType.BYTE -> {
+                    // sign-extend byte to word
+                    emitLoadD0(srcReg, IRDataType.BYTE)
+                    emitLine("ext.w  d0")
+                    emitStoreD0(dstReg, IRDataType.WORD)
+                }
+                IRDataType.WORD -> {
+                    // sign-extend word to long
+                    emitLoadD0(srcReg, IRDataType.WORD)
+                    emitLine("ext.l  d0")
+                    emitStoreD0(dstReg, IRDataType.LONG)
+                }
+                IRDataType.LONG -> {
+                    // no extension needed
+                    emitLoadD0(srcReg, IRDataType.LONG)
+                    emitStoreD0(dstReg, IRDataType.LONG)
+                }
+                else -> TODO("EXTS for ${type.name}")
+            }
+        }
+
+        Opcode.EXTL -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            // zero-extend byte to long: move.b only writes the low byte, so clear d0 first
+            invalidateD0Cache()
+            emitLine("moveq  #0, d0")
+            emitLoadD0(srcReg, IRDataType.BYTE)
+            emitStoreD0(dstReg, IRDataType.LONG)
+        }
+
+        Opcode.EXTLS -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            // sign-extend byte to long
+            emitLoadD0(srcReg, IRDataType.BYTE)
+            emitSignExtendByteToLong("d0")
+            emitStoreD0(dstReg, IRDataType.LONG)
+        }
+
+        // === Concatenation ===
+
+        Opcode.CONCAT -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val srcReg2 = insn.requireIntSourceA().intNumber
+            val srcReg3 = insn.requireSrcB().intNumber
+            val type = insn.type ?: IRDataType.BYTE
+            when (type) {
+                IRDataType.BYTE -> {
+                    // r1 = WORD(r2 as MSB, r3 as LSB)
+                    emitLoadD0(srcReg2, IRDataType.BYTE)
+                    emitLine("lsl.w  #8, d0")
+                    invalidateD0Cache()
+                    emitLine("or.b  ${regAddr(srcReg3)}, d0")
+                    emitStoreD0(dstReg, IRDataType.WORD)
+                }
+                IRDataType.WORD -> {
+                    // r1 = LONG(r2 as MSW, r3 as LSW)
+                    emitLoadD0(srcReg2, IRDataType.WORD)
+                    emitLine("swap  d0")
+                    invalidateD0Cache()
+                    emitLine("clr.w  d0")
+                    emitLine("or.w  ${regAddr(srcReg3)}, d0")
+                    emitStoreD0(dstReg, IRDataType.LONG)
+                }
+                else -> TODO("CONCAT for ${type.name}")
+            }
+        }
+
+        // === Math ===
+
+        Opcode.SGN -> {
+            val dstReg = insn.requireIntDest().intNumber
+            if (insn.type == IRDataType.FLOAT) {
+                val srcFp = insn.requireFloatSourceA().floatNumber
+                invalidateD0Cache()
+                emitLine("ftst.s  ${floatRegFileAddr(srcFp)}")
+                emitLine("fslt  d0")
+                emitLine("fsgt  d1")
+                emitLine("neg.b  d1")
+                emitLine("or.b  d1, d0")
+                emitStoreD0(dstReg, IRDataType.BYTE)
+            } else {
+                val srcReg = insn.requireIntSourceA().intNumber
+                val type = insn.type ?: IRDataType.BYTE
+                val s = dtSuffix(type)
+                val zeroLabel = makeLabel("sgn_zero")
+                val doneLabel = makeLabel("sgn_done")
+                emitLoadD0(srcReg, type)
+                emitLine("tst$s  d0")
+                emitLine("beq  $zeroLabel")
+                emitLine("smi  d1")
+                emitLine("bmi  $doneLabel")
+                emitLine("moveq  #1, d1")
+                emitLine("bra  $doneLabel")
+                emitLabel(zeroLabel)
+                emitLine("moveq  #0, d1")
+                emitLabel(doneLabel)
+                emitLine("move.b  d1, ${regAddr(dstReg)}")
+                invalidateD0CacheForSlot(dstReg)    // raw store bypasses emitStoreD0
+            }
+        }
+
+        // === Floating point operations via 68881/68882 FPU ===
+
+        Opcode.FFROMUB -> {
+            val fpDst = insn.requireFloatDest().floatNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            emitLoadD0(srcReg, IRDataType.BYTE)
+            emitLine($$"and.l  #$ff, d0")
+            invalidateD0Cache()
+            emitLine("fmove.l  d0, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(fpDst)}")
+        }
+
+        Opcode.FFROMSB -> {
+            val fpDst = insn.requireFloatDest().floatNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            emitLoadD0(srcReg, IRDataType.BYTE)
+            emitSignExtendByteToLong("d0")
+            emitLine("fmove.l  d0, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(fpDst)}")
+        }
+
+        Opcode.FFROMUW -> {
+            val fpDst = insn.requireFloatDest().floatNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            emitLoadD0(srcReg, IRDataType.WORD)
+            emitLine($$"and.l  #$ffff, d0")
+            invalidateD0Cache()
+            emitLine("fmove.l  d0, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(fpDst)}")
+        }
+
+        Opcode.FFROMSW -> {
+            val fpDst = insn.requireFloatDest().floatNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            emitLoadD0(srcReg, IRDataType.WORD)
+            emitLine("ext.l  d0")
+            emitLine("fmove.l  d0, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(fpDst)}")
+        }
+
+        Opcode.FFROMSL -> {
+            val fpDst = insn.requireFloatDest().floatNumber
+            val srcReg = insn.requireIntSourceA().intNumber
+            emitLoadD0(srcReg, IRDataType.LONG)
+            emitLine("fmove.l  d0, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(fpDst)}")
+        }
+
+        Opcode.FTOUB -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val fpSrc = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(fpSrc)}, $FP_ACC")
+            invalidateD0Cache()
+            emitLine("fmove.b  $FP_ACC, d0")
+            emitLine($$"and.l  #$ff, d0")
+            emitStoreD0(dstReg, IRDataType.BYTE)
+        }
+
+        Opcode.FTOSB -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val fpSrc = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(fpSrc)}, $FP_ACC")
+            invalidateD0Cache()
+            emitLine("fmove.b  $FP_ACC, d0")
+            emitStoreD0(dstReg, IRDataType.BYTE)
+        }
+
+        Opcode.FTOUW -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val fpSrc = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(fpSrc)}, $FP_ACC")
+            invalidateD0Cache()
+            emitLine("fmove.w  $FP_ACC, d0")
+            emitStoreD0(dstReg, IRDataType.WORD)
+        }
+
+        Opcode.FTOSW -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val fpSrc = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(fpSrc)}, $FP_ACC")
+            invalidateD0Cache()
+            emitLine("fmove.w  $FP_ACC, d0")
+            emitStoreD0(dstReg, IRDataType.WORD)
+        }
+
+        Opcode.FTOSL -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val fpSrc = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(fpSrc)}, $FP_ACC")
+            invalidateD0Cache()
+            emitLine("fmove.l  $FP_ACC, d0")
+            emitStoreD0(dstReg, IRDataType.LONG)
+        }
+
+        Opcode.FABS -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fabs  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FSIN -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fsin  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FCOS -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fcos  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FTAN -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("ftan  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FATAN -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fatan  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FPOW -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fpow  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FLN -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("flogn  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FLOG -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("flog2  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FROUND -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fround  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FFLOOR -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("ffloor  $FP_ACC, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FCEIL -> {
+            // 68881 has no fceil; implement as: if x == int(x) then x else x > 0 ? int(x)+1 : int(x)
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            val isIntLabel = makeLabel("fceil_is_int")
+            val doneLabel = makeLabel("fceil_done")
+            val posLabel = makeLabel(".fceil_pos")
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, $FP_SRC")
+            emitLine("fintrz  $FP_SRC, $FP_SRC")  // truncate toward zero
+            emitLine("fcmp  $FP_ACC, $FP_SRC")
+            emitLine("fbeq  $isIntLabel")              // if equal, already integer
+            emitLine("ftst  $FP_ACC")
+            emitLine("fbgt  $posLabel")                // if >0, need to add 1
+            emitLine("bra  $doneLabel")
+            emitLabel(posLabel)
+            emitLine("fadd.s  #1.0, $FP_SRC")
+            emitLine("bra  $doneLabel")
+            emitLabel(isIntLabel)
+            // dst already holds the integer (from fintrz)
+            emitLabel(doneLabel)
+            emitLine("fmove.s  $FP_SRC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.FCOMP -> {
+            val dstReg = insn.requireIntDest().intNumber
+            val fr1 = insn.requireFloatSourceA().floatNumber
+            val fr2 = insn.requireSrcB().floatNumber
+            val eqLabel = makeLabel("fcomp_eq")
+            val doneLabel = makeLabel("fcomp_done")
+            val gtLabel = makeLabel("fcomp_gt")
+            invalidateD0Cache()
+            emitLine("fmove.s  ${floatRegFileAddr(fr1)}, $FP_ACC")
+            emitLine("fcmp.s  ${floatRegFileAddr(fr2)}, $FP_ACC")
+            emitLine("fbeq  $eqLabel")
+            emitLine("fbgt  $gtLabel")
+            emitLine("moveq  #-1, d0")
+            emitLine("bra  $doneLabel")
+            emitLabel(gtLabel)
+            emitLine("moveq  #1, d0")
+            emitLine("bra  $doneLabel")
+            emitLabel(eqLabel)
+            emitLine("moveq  #0, d0")
+            emitLabel(doneLabel)
+            emitStoreD0(dstReg, IRDataType.BYTE)
+        }
+
+        Opcode.LOADHFACZERO, Opcode.LOADHFACONE, Opcode.STOREHFACZERO, Opcode.STOREHFACONE ->
+            error("${insn.opcode} should have been handled by translateLoadStore")
+
+        else -> error("Unknown control opcode: ${insn.opcode}")
+    }
+}
+
+// === CALL translation with argument handling ===
+
+private fun AsmGen.translateCall(fnLabel: String, callSite: CallSite, forwardedImmediateCall: ImmediateCallOptimization? = null) {
+    if (fnLabel == "sys.memcopy" && emitInlineMemcopyCall(callSite, forwardedImmediateCall))
+        return
+
+    // Emit non-constant arguments first (from regfile), then constant immediates
+    // just before the BSR/JSR to avoid clobbering the constant registers while
+    // evaluating non-const arguments (e.g. address calculations that may use D0/A0 as scratch).
+    // Parameter names are resolved by argument index so each stack push can be
+    // annotated with the parameter it fills; unknown callees get no annotation.
+    val paramNames = program.allSubs().find { it.label == fnLabel }?.parameters?.map { it.name } ?: emptyList()
+    fun paramNameFor(index: Int): String? =
+        paramNames.getOrNull(index)?.substringAfterLast('.')?.removePrefix("p8v_")
+    val forwardedRegs = forwardedImmediateCall?.loads?.keys ?: emptySet()
+    for ((index, arg) in callSite.arguments.withIndex()) {
+        if (arg.source.register in forwardedRegs) continue
+        translateArgument(arg, fnLabel, forwardedImmediateCall, paramNameFor(index))
+    }
+    for ((index, arg) in callSite.arguments.withIndex()) {
+        if (arg.source.register !in forwardedRegs) continue
+        translateArgument(arg, fnLabel, forwardedImmediateCall, paramNameFor(index))
+    }
+
+    // Check if this call targets an inline asmsub — emit its body directly instead of jsr
+    val inlineTarget = program.allAsmSubs().find { it.label == fnLabel && it.isInline }
+    if (inlineTarget != null) {
+        emitRaw(inlineTarget.asmChunk.assembly)
+    } else {
+        // PC-relative call; vasm picks the optimal size and falls back to jsr if out of range
+        emitLine("bsr  $fnLabel")
+    }
+
+    // caller cleanup of the all-stack argument area:
+    // each argument was pushed into its own word-packed slot (2-byte slots for byte/word args,
+    // 4-byte slots for wide args); the total area size is always even
+    val pushedSlots = callSite.arguments.count { it.location is CallLocation.FrameSlot }
+    if (pushedSlots > 0) {
+        val totalSize = callSite.arguments
+            .filter { it.location is CallLocation.FrameSlot }
+            .sumOf { incomingSlotSize(it.source.type, target) }
+        val needsCcr = callSite.results.any { it.location is CallLocation.StatusFlag }
+        when {
+            // addq is compact but updates CCR; lea preserves CCR so status-flag returns
+            // remain visible to the IR's branch pattern.
+            totalSize in 1..8 && !needsCcr -> emitLine("addq.l  #$totalSize,sp")
+            else -> emitLine("lea  $totalSize(sp),sp")
+        }
+    }
+
+    // Move return values back to virtual registers.
+    // Skip status flag returns: always handled by IR's bsteq/bstneg/bstvs branch pattern.
+    // In multi-assign context (results.size > 1), also skip slot-based returns:
+    // the IR generates LOADHR for them, and emitting here would clobber CPU flags
+    // before the branch pattern can read them.
+    // In single-return expression context, process all slot returns normally
+    // (the IR doesn't generate LOADHR for single-return calls).
+    // Multiple status flag returns in one multi-assign are protected by the IR:
+    // it wraps each flag's extraction in PUSHST/POPST so they don't clobber each other.
+    // On the 68000 cpu this fails, because it cannot save/restore the CCR with a
+    // nonprivileged instruction (compile for 68010 or higher to use this).
+    val isMultiReturn = callSite.results.size > 1
+    for (ret in callSite.results) {
+        if (ret.location is CallLocation.StatusFlag)
+            continue
+        if (isMultiReturn)
+            continue
+        translateReturnValue(ret)
+    }
+}
+
+private fun AsmGen.emitInlineMemcopyCall(callSite: CallSite, forwardedImmediateCall: ImmediateCallOptimization?): Boolean {
+    if (callSite.arguments.size != 3) return false
+    val fwd = forwardedImmediateCall ?: return false
+    val srcArg = callSite.arguments[0]
+    val tgtArg = callSite.arguments[1]
+    val countArg = callSite.arguments[2]
+    // size must be a constant immediate forwarded to the call; non-constant -> always call
+    val countLoad = fwd.loads[countArg.source.register] ?: return false
+    val count = countLoad.immediate?.integerValue ?: return false
+    if (count <= 0) return true
+
+    // Only use .w/.l if we can 100% prove both pointers are aligned.
+    // At codegen we only know immediates; computed addresses (e.g. &arr[i]) are not immediate
+    // so we conservatively fall back to byte copies. A future IR alignment analysis could prove more.
+    val srcImm = fwd.loads[srcArg.source.register]?.immediate?.integerValue
+    val tgtImm = fwd.loads[tgtArg.source.register]?.immediate?.integerValue
+    val bothEven = srcImm != null && tgtImm != null && srcImm % 2 == 0 && tgtImm % 2 == 0
+    val bothLongAligned = srcImm != null && tgtImm != null && srcImm % 4 == 0 && tgtImm % 4 == 0
+    val useLong = bothLongAligned && count % 4 == 0
+    val useWord = bothEven && count % 2 == 0
+    // Heuristics: inline small copies; use a dbra loop for the medium "small" range
+    // to keep code size reasonable. The call to memcopy/CopyMem wins for large copies.
+    val threshold = when {
+        useLong -> 64
+        useWord -> 32
+        else -> 16
+    }
+    if (count > threshold) return false
+    // Load pointers into a0/a1 (benefit from forwarded immediates)
+    translateArgument(srcArg, "sys.memcopy", fwd)
+    translateArgument(tgtArg, "sys.memcopy", fwd)
+    // count is constant, no need to load d0
+    // For larger small counts, a dbra loop is smaller than unrolled moves.
+    // Threshold: a dbra loop needs ~10 bytes setup (move.w #count,dN + label + dbra),
+    // so it wins over more than ~5 unrolled moves of the same size.
+    val unrolledLimit = 5
+    when {
+        useLong -> {
+            val longs = count / 4
+            emitRaw("        ; inline memcopy $count bytes as $longs longwords")
+            if (longs <= unrolledLimit) {
+                repeat(longs) { emitLine("move.l  (a0)+,(a1)+") }
+            } else {
+                emitInlineCopyLoop(longs, 4)
+            }
+        }
+        useWord -> {
+            val words = count / 2
+            emitRaw("        ; inline memcopy $count bytes as $words words")
+            if (words <= unrolledLimit) {
+                repeat(words) { emitLine("move.w  (a0)+,(a1)+") }
+            } else {
+                emitInlineCopyLoop(words, 2)
+            }
+        }
+        else -> {
+            emitRaw("        ; inline memcopy $count bytes")
+            if (count <= unrolledLimit) {
+                repeat(count) { emitLine("move.b  (a0)+,(a1)+") }
+            } else {
+                emitInlineCopyLoop(count, 1)
+            }
+        }
+    }
+    return true
+}
+
+private fun AsmGen.emitInlineCopyLoop(iterations: Int, size: Int) {
+    val loopLabel = makeLabel("memcpy_inline")
+    val moveInsn = when (size) {
+        4 -> "move.l"
+        2 -> "move.w"
+        else -> "move.b"
+    }
+    emitLine("moveq  #$iterations, d0")
+    emitLabel(loopLabel)
+    emitLine("$moveInsn  (a0)+,(a1)+")
+    emitLine("dbra  d0,$loopLabel")
+}
+
+private fun AsmGen.translateArgument(
+    arg: CallArgument,
+    fnLabel: String? = null,
+    forwardedImmediateCall: ImmediateCallOptimization? = null,
+    paramName: String? = null
+) {
+    val argReg = arg.source
+    val forwarded = forwardedImmediateCall?.loads?.get(argReg.register)
+
+    // If the argument has a calling convention slot, load it into that hardware register
+    val slot = (arg.location as? CallLocation.HardwareRegister)?.slot
+    if (slot != null) {
+        val hwReg = m68kSlotRegister(slot)
+        if (argReg.isFloat) {
+            if (forwarded != null) {
+                emitFloadConstantTo(hwReg, forwarded.requireImmediateFloat())
+            } else {
+                emitLine("fmove.s  ${floatRegFileAddr(argReg.floatNumber)}, $hwReg")
+            }
+        } else {
+            if (forwarded != null) {
+                val value = forwarded.requireImmediateInt()
+                val s = dtSuffix(argReg.type)
+                if (hwReg.startsWith("a")) {
+                    // address registers only accept movea/suba, not clr or moveq
+                    when (value) {
+                        0 -> emitLine("suba.l  $hwReg, $hwReg")
+                        else -> emitLine("movea.l  #$value, $hwReg")
+                    }
+                } else {
+                    if (value == 0)
+                        emitLine("clr$s  $hwReg")
+                    else {
+                        // use moveq (2 bytes, 4 cycles) when the value fits in its signed 8-bit range;
+                        // for byte args the value is unsigned 0-255, so map 128-255 to -128--1 (low byte is the same)
+                        val moveqValue = if (argReg.type == IRDataType.BYTE && value in 128..255) value - 256 else value
+                        if (moveqValue in -128..127)
+                            emitLine("moveq  #$moveqValue, $hwReg")
+                        else
+                            emitLine("move$s  #$value, $hwReg")
+                    }
+                }
+            } else {
+                val source = regAddr(argReg.intNumber)
+                if (hwReg.startsWith("a") && argReg.type in setOf(IRDataType.LONG, IRDataType.POINTER)) {
+                    // loading a pointer into an address register; movea.l is the proper form and skips the +0 offset
+                    emitLine("movea.l  ${source.removeSuffix("+0")}, $hwReg")
+                } else {
+                    val s = dtSuffix(argReg.type)
+                    emitLine("move$s  $source, $hwReg")
+                }
+            }
+        }
+    } else {
+        // Stack argument: push into the parameter's word-packed slot, landing at the slot
+        // base byte (2-byte slots for byte/word args, 4-byte slots for wide args).
+        // The caller pops the whole argument area after the return, so the callee never needs to know the argument count.
+        if (arg.location is CallLocation.FrameSlot) {
+            if (argReg.isFloat) {
+                // the FPU has no absolute-long addressing mode, so the value has to travel
+                // through an FPU register first (fp0 is translator scratch, never holds a live vreg).
+                // The store itself reserves the slot via predecrement, so no separate subq is needed.
+                emitLine("fmove.s  ${floatRegFileAddr(argReg.floatNumber)},$FP_ACC")
+                emitLine("fmove.s  $FP_ACC,-(sp)", paramName.orEmpty())
+            } else {
+                when (argReg.type) {
+                    IRDataType.BYTE ->
+                        // one-instruction push: on the stack pointer a byte predecrement
+                        // steps by 2 (not 1), so this reserves the 2-byte slot in one go;
+                        // the byte lands at the slot's base byte, matching the zero
+                        // within-slot displacement that byte parameters use
+                        emitLine("move.b  ${regAddr(argReg.intNumber)},-(sp)",
+                            (if (paramName != null) "$paramName " else "") + "(note: pushed as a word)")
+                    // a 2-byte predecrement push reserves and stores the word slot in one instruction
+                    IRDataType.WORD ->
+                        emitLine("move.w  ${regAddr(argReg.intNumber)},-(sp)", paramName.orEmpty())
+                    // longs and pointers fill the whole slot, so reserve and store in one push
+                    else -> emitLine("move.l  ${regAddr(argReg.intNumber)},-(sp)", paramName.orEmpty())
+                }
+            }
+            return
+        }
+        // Store to the callee's parameter variable (if this is a named param)
+        val paramName = (arg.location as? CallLocation.ParameterMemory)?.name.orEmpty()
+        if (paramName.isNotEmpty() && fnLabel != null) {
+            val target = fixNameSymbols("$fnLabel.$paramName")
+            if (argReg.isFloat) {
+                emitLine("fmove.s  ${floatRegFileAddr(argReg.floatNumber)}, $FP_ACC")
+                emitLine("fmove.s  $FP_ACC, $target")
+            } else {
+                val s = dtSuffix(argReg.type)
+                emitLine("move$s  ${regAddr(argReg.intNumber)}, $target")
+            }
+        }
+    }
+}
+
+private fun AsmGen.translateReturnValue(ret: CallResult) {
+    // Status flag returns are handled by the IR's branch-based pattern (bsteq/bstneg/bstvs).
+    // Do NOT emit Scc here - it would clobber the flags before the branch can read them.
+    if (ret.location is CallLocation.StatusFlag)
+        return
+
+    val destination = ret.destination ?: return
+
+    // Otherwise, return value is in a hardware register
+    val slot = (ret.location as? CallLocation.HardwareRegister)?.slot
+    if (slot != null) {
+        val hwReg = m68kSlotRegister(slot)
+        if (destination.isFloat) {
+            emitLine("fmove.s  $hwReg, ${floatRegFileAddr(destination.floatNumber)}")
+        } else {
+            val s = dtSuffix(destination.type)
+            emitLine("move$s  $hwReg, ${regAddr(destination.intNumber)}")
+        }
+    } else {
+        // Default: return value in d0 (standard m68k calling convention)
+        // All non-float returns go through d0-d7 (not a0-a6) because the calling convention
+        // expects values in data registers. Using A0 for pointers would be inconsistent:
+        // callers read return values from data regs unless the slot annotation says otherwise.
+        // Slightly inefficient: m68k pointers ideally live in address registers, but returning
+        // them in d0 is simpler and avoids ambiguity. Explicit @A0 can be used for hot paths.
+        if (destination.isFloat) {
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(destination.floatNumber)}")
+        } else {
+            emitStoreD0(destination.intNumber, destination.type)
+        }
+    }
+}
+
+// === Slot to M68k hardware register mapping ===
+
+// Slots 0..7 are the 6502/cx16-style scalar registers (A, X, Y, AX, AY, XY,
+// FAC1, FAC2) used by target-independent builtins (e.g. divmod returns its
+// quotient in AY). Map them onto distinct M68k hardware registers so the
+// backend can emit code for them.
+fun m68kSlotRegister(slot: CallingConventionSlot): String = when (slot.value) {
+    in 0..7 -> error("slots 0-7 should never be used on the M68K they are 6502 cpu registers")
+    in 10..17 -> "d${slot.value - 10}"     // M68k slots: D0-D7
+    in 18..24 -> "a${slot.value - 18}"     // M68k slots: A0-A6
+    in 25..32 -> "fp${slot.value - 25}"    // M68k slots: FP0-FP7
+    else -> error("unknown calling convention slot: $slot")
+}

codeGenM68k/src/prog8/codegen/m68k/InstrLoadStore.kt

@@ -0,0 +1,255 @@
+package prog8.codegen.m68k
+
+import prog8.intermediate.IRDataType
+import prog8.intermediate.IRInstruction
+import prog8.intermediate.ImmediateOperand
+import prog8.intermediate.Opcode
+import prog8.intermediate.floatNumber
+import prog8.intermediate.intNumber
+import prog8.intermediate.requireFloatDest
+import prog8.intermediate.requireFloatSourceA
+import prog8.intermediate.requireHardwareSlot
+import prog8.intermediate.requireImmediate
+import prog8.intermediate.requireImmediateFloat
+import prog8.intermediate.requireImmediateInt
+import prog8.intermediate.requireIntDest
+import prog8.intermediate.requireIntSourceA
+import prog8.intermediate.requireMemory
+
+internal fun AsmGen.translateLoadStore(insn: IRInstruction, suppressRegfileStore: Boolean = false) {
+    val type = insn.type ?: IRDataType.BYTE
+
+    if (type == IRDataType.FLOAT) {
+        translateFloatLoadStore(insn, suppressRegfileStore)
+        return
+    }
+
+    val s = dtSuffix(type)
+
+    when (insn.opcode) {
+        Opcode.LOAD -> {
+            val dst = insn.requireIntDest().intNumber
+            when (val value = insn.requireImmediate()) {
+                is ImmediateOperand.Integer -> {
+                    if (suppressRegfileStore)
+                        return
+                    if(value.value == 0)
+                        emitLine("clr$s  ${regAddr(dst)}")
+                    else
+                        emitLine("move$s  #${value.value}, ${regAddr(dst)}")
+                    invalidateD0CacheForSlot(dst)
+                }
+                is ImmediateOperand.SymbolAddress -> {
+                    val resolved = resolveSymbolRef(value.symbol)
+                    val symOff = if (value.offset != 0) "$resolved+${value.offset}" else resolved
+                    emitLine("move.l  #$symOff, ${regAddr(dst)}")
+                    invalidateD0CacheForSlot(dst)
+                }
+                is ImmediateOperand.FloatValue -> error("LOAD: float immediate for an integer register")
+            }
+        }
+
+        Opcode.LOADM -> {
+            val dst = insn.requireIntDest().intNumber
+            val target = resolveMemory(insn.requireMemory())
+            emitLine("move$s  $target, ${regAddr(dst)}")
+            invalidateD0CacheForSlot(dst)
+        }
+
+        Opcode.LOADR -> {
+            val dst = insn.requireIntDest().intNumber
+            val src = insn.requireIntSourceA().intNumber
+            emitLine("move$s  ${regAddr(src)}, ${regAddr(dst)}")
+            invalidateD0CacheForSlot(dst)
+        }
+
+        Opcode.LOADX, Opcode.LOADI -> {
+            val dst = insn.requireIntDest().intNumber
+            val source = resolveMemory(insn.requireMemory())
+            emitLoadD0FromAddress(source, type)
+            emitStoreD0(dst, type)
+        }
+
+        Opcode.LOADHR -> {
+            val dst = insn.requireIntDest().intNumber
+            val hwReg = m68kSlotRegister(insn.requireHardwareSlot().slot)
+            emitLine("move$s  $hwReg, ${regAddr(dst)}")
+            invalidateD0CacheForSlot(dst)
+        }
+
+        Opcode.STOREM -> {
+            val src = insn.requireIntSourceA().intNumber
+            val target = resolveMemory(insn.requireMemory())
+            emitLine("move$s  ${regAddr(src)}, $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.STOREIM -> {
+            val value = insn.requireImmediateInt()
+            val target = resolveMemory(insn.requireMemory())
+            if(value == 0)
+                emitLine("clr$s  $target")
+            else
+                emitLine("move$s  #$value, $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.STOREX -> {
+            val value = insn.requireIntSourceA().intNumber
+            val target = resolveMemory(insn.requireMemory())
+            // the indexed effective address uses a0/d0 as scratch, so stage the value in d1
+            emitLine("move$s  ${regAddr(value)}, d1")
+            emitLine("move$s  d1, $target")
+        }
+
+        Opcode.STOREZM, Opcode.STOREZI, Opcode.STOREZX -> {
+            val target = resolveMemory(insn.requireMemory())
+            emitLine("clr$s  $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.STOREHR -> {
+            val src = insn.requireIntSourceA().intNumber
+            val hwReg = m68kSlotRegister(insn.requireHardwareSlot().slot)
+            emitLine("move$s  ${regAddr(src)}, $hwReg")
+        }
+
+        Opcode.STOREI -> {
+            val value = insn.requireIntSourceA().intNumber
+            val target = resolveMemory(insn.requireMemory())
+            emitLine("move$s  ${regAddr(value)}, $target")
+        }
+
+        Opcode.LOADP_INC -> {
+            val dst = insn.requireIntDest().intNumber
+            // r1 = *a; a += sizeof(type)   where a is a pointer variable (LONG) in memory
+            val target = resolveMemory(insn.requireMemory())
+            emitLine("move.l  $target, a0")
+            emitLoadD0FromAddress("(a0)+", type)
+            emitLine("move.l  a0, $target")
+            invalidateD0CacheForAddress(target)
+            emitStoreD0(dst, type)
+        }
+
+        Opcode.STOREP_INC -> {
+            val value = insn.requireIntSourceA().intNumber
+            // *a = r1; a += sizeof(type)
+            val target = resolveMemory(insn.requireMemory())
+            emitLine("move.l  $target, a0")
+            emitLine("move$s  ${regAddr(value)}, (a0)+")
+            emitLine("move.l  a0, $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        else -> error("Unknown load/store opcode: ${insn.opcode}")
+    }
+}
+
+// === Float load/store via FPU (68881/68882) ===
+
+private fun AsmGen.translateFloatLoadStore(insn: IRInstruction, suppressRegfileStore: Boolean = false) {
+    when (insn.opcode) {
+        Opcode.STOREZM, Opcode.STOREZI, Opcode.STOREZX -> {
+            val target = resolveMemory(insn.requireMemory(), forFloat = true)
+            emitLine($$"fmovecr  #$0f, fp0")
+            emitLine("fmove.s  fp0, $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.STOREIM -> {
+            val target = resolveMemory(insn.requireMemory(), forFloat = true)
+            emitFloadConstantTo("fp0", insn.requireImmediateFloat())
+            emitLine("fmove.s  fp0, $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.LOAD -> {
+            val dst = insn.requireFloatDest().floatNumber
+            when (val value = insn.requireImmediate()) {
+                is ImmediateOperand.FloatValue -> {
+                    if (suppressRegfileStore)
+                        return
+                    emitFloadConstantToAcc(value.value)
+                    emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+                }
+                is ImmediateOperand.SymbolAddress -> {
+                    val resolved = resolveSymbolRef(value.symbol)
+                    val symOff = if (value.offset != 0) "$resolved+${value.offset}" else resolved
+                    emitLine("lea  $symOff, a0")
+                    emitLine("fmove.s  (a0), $FP_ACC")
+                    emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+                }
+                is ImmediateOperand.Integer -> error("LOAD.f: integer immediate for a float register")
+            }
+        }
+
+        Opcode.LOADM, Opcode.LOADX, Opcode.LOADI -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val source = resolveMemory(insn.requireMemory(), forFloat = true)
+            emitLine("fmove.s  $source, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.LOADR -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.LOADHR -> {
+            val dst = insn.requireFloatDest().floatNumber
+            val hwReg = m68kSlotRegister(insn.requireHardwareSlot().slot)
+            emitLine("fmove  $hwReg, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+
+        Opcode.STOREM, Opcode.STOREX, Opcode.STOREI -> {
+            val src = insn.requireFloatSourceA().floatNumber
+            val target = resolveMemory(insn.requireMemory(), forFloat = true)
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, $target")
+            invalidateD0CacheForAddress(target)
+        }
+
+        Opcode.STOREHR -> {
+            val src = insn.requireFloatSourceA().floatNumber
+            val hwReg = m68kSlotRegister(insn.requireHardwareSlot().slot)
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+            emitLine("fmove  $FP_ACC, $hwReg")
+        }
+
+        // The 'FAC' hardware float registers only exist on the 6502 targets. On m68k they are
+        // modelled with the two FPU scratch registers; loading them yields the 0.0 / 1.0 constants
+        // that the 6502 targets keep in FAC1 / FAC2.
+        Opcode.LOADHFACZERO -> {
+            val dst = insn.requireFloatDest().floatNumber
+            emitLine($$"fmovecr  #$0f, $$FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+        Opcode.LOADHFACONE -> {
+            val dst = insn.requireFloatDest().floatNumber
+            emitLine("fmovecr  #$32, $FP_ACC")
+            emitLine("fmove.s  $FP_ACC, ${floatRegFileAddr(dst)}")
+        }
+        Opcode.STOREHFACZERO -> {
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_ACC")
+        }
+        Opcode.STOREHFACONE -> {
+            val src = insn.requireFloatSourceA().floatNumber
+            emitLine("fmove.s  ${floatRegFileAddr(src)}, $FP_SRC")
+        }
+
+        else -> error("Unknown float load/store opcode: ${insn.opcode}")
+    }
+}
+
+
+private var floatConstCounter = 0
+internal fun AsmGen.makeFloatConstLabel(value: Double): String {
+    floatConstCounter++
+    val label = "p8c_fconst_$floatConstCounter"
+    this.dataFloatConstants.add(Pair(label, value))
+    return label
+}

codeGenM68k/src/prog8/codegen/m68k/InstrSyscall.kt

@@ -0,0 +1,256 @@
+/*
+ * SYSCALL instruction translations for the M68k code generator.
+ *
+ * Handles inline expansion of intermediate representation syscall
+ * instructions into M68k assembly code.
+ */
+
+package prog8.codegen.m68k
+
+import prog8.intermediate.CallSite
+import prog8.intermediate.IMSyscall
+import prog8.intermediate.IRDataType
+import prog8.intermediate.RegisterOperand
+import prog8.intermediate.floatNumber
+import prog8.intermediate.intNumber
+
+
+// === SYSCALL dispatch ===
+
+internal fun AsmGen.translateSyscall(num: Int, site: CallSite) {
+    when (num) {
+        IMSyscall.STRING_CONTAINS.number -> translateSyscallStringContains(site)
+        IMSyscall.BYTEARRAY_CONTAINS.number -> translateSyscallBytearrayContains(site)
+        IMSyscall.WORDARRAY_CONTAINS.number -> translateSyscallWordarrayContains(site)
+        IMSyscall.SPLIT_WORDARRAY_CONTAINS.number -> translateSyscallSplitWordarrayContains(site)
+        IMSyscall.LONGARRAY_CONTAINS.number -> translateSyscallLongarrayContains(site)
+        IMSyscall.FLOATARRAY_CONTAINS.number -> translateSyscallFloatarrayContains(site)
+        IMSyscall.COMPARE_STRINGS.number -> translateSyscallStringCompare(site)
+        IMSyscall.CLAMP_BYTE.number -> translateSyscallClamp(site, IRDataType.BYTE, true)
+        IMSyscall.CLAMP_UBYTE.number -> translateSyscallClamp(site, IRDataType.BYTE, false)
+        IMSyscall.CLAMP_WORD.number -> translateSyscallClamp(site, IRDataType.WORD, true)
+        IMSyscall.CLAMP_UWORD.number -> translateSyscallClamp(site, IRDataType.WORD, false)
+        IMSyscall.CLAMP_LONG.number -> translateSyscallClamp(site, IRDataType.LONG, true)
+        else -> TODO("syscall $num on m68k")
+    }
+}
+
+private fun CallSite.argument(index: Int): RegisterOperand =
+    arguments.getOrNull(index)?.source ?: error("syscall is missing argument $index")
+
+private fun CallSite.resultRegister(index: Int): RegisterOperand =
+    results.getOrNull(index)?.destination ?: error("syscall is missing result $index")
+
+// Compare two strings by delegating to the library routine prog8_lib.strcmp,
+// which already implements an efficient case-sensitive comparison that
+// correctly handles 32 bits pointers (the pointer type on M68k targets).
+private fun AsmGen.translateSyscallStringCompare(site: CallSite) {
+    val arg1 = site.argument(0)
+    val arg2 = site.argument(1)
+    val resultReg = site.resultRegister(0).intNumber
+    emitLoadD0(arg1.intNumber, arg1.type)
+    loadStringArg(arg2.intNumber, arg2.type, "d1")
+    invalidateD0Cache()         // d0 is caller-saved across the subroutine call
+    emitLine("bsr  prog8_lib.strcmp")
+    emitStoreD0(resultReg, IRDataType.BYTE)
+}
+
+// Load a string pointer argument (16 or 32 bits) into the given data register,
+// matching the calling convention of strings.compare (full 32 bits address).
+private fun AsmGen.loadStringArg(reg: Int, dt: IRDataType, dreg: String) {
+    if (dt == IRDataType.LONG) {
+        emitLine("move.l  ${regAddr(reg)}, $dreg")
+    } else {
+        emitLine("move.w  ${regAddr(reg)}, $dreg")
+    }
+}
+
+// Shared helper for array contains syscalls that scan a contiguous array.
+// The length argument is a word, so arrays up to the target's ARRAY_SIZE_LIMIT
+// are scanned correctly (m68k: 32768 elements, VM: 65535 elements).
+private fun AsmGen.translateSyscallArrayContains(site: CallSite, dt: IRDataType) {
+    val regElem = site.argument(0).intNumber
+    val regArr = site.argument(1).intNumber
+    val regLen = site.argument(2).intNumber
+    val resultReg = site.resultRegister(0).intNumber
+
+    val labelLoop = makeLabel(".ac_loop")
+    val labelFound = makeLabel(".ac_found")
+    val labelDone = makeLabel(".ac_done")
+
+    emitLoadD0(regElem, dt)
+    emitLine("move.l  ${regAddr(regArr)}, a0")
+    emitLine("move.w  ${regAddr(regLen)}, d1")
+    emitLine("subq.w  #1, d1")
+    emitLine("bmi  $labelDone       ; length was 0 -> not found")
+    emitRaw("$labelLoop:")
+    val cmpInstr = when (dt) {
+        IRDataType.BYTE -> "cmp.b"
+        IRDataType.WORD -> "cmp.w"
+        IRDataType.LONG -> "cmp.l"
+        else -> error("array contains: unsupported element type $dt")
+    }
+    emitLine("$cmpInstr  (a0)+, d0")
+    emitLine("beq  $labelFound")
+    emitLine("dbra  d1, $labelLoop")
+    emitLine("move.b  #0, ${regAddrByte(resultReg, 0)}")
+    emitLine("bra  $labelDone")
+    emitRaw("$labelFound:")
+    emitLine("move.b  #1, ${regAddrByte(resultReg, 0)}")
+    emitRaw("$labelDone:")
+}
+
+private fun AsmGen.translateSyscallBytearrayContains(site: CallSite) {
+    translateSyscallArrayContains(site, IRDataType.BYTE)
+}
+
+private fun AsmGen.translateSyscallWordarrayContains(site: CallSite) {
+    translateSyscallArrayContains(site, IRDataType.WORD)
+}
+
+private fun AsmGen.translateSyscallLongarrayContains(site: CallSite) {
+    translateSyscallArrayContains(site, IRDataType.LONG)
+}
+
+private fun AsmGen.translateSyscallSplitWordarrayContains(site: CallSite) {
+    val regElem = site.argument(0).intNumber
+    val regArr = site.argument(1).intNumber
+    val regLen = site.argument(2).intNumber
+    val resultReg = site.resultRegister(0).intNumber
+
+    val labelLoop = makeLabel(".swc_loop")
+    val labelFound = makeLabel(".swc_found")
+    val labelDone = makeLabel(".swc_done")
+
+    emitLoadD0(regElem, IRDataType.WORD)
+    emitLine("move.l  ${regAddr(regArr)}, a0")     // a0 = lsb array start
+    emitLine("move.w  ${regAddr(regLen)}, d1")     // d1 = element count
+    emitLine("move.w  d1, d2")
+    emitLine("add.l  a0, d2")                      // d2 = msb array start (lsb + length)
+    emitLine("subq.w  #1, d1")
+    emitLine("bmi  $labelDone       ; length was 0 -> not found")
+    emitRaw("$labelLoop:")
+    emitLine("move.b  (a0,d1.w), d3")              // d3 = lsb of element at index d1
+    emitLine("move.b  (d2,d1.w), d4")              // d4 = msb of element at index d1
+    emitLine("lsl.w  #8, d4")
+    emitLine("or.w  d3, d4")                       // d4 = reconstructed word
+    emitLine("cmp.w  d0, d4")
+    emitLine("beq  $labelFound")
+    emitLine("dbra  d1, $labelLoop")
+    emitLine("move.b  #0, ${regAddrByte(resultReg, 0)}")
+    emitLine("bra  $labelDone")
+    emitRaw("$labelFound:")
+    emitLine("move.b  #1, ${regAddrByte(resultReg, 0)}")
+    emitRaw("$labelDone:")
+}
+
+private fun AsmGen.translateSyscallFloatarrayContains(site: CallSite) {
+    val regNeedle = site.argument(0).floatNumber
+    val regArr = site.argument(1).intNumber
+    val regLen = site.argument(2).intNumber
+    val resultReg = site.resultRegister(0).intNumber
+
+    val labelLoop = makeLabel(".fac_loop")
+    val labelFound = makeLabel(".fac_found")
+    val labelDone = makeLabel(".fac_done")
+
+    emitLine("fmove.s  ${floatRegFileAddr(regNeedle)}, fp0")
+    emitLine("move.l  ${regAddr(regArr)}, a0")
+    emitLine("move.w  ${regAddr(regLen)}, d1")
+    emitLine("subq.w  #1, d1")
+    emitLine("bmi  $labelDone       ; length was 0 -> not found")
+    emitRaw("$labelLoop:")
+    emitLine("fcmp.s  (a0)+, fp0")
+    emitLine("fbeq  $labelFound")
+    emitLine("dbra  d1, $labelLoop")
+    emitLine("move.b  #0, ${regAddrByte(resultReg, 0)}")
+    emitLine("bra  $labelDone")
+    emitRaw("$labelFound:")
+    emitLine("move.b  #1, ${regAddrByte(resultReg, 0)}")
+    emitRaw("$labelDone:")
+}
+
+private fun AsmGen.translateSyscallStringContains(site: CallSite) {
+    val regChar = site.argument(0).intNumber
+    val regStr = site.argument(1).intNumber
+    val resultReg = site.resultRegister(0).intNumber
+
+    val labelLoop = makeLabel(".sc_loop")
+    val labelFound = makeLabel(".sc_found")
+    val labelDone = makeLabel(".sc_done")
+
+    emitLine("move.b  ${regAddrByte(regChar, 0)}, d0")
+    emitLine("move.l  ${regAddr(regStr)}, a0")
+    emitRaw("$labelLoop:")
+    emitLine("move.b  (a0)+, d1")
+    emitLine("beq  $labelDone       ; reached null terminator -> not found")
+    emitLine("cmp.b  d0, d1")
+    emitLine("beq  $labelFound")
+    emitLine("bra  $labelLoop")
+    emitRaw("$labelFound:")
+    emitLine("move.b  #1, ${regAddrByte(resultReg, 0)}")
+    emitLine("bra  $labelDone")
+    emitRaw("$labelDone:")
+    emitLine("move.b  #0, ${regAddrByte(resultReg, 0)}")
+}
+
+private fun AsmGen.translateSyscallClamp(site: CallSite, dt: IRDataType, signed: Boolean) {
+    val valueReg = site.argument(0).intNumber
+    val minReg = site.argument(1).intNumber
+    val maxReg = site.argument(2).intNumber
+    val resultReg = site.resultRegister(0).intNumber
+
+    val labelCheckMax = makeLabel(".clamp_max")
+    val labelDone = makeLabel(".clamp_done")
+    val bge = if (signed) "bge" else "bhs"
+    val ble = if (signed) "ble" else "bls"
+
+    when (dt) {
+        IRDataType.BYTE -> {
+            emitLoadD0(valueReg, IRDataType.BYTE)
+            emitLine("cmp.b  ${regAddr(minReg)}, d0")
+            emitLine("$bge  $labelCheckMax")
+            emitLoadD0(minReg, IRDataType.BYTE)
+            emitLine("bra  $labelDone")
+            emitRaw("$labelCheckMax:")
+            invalidateD0Cache()             // branch target: d0 still holds valueReg, not minReg
+            emitLine("cmp.b  ${regAddr(maxReg)}, d0")
+            emitLine("$ble  $labelDone")
+            emitLoadD0(maxReg, IRDataType.BYTE)
+            emitRaw("$labelDone:")
+            invalidateD0Cache()             // branch target: d0 differs per incoming path
+            emitStoreD0(resultReg, IRDataType.BYTE)
+        }
+        IRDataType.WORD -> {
+            emitLoadD0(valueReg, IRDataType.WORD)
+            emitLine("cmp.w  ${regAddr(minReg)}, d0")
+            emitLine("$bge  $labelCheckMax")
+            emitLoadD0(minReg, IRDataType.WORD)
+            emitLine("bra  $labelDone")
+            emitRaw("$labelCheckMax:")
+            invalidateD0Cache()             // branch target: d0 still holds valueReg, not minReg
+            emitLine("cmp.w  ${regAddr(maxReg)}, d0")
+            emitLine("$ble  $labelDone")
+            emitLoadD0(maxReg, IRDataType.WORD)
+            emitRaw("$labelDone:")
+            invalidateD0Cache()             // branch target: d0 differs per incoming path
+            emitStoreD0(resultReg, IRDataType.WORD)
+        }
+        IRDataType.LONG -> {
+            emitLoadD0(valueReg, IRDataType.LONG)
+            emitLine("cmp.l  ${regAddr(minReg)}, d0")
+            emitLine("$bge  $labelCheckMax")
+            emitLoadD0(minReg, IRDataType.LONG)
+            emitLine("bra  $labelDone")
+            emitRaw("$labelCheckMax:")
+            invalidateD0Cache()             // branch target: d0 still holds valueReg, not minReg
+            emitLine("cmp.l  ${regAddr(maxReg)}, d0")
+            emitLine("$ble  $labelDone")
+            emitLoadD0(maxReg, IRDataType.LONG)
+            emitRaw("$labelDone:")
+            invalidateD0Cache()             // branch target: d0 differs per incoming path
+            emitStoreD0(resultReg, IRDataType.LONG)
+        }
+        else -> emitLine("; clamp: unsupported dt $dt")
+    }
+}

codeGenM68k/src/prog8/codegen/m68k/M68kCodeGenerator.kt

@@ -0,0 +1,46 @@
+package prog8.codegen.m68k
+
+import prog8.code.ICodeGeneratorBackend
+import prog8.code.SymbolTable
+import prog8.code.assembly.IAssemblyProgram
+import prog8.code.ast.PtProgram
+import prog8.code.core.CompilationOptions
+import prog8.code.core.IErrorReporter
+import prog8.codegen.intermediate.IRCodeGen
+import prog8.codegen.intermediate.StackFrameLayout
+import prog8.intermediate.IRDataType
+import prog8.intermediate.IRFileWriter
+import prog8.intermediate.VirtualRegister
+import prog8.intermediate.dumpVariables
+
+class M68kCodeGenerator(val retainSSA: Boolean): ICodeGeneratorBackend {
+    override fun generate(
+        program: PtProgram,
+        symbolTable: SymbolTable,
+        options: CompilationOptions,
+        errors: IErrorReporter
+    ): IAssemblyProgram {
+
+        val irCodeGen = IRCodeGen(program, symbolTable, options, errors, retainSSA)
+        val irProgram = irCodeGen.generate()
+        val virtualRegisterTypes = irCodeGen.registerTypes().entries.associate { (num, type) ->
+            val register: VirtualRegister = if (type == IRDataType.FLOAT) VirtualRegister.float(num.value) else VirtualRegister.int(num.value)
+            register to type
+        }
+        irProgram.verifyRegisterTypes(virtualRegisterTypes)
+
+        if (options.compTarget.supportsStackFrames)
+            StackFrameLayout(irProgram, errors).apply()
+
+        if (options.dumpVariables)
+            dumpVariables(irProgram)
+
+        IRFileWriter(irProgram, null).write()
+
+        val gen = AsmGen(irProgram, irProgram.options.compTarget, errors)
+        if (!gen.generate())
+            throw RuntimeException("M68k assembly generation failed")
+
+        return AssemblyProgramM68k(irProgram.name)
+    }
+}

codeGenM68k/src/prog8/codegen/m68k/Main.kt

@@ -0,0 +1,45 @@
+package prog8.codegen.m68k
+
+import com.github.ajalt.clikt.core.CliktCommand
+import com.github.ajalt.clikt.core.main
+import com.github.ajalt.clikt.parameters.arguments.argument
+import prog8.code.core.CpuType
+import prog8.code.core.OutputFiles
+import prog8.intermediate.IRFileReader
+import kotlin.io.path.Path
+import kotlin.io.path.readText
+import kotlin.system.exitProcess
+
+fun main(args: Array<String>) {
+    val cli = M68kCli()
+    cli.main(args)
+    val inputFile = cli.inputFile
+
+    val reader = IRFileReader()
+    val source = Path(inputFile).readText()
+    val program = reader.read(source)
+    val target = program.options.compTarget
+    println("Loaded IR program: ${program.name}")
+    println("Target system and CPU: ${target.name} / ${target.cpu}")
+
+    when(target.cpu) {
+        CpuType.M68000, CpuType.M68020 -> {
+            val gen = AsmGen(program, target)
+            val ok = gen.generate()
+            if(!ok) {
+                System.err.println("Assembly failed.")
+                exitProcess(1)
+            }
+            println("Generated assembly: ${OutputFiles.of(program.options, program.name).asm()}")
+        }
+        else -> {
+            error("This code generator only works for M68000 or M68020 CPU.")
+        }
+    }
+}
+
+private class M68kCli : CliktCommand(name = "prog8-m68kgen") {
+    val inputFile by argument(help = "path to .p8ir")
+
+    override fun run() = Unit
+}