don't always use pha/pla in pointer expression code

fixed %asmbinary relative path issues

.gitignore

@@ -29,3 +29,5 @@ parsetab.py
 
 .gradle
 /prog8compiler.jar
+sd*.img
+

.idea/inspectionProfiles/Project_Default.xml

@@ -3,14 +3,11 @@
     <option name="myName" value="Project Default" />
     <inspection_tool class="DuplicatedCode" enabled="true" level="WEAK WARNING" enabled_by_default="true">
       <Languages>
-        <language minSize="100" isEnabled="false" name="JavaScript" />
-        <language isEnabled="false" name="Groovy" />
-        <language isEnabled="false" name="Style Sheets" />
         <language minSize="70" name="Kotlin" />
-        <language isEnabled="false" name="TypeScript" />
-        <language isEnabled="false" name="ActionScript" />
+        <language isEnabled="false" name="Groovy" />
       </Languages>
     </inspection_tool>
+    <inspection_tool class="PyInterpreterInspection" enabled="false" level="WARNING" enabled_by_default="false" />
     <inspection_tool class="SpellCheckingInspection" enabled="true" level="TYPO" enabled_by_default="true">
       <option name="processCode" value="false" />
       <option name="processLiterals" value="true" />

.idea/kotlinc.xml

@@ -1,6 +1,6 @@
 <?xml version="1.0" encoding="UTF-8"?>
 <project version="4">
   <component name="Kotlin2JvmCompilerArguments">
-    <option name="jvmTarget" value="1.8" />
+    <option name="jvmTarget" value="11" />
   </component>
 </project>

.idea/libraries/antlr_4_7_2_complete.xml

@@ -1,9 +0,0 @@
-<component name="libraryTable">
-  <library name="antlr-4.7.2-complete">
-    <CLASSES>
-      <root url="jar://$PROJECT_DIR$/parser/antlr/lib/antlr-4.7.2-complete.jar!/" />
-    </CLASSES>
-    <JAVADOC />
-    <SOURCES />
-  </library>
-</component>

.idea/libraries/antlr_4_9_complete.xml

@@ -1,7 +1,7 @@
 <component name="libraryTable">
-  <library name="antlr-4.8-complete">
+  <library name="antlr-4.9-complete">
     <CLASSES>
-      <root url="jar://$PROJECT_DIR$/parser/antlr/lib/antlr-4.8-complete.jar!/" />
+      <root url="jar://$PROJECT_DIR$/parser/antlr/lib/antlr-4.9-complete.jar!/" />
     </CLASSES>
     <JAVADOC />
     <SOURCES />

.idea/libraries/antlr_runtime_4_7_2.xml

@@ -1,9 +0,0 @@
-<component name="libraryTable">
-  <library name="antlr-runtime-4.7.2">
-    <CLASSES>
-      <root url="jar://$PROJECT_DIR$/parser/antlr/lib/antlr-runtime-4.7.2.jar!/" />
-    </CLASSES>
-    <JAVADOC />
-    <SOURCES />
-  </library>
-</component>

.idea/libraries/antlr_runtime_4_9.xml

@@ -1,7 +1,7 @@
 <component name="libraryTable">
-  <library name="antlr-runtime-4.8">
+  <library name="antlr-runtime-4.9">
     <CLASSES>
-      <root url="jar://$PROJECT_DIR$/parser/antlr/lib/antlr-runtime-4.8.jar!/" />
+      <root url="jar://$PROJECT_DIR$/parser/antlr/lib/antlr-runtime-4.9.jar!/" />
     </CLASSES>
     <JAVADOC />
     <SOURCES />

.idea/libraries/dbus_java_3_2_4.xml

@@ -0,0 +1,9 @@
+<component name="libraryTable">
+  <library name="dbus-java-3.2.4">
+    <CLASSES>
+      <root url="jar://$PROJECT_DIR$/dbusCompilerService/lib/dbus-java-3.2.4.jar!/" />
+    </CLASSES>
+    <JAVADOC />
+    <SOURCES />
+  </library>
+</component>

.idea/libraries/javax_json_api_1_1_4.xml

@@ -0,0 +1,10 @@
+<component name="libraryTable">
+  <library name="javax.json-api-1.1.4">
+    <CLASSES>
+      <root url="jar://$PROJECT_DIR$/httpCompilerService/lib/javax.json-api-1.1.4.jar!/" />
+      <root url="jar://$PROJECT_DIR$/httpCompilerService/lib/javax.json-1.1.4.jar!/" />
+    </CLASSES>
+    <JAVADOC />
+    <SOURCES />
+  </library>
+</component>

.idea/libraries/kotlinx_cli_jvm.xml

@@ -1,7 +1,7 @@
 <component name="libraryTable">
-  <library name="kotlinx-cli-jvm-0.1.0-dev-5">
+  <library name="kotlinx-cli-jvm">
     <CLASSES>
-      <root url="jar://$PROJECT_DIR$/compiler/lib/kotlinx-cli-jvm-0.1.0-dev-5.jar!/" />
+      <root url="jar://$PROJECT_DIR$/compiler/lib/kotlinx-cli-jvm-0.3.1.jar!/" />
     </CLASSES>
     <JAVADOC />
     <SOURCES />

.idea/libraries/slf4j_api_1_7_30.xml

@@ -0,0 +1,10 @@
+<component name="libraryTable">
+  <library name="slf4j-api-1.7.30">
+    <CLASSES>
+      <root url="jar://$PROJECT_DIR$/httpCompilerService/lib/slf4j-api-1.7.30.jar!/" />
+      <root url="jar://$PROJECT_DIR$/httpCompilerService/lib/slf4j-simple-1.7.30.jar!/" />
+    </CLASSES>
+    <JAVADOC />
+    <SOURCES />
+  </library>
+</component>

.idea/libraries/takes_http.xml

@@ -0,0 +1,12 @@
+<component name="libraryTable">
+  <library name="takes-http">
+    <CLASSES>
+      <root url="jar://$PROJECT_DIR$/httpCompilerService/lib/cactoos-0.42.jar!/" />
+      <root url="jar://$PROJECT_DIR$/httpCompilerService/lib/commons-lang3-3.7.jar!/" />
+      <root url="jar://$PROJECT_DIR$/httpCompilerService/lib/commons-text-1.4.jar!/" />
+      <root url="jar://$PROJECT_DIR$/httpCompilerService/lib/takes-1.19.jar!/" />
+    </CLASSES>
+    <JAVADOC />
+    <SOURCES />
+  </library>
+</component>

.idea/misc.xml

@@ -16,7 +16,7 @@
       </list>
     </option>
   </component>
-  <component name="ProjectRootManager" version="2" languageLevel="JDK_1_8" default="false" project-jdk-name="Kotlin SDK" project-jdk-type="KotlinSDK">
+  <component name="ProjectRootManager" version="2" languageLevel="JDK_11" default="false" project-jdk-name="Kotlin SDK" project-jdk-type="KotlinSDK">
     <output url="file://$PROJECT_DIR$/out" />
   </component>
 </project>

.idea/modules.xml

@@ -3,8 +3,11 @@
   <component name="ProjectModuleManager">
     <modules>
       <module fileurl="file://$PROJECT_DIR$/compiler/compiler.iml" filepath="$PROJECT_DIR$/compiler/compiler.iml" />
+      <module fileurl="file://$PROJECT_DIR$/compilerAst/compilerAst.iml" filepath="$PROJECT_DIR$/compilerAst/compilerAst.iml" />
+      <module fileurl="file://$PROJECT_DIR$/dbusCompilerService/dbusCompilerService.iml" filepath="$PROJECT_DIR$/dbusCompilerService/dbusCompilerService.iml" />
       <module fileurl="file://$PROJECT_DIR$/docs/docs.iml" filepath="$PROJECT_DIR$/docs/docs.iml" />
       <module fileurl="file://$PROJECT_DIR$/examples/examples.iml" filepath="$PROJECT_DIR$/examples/examples.iml" />
+      <module fileurl="file://$PROJECT_DIR$/httpCompilerService/httpCompilerService.iml" filepath="$PROJECT_DIR$/httpCompilerService/httpCompilerService.iml" />
       <module fileurl="file://$PROJECT_DIR$/parser/parser.iml" filepath="$PROJECT_DIR$/parser/parser.iml" />
     </modules>
   </component>

.travis.yml

@@ -4,8 +4,8 @@ sudo: false
 # dist: xenial
 
 before_install:
-  - chmod +x gradlew
+  - chmod +x ./gradlew
 
 script:
-  - gradle test
+  - ./gradlew test
 

README.md

@@ -2,61 +2,74 @@
 [![Build Status](https://travis-ci.org/irmen/prog8.svg?branch=master)](https://travis-ci.org/irmen/prog8)
 [![Documentation](https://readthedocs.org/projects/prog8/badge/?version=latest)](https://prog8.readthedocs.io/)
 
-Prog8 - Structured Programming Language for 8-bit 6502/6510 microprocessors
-===========================================================================
+Prog8 - Structured Programming Language for 8-bit 6502/65c02 microprocessors
+============================================================================
 
 *Written by Irmen de Jong (irmen@razorvine.net)*
 
 *Software license: GNU GPL 3.0, see file LICENSE*
 
 
-This is a structured programming language for the 8-bit 6502/6510 microprocessor from the late 1970's and 1980's
+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,
-which aims to provide many conveniences over raw assembly code (even when using a macro assembler):
+which aims to provide many conveniences over raw assembly code (even when using a macro assembler).
 
-- reduction of source code length
+Documentation
+-------------
+Full documentation (syntax reference, how to use the language and the compiler, etc.) can be found at:
+https://prog8.readthedocs.io/
+
+
+What does Prog8 provide?
+------------------------
+
+- big reduction of source code length over raw assembly
 - modularity, symbol scoping, subroutines
 - various data types other than just bytes (16-bit words, floats, strings)
 - automatic variable allocations, automatic string and array variables and string sharing
-- subroutines with a input- and output parameter signature
-- constant folding in expressions
+- subroutines with an input- and output parameter signature
+- no stack frame allocations because parameters and local variables are automatically allocated statically
+- constant folding in expressions and other high-level program optimizations
 - conditional branches
+- floating point operations  (requires the C64 Basic ROM routines for this)
 - 'when' statement to provide a concise jump table alternative to if/elseif chains
 - structs to group together sets of variables and manipulate them at once
-- floating point operations  (requires the C64 Basic ROM routines for this)
-- abstracting away low level aspects such as ZeroPage handling, program startup, explicit memory addresses
-- various code optimizations (code structure, logical and numerical expressions, unused code removal...)
+- many built-in functions such as ``sin``, ``cos``, ``rnd``, ``abs``, ``min``, ``max``, ``sqrt``, ``msb``, ``rol``, ``ror``, ``swap``, ``sort`` and ``reverse``
+- various powerful built-in libraries to do I/O, number conversions, graphics and more  
+- convenience abstractions for low level aspects such as ZeroPage handling, program startup, explicit memory addresses
+- fast execution speed due to compilation to native assembly code
 - inline assembly allows you to have full control when every cycle or byte matters
-- many built-in functions such as ``sin``, ``cos``, ``rnd``, ``abs``, ``min``, ``max``, ``sqrt``, ``msb``, ``rol``, ``ror``, ``swap``, ``memset``, ``memcopy``, ``sort`` and ``reverse``
+- supports the sixteen 'virtual' 16-bit registers R0 .. R15 from the Commander X16, and provides them also on the C64.
 
-Rapid edit-compile-run-debug cycle:
+*Rapid edit-compile-run-debug cycle:*
 
-- use a modern PC to do the work on
-- very quick compilation times
+- use a modern PC to do the work on, use nice editors and enjoy quick compilation times
 - can automatically run the program in the Vice emulator after succesful compilation
 - breakpoints, that let the Vice emulator drop into the monitor if execution hits them
 - source code labels automatically loaded in Vice emulator so it can show them in disassembly
 
-Prog8 is mainly targeted at the Commodore-64 machine at this time.
-Contributions to add support for other 8-bit (or other?!) machines are welcome.
+*Two supported compiler targets* (contributions to improve these or to add support for other machines are welcome!):
 
-Documentation/manual
---------------------
-https://prog8.readthedocs.io/
+- "c64": Commodore-64  (6510 CPU = almost a 6502)
+- "cx16": [CommanderX16](https://www.commanderx16.com)  (65c02 CPU)
+- If you only use standard kernal and prog8 library routines, it is possible to compile the *exact same program* for both machines (just change the compiler target flag)!
 
-Required tools
---------------
+
+
+Additional required tools
+-------------------------
 
 [64tass](https://sourceforge.net/projects/tass64/) - cross assembler. Install this on your shell path.
 A recent .exe version of this tool for Windows can be obtained from my [clone](https://github.com/irmen/64tass/releases) of this project.
 For other platforms it is very easy to compile it yourself (make ; make install).
 
-A **Java runtime (jre or jdk), version 8 or newer**  is required to run a prepackaged version of the compiler.
+A **Java runtime (jre or jdk), version 11 or newer**  is required to run a prepackaged version of the compiler.
 If you want to build it from source, you'll need a Java SDK + Kotlin 1.3.x SDK (or for instance,
 IntelliJ IDEA with the Kotlin plugin).
 
-It's handy to have a C-64 emulator or a real C-64 to run the programs on. The compiler assumes the presence
-of the [Vice emulator](http://vice-emu.sourceforge.net/)
+It's handy to have an emulator (or a real machine perhaps!) to run the programs on. The compiler assumes the presence
+of the [Vice emulator](http://vice-emu.sourceforge.net/)  for the C64 target,
+and the [x16emu emulator](https://github.com/commanderx16/x16-emulator) for the CommanderX16 target.
 
 
 Example code
@@ -64,44 +77,43 @@ Example code
 
 This code calculates prime numbers using the Sieve of Eratosthenes algorithm::
 
-    %import c64utils
+    %import textio
     %zeropage basicsafe
-
+    
     main {
-
+    
         ubyte[256] sieve
-        ubyte candidate_prime = 2
-
+        ubyte candidate_prime = 2       ; is increased in the loop
+    
         sub start() {
-            memset(sieve, 256, false)
-
-            c64scr.print("prime numbers up to 255:\n\n")
+            sys.memset(sieve, 256, false)   ; clear the sieve
+            txt.print("prime numbers up to 255:\n\n")
             ubyte amount=0
-            while true {
+            repeat {
                 ubyte prime = find_next_prime()
                 if prime==0
                     break
-                c64scr.print_ub(prime)
-                c64scr.print(", ")
+                txt.print_ub(prime)
+                txt.print(", ")
                 amount++
             }
-            c64.CHROUT('\n')
-            c64scr.print("number of primes (expected 54): ")
-            c64scr.print_ub(amount)
-            c64.CHROUT('\n')
+            txt.nl()
+            txt.print("number of primes (expected 54): ")
+            txt.print_ub(amount)
+            txt.nl()
         }
-
-
+    
         sub find_next_prime() -> ubyte {
-
             while sieve[candidate_prime] {
                 candidate_prime++
                 if candidate_prime==0
-                    return 0
+                    return 0        ; we wrapped; no more primes
             }
-
+    
+            ; found next one, mark the multiples and return it.
             sieve[candidate_prime] = true
             uword multiple = candidate_prime
+    
             while multiple < len(sieve) {
                 sieve[lsb(multiple)] = true
                 multiple += candidate_prime
@@ -111,11 +123,12 @@ This code calculates prime numbers using the Sieve of Eratosthenes algorithm::
     }
 
 
+
+
 when compiled an ran on a C-64 you'll get:
 
 ![c64 screen](docs/source/_static/primes_example.png)
 
-
 One of the included examples (wizzine.p8) animates a bunch of sprite balloons and looks like this:
 
 ![wizzine screen](docs/source/_static/wizzine.png)
@@ -127,3 +140,9 @@ Another example (cube3d-sprites.p8) draws the vertices of a rotating 3d cube:
 If you want to play a video game, a fully working Tetris clone is included in the examples:
 
 ![tehtriz_screen](docs/source/_static/tehtriz.png)
+
+There are a couple of examples specially made for the CommanderX16 compiler target.
+For instance here's a well known space ship animated in 3D with hidden line removal,
+in the CommanderX16 emulator:
+
+![cobra3d](docs/source/_static/cobra3d.png)

compiler/build.gradle

@@ -1,41 +1,29 @@
-buildscript {
-    dependencies {
-        classpath "org.jetbrains.kotlin:kotlin-gradle-plugin:1.3.72"
-    }
-}
-
 plugins {
-    // id "org.jetbrains.kotlin.jvm" version "1.3.72"
-    id 'application'
-    id 'org.jetbrains.dokka' version "0.9.18"
-    id 'com.github.johnrengelman.shadow' version '5.2.0'
     id 'java'
+    id 'application'
+    id "org.jetbrains.kotlin.jvm" version "1.4.30"
+    id 'org.jetbrains.dokka' version "0.9.18"
+    id 'com.github.johnrengelman.shadow' version '6.1.0'
 }
 
-apply plugin: "kotlin"
-apply plugin: "java"
-
-targetCompatibility = 1.8
-sourceCompatibility = 1.8
+targetCompatibility = 11
+sourceCompatibility = 11
 
 repositories {
     mavenLocal()
     mavenCentral()
-    jcenter()
-    maven { url "https://dl.bintray.com/orangy/maven/" }
+    maven { url "https://kotlin.bintray.com/kotlinx" }
 }
 
 def prog8version = rootProject.file('compiler/res/version.txt').text.trim()
 
 dependencies {
-    implementation project(':parser')
+    implementation project(':compilerAst')
     implementation "org.jetbrains.kotlin:kotlin-stdlib-jdk8"
     // implementation "org.jetbrains.kotlin:kotlin-reflect"
-    implementation 'org.antlr:antlr4-runtime:4.8'
-    implementation 'org.jetbrains.kotlinx:kotlinx-cli-jvm:0.1.0-dev-5'
-    // implementation 'net.razorvine:ksim65:1.6'
-    // implementation "com.github.hypfvieh:dbus-java:3.2.0"
-    implementation project(':parser')
+    implementation 'org.jetbrains.kotlinx:kotlinx-cli:0.3.1'
+    // implementation 'net.razorvine:ksim65:1.8'
+    // implementation "com.github.hypfvieh:dbus-java:3.2.4"
 
     testImplementation "org.jetbrains.kotlin:kotlin-test-junit5"
     testImplementation 'org.junit.jupiter:junit-jupiter-api:5.3.2'
@@ -45,7 +33,8 @@ dependencies {
 
 compileKotlin {
     kotlinOptions {
-        jvmTarget = "1.8"
+        jvmTarget = "11"
+        useIR = true
         // verbose = true
         // freeCompilerArgs += "-XXLanguage:+NewInference"
     }
@@ -53,7 +42,8 @@ compileKotlin {
 
 compileTestKotlin {
     kotlinOptions {
-        jvmTarget = "1.8"
+        jvmTarget = "11"
+        useIR = true
     }
 }
 
@@ -76,7 +66,8 @@ sourceSets {
 startScripts.enabled = true
 
 application {
-    mainClassName = 'prog8.CompilerMainKt'
+    mainClass = 'prog8.CompilerMainKt'
+    mainClassName = 'prog8.CompilerMainKt'  // deprecated
     applicationName = 'p8compile'
 }
 
@@ -110,3 +101,7 @@ dokka {
     outputFormat = 'html'
     outputDirectory = "$buildDir/kdoc"
 }
+
+task wrapper(type: Wrapper) {
+    gradleVersion = '6.7'
+}

compiler/compiler.iml

@@ -8,12 +8,11 @@
       <sourceFolder url="file://$MODULE_DIR$/test" isTestSource="true" />
       <excludeFolder url="file://$MODULE_DIR$/build" />
     </content>
-    <orderEntry type="jdk" jdkName="openjdk-11" jdkType="JavaSDK" />
+    <orderEntry type="jdk" jdkName="11" jdkType="JavaSDK" />
     <orderEntry type="sourceFolder" forTests="false" />
     <orderEntry type="library" name="KotlinJavaRuntime" level="project" />
-    <orderEntry type="module" module-name="parser" />
     <orderEntry type="library" name="unittest-libs" level="project" />
-    <orderEntry type="library" name="kotlinx-cli-jvm-0.1.0-dev-5" level="project" />
-    <orderEntry type="library" name="antlr-runtime-4.8" level="project" />
+    <orderEntry type="library" name="kotlinx-cli-jvm" level="project" />
+    <orderEntry type="module" module-name="compilerAst" />
   </component>
 </module>

compiler/res/.editorconfig

@@ -0,0 +1,20 @@
+root = true
+
+[*]
+charset = utf-8
+end_of_line = lf
+indent_size = 4
+indent_style = space
+insert_final_newline = true
+max_line_length = 120
+tab_width = 8
+trim_trailing_whitespace = true
+ij_smart_tabs = true
+
+[*.p8]
+indent_size = 4
+indent_style = space
+
+[*.asm]
+indent_size = 8
+indent_style = tab

compiler/res/prog8lib/c64/floats.asm

@@ -0,0 +1,678 @@
+; --- low level floating point assembly routines for the C64
+
+FL_ONE_const	.byte  129     			; 1.0
+FL_ZERO_const	.byte  0,0,0,0,0		; 0.0
+FL_LOG2_const	.byte  $80, $31, $72, $17, $f8	; log(2)
+
+
+floats_store_reg	.byte  0		; temp storage
+
+
+ub2float	.proc
+		; -- convert ubyte in SCRATCH_ZPB1 to float at address A/Y
+		;    clobbers A, Y
+		stx  P8ZP_SCRATCH_REG
+		sta  P8ZP_SCRATCH_W2
+		sty  P8ZP_SCRATCH_W2+1
+		ldy  P8ZP_SCRATCH_B1
+		lda  #0
+		jsr  GIVAYF
+_fac_to_mem	ldx  P8ZP_SCRATCH_W2
+		ldy  P8ZP_SCRATCH_W2+1
+		jsr  MOVMF
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+b2float		.proc
+		; -- convert byte in SCRATCH_ZPB1 to float at address A/Y
+		;    clobbers A, Y
+		stx  P8ZP_SCRATCH_REG
+		sta  P8ZP_SCRATCH_W2
+		sty  P8ZP_SCRATCH_W2+1
+		lda  P8ZP_SCRATCH_B1
+		jsr  FREADSA
+		jmp  ub2float._fac_to_mem
+		.pend
+
+uw2float	.proc
+		; -- convert uword in SCRATCH_ZPWORD1 to float at address A/Y
+		stx  P8ZP_SCRATCH_REG
+		sta  P8ZP_SCRATCH_W2
+		sty  P8ZP_SCRATCH_W2+1
+		lda  P8ZP_SCRATCH_W1
+		ldy  P8ZP_SCRATCH_W1+1
+		jsr  GIVUAYFAY
+		jmp  ub2float._fac_to_mem
+		.pend
+
+w2float		.proc
+		; -- convert word in SCRATCH_ZPWORD1 to float at address A/Y
+		stx  P8ZP_SCRATCH_REG
+		sta  P8ZP_SCRATCH_W2
+		sty  P8ZP_SCRATCH_W2+1
+		ldy  P8ZP_SCRATCH_W1
+		lda  P8ZP_SCRATCH_W1+1
+		jsr  GIVAYF
+		jmp  ub2float._fac_to_mem
+		.pend
+
+
+cast_from_uw	.proc
+		; -- uword in A/Y into float var at (P8ZP_SCRATCH_W2)
+		stx  P8ZP_SCRATCH_REG
+		jsr  GIVUAYFAY
+		jmp  ub2float._fac_to_mem
+		.pend
+
+
+cast_from_w	.proc
+		; -- word in A/Y into float var at (P8ZP_SCRATCH_W2)
+		stx  P8ZP_SCRATCH_REG
+		jsr  GIVAYFAY
+		jmp  ub2float._fac_to_mem
+		.pend
+
+
+cast_from_ub	.proc
+		; -- ubyte in Y into float var at (P8ZP_SCRATCH_W2)
+		stx  P8ZP_SCRATCH_REG
+		jsr  FREADUY
+		jmp  ub2float._fac_to_mem
+		.pend
+
+
+cast_from_b	.proc
+		; -- byte in A into float var at (P8ZP_SCRATCH_W2)
+		stx  P8ZP_SCRATCH_REG
+		jsr  FREADSA
+		jmp  ub2float._fac_to_mem
+		.pend
+
+cast_as_uw_into_ya	.proc               ; also used for float 2 ub
+		; -- cast float at A/Y to uword into Y/A
+		jsr  MOVFM
+		jmp  cast_FAC1_as_uw_into_ya
+		.pend
+
+cast_as_w_into_ay	.proc               ; also used for float 2 b
+		; -- cast float at A/Y to word into A/Y
+		jsr  MOVFM
+		jmp  cast_FAC1_as_w_into_ay
+		.pend
+
+cast_FAC1_as_uw_into_ya	.proc               ; also used for float 2 ub
+		; -- cast fac1 to uword into Y/A
+		stx  P8ZP_SCRATCH_REG
+		jsr  GETADR     ; into Y/A
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+cast_FAC1_as_w_into_ay	.proc               ; also used for float 2 b
+		; -- cast fac1 to word into A/Y
+		stx  P8ZP_SCRATCH_REG
+		jsr  AYINT
+		ldy  $64
+		lda  $65
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+
+stack_b2float	.proc
+		; -- b2float operating on the stack
+		inx
+		lda  P8ESTACK_LO,x
+		stx  P8ZP_SCRATCH_REG
+		jsr  FREADSA
+		jmp  push_fac1._internal
+		.pend
+
+stack_w2float	.proc
+		; -- w2float operating on the stack
+		inx
+		ldy  P8ESTACK_LO,x
+		lda  P8ESTACK_HI,x
+		stx  P8ZP_SCRATCH_REG
+		jsr  GIVAYF
+		jmp  push_fac1._internal
+		.pend
+
+stack_ub2float	.proc
+		; -- ub2float operating on the stack
+		inx
+		lda  P8ESTACK_LO,x
+		stx  P8ZP_SCRATCH_REG
+		tay
+		lda  #0
+		jsr  GIVAYF
+		jmp  push_fac1._internal
+		.pend
+
+stack_uw2float	.proc
+		; -- uw2float operating on the stack
+		inx
+		lda  P8ESTACK_LO,x
+		ldy  P8ESTACK_HI,x
+		stx  P8ZP_SCRATCH_REG
+		jsr  GIVUAYFAY
+		jmp  push_fac1._internal
+		.pend
+
+stack_float2w	.proc               ; also used for float2b
+		jsr  pop_float_fac1
+		stx  P8ZP_SCRATCH_REG
+		jsr  AYINT
+		ldx  P8ZP_SCRATCH_REG
+		lda  $64
+		sta  P8ESTACK_HI,x
+		lda  $65
+		sta  P8ESTACK_LO,x
+		dex
+		rts
+		.pend
+
+stack_float2uw	.proc               ; also used for float2ub
+		jsr  pop_float_fac1
+		stx  P8ZP_SCRATCH_REG
+		jsr  GETADR
+		ldx  P8ZP_SCRATCH_REG
+		sta  P8ESTACK_HI,x
+		tya
+		sta  P8ESTACK_LO,x
+		dex
+		rts
+		.pend
+
+push_float	.proc
+		; ---- push mflpt5 in A/Y onto stack
+		; (taking 3 stack positions = 6 bytes of which 1 is padding)
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		ldy  #0
+		lda  (P8ZP_SCRATCH_W1),y
+		sta  P8ESTACK_LO,x
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		sta  P8ESTACK_HI,x
+		dex
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		sta  P8ESTACK_LO,x
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		sta  P8ESTACK_HI,x
+		dex
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		sta  P8ESTACK_LO,x
+		dex
+		rts
+		.pend
+
+pop_float	.proc
+		; ---- pops mflpt5 from stack to memory A/Y
+		; (frees 3 stack positions = 6 bytes of which 1 is padding)
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		ldy  #4
+		inx
+		lda  P8ESTACK_LO,x
+		sta  (P8ZP_SCRATCH_W1),y
+		dey
+		inx
+		lda  P8ESTACK_HI,x
+		sta  (P8ZP_SCRATCH_W1),y
+		dey
+		lda  P8ESTACK_LO,x
+		sta  (P8ZP_SCRATCH_W1),y
+		dey
+		inx
+		lda  P8ESTACK_HI,x
+		sta  (P8ZP_SCRATCH_W1),y
+		dey
+		lda  P8ESTACK_LO,x
+		sta  (P8ZP_SCRATCH_W1),y
+		rts
+		.pend
+
+pop_float_fac1	.proc
+		; -- pops float from stack into FAC1
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  pop_float
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jmp  MOVFM
+		.pend
+
+copy_float	.proc
+		; -- copies the 5 bytes of the mflt value pointed to by SCRATCH_ZPWORD1,
+		;    into the 5 bytes pointed to by A/Y.  Clobbers A,Y.
+		sta  _target+1
+		sty  _target+2
+		ldy  #4
+_loop		lda  (P8ZP_SCRATCH_W1),y
+_target		sta  $ffff,y			; modified
+		dey
+		bpl  _loop
+		rts
+		.pend
+
+inc_var_f	.proc
+		; -- add 1 to float pointed to by A/Y
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		stx  P8ZP_SCRATCH_REG
+		jsr  MOVFM
+		lda  #<FL_ONE_const
+		ldy  #>FL_ONE_const
+		jsr  FADD
+		ldx  P8ZP_SCRATCH_W1
+		ldy  P8ZP_SCRATCH_W1+1
+		jsr  MOVMF
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+dec_var_f	.proc
+		; -- subtract 1 from float pointed to by A/Y
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		stx  P8ZP_SCRATCH_REG
+		lda  #<FL_ONE_const
+		ldy  #>FL_ONE_const
+		jsr  MOVFM
+		lda  P8ZP_SCRATCH_W1
+		ldy  P8ZP_SCRATCH_W1+1
+		jsr  FSUB
+		ldx  P8ZP_SCRATCH_W1
+		ldy  P8ZP_SCRATCH_W1+1
+		jsr  MOVMF
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+
+pop_2_floats_f2_in_fac1	.proc
+		; -- pop 2 floats from stack, load the second one in FAC1 as well
+		lda  #<fmath_float2
+		ldy  #>fmath_float2
+		jsr  pop_float
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  pop_float
+		lda  #<fmath_float2
+		ldy  #>fmath_float2
+		jmp  MOVFM
+		.pend
+
+
+fmath_float1	.byte 0,0,0,0,0	; storage for a mflpt5 value
+fmath_float2	.byte 0,0,0,0,0	; storage for a mflpt5 value
+
+
+push_fac1	.proc
+		; -- push the float in FAC1 onto the stack
+		stx  P8ZP_SCRATCH_REG
+_internal	ldx  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  MOVMF
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		ldx  P8ZP_SCRATCH_REG
+		jmp  push_float
+		.pend
+
+
+pow_f		.proc
+		; -- push f1 ** f2 on stack
+		lda  #<fmath_float2
+		ldy  #>fmath_float2
+		jsr  pop_float
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  pop_float
+		stx  P8ZP_SCRATCH_REG
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  CONUPK		; fac2 = float1
+		lda  #<fmath_float2
+		ldy  #>fmath_float2
+		jsr  FPWR
+		jmp  push_fac1._internal
+		.pend
+
+div_f		.proc
+		; -- push f1/f2 on stack
+		jsr  pop_2_floats_f2_in_fac1
+		stx  P8ZP_SCRATCH_REG
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  FDIV
+		jmp  push_fac1._internal
+		.pend
+
+add_f		.proc
+		; -- push f1+f2 on stack
+		jsr  pop_2_floats_f2_in_fac1
+		stx  P8ZP_SCRATCH_REG
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  FADD
+		jmp  push_fac1._internal
+		.pend
+
+sub_f		.proc
+		; -- push f1-f2 on stack
+		jsr  pop_2_floats_f2_in_fac1
+		stx  P8ZP_SCRATCH_REG
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  FSUB
+		jmp  push_fac1._internal
+		.pend
+
+mul_f		.proc
+		; -- push f1*f2 on stack
+		jsr  pop_2_floats_f2_in_fac1
+		stx  P8ZP_SCRATCH_REG
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  FMULT
+		jmp  push_fac1._internal
+		.pend
+
+neg_f		.proc
+		; -- toggle the sign bit on the stack
+		lda  P8ESTACK_HI+3,x
+		eor  #$80
+		sta  P8ESTACK_HI+3,x
+		rts
+		.pend
+
+var_fac1_less_f	.proc
+		; -- is the float in FAC1 < the variable AY?
+		stx  P8ZP_SCRATCH_REG
+		jsr  FCOMP
+		ldx  P8ZP_SCRATCH_REG
+		cmp  #255
+		beq  +
+		lda  #0
+		rts
++		lda  #1
+		rts
+		.pend
+
+var_fac1_lesseq_f	.proc
+		; -- is the float in FAC1 <= the variable AY?
+		stx  P8ZP_SCRATCH_REG
+		jsr  FCOMP
+		ldx  P8ZP_SCRATCH_REG
+		cmp  #0
+		beq  +
+		cmp  #255
+		beq  +
+		lda  #0
+		rts
++		lda  #1
+		rts
+		.pend
+
+var_fac1_greater_f	.proc
+		; -- is the float in FAC1 > the variable AY?
+		stx  P8ZP_SCRATCH_REG
+		jsr  FCOMP
+		ldx  P8ZP_SCRATCH_REG
+		cmp  #1
+		beq  +
+		lda  #0
++		rts
+		.pend
+
+var_fac1_greatereq_f	.proc
+		; -- is the float in FAC1 >= the variable AY?
+		stx  P8ZP_SCRATCH_REG
+		jsr  FCOMP
+		ldx  P8ZP_SCRATCH_REG
+		cmp  #0
+		beq  +
+		cmp  #1
+		beq  +
+		lda  #0
+		rts
++		lda  #1
+		rts
+		.pend
+
+var_fac1_notequal_f	.proc
+		; -- are the floats numbers in FAC1 and the variable AY *not* identical?
+		stx  P8ZP_SCRATCH_REG
+		jsr  FCOMP
+		ldx  P8ZP_SCRATCH_REG
+		and  #1
+		rts
+		.pend
+
+vars_equal_f	.proc
+		; -- are the mflpt5 numbers in P8ZP_SCRATCH_W1 and AY identical?
+		sta  P8ZP_SCRATCH_W2
+		sty  P8ZP_SCRATCH_W2+1
+		ldy  #0
+		lda  (P8ZP_SCRATCH_W1),y
+		cmp  (P8ZP_SCRATCH_W2),y
+		bne  _false
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		cmp  (P8ZP_SCRATCH_W2),y
+		bne  _false
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		cmp  (P8ZP_SCRATCH_W2),y
+		bne  _false
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		cmp  (P8ZP_SCRATCH_W2),y
+		bne  _false
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		cmp  (P8ZP_SCRATCH_W2),y
+		bne  _false
+		lda  #1
+		rts
+_false		lda  #0
+		rts
+		.pend
+
+equal_f		.proc
+		; -- are the two mflpt5 numbers on the stack identical?
+		inx
+		inx
+		inx
+		inx
+		lda  P8ESTACK_LO-3,x
+		cmp  P8ESTACK_LO,x
+		bne  _equals_false
+		lda  P8ESTACK_LO-2,x
+		cmp  P8ESTACK_LO+1,x
+		bne  _equals_false
+		lda  P8ESTACK_LO-1,x
+		cmp  P8ESTACK_LO+2,x
+		bne  _equals_false
+		lda  P8ESTACK_HI-2,x
+		cmp  P8ESTACK_HI+1,x
+		bne  _equals_false
+		lda  P8ESTACK_HI-1,x
+		cmp  P8ESTACK_HI+2,x
+		bne  _equals_false
+_equals_true	lda  #1
+_equals_store	inx
+		sta  P8ESTACK_LO+1,x
+		rts
+_equals_false	lda  #0
+		beq  _equals_store
+		.pend
+
+notequal_f	.proc
+		; -- are the two mflpt5 numbers on the stack different?
+		jsr  equal_f
+		eor  #1		; invert the result
+		sta  P8ESTACK_LO+1,x
+		rts
+		.pend
+
+vars_less_f	.proc
+		; -- is float in AY < float in P8ZP_SCRATCH_W2 ?
+		jsr  MOVFM
+		lda  P8ZP_SCRATCH_W2
+		ldy  P8ZP_SCRATCH_W2+1
+		stx  P8ZP_SCRATCH_REG
+		jsr  FCOMP
+		ldx  P8ZP_SCRATCH_REG
+		cmp  #255
+		bne  +
+		lda  #1
+		rts
++		lda  #0
+		rts
+		.pend
+
+vars_lesseq_f	.proc
+		; -- is float in AY <= float in P8ZP_SCRATCH_W2 ?
+		jsr  MOVFM
+		lda  P8ZP_SCRATCH_W2
+		ldy  P8ZP_SCRATCH_W2+1
+		stx  P8ZP_SCRATCH_REG
+		jsr  FCOMP
+		ldx  P8ZP_SCRATCH_REG
+		cmp  #255
+		bne  +
+-		lda  #1
+		rts
++		cmp  #0
+		beq  -
+		lda  #0
+		rts
+		.pend
+
+less_f		.proc
+		; -- is f1 < f2?
+		jsr  compare_floats
+		cmp  #255
+		beq  compare_floats._return_true
+		bne  compare_floats._return_false
+		.pend
+
+
+lesseq_f	.proc
+		; -- is f1 <= f2?
+		jsr  compare_floats
+		cmp  #255
+		beq  compare_floats._return_true
+		cmp  #0
+		beq  compare_floats._return_true
+		bne  compare_floats._return_false
+		.pend
+
+greater_f	.proc
+		; -- is f1 > f2?
+		jsr  compare_floats
+		cmp  #1
+		beq  compare_floats._return_true
+		bne  compare_floats._return_false
+		.pend
+
+greatereq_f	.proc
+		; -- is f1 >= f2?
+		jsr  compare_floats
+		cmp  #1
+		beq  compare_floats._return_true
+		cmp  #0
+		beq  compare_floats._return_true
+		bne  compare_floats._return_false
+		.pend
+
+compare_floats	.proc
+		lda  #<fmath_float2
+		ldy  #>fmath_float2
+		jsr  pop_float
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  pop_float
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  MOVFM		; fac1 = flt1
+		lda  #<fmath_float2
+		ldy  #>fmath_float2
+		stx  P8ZP_SCRATCH_REG
+		jsr  FCOMP		; A = flt1 compared with flt2 (0=equal, 1=flt1>flt2, 255=flt1<flt2)
+		ldx  P8ZP_SCRATCH_REG
+		rts
+_return_false	lda  #0
+_return_result  sta  P8ESTACK_LO,x
+		dex
+		rts
+_return_true	lda  #1
+		bne  _return_result
+		.pend
+
+set_array_float_from_fac1	.proc
+		; -- set the float in FAC1 in the array (index in A, array in P8ZP_SCRATCH_W1)
+		sta  P8ZP_SCRATCH_B1
+		asl  a
+		asl  a
+		clc
+		adc  P8ZP_SCRATCH_B1
+		ldy  P8ZP_SCRATCH_W1+1
+		clc
+		adc  P8ZP_SCRATCH_W1
+		bcc  +
+		iny
++		stx  floats_store_reg
+		tax
+		jsr  MOVMF
+		ldx  floats_store_reg
+		rts
+		.pend
+
+
+set_0_array_float	.proc
+		; -- set a float in an array to zero (index in A, array in P8ZP_SCRATCH_W1)
+		sta  P8ZP_SCRATCH_B1
+		asl  a
+		asl  a
+		clc
+		adc  P8ZP_SCRATCH_B1
+		tay
+		lda  #0
+		sta  (P8ZP_SCRATCH_W1),y
+		iny
+		sta  (P8ZP_SCRATCH_W1),y
+		iny
+		sta  (P8ZP_SCRATCH_W1),y
+		iny
+		sta  (P8ZP_SCRATCH_W1),y
+		iny
+		sta  (P8ZP_SCRATCH_W1),y
+		rts
+		.pend
+
+
+set_array_float		.proc
+		; -- set a float in an array to a value (index in A, float in P8ZP_SCRATCH_W1, array in P8ZP_SCRATCH_W2)
+		sta  P8ZP_SCRATCH_B1
+		asl  a
+		asl  a
+		clc
+		adc  P8ZP_SCRATCH_B1
+		adc  P8ZP_SCRATCH_W2
+		ldy  P8ZP_SCRATCH_W2+1
+		bcc  +
+		iny
++		jmp  copy_float
+			; -- copies the 5 bytes of the mflt value pointed to by SCRATCH_ZPWORD1,
+			;    into the 5 bytes pointed to by A/Y.  Clobbers A,Y.
+		.pend
+
+

compiler/res/prog8lib/c64/floats.p8

@@ -4,14 +4,14 @@
 ;
 ; indent format: TABS, size=8
 
+%target c64
 %option enable_floats
 
-
-c64flt {
+floats {
 	; ---- this block contains C-64 floating point related functions ----
 
-        const  float  PI        = 3.141592653589793
-        const  float  TWOPI	    = 6.283185307179586
+        const float  PI     = 3.141592653589793
+        const float  TWOPI  = 6.283185307179586
 
 
 ; ---- C64 basic and kernal ROM float constants and functions ----
@@ -19,29 +19,10 @@ c64flt {
 		; note: the fac1 and fac2 are working registers and take 6 bytes each,
 		; floats in memory  (and rom) are stored in 5-byte MFLPT packed format.
 
-		; constants in five-byte "mflpt" format in the BASIC ROM
-		&float  FL_PIVAL	= $aea8  ; 3.1415926...
-		&float  FL_N32768	= $b1a5  ; -32768
-		&float  FL_FONE		= $b9bc  ; 1
-		&float  FL_SQRHLF	= $b9d6  ; SQR(2) / 2
-		&float  FL_SQRTWO	= $b9db  ; SQR(2)
-		&float  FL_NEGHLF	= $b9e0  ; -.5
-		&float  FL_LOG2		= $b9e5  ; LOG(2)
-		&float  FL_TENC		= $baf9  ; 10
-		&float  FL_NZMIL	= $bdbd  ; 1e9 (1 billion)
-		&float  FL_FHALF	= $bf11  ; .5
-		&float  FL_LOGEB2	= $bfbf  ; 1 / LOG(2)
-		&float  FL_PIHALF	= $e2e0  ; PI / 2
-		&float  FL_TWOPI	= $e2e5  ; 2 * PI
-		&float  FL_FR4		= $e2ea  ; .25
-		; oddly enough, 0.0 isn't available in the kernel.
-        float   FL_ZERO     = 0.0    ; oddly enough 0.0 isn't available in the kernel
-
 
 ; note: fac1/2 might get clobbered even if not mentioned in the function's name.
 ; note: for subtraction and division, the left operand is in fac2, the right operand in fac1.
 
-; checked functions below:
 romsub $bba2 = MOVFM(uword mflpt @ AY) clobbers(A,Y)        ; load mflpt value from memory  in A/Y into fac1
 romsub $bba6 = FREADMEM() clobbers(A,Y)                     ; load mflpt value from memory  in $22/$23 into fac1
 romsub $ba8c = CONUPK(uword mflpt @ AY) clobbers(A,Y)       ; load mflpt value from memory  in A/Y into fac2
@@ -52,22 +33,22 @@ romsub $bc0f = MOVEF() clobbers(A,X)                        ; copy fac1 to fac2
 romsub $bbd4 = MOVMF(uword mflpt @ XY) clobbers(A,Y)        ; store fac1 to memory  X/Y as 5-byte mflpt
 
 ; fac1-> signed word in Y/A (might throw ILLEGAL QUANTITY)
-; (tip: use c64flt.FTOSWRDAY to get A/Y output; lo/hi switched to normal little endian order)
+; (tip: use floats.FTOSWRDAY to get A/Y output; lo/hi switched to normal little endian order)
 romsub $b1aa = FTOSWORDYA() clobbers(X) -> ubyte @ Y, ubyte @ A       ; note: calls AYINT.
 
 ; fac1 -> unsigned word in Y/A (might throw ILLEGAL QUANTITY) (result also in $14/15)
-; (tip: use c64flt.GETADRAY to get A/Y output; lo/hi switched to normal little endian order)
+; (tip: use floats.GETADRAY to get A/Y output; lo/hi switched to normal little endian order)
 romsub $b7f7 = GETADR() clobbers(X) -> ubyte @ Y, ubyte @ A
 
 romsub $bc9b = QINT() clobbers(A,X,Y)           ; fac1 -> 4-byte signed integer in 98-101 ($62-$65), with the MSB FIRST.
 romsub $b1bf = AYINT() clobbers(A,X,Y)          ; fac1-> signed word in 100-101 ($64-$65) MSB FIRST. (might throw ILLEGAL QUANTITY)
 
 ; GIVAYF: signed word in Y/A (note different lsb/msb order) -> float in fac1
-; (tip: use c64flt.GIVAYFAY to use A/Y input; lo/hi switched to normal order)
-; there is also c64flt.GIVUAYFAY - unsigned word in A/Y (lo/hi) to fac1
-; there is also c64flt.FREADS32  that reads from 98-101 ($62-$65) MSB FIRST
-; there is also c64flt.FREADUS32  that reads from 98-101 ($62-$65) MSB FIRST
-; there is also c64flt.FREADS24AXY  that reads signed int24 into fac1 from A/X/Y (lo/mid/hi bytes)
+; (tip: use floats.GIVAYFAY to use A/Y input; lo/hi switched to normal order)
+; there is also floats.GIVUAYFAY - unsigned word in A/Y (lo/hi) to fac1
+; there is also floats.FREADS32  that reads from 98-101 ($62-$65) MSB FIRST
+; there is also floats.FREADUS32  that reads from 98-101 ($62-$65) MSB FIRST
+; there is also floats.FREADS24AXY  that reads signed int24 into fac1 from A/X/Y (lo/mid/hi bytes)
 romsub $b391 = GIVAYF(ubyte lo @ Y, ubyte hi @ A) clobbers(A,X,Y)
 
 romsub $b3a2 = FREADUY(ubyte value @ Y) clobbers(A,X,Y)     ; 8 bit unsigned Y -> float in fac1
@@ -91,6 +72,7 @@ romsub $bb12 = FDIVT() clobbers(A,X,Y)                      ; fac1 = fac2/fac1
 romsub $bb0f = FDIV(uword mflpt @ AY) clobbers(A,X,Y)       ; fac1 = mflpt in A/Y / fac1  (remainder in fac2)
 romsub $bf7b = FPWRT() clobbers(A,X,Y)                      ; fac1 = fac2 ** fac1
 romsub $bf78 = FPWR(uword mflpt @ AY) clobbers(A,X,Y)       ; fac1 = fac2 ** mflpt from A/Y
+romsub $bd7e = FINLOG(byte value @A) clobbers (A, X, Y)     ; fac1 += signed byte in A
 
 romsub $aed4 = NOTOP() clobbers(A,X,Y)                      ; fac1 = NOT(fac1)
 romsub $bccc = INT() clobbers(A,X,Y)                        ; INT() truncates, use FADDH first to round instead of trunc
@@ -163,9 +145,9 @@ asmsub  GIVUAYFAY  (uword value @ AY) clobbers(A,X,Y)  {
 asmsub  GIVAYFAY  (uword value @ AY) clobbers(A,X,Y)  {
 	; ---- signed 16 bit word in A/Y (lo/hi) to float in fac1
 	%asm {{
-		sta  c64.SCRATCH_ZPREG
+		sta  P8ZP_SCRATCH_REG
 		tya
-		ldy  c64.SCRATCH_ZPREG
+		ldy  P8ZP_SCRATCH_REG
 		jmp  GIVAYF		; this uses the inverse order, Y/A
 	}}
 }
@@ -174,9 +156,9 @@ asmsub  FTOSWRDAY  () clobbers(X) -> uword @ AY  {
 	; ---- fac1 to signed word in A/Y
 	%asm {{
 		jsr  FTOSWORDYA	; note the inverse Y/A order
-		sta  c64.SCRATCH_ZPREG
+		sta  P8ZP_SCRATCH_REG
 		tya
-		ldy  c64.SCRATCH_ZPREG
+		ldy  P8ZP_SCRATCH_REG
 		rts
 	}}
 }
@@ -185,41 +167,35 @@ asmsub  GETADRAY  () clobbers(X) -> uword @ AY  {
 	; ---- fac1 to unsigned word in A/Y
 	%asm {{
 		jsr  GETADR		; this uses the inverse order, Y/A
-		sta  c64.SCRATCH_ZPB1
+		sta  P8ZP_SCRATCH_B1
 		tya
-		ldy  c64.SCRATCH_ZPB1
+		ldy  P8ZP_SCRATCH_B1
 		rts
 	}}
 }
 
 sub  print_f  (float value) {
-	; ---- prints the floating point value (without a newline) using basic rom routines.
+	; ---- prints the floating point value (without a newline).
 	%asm {{
-		stx  c64.SCRATCH_ZPREGX
+		stx  floats_store_reg
 		lda  #<value
 		ldy  #>value
 		jsr  MOVFM		; load float into fac1
 		jsr  FOUT		; fac1 to string in A/Y
-		jsr  c64.STROUT			; print string in A/Y
-		ldx  c64.SCRATCH_ZPREGX
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		ldy  #0
+-		lda  (P8ZP_SCRATCH_W1),y
+		beq  +
+		jsr  c64.CHROUT
+		iny
+		bne  -
++		ldx  floats_store_reg
 		rts
 	}}
 }
 
-sub  print_fln  (float value) {
-	; ---- prints the floating point value (with a newline at the end) using basic rom routines
-	%asm {{
-		stx  c64.SCRATCH_ZPREGX
-		lda  #<value
-		ldy  #>value
-		jsr  MOVFM		; load float into fac1
-		jsr  FPRINTLN		; print fac1 with newline
-		ldx  c64.SCRATCH_ZPREGX
-		rts
-	}}
+%asminclude "library:c64/floats.asm", ""
+%asminclude "library:c64/floats_funcs.asm", ""
 
 }
-
-%asminclude "library:c64floats.asm", ""
-
-}  ; ------ end of block c64flt

compiler/res/prog8lib/c64/floats_funcs.asm

@@ -0,0 +1,437 @@
+; --- floating point builtin functions
+
+
+abs_f_stack	.proc
+		; -- push abs(AY) on stack
+		jsr  floats.MOVFM
+		jsr  floats.ABS
+		jmp  push_fac1
+		.pend
+
+abs_f_fac1	.proc
+		; -- FAC1 = abs(AY)
+		jsr  floats.MOVFM
+		jmp  floats.ABS
+		.pend
+
+func_atan_stack	.proc
+		jsr  func_atan_fac1
+		jmp  push_fac1
+		.pend
+
+func_atan_fac1	.proc
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		jsr  ATN
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_ceil_stack	.proc
+		jsr  func_ceil_fac1
+		jmp  push_fac1
+		.pend
+
+func_ceil_fac1	.proc
+		; -- ceil: tr = int(f); if tr==f -> return  else return tr+1
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		ldx  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  MOVMF
+		jsr  INT
+		lda  #<fmath_float1
+		ldy  #>fmath_float1
+		jsr  FCOMP
+		cmp  #0
+		beq  +
+		lda  #<FL_ONE_const
+		ldy  #>FL_ONE_const
+		jsr  FADD
++		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_floor_stack	.proc
+		jsr  func_floor_fac1
+		jmp  push_fac1
+		.pend
+
+func_floor_fac1	.proc
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		jsr  INT
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_round_stack	.proc
+		jsr  func_round_fac1
+		jmp  push_fac1
+		.pend
+
+func_round_fac1	.proc
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		jsr  FADDH
+		jsr  INT
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_sin_stack	.proc
+		jsr  func_sin_fac1
+		jmp  push_fac1
+		.pend
+
+func_sin_fac1	.proc
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		jsr  SIN
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_cos_stack	.proc
+		jsr  func_cos_fac1
+		jmp  push_fac1
+		.pend
+
+func_cos_fac1	.proc
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		jsr  COS
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_tan_stack	.proc
+		jsr  func_tan_fac1
+		jmp  push_fac1
+		.pend
+
+func_tan_fac1	.proc
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		jsr  TAN
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_rad_stack	.proc
+		jsr  func_rad_fac1
+		jmp  push_fac1
+		.pend
+
+func_rad_fac1	.proc
+		; -- convert degrees to radians (d * pi / 180)
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		lda  #<_pi_div_180
+		ldy  #>_pi_div_180
+		jsr  FMULT
+		ldx  P8ZP_SCRATCH_REG
+		rts
+_pi_div_180	.byte 123, 14, 250, 53, 18		; pi / 180
+		.pend
+
+func_deg_stack	.proc
+		jsr  func_deg_fac1
+		jmp  push_fac1
+		.pend
+
+func_deg_fac1	.proc
+		; -- convert radians to degrees (d * (1/ pi * 180))
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		lda  #<_one_over_pi_div_180
+		ldy  #>_one_over_pi_div_180
+		jsr  FMULT
+		ldx  P8ZP_SCRATCH_REG
+		rts
+_one_over_pi_div_180	.byte 134, 101, 46, 224, 211		; 1 / (pi * 180)
+		.pend
+
+func_ln_stack	.proc
+		jsr  func_ln_fac1
+		jmp  push_fac1
+		.pend
+
+func_ln_fac1	.proc
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		jsr  LOG
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_log2_stack	.proc
+		jsr  func_log2_fac1
+		jmp  push_fac1
+		.pend
+
+func_log2_fac1	.proc
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		jsr  LOG
+		jsr  MOVEF
+		lda  #<FL_LOG2_const
+		ldy  #>FL_LOG2_const
+		jsr  MOVFM
+		jsr  FDIVT
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_sign_f_stack	.proc
+		jsr  func_sign_f_into_A
+		sta  P8ESTACK_LO,x
+		dex
+		rts
+		.pend
+
+func_sign_f_into_A	.proc
+		jsr  MOVFM
+		jmp  SIGN
+		.pend
+
+func_sqrt_stack	.proc
+		jsr  func_sqrt_fac1
+		jmp  push_fac1
+		.pend
+
+func_sqrt_fac1	.proc
+		jsr  MOVFM
+		stx  P8ZP_SCRATCH_REG
+		jsr  SQR
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_rndf_stack	.proc
+		jsr  func_rndf_fac1
+		jmp  push_fac1
+		.pend
+
+func_rndf_fac1	.proc
+		stx  P8ZP_SCRATCH_REG
+		lda  #1
+		jsr  FREADSA
+		jsr  RND		; rng into fac1
+		ldx  P8ZP_SCRATCH_REG
+		rts
+		.pend
+
+func_swap_f	.proc
+		; -- swap floats pointed to by SCRATCH_ZPWORD1, SCRATCH_ZPWORD2
+		ldy  #4
+-               lda  (P8ZP_SCRATCH_W1),y
+		pha
+		lda  (P8ZP_SCRATCH_W2),y
+		sta  (P8ZP_SCRATCH_W1),y
+		pla
+		sta  (P8ZP_SCRATCH_W2),y
+		dey
+		bpl  -
+		rts
+		.pend
+
+func_reverse_f	.proc
+		; --- reverse an array of floats (array in P8ZP_SCRATCH_W1, num elements in A)
+_left_index = P8ZP_SCRATCH_W2
+_right_index = P8ZP_SCRATCH_W2+1
+_loop_count = P8ZP_SCRATCH_REG
+		pha
+		jsr  a_times_5
+		sec
+		sbc  #5
+		sta  _right_index
+		lda  #0
+		sta  _left_index
+		pla
+		lsr  a
+		sta  _loop_count
+_loop		; push the left indexed float on the stack
+		ldy  _left_index
+		lda  (P8ZP_SCRATCH_W1),y
+		pha
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		pha
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		pha
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		pha
+		iny
+		lda  (P8ZP_SCRATCH_W1),y
+		pha
+		; copy right index float to left index float
+		ldy  _right_index
+		lda  (P8ZP_SCRATCH_W1),y
+		ldy  _left_index
+		sta  (P8ZP_SCRATCH_W1),y
+		inc  _left_index
+		inc  _right_index
+		ldy  _right_index
+		lda  (P8ZP_SCRATCH_W1),y
+		ldy  _left_index
+		sta  (P8ZP_SCRATCH_W1),y
+		inc  _left_index
+		inc  _right_index
+		ldy  _right_index
+		lda  (P8ZP_SCRATCH_W1),y
+		ldy  _left_index
+		sta  (P8ZP_SCRATCH_W1),y
+		inc  _left_index
+		inc  _right_index
+		ldy  _right_index
+		lda  (P8ZP_SCRATCH_W1),y
+		ldy  _left_index
+		sta  (P8ZP_SCRATCH_W1),y
+		inc  _left_index
+		inc  _right_index
+		ldy  _right_index
+		lda  (P8ZP_SCRATCH_W1),y
+		ldy  _left_index
+		sta  (P8ZP_SCRATCH_W1),y
+		; pop the float off the stack into the right index float
+		ldy  _right_index
+		pla
+		sta  (P8ZP_SCRATCH_W1),y
+		dey
+		pla
+		sta  (P8ZP_SCRATCH_W1),y
+		dey
+		pla
+		sta  (P8ZP_SCRATCH_W1),y
+		dey
+		pla
+		sta  (P8ZP_SCRATCH_W1),y
+		dey
+		pla
+		sta  (P8ZP_SCRATCH_W1),y
+		inc  _left_index
+		lda  _right_index
+		sec
+		sbc  #9
+		sta  _right_index
+		dec  _loop_count
+		bne  _loop
+		rts
+
+		.pend
+
+
+
+a_times_5	.proc
+		sta  P8ZP_SCRATCH_B1
+		asl  a
+		asl  a
+		clc
+		adc  P8ZP_SCRATCH_B1
+		rts
+		.pend
+
+func_any_f_into_A	.proc
+		jsr  a_times_5
+		jmp  prog8_lib.func_any_b_into_A
+		.pend
+
+func_all_f_into_A	.proc
+		jsr  a_times_5
+		jmp  prog8_lib.func_all_b_into_A
+		.pend
+
+func_any_f_stack	.proc
+		jsr  a_times_5
+		jmp  prog8_lib.func_any_b_stack
+		.pend
+
+func_all_f_stack	.proc
+		jsr  a_times_5
+		jmp  prog8_lib.func_all_b_stack
+		.pend
+
+func_max_f_stack	.proc
+		jsr  func_max_f_fac1
+		jmp  push_fac1
+		.pend
+
+func_max_f_fac1	.proc
+		; -- max(array) -> fac1,  array in P8ZP_SCRATCH_W1, num elts in A
+_loop_count = P8ZP_SCRATCH_REG
+		stx  floats_store_reg
+		sta  _loop_count
+		lda  #255
+		sta  _minmax_cmp+1		; modifying
+		lda  #<_largest_neg_float
+		ldy  #>_largest_neg_float
+_minmax_entry	jsr  MOVFM
+-		lda  P8ZP_SCRATCH_W1
+		ldy  P8ZP_SCRATCH_W1+1
+		jsr  FCOMP
+_minmax_cmp	cmp  #255			; modified
+		bne  +
+		lda  P8ZP_SCRATCH_W1
+		ldy  P8ZP_SCRATCH_W1+1
+		jsr  MOVFM
++		lda  P8ZP_SCRATCH_W1
+		clc
+		adc  #5
+		sta  P8ZP_SCRATCH_W1
+		bcc  +
+		inc  P8ZP_SCRATCH_W1+1
++		dec  _loop_count
+		bne  -
+		ldx  floats_store_reg
+		rts
+_largest_neg_float	.byte 255,255,255,255,255		; largest negative float -1.7014118345e+38
+		.pend
+
+func_min_f_stack	.proc
+		jsr  func_min_f_fac1
+		jmp  push_fac1
+		.pend
+
+func_min_f_fac1	.proc
+		; -- min(array) -> fac1,  array in P8ZP_SCRATCH_W1, num elts in A
+		sta  func_max_f_fac1._loop_count
+		lda  #1
+		sta  func_max_f_fac1._minmax_cmp+1
+		lda  #<_largest_pos_float
+		ldy  #>_largest_pos_float
+		jmp  func_max_f_fac1._minmax_entry
+_largest_pos_float	.byte  255,127,255,255,255		; largest positive float
+		rts
+		.pend
+
+
+func_sum_f_stack	.proc
+		jsr  func_sum_f_fac1
+		jmp  push_fac1
+		.pend
+
+func_sum_f_fac1	.proc
+		; -- sum(array) -> fac1,  array in P8ZP_SCRATCH_W1, num elts in A
+_loop_count = P8ZP_SCRATCH_REG
+		stx  floats_store_reg
+		sta  _loop_count
+		lda  #<FL_ZERO_const
+		ldy  #>FL_ZERO_const
+		jsr  MOVFM
+-		lda  P8ZP_SCRATCH_W1
+		ldy  P8ZP_SCRATCH_W1+1
+		jsr  FADD
+		lda  P8ZP_SCRATCH_W1
+		clc
+		adc  #5
+		sta  P8ZP_SCRATCH_W1
+		bcc  +
+		inc  P8ZP_SCRATCH_W1+1
++		dec  _loop_count
+		bne  -
+		ldx  floats_store_reg
+		rts
+		.pend

compiler/res/prog8lib/c64/graphics.p8

@@ -0,0 +1,373 @@
+%target c64
+%import textio
+
+; bitmap pixel graphics module for the C64
+; only black/white monchrome 320x200 for now
+; assumes bitmap screen memory is $2000-$3fff
+
+graphics {
+    const uword BITMAP_ADDRESS = $2000
+    const uword WIDTH = 320
+    const ubyte HEIGHT = 200
+
+    sub enable_bitmap_mode() {
+        ; enable bitmap screen, erase it and set colors to black/white.
+        c64.SCROLY = %00111011
+        c64.SCROLX = %00001000
+        c64.VMCSB = (c64.VMCSB & %11110000) | %00001000   ; $2000-$3fff
+        clear_screen(1, 0)
+    }
+
+    sub disable_bitmap_mode() {
+        ; enables text mode, erase the text screen, color white
+        c64.SCROLY = %00011011
+        c64.SCROLX = %00001000
+        c64.VMCSB = (c64.VMCSB & %11110000) | %00000100   ; $1000-$2fff
+        txt.fill_screen(' ', 1)
+    }
+
+    sub clear_screen(ubyte pixelcolor, ubyte bgcolor) {
+        sys.memset(BITMAP_ADDRESS, 320*200/8, 0)
+        txt.fill_screen(pixelcolor << 4 | bgcolor, 0)
+    }
+
+    sub line(uword @zp x1, ubyte @zp y1, uword @zp x2, ubyte @zp y2) {
+        ; Bresenham algorithm.
+        ; This code special-cases various quadrant loops to allow simple ++ and -- operations.
+        ; TODO there are some slight errors at the first/last pixels in certain slopes...??
+        if y1>y2 {
+            ; make sure dy is always positive to have only 4 instead of 8 special cases
+            swap(x1, x2)
+            swap(y1, y2)
+        }
+        word @zp dx = x2-x1 as word
+        word @zp dy = y2-y1
+
+        if dx==0 {
+            vertical_line(x1, y1, abs(dy)+1 as ubyte)
+            return
+        }
+        if dy==0 {
+            if x1>x2
+                x1=x2
+            horizontal_line(x1, y1, abs(dx)+1 as uword)
+            return
+        }
+
+        ; TODO rewrite the rest in optimized assembly
+
+        word @zp d = 0
+        ubyte positive_ix = true
+        if dx < 0 {
+            dx = -dx
+            positive_ix = false
+        }
+        dx *= 2
+        dy *= 2
+        internal_plotx = x1
+
+        if dx >= dy {
+            if positive_ix {
+                repeat {
+                    internal_plot(y1)
+                    if internal_plotx==x2
+                        return
+                    internal_plotx++
+                    d += dy
+                    if d > dx {
+                        y1++
+                        d -= dx
+                    }
+                }
+            } else {
+                repeat {
+                    internal_plot(y1)
+                    if internal_plotx==x2
+                        return
+                    internal_plotx--
+                    d += dy
+                    if d > dx {
+                        y1++
+                        d -= dx
+                    }
+                }
+            }
+        }
+        else {
+            if positive_ix {
+                repeat {
+                    internal_plot(y1)
+                    if y1 == y2
+                        return
+                    y1++
+                    d += dx
+                    if d > dy {
+                        internal_plotx++
+                        d -= dy
+                    }
+                }
+            } else {
+                repeat {
+                    internal_plot(y1)
+                    if y1 == y2
+                        return
+                    y1++
+                    d += dx
+                    if d > dy {
+                        internal_plotx--
+                        d -= dy
+                    }
+                }
+            }
+        }
+    }
+
+    sub rect(uword x, ubyte y, uword width, ubyte height) {
+        if width==0 or height==0
+            return
+        horizontal_line(x, y, width)
+        if height==1
+            return
+        horizontal_line(x, y+height-1, width)
+        vertical_line(x, y+1, height-2)
+        if width==1
+            return
+        vertical_line(x+width-1, y+1, height-2)
+    }
+
+    sub fillrect(uword x, ubyte y, uword width, ubyte height) {
+        if width==0
+            return
+        repeat height {
+            horizontal_line(x, y, width)
+            y++
+        }
+    }
+
+    sub horizontal_line(uword x, ubyte y, uword length) {
+        if length<8 {
+            internal_plotx=x
+            repeat lsb(length) {
+                internal_plot(y)
+                internal_plotx++
+            }
+            return
+        }
+
+        ubyte separate_pixels = lsb(x) & 7
+        uword addr = get_y_lookup(y) + (x&$fff8)
+
+        if separate_pixels {
+            %asm {{
+                lda  addr
+                sta  P8ZP_SCRATCH_W1
+                lda  addr+1
+                sta  P8ZP_SCRATCH_W1+1
+                ldy  separate_pixels
+                lda  _filled_right,y
+                eor  #255
+                ldy  #0
+                ora  (P8ZP_SCRATCH_W1),y
+                sta  (P8ZP_SCRATCH_W1),y
+            }}
+            addr += 8
+            length += separate_pixels
+            length -= 8
+        }
+
+        if length {
+            %asm {{
+                lda  length
+                and  #7
+                sta  separate_pixels
+                stx  P8ZP_SCRATCH_REG
+                lsr  length+1
+                ror  length
+                lsr  length+1
+                ror  length
+                lsr  length+1
+                ror  length
+                lda  addr
+                sta  _modified+1
+                lda  addr+1
+                sta  _modified+2
+                lda  length
+                ora  length+1
+                beq  _zero
+                ldy  length
+                ldx  #$ff
+_modified       stx  $ffff      ; modified
+                lda  _modified+1
+                clc
+                adc  #8
+                sta  _modified+1
+                bcc  +
+                inc  _modified+2
++               dey
+                bne  _modified
+_zero           ldx  P8ZP_SCRATCH_REG
+
+                ldy  separate_pixels
+                beq  _zero2
+                lda  _modified+1
+                sta  P8ZP_SCRATCH_W1
+                lda  _modified+2
+                sta  P8ZP_SCRATCH_W1+1
+                lda  _filled_right,y
+                ldy  #0
+                ora  (P8ZP_SCRATCH_W1),y
+                sta  (P8ZP_SCRATCH_W1),y
+                jmp  _zero2
+_filled_right   .byte  0, %10000000, %11000000, %11100000, %11110000, %11111000, %11111100, %11111110
+_zero2
+            }}
+        }
+    }
+
+    sub vertical_line(uword x, ubyte y, ubyte height) {
+        internal_plotx = x
+        repeat height {
+            internal_plot(y)
+            y++
+        }
+    }
+
+    sub circle(uword xcenter, ubyte ycenter, ubyte radius) {
+        ; Midpoint algorithm
+        if radius==0
+            return
+        ubyte @zp ploty
+        ubyte @zp yy = 0
+        word @zp decisionOver2 = (1 as word)-radius
+
+        while radius>=yy {
+            internal_plotx = xcenter + radius
+            ploty = ycenter + yy
+            internal_plot(ploty)
+            internal_plotx = xcenter - radius
+            internal_plot(ploty)
+            internal_plotx = xcenter + radius
+            ploty = ycenter - yy
+            internal_plot(ploty)
+            internal_plotx = xcenter - radius
+            internal_plot(ploty)
+            internal_plotx = xcenter + yy
+            ploty = ycenter + radius
+            internal_plot(ploty)
+            internal_plotx = xcenter - yy
+            internal_plot(ploty)
+            internal_plotx = xcenter + yy
+            ploty = ycenter - radius
+            internal_plot(ploty)
+            internal_plotx = xcenter - yy
+            internal_plot(ploty)
+            yy++
+            if decisionOver2<=0
+                decisionOver2 += (yy as word)*2+1
+            else {
+                radius--
+                decisionOver2 += (yy as word -radius)*2+1
+            }
+        }
+    }
+
+    sub disc(uword xcenter, ubyte ycenter, ubyte radius) {
+        ; Midpoint algorithm, filled
+        if radius==0
+            return
+        ubyte @zp yy = 0
+        word decisionOver2 = (1 as word)-radius
+
+        while radius>=yy {
+            horizontal_line(xcenter-radius, ycenter+yy, radius*2+1)
+            horizontal_line(xcenter-radius, ycenter-yy, radius*2+1)
+            horizontal_line(xcenter-yy, ycenter+radius, yy*2+1)
+            horizontal_line(xcenter-yy, ycenter-radius, yy*2+1)
+            yy++
+            if decisionOver2<=0
+                decisionOver2 += (yy as word)*2+1
+            else {
+                radius--
+                decisionOver2 += (yy as word -radius)*2+1
+            }
+        }
+    }
+
+
+; here is the non-asm code for the plot routine below:
+;    sub plot_nonasm(uword px, ubyte py) {
+;        ubyte[] ormask = [128, 64, 32, 16, 8, 4, 2, 1]
+;        uword addr = BITMAP_ADDRESS + 320*(py>>3) + (py & 7) + (px & %0000000111111000)
+;        @(addr) |= ormask[lsb(px) & 7]
+;    }
+
+    inline asmsub  plot(uword plotx @XY, ubyte ploty @A) clobbers (A, X, Y) {
+        %asm {{
+            stx  graphics.internal_plotx
+            sty  graphics.internal_plotx+1
+            jsr  graphics.internal_plot
+        }}
+    }
+
+    ; for efficiency of internal algorithms here is the internal plot routine
+    ; that takes the plotx coordinate in a separate variable instead of the XY register pair:
+
+    uword internal_plotx     ; 0..319        ; separate 'parameter' for internal_plot()
+
+    asmsub  internal_plot(ubyte ploty @A) clobbers (A, X, Y) {      ; internal_plotx is 16 bits 0 to 319... doesn't fit in a register
+        %asm {{
+        tay
+        lda  internal_plotx+1
+        sta  P8ZP_SCRATCH_W2+1
+        lsr  a            ; 0
+        sta  P8ZP_SCRATCH_W2
+        lda  internal_plotx
+        pha
+        and  #7
+        tax
+
+        lda  _y_lookup_lo,y
+        clc
+        adc  P8ZP_SCRATCH_W2
+        sta  P8ZP_SCRATCH_W2
+        lda  _y_lookup_hi,y
+        adc  P8ZP_SCRATCH_W2+1
+        sta  P8ZP_SCRATCH_W2+1
+
+        pla     ; internal_plotx
+        and  #%11111000
+        tay
+        lda  (P8ZP_SCRATCH_W2),y
+        ora  _ormask,x
+        sta  (P8ZP_SCRATCH_W2),y
+        rts
+
+_ormask     .byte 128, 64, 32, 16, 8, 4, 2, 1
+
+; note: this can be even faster if we also have a 256 byte x-lookup table, but hey.
+; see http://codebase64.org/doku.php?id=base:various_techniques_to_calculate_adresses_fast_common_screen_formats_for_pixel_graphics
+; the y lookup tables encodes this formula:  BITMAP_ADDRESS + 320*(py>>3) + (py & 7)    (y from 0..199)
+; We use the 64tass syntax for range expressions to calculate this table on assembly time.
+
+_plot_y_values := $2000 + 320*(range(200)>>3) + (range(200) & 7)
+
+_y_lookup_lo    .byte  <_plot_y_values
+_y_lookup_hi    .byte  >_plot_y_values
+
+        }}
+    }
+
+    asmsub get_y_lookup(ubyte y @Y) -> uword @AY {
+        %asm {{
+            lda  internal_plot._y_lookup_lo,y
+            pha
+            lda  internal_plot._y_lookup_hi,y
+            tay
+            pla
+            rts
+        }}
+    }
+
+}
+
+

compiler/res/prog8lib/c64/syslib.p8

@@ -5,20 +5,13 @@
 ;
 ; indent format: TABS, size=8
 
+%target c64
 
 c64 {
-        const  uword  ESTACK_LO = $ce00     ; evaluation stack (lsb)
-        const  uword  ESTACK_HI = $cf00     ; evaluation stack (msb)
-        &ubyte  SCRATCH_ZPB1    = $02       ; scratch byte 1 in ZP
-        &ubyte  SCRATCH_ZPREG   = $03       ; scratch register in ZP
-        &ubyte  SCRATCH_ZPREGX  = $fa       ; temp storage for X register (stack pointer)
-        &uword  SCRATCH_ZPWORD1 = $fb       ; scratch word in ZP ($fb/$fc)
-        &uword  SCRATCH_ZPWORD2 = $fd       ; scratch word in ZP ($fd/$fe)
-
-
         &ubyte  TIME_HI         = $a0       ; software jiffy clock, hi byte
         &ubyte  TIME_MID        = $a1       ;  .. mid byte
         &ubyte  TIME_LO         = $a2       ;    .. lo byte. Updated by IRQ every 1/60 sec
+        &ubyte  STATUS          = $90       ; kernal status variable for I/O
         &ubyte  STKEY           = $91       ; various keyboard statuses (updated by IRQ)
         &ubyte  SFDX            = $cb       ; current key pressed (matrix value) (updated by IRQ)
 
@@ -183,19 +176,11 @@ c64 {
 ; ---- end of SID registers ----
 
 
+; ---- C64 ROM kernal routines ----
 
-; ---- C64 basic routines ----
-
-romsub $E544 = CLEARSCR() clobbers(A,X,Y)       ; clear the screen
-romsub $E566 = HOMECRSR() clobbers(A,X,Y)       ; cursor to top left of screen
-
-
-; ---- end of C64 basic routines ----
-
-
-; ---- C64 kernal routines ----
-
-romsub $AB1E = STROUT(uword strptr @ AY) clobbers(A, X, Y)      ; print null-terminated string (use c64scr.print instead)
+romsub $AB1E = STROUT(uword strptr @ AY) clobbers(A, X, Y)      ; print null-terminated string (use txt.print instead)
+romsub $E544 = CLEARSCR() clobbers(A,X,Y)                       ; clear the screen
+romsub $E566 = HOMECRSR() clobbers(A,X,Y)                       ; cursor to top left of screen
 romsub $EA31 = IRQDFRT() clobbers(A,X,Y)                        ; default IRQ routine
 romsub $EA81 = IRQDFEND() clobbers(A,X,Y)                       ; default IRQ end/cleanup
 romsub $FF81 = CINT() clobbers(A,X,Y)                           ; (alias: SCINIT) initialize screen editor and video chip
@@ -219,25 +204,479 @@ romsub $FFB4 = TALK(ubyte device @ A) clobbers(A)               ; command serial
 romsub $FFB7 = READST() -> ubyte @ A                            ; read I/O status word
 romsub $FFBA = SETLFS(ubyte logical @ A, ubyte device @ X, ubyte address @ Y)   ; set logical file parameters
 romsub $FFBD = SETNAM(ubyte namelen @ A, str filename @ XY)     ; set filename parameters
-romsub $FFC0 = OPEN() clobbers(A,X,Y)                           ; (via 794 ($31A)) open a logical file
+romsub $FFC0 = OPEN() clobbers(X,Y) -> ubyte @Pc, ubyte @A      ; (via 794 ($31A)) open a logical file
 romsub $FFC3 = CLOSE(ubyte logical @ A) clobbers(A,X,Y)         ; (via 796 ($31C)) close a logical file
-romsub $FFC6 = CHKIN(ubyte logical @ X) clobbers(A,X)           ; (via 798 ($31E)) define an input channel
+romsub $FFC6 = CHKIN(ubyte logical @ X) clobbers(A,X) -> ubyte @Pc    ; (via 798 ($31E)) define an input channel
 romsub $FFC9 = CHKOUT(ubyte logical @ X) clobbers(A,X)          ; (via 800 ($320)) define an output channel
 romsub $FFCC = CLRCHN() clobbers(A,X)                           ; (via 802 ($322)) restore default devices
-romsub $FFCF = CHRIN() clobbers(Y) -> ubyte @ A                 ; (via 804 ($324)) input a character (for keyboard, read a whole line from the screen) A=byte read.
+romsub $FFCF = CHRIN() clobbers(X, Y) -> ubyte @ A   ; (via 804 ($324)) input a character (for keyboard, read a whole line from the screen) A=byte read.
 romsub $FFD2 = CHROUT(ubyte char @ A)                           ; (via 806 ($326)) output a character
 romsub $FFD5 = LOAD(ubyte verify @ A, uword address @ XY) -> ubyte @Pc, ubyte @ A, ubyte @ X, ubyte @ Y     ; (via 816 ($330)) load from device
 romsub $FFD8 = SAVE(ubyte zp_startaddr @ A, uword endaddr @ XY) -> ubyte @ Pc, ubyte @ A                    ; (via 818 ($332)) save to a device
 romsub $FFDB = SETTIM(ubyte low @ A, ubyte middle @ X, ubyte high @ Y)      ; set the software clock
 romsub $FFDE = RDTIM() -> ubyte @ A, ubyte @ X, ubyte @ Y       ; read the software clock
-romsub $FFE1 = STOP() clobbers(A,X) -> ubyte @ Pz, ubyte @ Pc   ; (via 808 ($328)) check the STOP key
-romsub $FFE4 = GETIN() clobbers(X,Y) -> ubyte @ A               ; (via 810 ($32A)) get a character
+romsub $FFE1 = STOP() clobbers(X) -> ubyte @ Pz, ubyte @ A      ; (via 808 ($328)) check the STOP key (and some others in A)
+romsub $FFE4 = GETIN() clobbers(X,Y) -> ubyte @Pc, ubyte @ A    ; (via 810 ($32A)) get a character
 romsub $FFE7 = CLALL() clobbers(A,X)                            ; (via 812 ($32C)) close all files
 romsub $FFEA = UDTIM() clobbers(A,X)                            ; update the software clock
 romsub $FFED = SCREEN() -> ubyte @ X, ubyte @ Y                 ; read number of screen rows and columns
-romsub $FFF0 = PLOT(ubyte col @ Y, ubyte row @ X, ubyte dir @ Pc) -> ubyte @ X, ubyte @ Y       ; read/set position of cursor on screen.  Use c64scr.plot for a 'safe' wrapper that preserves X.
+romsub $FFF0 = PLOT(ubyte col @ Y, ubyte row @ X, ubyte dir @ Pc) -> ubyte @ X, ubyte @ Y       ; read/set position of cursor on screen.  Use txt.plot for a 'safe' wrapper that preserves X.
 romsub $FFF3 = IOBASE() -> uword @ XY                           ; read base address of I/O devices
 
-; ---- end of C64 kernal routines ----
+; ---- end of C64 ROM kernal routines ----
+
+; ---- utilities -----
+
+asmsub STOP2() -> ubyte @A  {
+    ; -- check if STOP key was pressed, returns true if so.  More convenient to use than STOP() because that only sets the carry status flag.
+    %asm {{
+        txa
+        pha
+        jsr  c64.STOP
+        beq  +
+        pla
+        tax
+        lda  #0
+        rts
++       pla
+        tax
+        lda  #1
+        rts
+    }}
+}
+
+asmsub RDTIM16() -> uword @AY {
+    ; --  like RDTIM() but only returning the lower 16 bits for convenience
+    %asm {{
+        stx  P8ZP_SCRATCH_REG
+        jsr  c64.RDTIM
+        pha
+        txa
+        tay
+        pla
+        ldx  P8ZP_SCRATCH_REG
+        rts
+    }}
+}
+
+
+
+; ---- C64 specific system utility routines: ----
+
+asmsub  init_system()  {
+    ; Initializes the machine to a sane starting state.
+    ; Called automatically by the loader program logic.
+    ; This means that the BASIC, KERNAL and CHARGEN ROMs are banked in,
+    ; the VIC, SID and CIA chips are reset, screen is cleared, and the default IRQ is set.
+    ; Also a different color scheme is chosen to identify ourselves a little.
+    ; Uppercase charset is activated, and all three registers set to 0, status flags cleared.
+    %asm {{
+        sei
+        cld
+        lda  #%00101111
+        sta  $00
+        lda  #%00100111
+        sta  $01
+        jsr  c64.IOINIT
+        jsr  c64.RESTOR
+        jsr  c64.CINT
+        lda  #6
+        sta  c64.EXTCOL
+        lda  #7
+        sta  c64.COLOR
+        lda  #0
+        sta  c64.BGCOL0
+        jsr  disable_runstop_and_charsetswitch
+        clc
+        clv
+        cli
+        rts
+    }}
+}
+
+asmsub  init_system_phase2()  {
+    %asm {{
+        rts     ; no phase 2 steps on the C64
+    }}
+}
+
+asmsub  disable_runstop_and_charsetswitch() clobbers(A) {
+    %asm {{
+        lda  #$80
+        sta  657    ; disable charset switching
+        lda  #239
+        sta  808    ; disable run/stop key
+        rts
+    }}
+}
+
+asmsub  set_irq(uword handler @AY, ubyte useKernal @Pc) clobbers(A)  {
+	%asm {{
+	        sta  _modified+1
+	        sty  _modified+2
+	        lda  #0
+	        adc  #0
+	        sta  _use_kernal
+		sei
+		lda  #<_irq_handler
+		sta  c64.CINV
+		lda  #>_irq_handler
+		sta  c64.CINV+1
+		cli
+		rts
+_irq_handler    jsr  _irq_handler_init
+_modified	jsr  $ffff                      ; modified
+		jsr  _irq_handler_end
+		lda  _use_kernal
+		bne  +
+		lda  #$ff
+		sta  c64.VICIRQ			; acknowledge raster irq
+		lda  c64.CIA1ICR		; acknowledge CIA1 interrupt
+		; end irq processing - don't use kernal's irq handling
+		pla
+		tay
+		pla
+		tax
+		pla
+		rti
++		jmp  c64.IRQDFRT		; continue with normal kernal irq routine
+
+_use_kernal     .byte  0
+
+_irq_handler_init
+		; save all zp scratch registers and the X register as these might be clobbered by the irq routine
+		stx  IRQ_X_REG
+		lda  P8ZP_SCRATCH_B1
+		sta  IRQ_SCRATCH_ZPB1
+		lda  P8ZP_SCRATCH_REG
+		sta  IRQ_SCRATCH_ZPREG
+		lda  P8ZP_SCRATCH_W1
+		sta  IRQ_SCRATCH_ZPWORD1
+		lda  P8ZP_SCRATCH_W1+1
+		sta  IRQ_SCRATCH_ZPWORD1+1
+		lda  P8ZP_SCRATCH_W2
+		sta  IRQ_SCRATCH_ZPWORD2
+		lda  P8ZP_SCRATCH_W2+1
+		sta  IRQ_SCRATCH_ZPWORD2+1
+		; stack protector; make sure we don't clobber the top of the evaluation stack
+		dex
+		dex
+		dex
+		dex
+		dex
+		dex
+		cld
+		rts
+
+_irq_handler_end
+		; restore all zp scratch registers and the X register
+		lda  IRQ_SCRATCH_ZPB1
+		sta  P8ZP_SCRATCH_B1
+		lda  IRQ_SCRATCH_ZPREG
+		sta  P8ZP_SCRATCH_REG
+		lda  IRQ_SCRATCH_ZPWORD1
+		sta  P8ZP_SCRATCH_W1
+		lda  IRQ_SCRATCH_ZPWORD1+1
+		sta  P8ZP_SCRATCH_W1+1
+		lda  IRQ_SCRATCH_ZPWORD2
+		sta  P8ZP_SCRATCH_W2
+		lda  IRQ_SCRATCH_ZPWORD2+1
+		sta  P8ZP_SCRATCH_W2+1
+		ldx  IRQ_X_REG
+		rts
+
+IRQ_X_REG		.byte  0
+IRQ_SCRATCH_ZPB1	.byte  0
+IRQ_SCRATCH_ZPREG	.byte  0
+IRQ_SCRATCH_ZPWORD1	.word  0
+IRQ_SCRATCH_ZPWORD2	.word  0
+
+		}}
+}
+
+asmsub  restore_irq() clobbers(A) {
+	%asm {{
+		sei
+		lda  #<c64.IRQDFRT
+		sta  c64.CINV
+		lda  #>c64.IRQDFRT
+		sta  c64.CINV+1
+		lda  #0
+		sta  c64.IREQMASK	; disable raster irq
+		lda  #%10000001
+		sta  c64.CIA1ICR	; restore CIA1 irq
+		cli
+		rts
+	}}
+}
+
+asmsub  set_rasterirq(uword handler @AY, uword rasterpos @R0, ubyte useKernal @Pc) clobbers(A) {
+	%asm {{
+	        sta  _modified+1
+	        sty  _modified+2
+	        lda  #0
+	        adc  #0
+	        sta  set_irq._use_kernal
+		lda  cx16.r0
+		ldy  cx16.r0+1
+		sei
+		jsr  _setup_raster_irq
+		lda  #<_raster_irq_handler
+		sta  c64.CINV
+		lda  #>_raster_irq_handler
+		sta  c64.CINV+1
+		cli
+		rts
+
+_raster_irq_handler
+		jsr  set_irq._irq_handler_init
+_modified	jsr  $ffff              ; modified
+		jsr  set_irq._irq_handler_end
+                lda  #$ff
+                sta  c64.VICIRQ			; acknowledge raster irq
+		lda  set_irq._use_kernal
+		bne  +
+		; end irq processing - don't use kernal's irq handling
+		pla
+		tay
+		pla
+		tax
+		pla
+		rti
++		jmp  c64.IRQDFRT                ; continue with kernal irq routine
+
+_setup_raster_irq
+		pha
+		lda  #%01111111
+		sta  c64.CIA1ICR    ; "switch off" interrupts signals from cia-1
+		sta  c64.CIA2ICR    ; "switch off" interrupts signals from cia-2
+		and  c64.SCROLY
+		sta  c64.SCROLY     ; clear most significant bit of raster position
+		lda  c64.CIA1ICR    ; ack previous irq
+		lda  c64.CIA2ICR    ; ack previous irq
+		pla
+		sta  c64.RASTER     ; set the raster line number where interrupt should occur
+		cpy  #0
+		beq  +
+		lda  c64.SCROLY
+		ora  #%10000000
+		sta  c64.SCROLY     ; set most significant bit of raster position
++		lda  #%00000001
+		sta  c64.IREQMASK   ;enable raster interrupt signals from vic
+		rts
+	}}
+}
+
+; ---- end of C64 specific system utility routines ----
+
+}
+
+sys {
+    ; ------- lowlevel system routines --------
+
+    const ubyte target = 64         ;  compilation target specifier.  64 = C64,  16 = CommanderX16.
+
+
+    asmsub  reset_system()  {
+        ; Soft-reset the system back to Basic prompt.
+        %asm {{
+            sei
+            lda  #14
+            sta  $01        ; bank the kernal in
+            jmp  (c64.RESET_VEC)
+        }}
+    }
+
+    sub wait(uword jiffies) {
+        ; --- wait approximately the given number of jiffies (1/60th seconds)
+        repeat jiffies {
+            ubyte jiff = lsb(c64.RDTIM16())
+            while jiff==lsb(c64.RDTIM16()) {
+                ; wait until 1 jiffy has passed
+            }
+        }
+    }
+
+    asmsub memcopy(uword source @R0, uword target @R1, uword count @AY) clobbers(A,X,Y) {
+        %asm {{
+            ldx  cx16.r0
+            stx  P8ZP_SCRATCH_W1        ; source in ZP
+            ldx  cx16.r0+1
+            stx  P8ZP_SCRATCH_W1+1
+            ldx  cx16.r1
+            stx  P8ZP_SCRATCH_W2        ; target in ZP
+            ldx  cx16.r1+1
+            stx  P8ZP_SCRATCH_W2+1
+            cpy  #0
+            bne  _longcopy
+
+            ; copy <= 255 bytes
+            tay
+            bne  _copyshort
+            rts     ; nothing to copy
+
+_copyshort
+            ; decrease source and target pointers so we can simply index by Y
+            lda  P8ZP_SCRATCH_W1
+            bne  +
+            dec  P8ZP_SCRATCH_W1+1
++           dec  P8ZP_SCRATCH_W1
+            lda  P8ZP_SCRATCH_W2
+            bne  +
+            dec  P8ZP_SCRATCH_W2+1
++           dec  P8ZP_SCRATCH_W2
+
+-           lda  (P8ZP_SCRATCH_W1),y
+            sta  (P8ZP_SCRATCH_W2),y
+            dey
+            bne  -
+            rts
+
+_longcopy
+            sta  P8ZP_SCRATCH_B1        ; lsb(count) = remainder in last page
+            tya
+            tax                         ; x = num pages (1+)
+            ldy  #0
+-           lda  (P8ZP_SCRATCH_W1),y
+            sta  (P8ZP_SCRATCH_W2),y
+            iny
+            bne  -
+            inc  P8ZP_SCRATCH_W1+1
+            inc  P8ZP_SCRATCH_W2+1
+            dex
+            bne  -
+            ldy  P8ZP_SCRATCH_B1
+            bne  _copyshort
+            rts
+        }}
+    }
+
+    asmsub memset(uword mem @R0, uword numbytes @R1, ubyte value @A) clobbers(A,X,Y) {
+        %asm {{
+            ldy  cx16.r0
+            sty  P8ZP_SCRATCH_W1
+            ldy  cx16.r0+1
+            sty  P8ZP_SCRATCH_W1+1
+            ldx  cx16.r1
+            ldy  cx16.r1+1
+            jmp  prog8_lib.memset
+        }}
+    }
+
+    asmsub memsetw(uword mem @R0, uword numwords @R1, uword value @AY) clobbers(A,X,Y) {
+        %asm {{
+            ldx  cx16.r0
+            stx  P8ZP_SCRATCH_W1
+            ldx  cx16.r0+1
+            stx  P8ZP_SCRATCH_W1+1
+            ldx  cx16.r1
+            stx  P8ZP_SCRATCH_W2
+            ldx  cx16.r1+1
+            stx  P8ZP_SCRATCH_W2+1
+            jmp  prog8_lib.memsetw
+        }}
+    }
+
+
+    inline asmsub rsave() {
+        ; save cpu status flag and all registers A, X, Y.
+        ; see http://6502.org/tutorials/register_preservation.html
+        %asm {{
+            php
+            sta  P8ZP_SCRATCH_REG
+            pha
+            txa
+            pha
+            tya
+            pha
+            lda  P8ZP_SCRATCH_REG
+        }}
+    }
+
+    inline asmsub rrestore() {
+        ; restore all registers and cpu status flag
+        %asm {{
+            pla
+            tay
+            pla
+            tax
+            pla
+            plp
+        }}
+    }
+
+    inline asmsub read_flags() -> ubyte @A {
+        %asm {{
+            php
+            pla
+        }}
+    }
+
+    inline asmsub clear_carry() {
+        %asm {{
+            clc
+        }}
+    }
+
+    inline asmsub set_carry() {
+        %asm {{
+            sec
+        }}
+    }
+
+    inline asmsub clear_irqd() {
+        %asm {{
+            cli
+        }}
+    }
+
+    inline asmsub set_irqd() {
+        %asm {{
+            sei
+        }}
+    }
+
+    inline asmsub exit(ubyte returnvalue @A) {
+        ; -- immediately exit the program with a return code in the A register
+        %asm {{
+            jsr  c64.CLRCHN		; reset i/o channels
+            ldx  prog8_lib.orig_stackpointer
+            txs
+            rts		; return to original caller
+        }}
+    }
+
+    inline asmsub progend() -> uword @AY {
+        %asm {{
+            lda  #<prog8_program_end
+            ldy  #>prog8_program_end
+        }}
+    }
+
+}
+
+cx16 {
+
+    ; the sixteen virtual 16-bit registers that the CX16 has defined in the zeropage
+    ; they are simulated on the C64 as well but their location in memory is different
+    ; (because there's no room for them in the zeropage)
+    ; they are allocated at the bottom of the eval-stack (should be ample space unless
+    ; you're doing insane nesting of expressions...)
+    &uword r0  = $cf00
+    &uword r1  = $cf02
+    &uword r2  = $cf04
+    &uword r3  = $cf06
+    &uword r4  = $cf08
+    &uword r5  = $cf0a
+    &uword r6  = $cf0c
+    &uword r7  = $cf0e
+    &uword r8  = $cf10
+    &uword r9  = $cf12
+    &uword r10 = $cf14
+    &uword r11 = $cf16
+    &uword r12 = $cf18
+    &uword r13 = $cf1a
+    &uword r14 = $cf1c
+    &uword r15 = $cf1e
 
 }

compiler/res/prog8lib/c64/textio.p8

@@ -0,0 +1,618 @@
+; Prog8 definitions for the Text I/O and Screen routines for the Commodore-64
+;
+; Written by Irmen de Jong (irmen@razorvine.net) - license: GNU GPL 3.0
+;
+; indent format: TABS, size=8
+
+%target c64
+%import syslib
+%import conv
+
+
+txt {
+
+const ubyte DEFAULT_WIDTH = 40
+const ubyte DEFAULT_HEIGHT = 25
+
+
+sub  clear_screen() {
+    txt.chrout(147)
+}
+
+sub home() {
+    txt.chrout(19)
+}
+
+sub nl() {
+    txt.chrout('\n')
+}
+
+sub spc() {
+    txt.chrout(' ')
+}
+
+asmsub column(ubyte col @A) clobbers(A, X, Y) {
+    ; ---- set the cursor on the given column (starting with 0) on the current line
+    %asm {{
+        sec
+        jsr  c64.PLOT
+        tay
+        clc
+        jmp  c64.PLOT
+    }}
+}
+
+asmsub  fill_screen (ubyte char @ A, ubyte color @ Y) clobbers(A)  {
+	; ---- fill the character screen with the given fill character and character color.
+	;      (assumes screen and color matrix are at their default addresses)
+
+	%asm {{
+		pha
+		tya
+		jsr  clear_screencolors
+		pla
+		jsr  clear_screenchars
+		rts
+        }}
+
+}
+
+asmsub  clear_screenchars (ubyte char @ A) clobbers(Y)  {
+	; ---- clear the character screen with the given fill character (leaves colors)
+	;      (assumes screen matrix is at the default address)
+	%asm {{
+		ldy  #250
+-		sta  c64.Screen+250*0-1,y
+		sta  c64.Screen+250*1-1,y
+		sta  c64.Screen+250*2-1,y
+		sta  c64.Screen+250*3-1,y
+		dey
+		bne  -
+		rts
+        }}
+}
+
+asmsub  clear_screencolors (ubyte color @ A) clobbers(Y)  {
+	; ---- clear the character screen colors with the given color (leaves characters).
+	;      (assumes color matrix is at the default address)
+	%asm {{
+		ldy  #250
+-		sta  c64.Colors+250*0-1,y
+		sta  c64.Colors+250*1-1,y
+		sta  c64.Colors+250*2-1,y
+		sta  c64.Colors+250*3-1,y
+		dey
+		bne  -
+		rts
+        }}
+}
+
+sub color (ubyte txtcol) {
+    c64.COLOR = txtcol
+}
+
+sub lowercase() {
+    c64.VMCSB |= 2
+}
+
+sub uppercase() {
+    c64.VMCSB &= ~2
+}
+
+asmsub  scroll_left  (ubyte alsocolors @ Pc) clobbers(A, Y)  {
+	; ---- scroll the whole screen 1 character to the left
+	;      contents of the rightmost column are unchanged, you should clear/refill this yourself
+	;      Carry flag determines if screen color data must be scrolled too
+
+	%asm {{
+		stx  P8ZP_SCRATCH_REG
+		bcc _scroll_screen
+
++               ; scroll the screen and the color memory
+		ldx  #0
+		ldy  #38
+-
+        .for row=0, row<=24, row+=1
+            lda  c64.Screen + 40*row + 1,x
+            sta  c64.Screen + 40*row + 0,x
+            lda  c64.Colors + 40*row + 1,x
+            sta  c64.Colors + 40*row + 0,x
+        .next
+		inx
+		dey
+		bpl  -
+		rts
+
+_scroll_screen  ; scroll only the screen memory
+		ldx  #0
+		ldy  #38
+-
+        .for row=0, row<=24, row+=1
+            lda  c64.Screen + 40*row + 1,x
+            sta  c64.Screen + 40*row + 0,x
+        .next
+		inx
+		dey
+		bpl  -
+
+		ldx  P8ZP_SCRATCH_REG
+		rts
+	}}
+}
+
+asmsub  scroll_right  (ubyte alsocolors @ Pc) clobbers(A)  {
+	; ---- scroll the whole screen 1 character to the right
+	;      contents of the leftmost column are unchanged, you should clear/refill this yourself
+	;      Carry flag determines if screen color data must be scrolled too
+	%asm {{
+		stx  P8ZP_SCRATCH_REG
+		bcc  _scroll_screen
+
++               ; scroll the screen and the color memory
+		ldx  #38
+-
+        .for row=0, row<=24, row+=1
+            lda  c64.Screen + 40*row + 0,x
+            sta  c64.Screen + 40*row + 1,x
+            lda  c64.Colors + 40*row + 0,x
+            sta  c64.Colors + 40*row + 1,x
+        .next
+		dex
+		bpl  -
+		rts
+
+_scroll_screen  ; scroll only the screen memory
+		ldx  #38
+-
+        .for row=0, row<=24, row+=1
+            lda  c64.Screen + 40*row + 0,x
+            sta  c64.Screen + 40*row + 1,x
+        .next
+		dex
+		bpl  -
+
+		ldx  P8ZP_SCRATCH_REG
+		rts
+	}}
+}
+
+asmsub  scroll_up  (ubyte alsocolors @ Pc) clobbers(A)  {
+	; ---- scroll the whole screen 1 character up
+	;      contents of the bottom row are unchanged, you should refill/clear this yourself
+	;      Carry flag determines if screen color data must be scrolled too
+	%asm {{
+		stx  P8ZP_SCRATCH_REG
+		bcc  _scroll_screen
+
++               ; scroll the screen and the color memory
+		ldx #39
+-
+        .for row=1, row<=24, row+=1
+            lda  c64.Screen + 40*row,x
+            sta  c64.Screen + 40*(row-1),x
+            lda  c64.Colors + 40*row,x
+            sta  c64.Colors + 40*(row-1),x
+        .next
+		dex
+		bpl  -
+		rts
+
+_scroll_screen  ; scroll only the screen memory
+		ldx #39
+-
+        .for row=1, row<=24, row+=1
+            lda  c64.Screen + 40*row,x
+            sta  c64.Screen + 40*(row-1),x
+        .next
+		dex
+		bpl  -
+
+		ldx  P8ZP_SCRATCH_REG
+		rts
+	}}
+}
+
+asmsub  scroll_down  (ubyte alsocolors @ Pc) clobbers(A)  {
+	; ---- scroll the whole screen 1 character down
+	;      contents of the top row are unchanged, you should refill/clear this yourself
+	;      Carry flag determines if screen color data must be scrolled too
+	%asm {{
+		stx  P8ZP_SCRATCH_REG
+		bcc  _scroll_screen
+
++               ; scroll the screen and the color memory
+		ldx #39
+-
+        .for row=23, row>=0, row-=1
+            lda  c64.Colors + 40*row,x
+            sta  c64.Colors + 40*(row+1),x
+            lda  c64.Screen + 40*row,x
+            sta  c64.Screen + 40*(row+1),x
+        .next
+		dex
+		bpl  -
+		rts
+
+_scroll_screen  ; scroll only the screen memory
+		ldx #39
+-
+        .for row=23, row>=0, row-=1
+            lda  c64.Screen + 40*row,x
+            sta  c64.Screen + 40*(row+1),x
+        .next
+		dex
+		bpl  -
+
+		ldx  P8ZP_SCRATCH_REG
+		rts
+	}}
+}
+
+romsub $FFD2 = chrout(ubyte char @ A)    ; for consistency. You can also use c64.CHROUT directly ofcourse.
+
+asmsub  print (str text @ AY) clobbers(A,Y)  {
+	; ---- print null terminated string from A/Y
+	; note: the compiler contains an optimization that will replace
+	;       a call to this subroutine with a string argument of just one char,
+	;       by just one call to c64.CHROUT of that single char.
+	%asm {{
+		sta  P8ZP_SCRATCH_B1
+		sty  P8ZP_SCRATCH_REG
+		ldy  #0
+-		lda  (P8ZP_SCRATCH_B1),y
+		beq  +
+		jsr  c64.CHROUT
+		iny
+		bne  -
++		rts
+	}}
+}
+
+asmsub  print_ub0  (ubyte value @ A) clobbers(A,Y)  {
+	; ---- print the ubyte in A in decimal form, with left padding 0s (3 positions total)
+	%asm {{
+		stx  P8ZP_SCRATCH_REG
+		jsr  conv.ubyte2decimal
+		pha
+		tya
+		jsr  c64.CHROUT
+		pla
+		jsr  c64.CHROUT
+		txa
+		jsr  c64.CHROUT
+		ldx  P8ZP_SCRATCH_REG
+		rts
+	}}
+}
+
+asmsub  print_ub  (ubyte value @ A) clobbers(A,Y)  {
+	; ---- print the ubyte in A in decimal form, without left padding 0s
+	%asm {{
+		stx  P8ZP_SCRATCH_REG
+		jsr  conv.ubyte2decimal
+_print_byte_digits
+		pha
+		cpy  #'0'
+		beq  +
+		tya
+		jsr  c64.CHROUT
+		pla
+		jsr  c64.CHROUT
+		jmp  _ones
++       pla
+        cmp  #'0'
+        beq  _ones
+        jsr  c64.CHROUT
+_ones   txa
+		jsr  c64.CHROUT
+		ldx  P8ZP_SCRATCH_REG
+		rts
+	}}
+}
+
+asmsub  print_b  (byte value @ A) clobbers(A,Y)  {
+	; ---- print the byte in A in decimal form, without left padding 0s
+	%asm {{
+		stx  P8ZP_SCRATCH_REG
+		pha
+		cmp  #0
+		bpl  +
+		lda  #'-'
+		jsr  c64.CHROUT
++		pla
+		jsr  conv.byte2decimal
+		jmp  print_ub._print_byte_digits
+	}}
+}
+
+asmsub  print_ubhex  (ubyte value @ A, ubyte prefix @ Pc) clobbers(A,Y)  {
+	; ---- print the ubyte in A in hex form (if Carry is set, a radix prefix '$' is printed as well)
+	%asm {{
+		stx  P8ZP_SCRATCH_REG
+		bcc  +
+		pha
+		lda  #'$'
+		jsr  c64.CHROUT
+		pla
++		jsr  conv.ubyte2hex
+		jsr  c64.CHROUT
+		tya
+		jsr  c64.CHROUT
+		ldx  P8ZP_SCRATCH_REG
+		rts
+	}}
+}
+
+asmsub  print_ubbin  (ubyte value @ A, ubyte prefix @ Pc) clobbers(A,Y)  {
+	; ---- print the ubyte in A in binary form (if Carry is set, a radix prefix '%' is printed as well)
+	%asm {{
+		stx  P8ZP_SCRATCH_REG
+		sta  P8ZP_SCRATCH_B1
+		bcc  +
+		lda  #'%'
+		jsr  c64.CHROUT
++		ldy  #8
+-		lda  #'0'
+		asl  P8ZP_SCRATCH_B1
+		bcc  +
+		lda  #'1'
++		jsr  c64.CHROUT
+		dey
+		bne  -
+		ldx  P8ZP_SCRATCH_REG
+		rts
+	}}
+}
+
+asmsub  print_uwbin  (uword value @ AY, ubyte prefix @ Pc) clobbers(A,Y)  {
+	; ---- print the uword in A/Y in binary form (if Carry is set, a radix prefix '%' is printed as well)
+	%asm {{
+		pha
+		tya
+		jsr  print_ubbin
+		pla
+		clc
+		jmp  print_ubbin
+	}}
+}
+
+asmsub  print_uwhex  (uword value @ AY, ubyte prefix @ Pc) clobbers(A,Y)  {
+	; ---- print the uword in A/Y in hexadecimal form (4 digits)
+	;      (if Carry is set, a radix prefix '$' is printed as well)
+	%asm {{
+		pha
+		tya
+		jsr  print_ubhex
+		pla
+		clc
+		jmp  print_ubhex
+	}}
+}
+
+asmsub  print_uw0  (uword value @ AY) clobbers(A,Y)  {
+	; ---- print the uword in A/Y in decimal form, with left padding 0s (5 positions total)
+	%asm {{
+	    stx  P8ZP_SCRATCH_REG
+		jsr  conv.uword2decimal
+		ldy  #0
+-		lda  conv.uword2decimal.decTenThousands,y
+        beq  +
+		jsr  c64.CHROUT
+		iny
+		bne  -
++		ldx  P8ZP_SCRATCH_REG
+		rts
+	}}
+}
+
+asmsub  print_uw  (uword value @ AY) clobbers(A,Y)  {
+	; ---- print the uword in A/Y in decimal form, without left padding 0s
+	%asm {{
+	    stx  P8ZP_SCRATCH_REG
+		jsr  conv.uword2decimal
+		ldx  P8ZP_SCRATCH_REG
+		ldy  #0
+-		lda  conv.uword2decimal.decTenThousands,y
+		beq  _allzero
+		cmp  #'0'
+		bne  _gotdigit
+		iny
+		bne  -
+
+_gotdigit
+		jsr  c64.CHROUT
+		iny
+		lda  conv.uword2decimal.decTenThousands,y
+		bne  _gotdigit
+		rts
+_allzero
+        lda  #'0'
+        jmp  c64.CHROUT
+	}}
+}
+
+asmsub  print_w  (word value @ AY) clobbers(A,Y)  {
+	; ---- print the (signed) word in A/Y in decimal form, without left padding 0's
+	%asm {{
+		cpy  #0
+		bpl  +
+		pha
+		lda  #'-'
+		jsr  c64.CHROUT
+		tya
+		eor  #255
+		tay
+		pla
+		eor  #255
+		clc
+		adc  #1
+		bcc  +
+		iny
++		jmp  print_uw
+	}}
+}
+
+asmsub  input_chars  (uword buffer @ AY) clobbers(A) -> ubyte @ Y  {
+	; ---- Input a string (max. 80 chars) from the keyboard. Returns length in Y. (string is terminated with a 0 byte as well)
+	;      It assumes the keyboard is selected as I/O channel!
+
+	%asm {{
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		ldy  #0				; char counter = 0
+-		jsr  c64.CHRIN
+		cmp  #$0d			; return (ascii 13) pressed?
+		beq  +				; yes, end.
+		sta  (P8ZP_SCRATCH_W1),y	; else store char in buffer
+		iny
+		bne  -
++		lda  #0
+		sta  (P8ZP_SCRATCH_W1),y	; finish string with 0 byte
+		rts
+
+	}}
+}
+
+asmsub  setchr  (ubyte col @X, ubyte row @Y, ubyte character @A) clobbers(A, Y)  {
+	; ---- sets the character in the screen matrix at the given position
+	%asm {{
+		pha
+		tya
+		asl  a
+		tay
+		lda  _screenrows+1,y
+		sta  _mod+2
+		txa
+		clc
+		adc  _screenrows,y
+		sta  _mod+1
+		bcc  +
+		inc  _mod+2
++		pla
+_mod		sta  $ffff		; modified
+		rts
+
+_screenrows	.word  $0400 + range(0, 1000, 40)
+	}}
+}
+
+asmsub  getchr  (ubyte col @A, ubyte row @Y) clobbers(Y) -> ubyte @ A {
+	; ---- get the character in the screen matrix at the given location
+	%asm  {{
+		pha
+		tya
+		asl  a
+		tay
+		lda  setchr._screenrows+1,y
+		sta  _mod+2
+		pla
+		clc
+		adc  setchr._screenrows,y
+		sta  _mod+1
+		bcc  _mod
+		inc  _mod+2
+_mod		lda  $ffff		; modified
+		rts
+	}}
+}
+
+asmsub  setclr  (ubyte col @X, ubyte row @Y, ubyte color @A) clobbers(A, Y)  {
+	; ---- set the color in A on the screen matrix at the given position
+	%asm {{
+		pha
+		tya
+		asl  a
+		tay
+		lda  _colorrows+1,y
+		sta  _mod+2
+		txa
+		clc
+		adc  _colorrows,y
+		sta  _mod+1
+		bcc  +
+		inc  _mod+2
++		pla
+_mod		sta  $ffff		; modified
+		rts
+
+_colorrows	.word  $d800 + range(0, 1000, 40)
+	}}
+}
+
+asmsub  getclr  (ubyte col @A, ubyte row @Y) clobbers(Y) -> ubyte @ A {
+	; ---- get the color in the screen color matrix at the given location
+	%asm  {{
+		pha
+		tya
+		asl  a
+		tay
+		lda  setclr._colorrows+1,y
+		sta  _mod+2
+		pla
+		clc
+		adc  setclr._colorrows,y
+		sta  _mod+1
+		bcc  _mod
+		inc  _mod+2
+_mod		lda  $ffff		; modified
+		rts
+	}}
+}
+
+sub  setcc  (ubyte column, ubyte row, ubyte char, ubyte charcolor)  {
+	; ---- set char+color at the given position on the screen
+	%asm {{
+		lda  row
+		asl  a
+		tay
+		lda  setchr._screenrows+1,y
+		sta  _charmod+2
+		adc  #$d4
+		sta  _colormod+2
+		lda  setchr._screenrows,y
+		clc
+		adc  column
+		sta  _charmod+1
+		sta  _colormod+1
+		bcc  +
+		inc  _charmod+2
+		inc  _colormod+2
++		lda  char
+_charmod	sta  $ffff		; modified
+		lda  charcolor
+_colormod	sta  $ffff		; modified
+		rts
+	}}
+}
+
+asmsub  plot  (ubyte col @ Y, ubyte row @ A) clobbers(A) {
+	; ---- safe wrapper around PLOT kernal routine, to save the X register.
+	%asm  {{
+		stx  P8ZP_SCRATCH_REG
+		tax
+		clc
+		jsr  c64.PLOT
+		ldx  P8ZP_SCRATCH_REG
+		rts
+	}}
+}
+
+asmsub width() clobbers(X,Y) -> ubyte @A {
+    ; -- returns the text screen width (number of columns)
+    %asm {{
+        jsr  c64.SCREEN
+        txa
+        rts
+    }}
+}
+
+asmsub height() clobbers(X, Y) -> ubyte @A {
+    ; -- returns the text screen height (number of rows)
+    %asm {{
+        jsr  c64.SCREEN
+        tya
+        rts
+    }}
+}
+
+}

compiler/res/prog8lib/c64floats.asm

@@ -1,783 +0,0 @@
-; --- low level floating point assembly routines for the C64
-
-ub2float	.proc
-		; -- convert ubyte in SCRATCH_ZPB1 to float at address A/Y
-		;    clobbers A, Y
-		stx  c64.SCRATCH_ZPREGX
-		sta  c64.SCRATCH_ZPWORD2
-		sty  c64.SCRATCH_ZPWORD2+1
-		ldy  c64.SCRATCH_ZPB1
-		jsr  FREADUY
-_fac_to_mem	ldx  c64.SCRATCH_ZPWORD2
-		ldy  c64.SCRATCH_ZPWORD2+1
-		jsr  MOVMF
-		ldx  c64.SCRATCH_ZPREGX
-		rts
-		.pend
-
-b2float		.proc
-		; -- convert byte in SCRATCH_ZPB1 to float at address A/Y
-		;    clobbers A, Y
-		stx  c64.SCRATCH_ZPREGX
-		sta  c64.SCRATCH_ZPWORD2
-		sty  c64.SCRATCH_ZPWORD2+1
-		lda  c64.SCRATCH_ZPB1
-		jsr  FREADSA
-		jmp  ub2float._fac_to_mem
-		.pend
-
-uw2float	.proc
-		; -- convert uword in SCRATCH_ZPWORD1 to float at address A/Y
-		stx  c64.SCRATCH_ZPREGX
-		sta  c64.SCRATCH_ZPWORD2
-		sty  c64.SCRATCH_ZPWORD2+1
-		lda  c64.SCRATCH_ZPWORD1
-		ldy  c64.SCRATCH_ZPWORD1+1
-		jsr  GIVUAYFAY
-		jmp  ub2float._fac_to_mem
-		.pend
-
-w2float		.proc
-		; -- convert word in SCRATCH_ZPWORD1 to float at address A/Y
-		stx  c64.SCRATCH_ZPREGX
-		sta  c64.SCRATCH_ZPWORD2
-		sty  c64.SCRATCH_ZPWORD2+1
-		ldy  c64.SCRATCH_ZPWORD1
-		lda  c64.SCRATCH_ZPWORD1+1
-		jsr  GIVAYF
-		jmp  ub2float._fac_to_mem
-		.pend
-
-stack_b2float	.proc
-		; -- b2float operating on the stack
-		inx
-		lda  c64.ESTACK_LO,x
-		stx  c64.SCRATCH_ZPREGX
-		jsr  FREADSA
-		jmp  push_fac1_as_result
-		.pend
-
-stack_w2float	.proc
-		; -- w2float operating on the stack
-		inx
-		ldy  c64.ESTACK_LO,x
-		lda  c64.ESTACK_HI,x
-		stx  c64.SCRATCH_ZPREGX
-		jsr  GIVAYF
-		jmp  push_fac1_as_result
-		.pend
-
-stack_ub2float	.proc
-		; -- ub2float operating on the stack
-		inx
-		lda  c64.ESTACK_LO,x
-		stx  c64.SCRATCH_ZPREGX
-		tay
-		jsr  FREADUY
-		jmp  push_fac1_as_result
-		.pend
-
-stack_uw2float	.proc
-		; -- uw2float operating on the stack
-		inx
-		lda  c64.ESTACK_LO,x
-		ldy  c64.ESTACK_HI,x
-		stx  c64.SCRATCH_ZPREGX
-		jsr  GIVUAYFAY
-		jmp  push_fac1_as_result
-		.pend
-
-stack_float2w	.proc               ; also used for float2b
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  AYINT
-		ldx  c64.SCRATCH_ZPREGX
-		lda  $64
-		sta  c64.ESTACK_HI,x
-		lda  $65
-		sta  c64.ESTACK_LO,x
-		dex
-		rts
-		.pend
-
-stack_float2uw	.proc               ; also used for float2ub
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  GETADR
-		ldx  c64.SCRATCH_ZPREGX
-		sta  c64.ESTACK_HI,x
-		tya
-		sta  c64.ESTACK_LO,x
-		dex
-		rts
-		.pend
-
-push_float	.proc
-		; ---- push mflpt5 in A/Y onto stack
-		; (taking 3 stack positions = 6 bytes of which 1 is padding)
-		sta  c64.SCRATCH_ZPWORD1
-		sty  c64.SCRATCH_ZPWORD1+1
-		ldy  #0
-		lda  (c64.SCRATCH_ZPWORD1),y
-		sta  c64.ESTACK_LO,x
-		iny
-		lda  (c64.SCRATCH_ZPWORD1),y
-		sta  c64.ESTACK_HI,x
-		dex
-		iny
-		lda  (c64.SCRATCH_ZPWORD1),y
-		sta  c64.ESTACK_LO,x
-		iny
-		lda  (c64.SCRATCH_ZPWORD1),y
-		sta  c64.ESTACK_HI,x
-		dex
-		iny
-		lda  (c64.SCRATCH_ZPWORD1),y
-		sta  c64.ESTACK_LO,x
-		dex
-		rts
-		.pend
-
-func_rndf	.proc
-		; -- put a random floating point value on the stack
-		stx  c64.SCRATCH_ZPREG
-		lda  #1
-		jsr  FREADSA
-		jsr  RND		; rng into fac1
-		ldx  #<_rndf_rnum5
-		ldy  #>_rndf_rnum5
-		jsr  MOVMF	; fac1 to mem X/Y
-		ldx  c64.SCRATCH_ZPREG
-		lda  #<_rndf_rnum5
-		ldy  #>_rndf_rnum5
-		jmp  push_float
-_rndf_rnum5	.byte  0,0,0,0,0
-		.pend
-
-push_float_from_indexed_var	.proc
-		; -- push the float from the array at A/Y with index on stack, onto the stack.
-		sta  c64.SCRATCH_ZPWORD1
-		sty  c64.SCRATCH_ZPWORD1+1
-		jsr  prog8_lib.pop_index_times_5
-		jsr  prog8_lib.add_a_to_zpword
-		lda  c64.SCRATCH_ZPWORD1
-		ldy  c64.SCRATCH_ZPWORD1+1
-		jmp  push_float
-		.pend
-
-pop_float	.proc
-		; ---- pops mflpt5 from stack to memory A/Y
-		; (frees 3 stack positions = 6 bytes of which 1 is padding)
-		sta  c64.SCRATCH_ZPWORD1
-		sty  c64.SCRATCH_ZPWORD1+1
-		ldy  #4
-		inx
-		lda  c64.ESTACK_LO,x
-		sta  (c64.SCRATCH_ZPWORD1),y
-		dey
-		inx
-		lda  c64.ESTACK_HI,x
-		sta  (c64.SCRATCH_ZPWORD1),y
-		dey
-		lda  c64.ESTACK_LO,x
-		sta  (c64.SCRATCH_ZPWORD1),y
-		dey
-		inx
-		lda  c64.ESTACK_HI,x
-		sta  (c64.SCRATCH_ZPWORD1),y
-		dey
-		lda  c64.ESTACK_LO,x
-		sta  (c64.SCRATCH_ZPWORD1),y
-		rts
-		.pend
-
-pop_float_fac1	.proc
-		; -- pops float from stack into FAC1
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  pop_float
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jmp  MOVFM
-		.pend
-
-pop_float_to_indexed_var	.proc
-		; -- pop the float on the stack, to the memory in the array at A/Y indexed by the byte on stack
-		sta  c64.SCRATCH_ZPWORD1
-		sty  c64.SCRATCH_ZPWORD1+1
-		jsr  prog8_lib.pop_index_times_5
-		jsr  prog8_lib.add_a_to_zpword
-		lda  c64.SCRATCH_ZPWORD1
-		ldy  c64.SCRATCH_ZPWORD1+1
-		jmp  pop_float
-		.pend
-
-copy_float	.proc
-		; -- copies the 5 bytes of the mflt value pointed to by SCRATCH_ZPWORD1,
-		;    into the 5 bytes pointed to by A/Y.  Clobbers A,Y.
-		sta  c64.SCRATCH_ZPWORD2
-		sty  c64.SCRATCH_ZPWORD2+1
-		ldy  #0
-		lda  (c64.SCRATCH_ZPWORD1),y
-		sta  (c64.SCRATCH_ZPWORD2),y
-		iny
-		lda  (c64.SCRATCH_ZPWORD1),y
-		sta  (c64.SCRATCH_ZPWORD2),y
-		iny
-		lda  (c64.SCRATCH_ZPWORD1),y
-		sta  (c64.SCRATCH_ZPWORD2),y
-		iny
-		lda  (c64.SCRATCH_ZPWORD1),y
-		sta  (c64.SCRATCH_ZPWORD2),y
-		iny
-		lda  (c64.SCRATCH_ZPWORD1),y
-		sta  (c64.SCRATCH_ZPWORD2),y
-		rts
-		.pend
-
-inc_var_f	.proc
-		; -- add 1 to float pointed to by A/Y
-		sta  c64.SCRATCH_ZPWORD1
-		sty  c64.SCRATCH_ZPWORD1+1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  MOVFM
-		lda  #<FL_FONE
-		ldy  #>FL_FONE
-		jsr  FADD
-		ldx  c64.SCRATCH_ZPWORD1
-		ldy  c64.SCRATCH_ZPWORD1+1
-		jsr  MOVMF
-		ldx  c64.SCRATCH_ZPREGX
-		rts
-		.pend
-
-dec_var_f	.proc
-		; -- subtract 1 from float pointed to by A/Y
-		sta  c64.SCRATCH_ZPWORD1
-		sty  c64.SCRATCH_ZPWORD1+1
-		stx  c64.SCRATCH_ZPREGX
-		lda  #<FL_FONE
-		ldy  #>FL_FONE
-		jsr  MOVFM
-		lda  c64.SCRATCH_ZPWORD1
-		ldy  c64.SCRATCH_ZPWORD1+1
-		jsr  FSUB
-		ldx  c64.SCRATCH_ZPWORD1
-		ldy  c64.SCRATCH_ZPWORD1+1
-		jsr  MOVMF
-		ldx  c64.SCRATCH_ZPREGX
-		rts
-		.pend
-
-inc_indexed_var_f	.proc
-		; -- add 1 to float in array pointed to by A/Y, at index X
-		pha
-		txa
-		sta  c64.SCRATCH_ZPB1
-		asl  a
-		asl  a
-		clc
-		adc  c64.SCRATCH_ZPB1
-		sta  c64.SCRATCH_ZPB1
-		pla
-		clc
-		adc  c64.SCRATCH_ZPB1
-		bcc  +
-		iny
-+		jmp  inc_var_f
-		.pend
-
-dec_indexed_var_f	.proc
-		; -- subtract 1 to float in array pointed to by A/Y, at index X
-		pha
-		txa
-		sta  c64.SCRATCH_ZPB1
-		asl  a
-		asl  a
-		clc
-		adc  c64.SCRATCH_ZPB1
-		sta  c64.SCRATCH_ZPB1
-		pla
-		clc
-		adc  c64.SCRATCH_ZPB1
-		bcc  +
-		iny
-+		jmp  dec_var_f
-		.pend
-
-
-pop_2_floats_f2_in_fac1	.proc
-		; -- pop 2 floats from stack, load the second one in FAC1 as well
-		lda  #<fmath_float2
-		ldy  #>fmath_float2
-		jsr  pop_float
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  pop_float
-		lda  #<fmath_float2
-		ldy  #>fmath_float2
-		jmp  MOVFM
-		.pend
-
-
-fmath_float1	.byte 0,0,0,0,0	; storage for a mflpt5 value
-fmath_float2	.byte 0,0,0,0,0	; storage for a mflpt5 value
-
-push_fac1_as_result	.proc
-		; -- push the float in FAC1 onto the stack, and return from calculation
-		ldx  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  MOVMF
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		ldx  c64.SCRATCH_ZPREGX
-		jmp  push_float
-		.pend
-
-pow_f		.proc
-		; -- push f1 ** f2 on stack
-		lda  #<fmath_float2
-		ldy  #>fmath_float2
-		jsr  pop_float
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  pop_float
-		stx  c64.SCRATCH_ZPREGX
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  CONUPK		; fac2 = float1
-		lda  #<fmath_float2
-		ldy  #>fmath_float2
-		jsr  FPWR
-		ldx  c64.SCRATCH_ZPREGX
-		jmp  push_fac1_as_result
-		.pend
-
-div_f		.proc
-		; -- push f1/f2 on stack
-		jsr  pop_2_floats_f2_in_fac1
-		stx  c64.SCRATCH_ZPREGX
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  FDIV
-		jmp  push_fac1_as_result
-		.pend
-
-add_f		.proc
-		; -- push f1+f2 on stack
-		jsr  pop_2_floats_f2_in_fac1
-		stx  c64.SCRATCH_ZPREGX
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  FADD
-		jmp  push_fac1_as_result
-		.pend
-
-sub_f		.proc
-		; -- push f1-f2 on stack
-		jsr  pop_2_floats_f2_in_fac1
-		stx  c64.SCRATCH_ZPREGX
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  FSUB
-		jmp  push_fac1_as_result
-		.pend
-
-mul_f		.proc
-		; -- push f1*f2 on stack
-		jsr  pop_2_floats_f2_in_fac1
-		stx  c64.SCRATCH_ZPREGX
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  FMULT
-		jmp  push_fac1_as_result
-		.pend
-
-neg_f		.proc
-		; -- push -flt back on stack
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  NEGOP
-		jmp  push_fac1_as_result
-		.pend
-
-abs_f		.proc
-		; -- push abs(float) on stack (as float)
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  ABS
-		jmp  push_fac1_as_result
-		.pend
-
-equal_f		.proc
-		; -- are the two mflpt5 numbers on the stack identical?
-		inx
-		inx
-		inx
-		inx
-		lda  c64.ESTACK_LO-3,x
-		cmp  c64.ESTACK_LO,x
-		bne  _equals_false
-		lda  c64.ESTACK_LO-2,x
-		cmp  c64.ESTACK_LO+1,x
-		bne  _equals_false
-		lda  c64.ESTACK_LO-1,x
-		cmp  c64.ESTACK_LO+2,x
-		bne  _equals_false
-		lda  c64.ESTACK_HI-2,x
-		cmp  c64.ESTACK_HI+1,x
-		bne  _equals_false
-		lda  c64.ESTACK_HI-1,x
-		cmp  c64.ESTACK_HI+2,x
-		bne  _equals_false
-_equals_true	lda  #1
-_equals_store	inx
-		sta  c64.ESTACK_LO+1,x
-		rts
-_equals_false	lda  #0
-		beq  _equals_store
-		.pend
-
-notequal_f	.proc
-		; -- are the two mflpt5 numbers on the stack different?
-		jsr  equal_f
-		eor  #1		; invert the result
-		sta  c64.ESTACK_LO+1,x
-		rts
-		.pend
-
-less_f		.proc
-		; -- is f1 < f2?
-		jsr  compare_floats
-		cmp  #255
-		beq  compare_floats._return_true
-		bne  compare_floats._return_false
-		.pend
-
-
-lesseq_f	.proc
-		; -- is f1 <= f2?
-		jsr  compare_floats
-		cmp  #255
-		beq  compare_floats._return_true
-		cmp  #0
-		beq  compare_floats._return_true
-		bne  compare_floats._return_false
-		.pend
-
-greater_f	.proc
-		; -- is f1 > f2?
-		jsr  compare_floats
-		cmp  #1
-		beq  compare_floats._return_true
-		bne  compare_floats._return_false
-		.pend
-
-greatereq_f	.proc
-		; -- is f1 >= f2?
-		jsr  compare_floats
-		cmp  #1
-		beq  compare_floats._return_true
-		cmp  #0
-		beq  compare_floats._return_true
-		bne  compare_floats._return_false
-		.pend
-
-compare_floats	.proc
-		lda  #<fmath_float2
-		ldy  #>fmath_float2
-		jsr  pop_float
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  pop_float
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  MOVFM		; fac1 = flt1
-		lda  #<fmath_float2
-		ldy  #>fmath_float2
-		stx  c64.SCRATCH_ZPREG
-		jsr  FCOMP		; A = flt1 compared with flt2 (0=equal, 1=flt1>flt2, 255=flt1<flt2)
-		ldx  c64.SCRATCH_ZPREG
-		rts
-_return_false	lda  #0
-_return_result  sta  c64.ESTACK_LO,x
-		dex
-		rts
-_return_true	lda  #1
-		bne  _return_result
-		.pend
-
-func_sin	.proc
-		; -- push sin(f) back onto stack
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  SIN
-		jmp  push_fac1_as_result
-		.pend
-
-func_cos	.proc
-		; -- push cos(f) back onto stack
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  COS
-		jmp  push_fac1_as_result
-		.pend
-
-func_tan	.proc
-		; -- push tan(f) back onto stack
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  TAN
-		jmp  push_fac1_as_result
-		.pend
-
-func_atan	.proc
-		; -- push atan(f) back onto stack
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  ATN
-		jmp  push_fac1_as_result
-		.pend
-
-func_ln		.proc
-		; -- push ln(f) back onto stack
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  LOG
-		jmp  push_fac1_as_result
-		.pend
-
-func_log2	.proc
-		; -- push log base 2, ln(f)/ln(2), back onto stack
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  LOG
-		jsr  MOVEF
-		lda  #<c64.FL_LOG2
-		ldy  #>c64.FL_LOG2
-		jsr  MOVFM
-		jsr  FDIVT
-		jmp  push_fac1_as_result
-		.pend
-
-func_sqrt	.proc
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  SQR
-		jmp  push_fac1_as_result
-		.pend
-
-func_rad	.proc
-		; -- convert degrees to radians (d * pi / 180)
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		lda  #<_pi_div_180
-		ldy  #>_pi_div_180
-		jsr  FMULT
-		jmp  push_fac1_as_result
-_pi_div_180	.byte 123, 14, 250, 53, 18		; pi / 180
-		.pend
-
-func_deg	.proc
-		; -- convert radians to degrees (d * (1/ pi * 180))
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		lda  #<_one_over_pi_div_180
-		ldy  #>_one_over_pi_div_180
-		jsr  FMULT
-		jmp  push_fac1_as_result
-_one_over_pi_div_180	.byte 134, 101, 46, 224, 211		; 1 / (pi * 180)
-		.pend
-
-func_round	.proc
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  FADDH
-		jsr  INT
-		jmp  push_fac1_as_result
-		.pend
-
-func_floor	.proc
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		jsr  INT
-		jmp  push_fac1_as_result
-		.pend
-
-func_ceil	.proc
-		; -- ceil: tr = int(f); if tr==f -> return  else return tr+1
-		jsr  pop_float_fac1
-		stx  c64.SCRATCH_ZPREGX
-		ldx  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  MOVMF
-		jsr  INT
-		lda  #<fmath_float1
-		ldy  #>fmath_float1
-		jsr  FCOMP
-		cmp  #0
-		beq  +
-		lda  #<FL_FONE
-		ldy  #>FL_FONE
-		jsr  FADD
-+		jmp  push_fac1_as_result
-		.pend
-
-func_any_f	.proc
-		inx
-		lda  c64.ESTACK_LO,x	; array size
-		sta  c64.SCRATCH_ZPB1
-		asl  a
-		asl  a
-		clc
-		adc  c64.SCRATCH_ZPB1	; times 5 because of float
-		jmp  prog8_lib.func_any_b._entry
-		.pend
-
-func_all_f	.proc
-		inx
-		jsr  prog8_lib.peek_address
-		lda  c64.ESTACK_LO,x	; array size
-		sta  c64.SCRATCH_ZPB1
-		asl  a
-		asl  a
-		clc
-		adc  c64.SCRATCH_ZPB1	; times 5 because of float
-		tay
-		dey
--		lda  (c64.SCRATCH_ZPWORD1),y
-		clc
-		dey
-		adc  (c64.SCRATCH_ZPWORD1),y
-		dey
-		adc  (c64.SCRATCH_ZPWORD1),y
-		dey
-		adc  (c64.SCRATCH_ZPWORD1),y
-		dey
-		adc  (c64.SCRATCH_ZPWORD1),y
-		dey
-		cmp  #0
-		beq  +
-		cpy  #255
-        	bne  -
-		lda  #1
-		sta  c64.ESTACK_LO+1,x
-		rts
-+		sta  c64.ESTACK_LO+1,x
-		rts
-		.pend
-
-func_max_f	.proc
-		lda  #255
-		sta  _minmax_cmp+1
-		lda  #<_largest_neg_float
-		ldy  #>_largest_neg_float
-_minmax_entry	jsr  MOVFM
-		jsr  prog8_lib.pop_array_and_lengthmin1Y
-		stx  c64.SCRATCH_ZPREGX
--		sty  c64.SCRATCH_ZPREG
-		lda  c64.SCRATCH_ZPWORD1
-		ldy  c64.SCRATCH_ZPWORD1+1
-		jsr  FCOMP
-_minmax_cmp	cmp  #255			; modified
-		bne  +
-		lda  c64.SCRATCH_ZPWORD1
-		ldy  c64.SCRATCH_ZPWORD1+1
-		jsr  MOVFM
-+		lda  c64.SCRATCH_ZPWORD1
-		clc
-		adc  #5
-		sta  c64.SCRATCH_ZPWORD1
-		bcc  +
-		inc  c64.SCRATCH_ZPWORD1+1
-+		ldy  c64.SCRATCH_ZPREG
-		dey
-		cpy  #255
-		bne  -
-		jmp  push_fac1_as_result
-_largest_neg_float	.byte 255,255,255,255,255		; largest negative float -1.7014118345e+38
-		.pend
-
-func_min_f	.proc
-		lda  #1
-		sta  func_max_f._minmax_cmp+1
-		lda  #<_largest_pos_float
-		ldy  #>_largest_pos_float
-		jmp  func_max_f._minmax_entry
-_largest_pos_float	.byte  255,127,255,255,255		; largest positive float
-		rts
-		.pend
-
-func_sum_f	.proc
-		lda  #<FL_ZERO
-		ldy  #>FL_ZERO
-		jsr  MOVFM
-		jsr  prog8_lib.pop_array_and_lengthmin1Y
-		stx  c64.SCRATCH_ZPREGX
--		sty  c64.SCRATCH_ZPREG
-		lda  c64.SCRATCH_ZPWORD1
-		ldy  c64.SCRATCH_ZPWORD1+1
-		jsr  FADD
-		ldy  c64.SCRATCH_ZPREG
-		dey
-		cpy  #255
-		beq  +
-		lda  c64.SCRATCH_ZPWORD1
-		clc
-		adc  #5
-		sta  c64.SCRATCH_ZPWORD1
-		bcc  -
-		inc  c64.SCRATCH_ZPWORD1+1
-		bne  -
-+		jmp  push_fac1_as_result
-		.pend
-
-sign_f		.proc
-		jsr  pop_float_fac1
-		jsr  SIGN
-		sta  c64.ESTACK_LO,x
-		dex
-		rts
-		.pend
-
-
-set_0_array_float	.proc
-		; -- set a float in an array to zero (index on stack, array in SCRATCH_ZPWORD1)
-		inx
-		lda  c64.ESTACK_LO,x
-		asl  a
-		asl  a
-		clc
-		adc  c64.ESTACK_LO,x
-		tay
-		lda  #0
-		sta  (c64.SCRATCH_ZPWORD1),y
-		iny
-		sta  (c64.SCRATCH_ZPWORD1),y
-		iny
-		sta  (c64.SCRATCH_ZPWORD1),y
-		iny
-		sta  (c64.SCRATCH_ZPWORD1),y
-		iny
-		sta  (c64.SCRATCH_ZPWORD1),y
-		rts
-		.pend
-
-
-set_array_float		.proc
-		; -- set a float in an array to a value (index on stack, float in SCRATCH_ZPWORD1, array in SCRATCH_ZPWORD2)
-		inx
-		lda  c64.ESTACK_LO,x
-		asl  a
-		asl  a
-		clc
-		adc  c64.ESTACK_LO,x
-		clc
-		adc  c64.SCRATCH_ZPWORD2
-		ldy  c64.SCRATCH_ZPWORD2+1
-		bcc  +
-		iny
-+		jmp  copy_float
-			; -- copies the 5 bytes of the mflt value pointed to by SCRATCH_ZPWORD1,
-			;    into the 5 bytes pointed to by A/Y.  Clobbers A,Y.
-		.pend

compiler/res/prog8lib/conv.p8

@@ -0,0 +1,523 @@
+; Number conversions routines.
+;
+; Written by Irmen de Jong (irmen@razorvine.net) - license: GNU GPL 3.0
+
+
+conv {
+
+; ----- number conversions to decimal strings ----
+
+asmsub  ubyte2decimal  (ubyte value @A) -> ubyte @Y, ubyte @A, ubyte @X  {
+	; ---- A to decimal string in Y/A/X  (100s in Y, 10s in A, 1s in X)
+	%asm {{
+		ldy  #uword2decimal.ASCII_0_OFFSET
+		bne  uword2decimal.hex_try200
+		rts
+	}}
+}
+
+asmsub  uword2decimal  (uword value @AY) -> ubyte @Y, ubyte @A, ubyte @X  {
+	;  ---- convert 16 bit uword in A/Y to decimal
+	;  output in uword2decimal.decTenThousands, decThousands, decHundreds, decTens, decOnes
+	;  (these are terminated by a zero byte so they can be easily printed)
+	;  also returns Y = 100's, A = 10's, X = 1's
+
+	%asm {{
+
+;Convert 16 bit Hex to Decimal (0-65535) Rev 2
+;By Omegamatrix    Further optimizations by tepples
+; routine from http://forums.nesdev.com/viewtopic.php?f=2&t=11341&start=15
+
+;HexToDec99
+; start in A
+; end with A = 10's, decOnes (also in X)
+
+;HexToDec255
+; start in A
+; end with Y = 100's, A = 10's, decOnes (also in X)
+
+;HexToDec999
+; start with A = high byte, Y = low byte
+; end with Y = 100's, A = 10's, decOnes (also in X)
+; requires 1 extra temp register on top of decOnes, could combine
+; these two if HexToDec65535 was eliminated...
+
+;HexToDec65535
+; start with A/Y (low/high) as 16 bit value
+; end with decTenThousand, decThousand, Y = 100's, A = 10's, decOnes (also in X)
+; (irmen: I store Y and A in decHundreds and decTens too, so all of it can be easily printed)
+
+
+ASCII_0_OFFSET 	= $30
+temp       	    = P8ZP_SCRATCH_B1	; byte in zeropage
+hexHigh      	= P8ZP_SCRATCH_W1	; byte in zeropage
+hexLow       	= P8ZP_SCRATCH_W1+1	; byte in zeropage
+
+
+HexToDec65535; SUBROUTINE
+    sty    hexHigh               ;3  @9
+    sta    hexLow                ;3  @12
+    tya
+    tax                          ;2  @14
+    lsr    a                     ;2  @16
+    lsr    a                     ;2  @18   integer divide 1024 (result 0-63)
+
+    cpx    #$A7                  ;2  @20   account for overflow of multiplying 24 from 43,000 ($A7F8) onward,
+    adc    #1                    ;2  @22   we can just round it to $A700, and the divide by 1024 is fine...
+
+    ;at this point we have a number 1-65 that we have to times by 24,
+    ;add to original sum, and Mod 1024 to get a remainder 0-999
+
+
+    sta    temp                  ;3  @25
+    asl    a                     ;2  @27
+    adc    temp                  ;3  @30  x3
+    tay                          ;2  @32
+    lsr    a                     ;2  @34
+    lsr    a                     ;2  @36
+    lsr    a                     ;2  @38
+    lsr    a                     ;2  @40
+    lsr    a                     ;2  @42
+    tax                          ;2  @44
+    tya                          ;2  @46
+    asl    a                     ;2  @48
+    asl    a                     ;2  @50
+    asl    a                     ;2  @52
+    clc                          ;2  @54
+    adc    hexLow                ;3  @57
+    sta    hexLow                ;3  @60
+    txa                          ;2  @62
+    adc    hexHigh               ;3  @65
+    sta    hexHigh               ;3  @68
+    ror    a                     ;2  @70
+    lsr    a                     ;2  @72
+    tay                          ;2  @74    integer divide 1,000 (result 0-65)
+
+    lsr    a                     ;2  @76    split the 1,000 and 10,000 digit
+    tax                          ;2  @78
+    lda    ShiftedBcdTab,x       ;4  @82
+    tax                          ;2  @84
+    rol    a                     ;2  @86
+    and    #$0F                  ;2  @88
+    ora    #ASCII_0_OFFSET
+    sta    decThousands          ;3  @91
+    txa                          ;2  @93
+    lsr    a                     ;2  @95
+    lsr    a                     ;2  @97
+    lsr    a                     ;2  @99
+    ora    #ASCII_0_OFFSET
+    sta    decTenThousands       ;3  @102
+
+    lda    hexLow                ;3  @105
+    cpy    temp                  ;3  @108
+    bmi    _doSubtract           ;2³ @110/111
+    beq    _useZero               ;2³ @112/113
+    adc    #23 + 24              ;2  @114
+_doSubtract
+    sbc    #23                   ;2  @116
+    sta    hexLow                ;3  @119
+_useZero
+    lda    hexHigh               ;3  @122
+    sbc    #0                    ;2  @124
+
+Start100s
+    and    #$03                  ;2  @126
+    tax                          ;2  @128   0,1,2,3
+    cmp    #2                    ;2  @130
+    rol    a                     ;2  @132   0,2,5,7
+    ora    #ASCII_0_OFFSET
+    tay                          ;2  @134   Y = Hundreds digit
+
+    lda    hexLow                ;3  @137
+    adc    Mod100Tab,x           ;4  @141    adding remainder of 256, 512, and 256+512 (all mod 100)
+    bcs    hex_doSub200             ;2³ @143/144
+
+hex_try200
+    cmp    #200                  ;2  @145
+    bcc    hex_try100               ;2³ @147/148
+hex_doSub200
+    iny                          ;2  @149
+    iny                          ;2  @151
+    sbc    #200                  ;2  @153
+hex_try100
+    cmp    #100                  ;2  @155
+    bcc    HexToDec99            ;2³ @157/158
+    iny                          ;2  @159
+    sbc    #100                  ;2  @161
+
+HexToDec99; SUBROUTINE
+    lsr    a                     ;2  @163
+    tax                          ;2  @165
+    lda    ShiftedBcdTab,x       ;4  @169
+    tax                          ;2  @171
+    rol    a                     ;2  @173
+    and    #$0F                  ;2  @175
+    ora    #ASCII_0_OFFSET
+    sta    decOnes               ;3  @178
+    txa                          ;2  @180
+    lsr    a                     ;2  @182
+    lsr    a                     ;2  @184
+    lsr    a                     ;2  @186
+    ora    #ASCII_0_OFFSET
+
+    ; irmen: load X with ones, and store Y and A too, for easy printing afterwards
+    sty  decHundreds
+    sta  decTens
+    ldx  decOnes
+    rts                          ;6  @192   Y=hundreds, A = tens digit, X=ones digit
+
+
+HexToDec999; SUBROUTINE
+    sty    hexLow                ;3  @9
+    jmp    Start100s             ;3  @12
+
+Mod100Tab
+    .byte 0,56,12,56+12
+
+ShiftedBcdTab
+    .byte $00,$01,$02,$03,$04,$08,$09,$0A,$0B,$0C
+    .byte $10,$11,$12,$13,$14,$18,$19,$1A,$1B,$1C
+    .byte $20,$21,$22,$23,$24,$28,$29,$2A,$2B,$2C
+    .byte $30,$31,$32,$33,$34,$38,$39,$3A,$3B,$3C
+    .byte $40,$41,$42,$43,$44,$48,$49,$4A,$4B,$4C
+
+decTenThousands   	.byte  0
+decThousands    	.byte  0
+decHundreds		.byte  0
+decTens			.byte  0
+decOnes   		.byte  0
+			.byte  0		; zero-terminate the decimal output string
+
+    }}
+}
+
+asmsub  byte2decimal  (byte value @A) -> ubyte @Y, ubyte @A, ubyte @X  {
+	; ---- A (signed byte) to decimal string in Y/A/X  (100s in Y, 10s in A, 1s in X)
+	;      note: if the number is negative, you have to deal with the '-' yourself!
+	%asm {{
+		cmp  #0
+		bpl  +
+		eor  #255
+		clc
+		adc  #1
++		jmp  ubyte2decimal
+	}}
+}
+
+asmsub  ubyte2hex  (ubyte value @A) -> ubyte @A, ubyte @Y  {
+	; ---- A to hex petscii string in AY (first hex char in A, second hex char in Y)
+	%asm {{
+		stx  P8ZP_SCRATCH_REG
+		pha
+		and  #$0f
+		tax
+		ldy  _hex_digits,x
+		pla
+		lsr  a
+		lsr  a
+		lsr  a
+		lsr  a
+		tax
+		lda  _hex_digits,x
+		ldx  P8ZP_SCRATCH_REG
+		rts
+
+_hex_digits	.text "0123456789abcdef"	; can probably be reused for other stuff as well
+	}}
+}
+
+asmsub  uword2hex  (uword value @AY) clobbers(A,Y)  {
+	; ---- convert 16 bit uword in A/Y into 4-character hexadecimal string 'uword2hex.output' (0-terminated)
+	%asm {{
+		sta  P8ZP_SCRATCH_REG
+		tya
+		jsr  ubyte2hex
+		sta  output
+		sty  output+1
+		lda  P8ZP_SCRATCH_REG
+		jsr  ubyte2hex
+		sta  output+2
+		sty  output+3
+		rts
+output		.text  "0000", $00      ; 0-terminated output buffer (to make printing easier)
+	}}
+}
+
+
+; ---- string conversion to numbers -----
+
+asmsub  any2uword(str string @AY) clobbers(Y) -> ubyte @A {
+	; -- parses a string into a 16 bit unsigned number. String may be in decimal, hex or binary format.
+	;    (the latter two require a $ or % prefix to be recognised)
+	;    (any non-digit character will terminate the number string that is parsed)
+	;    returns amount of processed characters in A, and the parsed number will be in cx16.r15.
+	;    if the string was invalid, 0 will be returned in A.
+	%asm {{
+	pha
+	sta  P8ZP_SCRATCH_W1
+	sty  P8ZP_SCRATCH_W1+1
+	ldy  #0
+	lda  (P8ZP_SCRATCH_W1),y
+	ldy  P8ZP_SCRATCH_W1+1
+	cmp  #'$'
+	beq  _hex
+	cmp  #'%'
+	beq  _bin
+	pla
+	jsr  str2uword
+	jmp  _result
+_hex	pla
+	jsr  hex2uword
+	jmp  _result
+_bin	pla
+	jsr  bin2uword
+_result
+        pha
+        lda  cx16.r15
+        sta  P8ZP_SCRATCH_B1        ; result value
+        pla
+        sta  cx16.r15
+        sty  cx16.r15+1
+        lda  P8ZP_SCRATCH_B1
+        rts
+	}}
+}
+
+inline asmsub  str2ubyte(str string @AY) clobbers(Y) -> ubyte @A {
+	; -- returns in A the unsigned byte value of the string number argument in AY
+	;    the number may NOT be preceded by a + sign and may NOT contain spaces
+	;    (any non-digit character will terminate the number string that is parsed)
+	;    result in A,  number of characters processed also remains in cx16.r15 if you want to use it!! (0 = error)
+	%asm {{
+		jsr  conv.str2uword
+	}}
+}
+
+inline asmsub  str2byte(str string @AY) clobbers(Y) -> ubyte @A {
+	; -- returns in A the signed byte value of the string number argument in AY
+	;    the number may be preceded by a + or - sign but may NOT contain spaces
+	;    (any non-digit character will terminate the number string that is parsed)
+	;    result in A,  number of characters processed also remains in cx16.r15 if you want to use it!! (0 = error)
+	%asm {{
+		jsr  conv.str2word
+	}}
+}
+
+asmsub  str2uword(str string @AY) -> uword @AY {
+	; -- returns the unsigned word value of the string number argument in AY
+	;    the number may NOT be preceded by a + sign and may NOT contain spaces
+	;    (any non-digit character will terminate the number string that is parsed)
+	;    result in AY,  number of characters processed also remains in cx16.r15 if you want to use it!! (0 = error)
+	%asm {{
+_result = P8ZP_SCRATCH_W1
+        	sta  P8ZP_SCRATCH_W2
+        	sty  P8ZP_SCRATCH_W2+1
+		ldy  #0
+		sty  _result
+		sty  _result+1
+		sty  cx16.r15+1
+_loop
+		lda  (P8ZP_SCRATCH_W2),y
+		sec
+		sbc  #48
+		bpl  _digit
+_done
+		sty  cx16.r15
+		lda  _result
+		ldy  _result+1
+		rts
+_digit
+		cmp  #10
+		bcs  _done
+		; add digit to result
+		pha
+		jsr  _result_times_10
+		pla
+		clc
+		adc  _result
+		sta  _result
+		bcc  +
+		inc  _result+1
++		iny
+		bne  _loop
+		; never reached
+
+_result_times_10     ; (W*4 + W)*2
+		lda  _result+1
+		sta  P8ZP_SCRATCH_REG
+		lda  _result
+		asl  a
+		rol  P8ZP_SCRATCH_REG
+		asl  a
+		rol  P8ZP_SCRATCH_REG
+		clc
+		adc  _result
+		sta  _result
+		lda  P8ZP_SCRATCH_REG
+		adc  _result+1
+		asl  _result
+		rol  a
+		sta  _result+1
+		rts
+	}}
+}
+
+asmsub  str2word(str string @AY) -> word @AY {
+	; -- returns the signed word value of the string number argument in AY
+	;    the number may be preceded by a + or - sign but may NOT contain spaces
+	;    (any non-digit character will terminate the number string that is parsed)
+	;    result in AY,  number of characters processed also remains in cx16.r15 if you want to use it!! (0 = error)
+	%asm {{
+_result = P8ZP_SCRATCH_W1
+		sta  P8ZP_SCRATCH_W2
+		sty  P8ZP_SCRATCH_W2+1
+		ldy  #0
+		sty  _result
+		sty  _result+1
+		sty  _negative
+		sty  cx16.r15+1
+		lda  (P8ZP_SCRATCH_W2),y
+		cmp  #'+'
+		bne  +
+		iny
++		cmp  #'-'
+		bne  _parse
+		inc  _negative
+		iny
+_parse		lda  (P8ZP_SCRATCH_W2),y
+		sec
+		sbc  #48
+		bpl  _digit
+_done
+		sty  cx16.r15
+		lda  _negative
+		beq  +
+		sec
+		lda  #0
+		sbc  _result
+		sta  _result
+		lda  #0
+		sbc  _result+1
+		sta  _result+1
++		lda  _result
+		ldy  _result+1
+		rts
+_digit
+		cmp  #10
+		bcs  _done
+		; add digit to result
+		pha
+		jsr  str2uword._result_times_10
+		pla
+		clc
+		adc  _result
+		sta  _result
+		bcc  +
+		inc  _result+1
++		iny
+		bne  _parse
+		; never reached
+_negative	.byte  0
+	}}
+}
+
+asmsub  hex2uword(str string @AY) -> uword @AY {
+	; -- hexadecimal string (with or without '$') to uword.
+	;    string may be in petscii or c64-screencode encoding.
+	;    stops parsing at the first character that's not a hex digit (except leading $)
+	;    result in AY,  number of characters processed also remains in cx16.r15 if you want to use it!! (0 = error)
+	%asm {{
+	sta  P8ZP_SCRATCH_W2
+	sty  P8ZP_SCRATCH_W2+1
+	ldy  #0
+	sty  P8ZP_SCRATCH_W1
+	sty  P8ZP_SCRATCH_W1+1
+	sty  cx16.r15+1
+	lda  (P8ZP_SCRATCH_W2),y
+	beq  _stop
+	cmp  #'$'
+	bne  _loop
+	iny
+_loop
+	lda  #0
+	sta  P8ZP_SCRATCH_B1
+	lda  (P8ZP_SCRATCH_W2),y
+	beq  _stop
+	cmp  #7                 ; screencode letters A-F are 1-6
+	bcc  _add_letter
+	cmp  #'g'
+	bcs  _stop
+	cmp  #'a'
+	bcs  _add_letter
+	cmp  #'0'
+	bcc  _stop
+	cmp  #'9'+1
+	bcs  _stop
+_calc
+	asl  P8ZP_SCRATCH_W1
+	rol  P8ZP_SCRATCH_W1+1
+	asl  P8ZP_SCRATCH_W1
+	rol  P8ZP_SCRATCH_W1+1
+	asl  P8ZP_SCRATCH_W1
+	rol  P8ZP_SCRATCH_W1+1
+	asl  P8ZP_SCRATCH_W1
+	rol  P8ZP_SCRATCH_W1+1
+	and  #$0f
+	clc
+	adc  P8ZP_SCRATCH_B1
+	ora  P8ZP_SCRATCH_W1
+	sta  P8ZP_SCRATCH_W1
+	iny
+	bne  _loop
+_stop
+	sty  cx16.r15
+	lda  P8ZP_SCRATCH_W1
+	ldy  P8ZP_SCRATCH_W1+1
+	rts
+_add_letter
+	pha
+	lda  #9
+	sta  P8ZP_SCRATCH_B1
+	pla
+	jmp  _calc
+	}}
+}
+
+asmsub  bin2uword(str string @AY) -> uword @AY {
+	; -- binary string (with or without '%') to uword.
+	;    stops parsing at the first character that's not a 0 or 1. (except leading %)
+	;    result in AY,  number of characters processed also remains in cx16.r15 if you want to use it!! (0 = error)
+	%asm {{
+	sta  P8ZP_SCRATCH_W2
+	sty  P8ZP_SCRATCH_W2+1
+	ldy  #0
+	sty  P8ZP_SCRATCH_W1
+	sty  P8ZP_SCRATCH_W1+1
+	sty  cx16.r15+1
+	lda  (P8ZP_SCRATCH_W2),y
+	beq  _stop
+	cmp  #'%'
+	bne  _loop
+	iny
+_loop
+	lda  (P8ZP_SCRATCH_W2),y
+	cmp  #'0'
+	bcc  _stop
+	cmp  #'2'
+	bcs  _stop
+_first  asl  P8ZP_SCRATCH_W1
+	rol  P8ZP_SCRATCH_W1+1
+	and  #1
+	ora  P8ZP_SCRATCH_W1
+	sta  P8ZP_SCRATCH_W1
+	iny
+	bne  _loop
+_stop
+	sty  cx16.r15
+	lda  P8ZP_SCRATCH_W1
+	ldy  P8ZP_SCRATCH_W1+1
+	rts
+	}}
+}
+
+}

compiler/res/prog8lib/cx16/floats.p8

@@ -0,0 +1,155 @@
+; Prog8 definitions for floating point handling on the CommanderX16
+;
+; Written by Irmen de Jong (irmen@razorvine.net) - license: GNU GPL 3.0
+;
+; indent format: TABS, size=8
+
+%target cx16
+%option enable_floats
+
+floats {
+	; ---- this block contains C-64 floating point related functions ----
+
+        const float  PI     = 3.141592653589793
+        const float  TWOPI  = 6.283185307179586
+
+
+; ---- ROM float functions ----
+
+		; note: the fac1 and fac2 are working registers and take 6 bytes each,
+		; floats in memory  (and rom) are stored in 5-byte MFLPT packed format.
+
+; note: fac1/2 might get clobbered even if not mentioned in the function's name.
+; note: for subtraction and division, the left operand is in fac2, the right operand in fac1.
+
+romsub $fe00 = AYINT() clobbers(A,X,Y)          ; fac1-> signed word in 100-101 ($64-$65) MSB FIRST. (might throw ILLEGAL QUANTITY)
+
+; GIVAYF: signed word in Y/A (note different lsb/msb order) -> float in fac1
+; there is also floats.GIVUAYFAY - unsigned word in A/Y (lo/hi) to fac1
+; (tip: use GIVAYFAY to use A/Y input; lo/hi switched to normal order)
+romsub $fe03 = GIVAYF(ubyte lo @ Y, ubyte hi @ A) clobbers(A,X,Y)
+
+; fac1 -> unsigned word in Y/A (might throw ILLEGAL QUANTITY) (result also in $14/15)
+; (tip: use GETADRAY to get A/Y output; lo/hi switched to normal little endian order)
+romsub $fe06 = GETADR() clobbers(X) -> ubyte @ Y, ubyte @ A
+
+romsub $fe09 = FADDH() clobbers(A,X,Y)                      ; fac1 += 0.5, for rounding- call this before INT
+romsub $fe0c = FSUB(uword mflpt @ AY) clobbers(A,X,Y)       ; fac1 = mflpt from A/Y - fac1
+romsub $fe0f = FSUBT() clobbers(A,X,Y)                      ; fac1 = fac2-fac1   mind the order of the operands
+romsub $fe12 = FADD(uword mflpt @ AY) clobbers(A,X,Y)       ; fac1 += mflpt value from A/Y
+romsub $fe15 = FADDT() clobbers(A,X,Y)                      ; fac1 += fac2
+romsub $fe1b = ZEROFC() clobbers(A,X,Y)                     ; fac1 = 0
+romsub $fe1e = NORMAL() clobbers(A,X,Y)                     ; normalize fac1 (?)
+romsub $fe24 = LOG() clobbers(A,X,Y)                        ; fac1 = LN(fac1)  (natural log)
+romsub $fe27 = FMULT(uword mflpt @ AY) clobbers(A,X,Y)      ; fac1 *= mflpt value from A/Y
+romsub $fe2a = FMULTT() clobbers(A,X,Y)                     ; fac1 *= fac2
+romsub $fe33 = CONUPK(uword mflpt @ AY) clobbers(A,Y)       ; load mflpt value from memory  in A/Y into fac2
+romsub $fe36 = MUL10() clobbers(A,X,Y)                      ; fac1 *= 10
+romsub $fe3c = DIV10() clobbers(A,X,Y)                      ; fac1 /= 10 , CAUTION: result is always positive!
+romsub $fe3f = FDIV(uword mflpt @ AY) clobbers(A,X,Y)       ; fac1 = mflpt in A/Y / fac1  (remainder in fac2)
+romsub $fe42 = FDIVT() clobbers(A,X,Y)                      ; fac1 = fac2/fac1  (remainder in fac2)  mind the order of the operands
+
+romsub $fe48 = MOVFM(uword mflpt @ AY) clobbers(A,Y)        ; load mflpt value from memory  in A/Y into fac1
+romsub $fe4b = MOVMF(uword mflpt @ XY) clobbers(A,Y)        ; store fac1 to memory  X/Y as 5-byte mflpt
+romsub $fe4e = MOVFA() clobbers(A,X)                        ; copy fac2 to fac1
+romsub $fe51 = MOVAF() clobbers(A,X)                        ; copy fac1 to fac2  (rounded)
+romsub $fe54 = MOVEF() clobbers(A,X)                        ; copy fac1 to fac2
+romsub $fe5a = SIGN() -> ubyte @ A                          ; SIGN(fac1) to A, $ff, $0, $1 for negative, zero, positive
+romsub $fe5d = SGN() clobbers(A,X,Y)                        ; fac1 = SGN(fac1), result of SIGN (-1, 0 or 1)
+romsub $fe60 = FREADSA(byte value @ A) clobbers(A,X,Y)      ; 8 bit signed A -> float in fac1
+romsub $fe6c = ABS()                                        ; fac1 = ABS(fac1)
+romsub $fe6f = FCOMP(uword mflpt @ AY) clobbers(X,Y) -> ubyte @ A   ; A = compare fac1 to mflpt in A/Y, 0=equal 1=fac1 is greater, 255=fac1 is less than
+romsub $fe78 = INT() clobbers(A,X,Y)                        ; INT() truncates, use FADDH first to round instead of trunc
+romsub $fe7e = FINLOG(byte value @A) clobbers (A, X, Y)           ; fac1 += signed byte in A
+romsub $fe81 = FOUT() clobbers(X) -> uword @ AY             ; fac1 -> string, address returned in AY
+romsub $fe8a = SQR() clobbers(A,X,Y)                        ; fac1 = SQRT(fac1)
+romsub $fe8d = FPWRT() clobbers(A,X,Y)                      ; fac1 = fac2 ** fac1
+; note: there is no FPWR() on the Cx16
+romsub $fe93 = NEGOP() clobbers(A)                          ; switch the sign of fac1
+romsub $fe96 = EXP() clobbers(A,X,Y)                        ; fac1 = EXP(fac1)  (e ** fac1)
+romsub $fe9f = RND2(byte value @A) clobbers(A,X,Y)                ; fac1 = RND(A) float random number generator
+romsub $fea2 = RND() clobbers(A,X,Y)                        ; fac1 = RND(fac1) float random number generator
+romsub $fea5 = COS() clobbers(A,X,Y)                        ; fac1 = COS(fac1)
+romsub $fea8 = SIN() clobbers(A,X,Y)                        ; fac1 = SIN(fac1)
+romsub $feab = TAN() clobbers(A,X,Y)                        ; fac1 = TAN(fac1)
+romsub $feae = ATN() clobbers(A,X,Y)                        ; fac1 = ATN(fac1)
+
+
+asmsub  GIVUAYFAY  (uword value @ AY) clobbers(A,X,Y)  {
+	; ---- unsigned 16 bit word in A/Y (lo/hi) to fac1
+	%asm {{
+        phx
+        sta  P8ZP_SCRATCH_W2
+        sty  P8ZP_SCRATCH_B1
+        tya
+        ldy  P8ZP_SCRATCH_W2
+        jsr  GIVAYF                 ; load it as signed... correct afterwards
+        lda  P8ZP_SCRATCH_B1
+	    bpl  +
+	    lda  #<_flt65536
+	    ldy  #>_flt65536
+	    jsr  FADD
++	    plx
+        rts
+_flt65536    .byte 145,0,0,0,0       ; 65536.0
+	}}
+}
+
+
+asmsub  GIVAYFAY  (uword value @ AY) clobbers(A,X,Y)  {
+	; ---- signed 16 bit word in A/Y (lo/hi) to float in fac1
+	%asm {{
+		sta  P8ZP_SCRATCH_B1
+		tya
+		ldy  P8ZP_SCRATCH_B1
+		jmp  GIVAYF		; this uses the inverse order, Y/A
+	}}
+}
+
+asmsub  FTOSWRDAY  () clobbers(X) -> uword @ AY  {
+	; ---- fac1 to signed word in A/Y
+	%asm {{
+		jsr  FTOSWORDYA	; note the inverse Y/A order
+		sta  P8ZP_SCRATCH_B1
+		tya
+		ldy  P8ZP_SCRATCH_B1
+		rts
+	}}
+}
+
+asmsub  GETADRAY  () clobbers(X) -> uword @ AY  {
+	; ---- fac1 to unsigned word in A/Y
+	%asm {{
+		jsr  GETADR		; this uses the inverse order, Y/A
+		sta  P8ZP_SCRATCH_B1
+		tya
+		ldy  P8ZP_SCRATCH_B1
+		rts
+	}}
+}
+
+sub  print_f  (float value) {
+	; ---- prints the floating point value (without a newline).
+	%asm {{
+		phx
+		lda  #<value
+		ldy  #>value
+		jsr  MOVFM		; load float into fac1
+		jsr  FOUT		; fac1 to string in A/Y
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		ldy  #0
+-		lda  (P8ZP_SCRATCH_W1),y
+		beq  +
+		jsr  c64.CHROUT
+		iny
+		bne  -
++		plx
+		rts
+	}}
+}
+
+%asminclude "library:c64/floats.asm", ""
+%asminclude "library:c64/floats_funcs.asm", ""
+
+}

compiler/res/prog8lib/cx16/gfx2.p8

@@ -0,0 +1,976 @@
+%target cx16
+
+; Bitmap pixel graphics routines for the CommanderX16
+; Custom routines to use the full-screen 640x480 and 320x240 screen modes.
+; (These modes are not supported by the documented GRAPH_xxxx kernal routines)
+;
+; No text layer is currently shown, text can be drawn as part of the bitmap itself.
+; Note: for similar graphics routines that also work on the C-64, use the "graphics" module instead.
+; Note: for color palette manipulation, use the "palette" module or write Vera registers yourself.
+; Note: this library implements code for various resolutions and color depths. This takes up memory.
+;       If you're memory constrained you should probably not use this built-in library,
+;       but make a copy in your project only containing the code for the required resolution.
+;
+;
+; SCREEN MODE LIST:
+;   mode 0 = reset back to default text mode
+;   mode 1 = bitmap 320 x 240 monochrome
+;   mode 2 = bitmap 320 x 240 x 4c (TODO not yet implemented)
+;   mode 3 = bitmap 320 x 240 x 16c (TODO not yet implemented)
+;   mode 4 = bitmap 320 x 240 x 256c
+;   mode 5 = bitmap 640 x 480 monochrome
+;   mode 6 = bitmap 640 x 480 x 4c
+;   higher color dephts in highres are not supported due to lack of VRAM
+
+; TODO can we make a FB vector table and emulation routines for the Cx16s' GRAPH_init() call? to replace the builtin 320x200 fb driver?
+
+gfx2 {
+
+    ; read-only control variables:
+    ubyte active_mode = 0
+    uword width = 0
+    uword height = 0
+    ubyte bpp = 0
+    ubyte monochrome_dont_stipple_flag = false            ; set to false to enable stippling mode in monochrome displaymodes
+
+    sub screen_mode(ubyte mode) {
+        when mode {
+            1 -> {
+                ; lores monchrome
+                cx16.VERA_DC_VIDEO = (cx16.VERA_DC_VIDEO & %11001111) | %00100000      ; enable only layer 1
+                cx16.VERA_DC_HSCALE = 64
+                cx16.VERA_DC_VSCALE = 64
+                cx16.VERA_L1_CONFIG = %00000100
+                cx16.VERA_L1_MAPBASE = 0
+                cx16.VERA_L1_TILEBASE = 0
+                width = 320
+                height = 240
+                bpp = 1
+            }
+            ; TODO modes 2, 3 not yet implemented
+            4 -> {
+                ; lores 256c
+                cx16.VERA_DC_VIDEO = (cx16.VERA_DC_VIDEO & %11001111) | %00100000      ; enable only layer 1
+                cx16.VERA_DC_HSCALE = 64
+                cx16.VERA_DC_VSCALE = 64
+                cx16.VERA_L1_CONFIG = %00000111
+                cx16.VERA_L1_MAPBASE = 0
+                cx16.VERA_L1_TILEBASE = 0
+                width = 320
+                height = 240
+                bpp = 8
+            }
+            5 -> {
+                ; highres monochrome
+                cx16.VERA_DC_VIDEO = (cx16.VERA_DC_VIDEO & %11001111) | %00100000      ; enable only layer 1
+                cx16.VERA_DC_HSCALE = 128
+                cx16.VERA_DC_VSCALE = 128
+                cx16.VERA_L1_CONFIG = %00000100
+                cx16.VERA_L1_MAPBASE = 0
+                cx16.VERA_L1_TILEBASE = %00000001
+                width = 640
+                height = 480
+                bpp = 1
+            }
+            6 -> {
+                ; highres 4c
+                cx16.VERA_DC_VIDEO = (cx16.VERA_DC_VIDEO & %11001111) | %00100000      ; enable only layer 1
+                cx16.VERA_DC_HSCALE = 128
+                cx16.VERA_DC_VSCALE = 128
+                cx16.VERA_L1_CONFIG = %00000101
+                cx16.VERA_L1_MAPBASE = 0
+                cx16.VERA_L1_TILEBASE = %00000001
+                width = 640
+                height = 480
+                bpp = 2
+            }
+            else -> {
+                ; back to default text mode and colors
+                cx16.VERA_CTRL = %10000000      ; reset VERA and palette
+                c64.CINT()      ; back to text mode
+                width = 0
+                height = 0
+                bpp = 0
+                mode = 0
+            }
+        }
+
+        active_mode = mode
+        if bpp
+            clear_screen()
+    }
+
+    sub clear_screen() {
+        monochrome_stipple(false)
+        position(0, 0)
+        when active_mode {
+            1 -> {
+                ; lores monochrome
+                repeat 240/2/8
+                    cs_innerloop640()
+            }
+            ; TODO mode 2, 3
+            4 -> {
+                ; lores 256c
+                repeat 240/2
+                    cs_innerloop640()
+            }
+            5 -> {
+                ; highres monochrome
+                repeat 480/8
+                    cs_innerloop640()
+            }
+            6 -> {
+                ; highres 4c
+                repeat 480/4
+                    cs_innerloop640()
+            }
+            ; modes 7 and 8 not supported due to lack of VRAM
+        }
+        position(0, 0)
+    }
+
+    sub monochrome_stipple(ubyte enable) {
+        monochrome_dont_stipple_flag = not enable
+    }
+
+    sub rect(uword x, uword y, uword width, uword height, ubyte color) {
+        if width==0 or height==0
+            return
+        horizontal_line(x, y, width, color)
+        if height==1
+            return
+        horizontal_line(x, y+height-1, width, color)
+        vertical_line(x, y+1, height-2, color)
+        if width==1
+            return
+        vertical_line(x+width-1, y+1, height-2, color)
+    }
+
+    sub fillrect(uword x, uword y, uword width, uword height, ubyte color) {
+        if width==0
+            return
+        repeat height {
+            horizontal_line(x, y, width, color)
+            y++
+        }
+    }
+
+    sub horizontal_line(uword x, uword y, uword length, ubyte color) {
+        if length==0
+            return
+        when active_mode {
+            1, 5 -> {
+                ; monochrome modes, either resolution
+                ubyte separate_pixels = (8-lsb(x)) & 7
+                if separate_pixels as uword > length
+                    separate_pixels = lsb(length)
+                repeat separate_pixels {
+                    ; TODO optimize this by writing a masked byte in 1 go
+                    plot(x, y, color)
+                    x++
+                }
+                length -= separate_pixels
+                if length {
+                    position(x, y)
+                    separate_pixels = lsb(length) & 7
+                    x += length & $fff8
+                    %asm {{
+                        lsr  length+1
+                        ror  length
+                        lsr  length+1
+                        ror  length
+                        lsr  length+1
+                        ror  length
+                        lda  color
+                        bne  +
+                        ldy  #0     ; black
+                        bra  _loop
++                       lda  monochrome_dont_stipple_flag
+                        beq  _stipple
+                        ldy  #255       ; don't stipple
+                        bra  _loop
+_stipple                lda  y
+                        and  #1         ; determine stipple pattern to use
+                        bne  +
+                        ldy  #%01010101
+                        bra  _loop
++                       ldy  #%10101010
+_loop                   lda  length
+                        ora  length+1
+                        beq  _done
+                        sty  cx16.VERA_DATA0
+                        lda  length
+                        bne  +
+                        dec  length+1
++                       dec  length
+                        bra  _loop
+_done
+                    }}
+                    repeat separate_pixels {
+                        ; TODO optimize this by writing a masked byte in 1 go
+                        plot(x, y, color)
+                        x++
+                    }
+                }
+                cx16.VERA_ADDR_H &= %00000111   ; vera auto-increment off again
+            }
+            4 -> {
+                ; lores 256c
+                position(x, y)
+                %asm {{
+                    lda  color
+                    phx
+                    ldx  length+1
+                    beq  +
+                    ldy  #0
+-                   sta  cx16.VERA_DATA0
+                    iny
+                    bne  -
+                    dex
+                    bne  -
++                   ldy  length     ; remaining
+                    beq  +
+-                   sta  cx16.VERA_DATA0
+                    dey
+                    bne  -
++                   plx
+                }}
+            }
+            6 -> {
+                ; highres 4c
+                ; TODO also mostly usable for lores 4c?
+                color &= 3
+                ubyte[4] colorbits
+                ubyte ii
+                for ii in 3 downto 0 {
+                    colorbits[ii] = color
+                    color <<= 2
+                }
+                void addr_mul_24_for_highres_4c(y, x)      ; 24 bits result is in r0 and r1L (highest byte)
+                %asm {{
+                    lda  cx16.VERA_ADDR_H
+                    and  #%00000111         ; no auto advance
+                    sta  cx16.VERA_ADDR_H
+                    stz  cx16.VERA_CTRL     ; setup vera addr 0
+                    lda  cx16.r1
+                    and  #1
+                    sta  cx16.VERA_ADDR_H
+                    lda  cx16.r0
+                    sta  cx16.VERA_ADDR_L
+                    lda  cx16.r0+1
+                    sta  cx16.VERA_ADDR_M
+                    phx
+                    ldx  x
+                }}
+
+                repeat length {
+                    %asm {{
+                        txa
+                        and  #3
+                        tay
+                        lda  cx16.VERA_DATA0
+                        and  gfx2.plot.mask4c,y
+                        ora  colorbits,y
+                        sta  cx16.VERA_DATA0
+                        cpy  #%00000011         ; next vera byte?
+                        bne  +
+                        inc  cx16.VERA_ADDR_L
+                        bne  +
+                        inc  cx16.VERA_ADDR_M
++                       bne  +
+                        inc  cx16.VERA_ADDR_H
++                       inx                     ; next pixel
+                    }}
+                }
+
+                %asm {{
+                    plx
+                }}
+            }
+        }
+    }
+
+    sub vertical_line(uword x, uword y, uword height, ubyte color) {
+        position(x,y)
+        when active_mode {
+            1, 5 -> {
+                ; monochrome, either resolution
+                ; note for the 1 bpp modes we can't use vera's auto increment mode because we have to 'or' the pixel data in place.
+                ; TODO use TWO vera adress pointers simultaneously one for reading, one for writing, so auto-increment IS possible
+                cx16.VERA_ADDR_H &= %00000111   ; no auto advance
+                cx16.r15 = gfx2.plot.bits[x as ubyte & 7]           ; bitmask
+                if active_mode>=5
+                    cx16.r14 = 640/8
+                else
+                    cx16.r14 = 320/8
+                if color {
+                    if monochrome_dont_stipple_flag {
+                        repeat height {
+                            %asm {{
+                                lda  cx16.VERA_DATA0
+                                ora  cx16.r15
+                                sta  cx16.VERA_DATA0
+                                lda  cx16.VERA_ADDR_L
+                                clc
+                                adc  cx16.r14                 ; advance vera ptr to go to the next line
+                                sta  cx16.VERA_ADDR_L
+                                lda  cx16.VERA_ADDR_M
+                                adc  #0
+                                sta  cx16.VERA_ADDR_M
+                                ; lda  cx16.VERA_ADDR_H     ; the bitmap size is small enough to not have to deal with the _H part.
+                                ; adc  #0
+                                ; sta  cx16.VERA_ADDR_H
+                            }}
+                        }
+                    } else {
+                        ; stippling.
+                        height = (height+1)/2       ; TODO is the line sometimes 1 pixel too long now because of rounding?
+                        %asm {{
+                            lda x
+                            eor y
+                            and #1
+                            bne +
+                            lda  cx16.VERA_ADDR_L
+                            clc
+                            adc  cx16.r14                ; advance vera ptr to go to the next line for correct stipple pattern
+                            sta  cx16.VERA_ADDR_L
+                            lda  cx16.VERA_ADDR_M
+                            adc  #0
+                            sta  cx16.VERA_ADDR_M
+        +
+                            asl  cx16.r14
+                            ldy  height
+                            beq  +
+        -                   lda  cx16.VERA_DATA0
+                            ora  cx16.r15
+                            sta  cx16.VERA_DATA0
+                            lda  cx16.VERA_ADDR_L
+                            clc
+                            adc  cx16.r14               ; advance vera data ptr to go to the next-next line
+                            sta  cx16.VERA_ADDR_L
+                            lda  cx16.VERA_ADDR_M
+                            adc  #0
+                            sta  cx16.VERA_ADDR_M
+                            ; lda  cx16.VERA_ADDR_H      ; the bitmap size is small enough to not have to deal with the _H part.
+                            ; adc  #0
+                            ; sta  cx16.VERA_ADDR_H
+                            dey
+                            bne  -
+        +
+                        }}
+                    }
+                } else {
+                    cx16.r15 = ~cx16.r15
+                    repeat height {
+                        %asm {{
+                            lda  cx16.VERA_DATA0
+                            and  cx16.r15
+                            sta  cx16.VERA_DATA0
+                            lda  cx16.VERA_ADDR_L
+                            clc
+                            adc  cx16.r14             ; advance vera data ptr to go to the next line
+                            sta  cx16.VERA_ADDR_L
+                            lda  cx16.VERA_ADDR_M
+                            adc  #0
+                            sta  cx16.VERA_ADDR_M
+                            ; the bitmap size is small enough to not have to deal with the _H part:
+                            ; lda  cx16.VERA_ADDR_H
+                            ; adc  #0
+                            ; sta  cx16.VERA_ADDR_H
+                        }}
+                    }
+                }
+            }
+            4 -> {
+                ; lores 256c
+                ; set vera auto-increment to 320 pixel increment (=next line)
+                cx16.VERA_ADDR_H = cx16.VERA_ADDR_H & %00000111 | (14<<4)
+                %asm {{
+                    ldy  height
+                    beq  +
+                    lda  color
+-                   sta  cx16.VERA_DATA0
+                    dey
+                    bne  -
++
+                }}
+            }
+            6 -> {
+                ; highres 4c
+                ; note for this mode we can't use vera's auto increment mode because we have to 'or' the pixel data in place.
+                ; TODO use TWO vera adress pointers simultaneously one for reading, one for writing, so auto-increment IS possible
+                cx16.VERA_ADDR_H &= %00000111   ; no auto advance
+                ; TODO also mostly usable for lores 4c?
+                void addr_mul_24_for_highres_4c(y, x)      ; 24 bits result is in r0 and r1L (highest byte)
+
+                ; TODO optimize the loop in pure assembly
+                color &= 3
+                color <<= gfx2.plot.shift4c[lsb(x) & 3]
+                ubyte mask = gfx2.plot.mask4c[lsb(x) & 3]
+                repeat height {
+                    ubyte value = cx16.vpeek(lsb(cx16.r1), cx16.r0) & mask | color
+                    cx16.vpoke(lsb(cx16.r1), cx16.r0, value)
+                    %asm {{
+                        ; 24 bits add 160 (640/4)
+                        clc
+                        lda  cx16.r0
+                        adc  #640/4
+                        sta  cx16.r0
+                        lda  cx16.r0+1
+                        adc  #0
+                        sta  cx16.r0+1
+                        bcc  +
+                        inc  cx16.r1
++
+                    }}
+                }
+            }
+        }
+
+    }
+
+    sub line(uword @zp x1, uword @zp y1, uword @zp x2, uword @zp y2, ubyte color) {
+        ; Bresenham algorithm.
+        ; This code special-cases various quadrant loops to allow simple ++ and -- operations.
+        ; TODO there are some slight errors at the first/last pixels in certain slopes...
+        if y1>y2 {
+            ; make sure dy is always positive to have only 4 instead of 8 special cases
+            swap(x1, x2)
+            swap(y1, y2)
+        }
+        word @zp dx = x2-x1 as word
+        word @zp dy = y2-y1 as word
+
+        if dx==0 {
+            vertical_line(x1, y1, abs(dy)+1 as uword, color)
+            return
+        }
+        if dy==0 {
+            if x1>x2
+                x1=x2
+            horizontal_line(x1, y1, abs(dx)+1 as uword, color)
+            return
+        }
+
+        ; TODO rewrite the rest in optimized assembly (or reuse GRAPH_draw_line if we can get the FB replacement vector layer working)
+        word @zp d = 0
+        ubyte positive_ix = true
+        if dx < 0 {
+            dx = -dx
+            positive_ix = false
+        }
+        dx *= 2
+        dy *= 2
+        cx16.r14 = x1       ; internal plot X
+
+        if dx >= dy {
+            if positive_ix {
+                repeat {
+                    plot(cx16.r14, y1, color)
+                    if cx16.r14==x2
+                        return
+                    cx16.r14++
+                    d += dy
+                    if d > dx {
+                        y1++
+                        d -= dx
+                    }
+                }
+            } else {
+                repeat {
+                    plot(cx16.r14, y1, color)
+                    if cx16.r14==x2
+                        return
+                    cx16.r14--
+                    d += dy
+                    if d > dx {
+                        y1++
+                        d -= dx
+                    }
+                }
+            }
+        }
+        else {
+            if positive_ix {
+                repeat {
+                    plot(cx16.r14, y1, color)
+                    if y1 == y2
+                        return
+                    y1++
+                    d += dx
+                    if d > dy {
+                        cx16.r14++
+                        d -= dy
+                    }
+                }
+            } else {
+                repeat {
+                    plot(cx16.r14, y1, color)
+                    if y1 == y2
+                        return
+                    y1++
+                    d += dx
+                    if d > dy {
+                        cx16.r14--
+                        d -= dy
+                    }
+                }
+            }
+        }
+    }
+
+    sub circle(uword @zp xcenter, uword @zp ycenter, ubyte radius, ubyte color) {
+        ; Midpoint algorithm.
+        if radius==0
+            return
+
+        ubyte @zp xx = radius
+        ubyte @zp yy = 0
+        word @zp decisionOver2 = (1 as word)-xx
+        ; R14 = internal plot X
+        ; R15 = internal plot Y
+
+        while xx>=yy {
+            cx16.r14 = xcenter + xx
+            cx16.r15 = ycenter + yy
+            plot(cx16.r14, cx16.r15, color)
+            cx16.r14 = xcenter - xx
+            plot(cx16.r14, cx16.r15, color)
+            cx16.r14 = xcenter + xx
+            cx16.r15 = ycenter - yy
+            plot(cx16.r14, cx16.r15, color)
+            cx16.r14 = xcenter - xx
+            plot(cx16.r14, cx16.r15, color)
+            cx16.r14 = xcenter + yy
+            cx16.r15 = ycenter + xx
+            plot(cx16.r14, cx16.r15, color)
+            cx16.r14 = xcenter - yy
+            plot(cx16.r14, cx16.r15, color)
+            cx16.r14 = xcenter + yy
+            cx16.r15 = ycenter - xx
+            plot(cx16.r14, cx16.r15, color)
+            cx16.r14 = xcenter - yy
+            plot(cx16.r14, cx16.r15, color)
+
+            yy++
+            if decisionOver2<=0
+                decisionOver2 += (yy as word)*2+1
+            else {
+                xx--
+                decisionOver2 += (yy as word -xx)*2+1
+            }
+        }
+    }
+
+    sub disc(uword @zp xcenter, uword @zp ycenter, ubyte @zp radius, ubyte color) {
+        ; Midpoint algorithm, filled
+        if radius==0
+            return
+        ubyte @zp yy = 0
+        word @zp decisionOver2 = (1 as word)-radius
+
+        while radius>=yy {
+            horizontal_line(xcenter-radius, ycenter+yy, radius*$0002+1, color)
+            horizontal_line(xcenter-radius, ycenter-yy, radius*$0002+1, color)
+            horizontal_line(xcenter-yy, ycenter+radius, yy*$0002+1, color)
+            horizontal_line(xcenter-yy, ycenter-radius, yy*$0002+1, color)
+            yy++
+            if decisionOver2<=0
+                decisionOver2 += (yy as word)*2+1
+            else {
+                radius--
+                decisionOver2 += (yy as word -radius)*2+1
+            }
+        }
+    }
+
+    sub plot(uword @zp x, uword y, ubyte color) {
+        ubyte[8] bits = [128, 64, 32, 16, 8, 4, 2, 1]
+        ubyte[4] mask4c = [%00111111, %11001111, %11110011, %11111100]
+        ubyte[4] shift4c = [6,4,2,0]
+        uword addr
+        ubyte value
+
+        when active_mode {
+            1 -> {
+                ; lores monochrome
+                %asm {{
+                    lda  x
+                    eor  y
+                    ora  monochrome_dont_stipple_flag
+                    and  #1
+                }}
+                if_nz {
+                    addr = x/8 + y*(320/8)
+                    value = bits[lsb(x)&7]
+                    if color
+                        cx16.vpoke_or(0, addr, value)
+                    else {
+                        value = ~value
+                        cx16.vpoke_and(0, addr, value)
+                    }
+                }
+            }
+            4 -> {
+                ; lores 256c
+                void addr_mul_24_for_lores_256c(y, x)      ; 24 bits result is in r0 and r1L (highest byte)
+                cx16.vpoke(lsb(cx16.r1), cx16.r0, color)
+                ; activate vera auto-increment mode so next_pixel() can be used after this
+                cx16.VERA_ADDR_H = cx16.VERA_ADDR_H & %00000111 | %00010000
+                color = cx16.VERA_DATA0
+            }
+            5 -> {
+                ; highres monochrome
+                %asm {{
+                    lda  x
+                    eor  y
+                    ora  monochrome_dont_stipple_flag
+                    and  #1
+                }}
+                if_nz {
+                    addr = x/8 + y*(640/8)
+                    value = bits[lsb(x)&7]
+                    if color
+                        cx16.vpoke_or(0, addr, value)
+                    else {
+                        value = ~value
+                        cx16.vpoke_and(0, addr, value)
+                    }
+                }
+            }
+            6 -> {
+                ; highres 4c
+                ; TODO also mostly usable for lores 4c?
+                void addr_mul_24_for_highres_4c(y, x)      ; 24 bits result is in r0 and r1L (highest byte)
+                color &= 3
+                color <<= shift4c[lsb(x) & 3]
+                ; TODO optimize the vera memory manipulation in pure assembly
+                cx16.VERA_ADDR_H &= %00000111   ; no auto advance
+                value = cx16.vpeek(lsb(cx16.r1), cx16.r0) & mask4c[lsb(x) & 3] | color
+                cx16.vpoke(lsb(cx16.r1), cx16.r0, value)
+            }
+        }
+    }
+
+    sub position(uword @zp x, uword y) {
+        ubyte bank
+        when active_mode {
+            1 -> {
+                ; lores monochrome
+                cx16.r0 = y*(320/8) + x/8
+                cx16.vaddr(0, cx16.r0, 0, 1)
+            }
+            ; TODO modes 2,3
+            4 -> {
+                ; lores 256c
+                void addr_mul_24_for_lores_256c(y, x)      ; 24 bits result is in r0 and r1L (highest byte)
+                bank = lsb(cx16.r1)
+                cx16.vaddr(bank, cx16.r0, 0, 1)
+            }
+            5 -> {
+                ; highres monochrome
+                cx16.r0 = y*(640/8) + x/8
+                cx16.vaddr(0, cx16.r0, 0, 1)
+            }
+            6 -> {
+                ; highres 4c
+                void addr_mul_24_for_highres_4c(y, x)      ; 24 bits result is in r0 and r1L (highest byte)
+                bank = lsb(cx16.r1)
+                cx16.vaddr(bank, cx16.r0, 0, 1)
+            }
+        }
+    }
+
+    inline asmsub next_pixel(ubyte color @A) {
+        ; -- sets the next pixel byte to the graphics chip.
+        ;    for 8 bpp screens this will plot 1 pixel.
+        ;    for 1 bpp screens it will plot 8 pixels at once (color = bit pattern).
+        ;    for 2 bpp screens it will plot 4 pixels at once (color = bit pattern).
+        %asm {{
+            sta  cx16.VERA_DATA0
+        }}
+    }
+
+    asmsub next_pixels(uword pixels @AY, uword amount @R0) clobbers(A, Y)  {
+        ; -- sets the next bunch of pixels from a prepared array of bytes.
+        ;    for 8 bpp screens this will plot 1 pixel per byte.
+        ;    for 1 bpp screens it will plot 8 pixels at once (colors are the bit patterns per byte).
+        ;    for 2 bpp screens it will plot 4 pixels at once (colors are the bit patterns per byte).
+        %asm {{
+            phx
+            sta  P8ZP_SCRATCH_W1
+            sty  P8ZP_SCRATCH_W1+1
+            ldx  cx16.r0+1
+            beq  +
+            ldy  #0
+-           lda  (P8ZP_SCRATCH_W1),y
+            sta  cx16.VERA_DATA0
+            iny
+            bne  -
+            inc  P8ZP_SCRATCH_W1+1       ; next page of 256 pixels
+            dex
+            bne  -
+
++           ldx  cx16.r0           ; remaining pixels
+            beq  +
+            ldy  #0
+-           lda  (P8ZP_SCRATCH_W1),y
+            sta  cx16.VERA_DATA0
+            iny
+            dex
+            bne  -
++           plx
+        }}
+    }
+
+    asmsub set_8_pixels_from_bits(ubyte bits @R0, ubyte oncolor @A, ubyte offcolor @Y) {
+        ; this is only useful in 256 color mode where one pixel equals one byte value.
+        %asm {{
+            phx
+            ldx  #8
+-           asl  cx16.r0
+            bcc  +
+            sta  cx16.VERA_DATA0
+            bra  ++
++           sty  cx16.VERA_DATA0
++           dex
+            bne  -
+            plx
+            rts
+        }}
+    }
+
+    const ubyte charset_orig_bank = $0
+    const uword charset_orig_addr = $f800        ; in bank 0, so $0f800
+    const ubyte charset_bank = $1
+    const uword charset_addr = $f000       ; in bank 1, so $1f000
+
+    sub text_charset(ubyte charset) {
+        ; -- make a copy of the selected character set to use with text()
+        ;    the charset number is the same as for the cx16.screen_set_charset() ROM function.
+        ;    1 = ISO charset, 2 = PETSCII uppercase+graphs, 3= PETSCII uppercase+lowercase.
+        cx16.screen_set_charset(charset, 0)
+        cx16.vaddr(charset_orig_bank, charset_orig_addr, 0, 1)
+        cx16.vaddr(charset_bank, charset_addr, 1, 1)
+        repeat 256*8 {
+            cx16.VERA_DATA1 = cx16.VERA_DATA0
+        }
+    }
+
+    sub text(uword @zp x, uword y, ubyte color, uword sctextptr) {
+        ; -- Write some text at the given pixel position. The text string must be in screencode encoding (not petscii!).
+        ;    You must also have called text_charset() first to select and prepare the character set to use.
+        ;    NOTE: in monochrome (1bpp) screen modes, x position is currently constrained to mulitples of 8 !  TODO allow per-pixel horizontal positioning
+        uword chardataptr
+        when active_mode {
+            1, 5 -> {
+                ; monochrome mode, either resolution
+                cx16.r2 = 40
+                if active_mode>=5
+                    cx16.r2 = 80
+                while @(sctextptr) {
+                    chardataptr = charset_addr + (@(sctextptr) as uword)*8
+                    cx16.vaddr(charset_bank, chardataptr, 1, 1)
+                    position(x,y)
+                    %asm {{
+                        lda  cx16.VERA_ADDR_H
+                        and  #%111              ; don't auto-increment, we have to do that manually because of the ora
+                        sta  cx16.VERA_ADDR_H
+                        lda  color
+                        sta  P8ZP_SCRATCH_B1
+                        ldy  #8
+-                       lda  P8ZP_SCRATCH_B1
+                        bne  +                  ; white color, plot normally
+                        lda  cx16.VERA_DATA1
+                        eor  #255               ; black color, keep only the other pixels
+                        and  cx16.VERA_DATA0
+                        bra  ++
++                       lda  cx16.VERA_DATA0
+                        ora  cx16.VERA_DATA1
++                       sta  cx16.VERA_DATA0
+                        lda  cx16.VERA_ADDR_L
+                        clc
+                        adc  cx16.r2
+                        sta  cx16.VERA_ADDR_L
+                        bcc  +
+                        inc  cx16.VERA_ADDR_M
++                       lda  x
+                        clc
+                        adc  #1
+                        sta  x
+                        bcc  +
+                        inc  x+1
++                       dey
+                        bne  -
+                    }}
+                    sctextptr++
+                }
+            }
+            4 -> {
+                ; lores 256c
+                while @(sctextptr) {
+                    chardataptr = charset_addr + (@(sctextptr) as uword)*8
+                    cx16.vaddr(charset_bank, chardataptr, 1, 1)
+                    repeat 8 {
+                        ; TODO rewrite this inner loop fully in assembly
+                        position(x,y)
+                        y++
+                        %asm {{
+                            phx
+                            ldx  #1
+                            lda  cx16.VERA_DATA1
+                            sta  P8ZP_SCRATCH_B1
+                            ldy  #8
+-                           asl  P8ZP_SCRATCH_B1
+                            bcc  +
+                            stx  cx16.VERA_DATA0    ; write a pixel
+                            bra  ++
++                           lda  cx16.VERA_DATA0    ; don't write a pixel, but do advance to the next address
++                           dey
+                            bne  -
+                            plx
+                        }}
+                    }
+                    x+=8
+                    y-=8
+                    sctextptr++
+                }
+            }
+            6 -> {
+                ; hires 4c
+                while @(sctextptr) {
+                    chardataptr = charset_addr + (@(sctextptr) as uword)*8
+                    repeat 8 {
+                        ; TODO rewrite this inner loop fully in assembly
+                        ubyte charbits = cx16.vpeek(charset_bank, chardataptr)
+                        repeat 8 {
+                            charbits <<= 1
+                            if_cs
+                                plot(x, y, color)
+                            x++
+                        }
+                        x-=8
+                        chardataptr++
+                        y++
+                    }
+                    x+=8
+                    y-=8
+                    sctextptr++
+                }
+            }
+        }
+    }
+
+    asmsub cs_innerloop640() clobbers(Y) {
+        %asm {{
+            ldy  #80
+-           stz  cx16.VERA_DATA0
+            stz  cx16.VERA_DATA0
+            stz  cx16.VERA_DATA0
+            stz  cx16.VERA_DATA0
+            stz  cx16.VERA_DATA0
+            stz  cx16.VERA_DATA0
+            stz  cx16.VERA_DATA0
+            stz  cx16.VERA_DATA0
+            dey
+            bne  -
+            rts
+        }}
+    }
+
+    asmsub addr_mul_24_for_highres_4c(uword yy @R2, uword xx @R3)  clobbers(A, Y) -> uword @R0, uword @R1 {
+        ; yy * 160 + xx/4  (24 bits calculation)
+        ; 24 bits result is in r0 and r1L (highest byte)
+        %asm {{
+            ldy  #5
+-           asl  cx16.r2
+            rol  cx16.r2+1
+            dey
+            bne  -
+            lda  cx16.r2
+            sta  cx16.r0
+            lda  cx16.r2+1
+            sta  cx16.r0+1
+            asl  cx16.r0
+            rol  cx16.r0+1
+            asl  cx16.r0
+            rol  cx16.r0+1
+
+            ; xx >>= 2  (xx=R3)
+            lsr  cx16.r3+1
+            ror  cx16.r3
+            lsr  cx16.r3+1
+            ror  cx16.r3
+
+            ; add r2 and xx (r3) to r0 (24-bits)
+            stz  cx16.r1
+            clc
+            lda  cx16.r0
+            adc  cx16.r2
+            sta  cx16.r0
+            lda  cx16.r0+1
+            adc  cx16.r2+1
+            sta  cx16.r0+1
+            bcc  +
+            inc  cx16.r1
++           clc
+            lda  cx16.r0
+            adc  cx16.r3
+            sta  cx16.r0
+            lda  cx16.r0+1
+            adc  cx16.r3+1
+            sta  cx16.r0+1
+            bcc  +
+            inc  cx16.r1
++
+            rts
+        }}
+    }
+
+    asmsub addr_mul_24_for_lores_256c(uword yy @R0, uword xx @AY) clobbers(A) -> uword @R0, ubyte @R1  {
+        ; yy * 320 + xx (24 bits calculation)
+        %asm {{
+            sta  P8ZP_SCRATCH_W1
+            sty  P8ZP_SCRATCH_W1+1
+            lda  cx16.r0
+            sta  P8ZP_SCRATCH_B1
+            lda  cx16.r0+1
+            sta  cx16.r1
+            sta  P8ZP_SCRATCH_REG
+            lda  cx16.r0
+            asl  a
+            rol  P8ZP_SCRATCH_REG
+            asl  a
+            rol  P8ZP_SCRATCH_REG
+            asl  a
+            rol  P8ZP_SCRATCH_REG
+            asl  a
+            rol  P8ZP_SCRATCH_REG
+            asl  a
+            rol  P8ZP_SCRATCH_REG
+            asl  a
+            rol  P8ZP_SCRATCH_REG
+            sta  cx16.r0
+            lda  P8ZP_SCRATCH_B1
+            clc
+            adc  P8ZP_SCRATCH_REG
+            sta  cx16.r0+1
+            bcc  +
+            inc  cx16.r1
++           ; now add the value to this 24-bits number
+            lda  cx16.r0
+            clc
+            adc  P8ZP_SCRATCH_W1
+            sta  cx16.r0
+            lda  cx16.r0+1
+            adc  P8ZP_SCRATCH_W1+1
+            sta  cx16.r0+1
+            bcc  +
+            inc  cx16.r1
++           lda  cx16.r1
+            rts
+        }}
+    }
+
+}

compiler/res/prog8lib/cx16/graphics.p8

@@ -0,0 +1,138 @@
+%target cx16
+%import syslib
+%import textio
+
+; Bitmap pixel graphics module for the CommanderX16
+; wraps the graphics functions that are in ROM.
+; only black/white monchrome 320x200 for now. (i.e. truncated at the bottom)
+; For full-screen 640x480 or 320x240 graphics, use the "gfx2" module instead. (but that is Cx16-specific)
+; Note: there is no color palette manipulation here, you have to do that yourself or use the "palette" module.
+
+
+graphics {
+    const uword WIDTH = 320
+    const ubyte HEIGHT = 200
+
+    sub enable_bitmap_mode() {
+        ; enable bitmap screen, erase it and set colors to black/white.
+        void cx16.screen_set_mode($80)
+        cx16.GRAPH_init(0)
+        clear_screen(1, 0)
+    }
+
+    sub disable_bitmap_mode() {
+        ; enables text mode, erase the text screen, color white
+        void cx16.screen_set_mode(2)
+        txt.fill_screen(' ', 1)     ; doesn't seem to fully clear the text screen after returning from gfx mode
+    }
+
+
+    sub clear_screen(ubyte pixelcolor, ubyte bgcolor) {
+        cx16.GRAPH_set_colors(pixelcolor, pixelcolor, bgcolor)
+        cx16.GRAPH_clear()
+    }
+
+    sub line(uword @zp x1, ubyte @zp y1, uword @zp x2, ubyte @zp y2) {
+        cx16.GRAPH_draw_line(x1, y1, x2, y2)
+    }
+
+    sub fillrect(uword x, uword y, uword width, uword height) {
+        cx16.GRAPH_draw_rect(x, y, width, height, 0, 1)
+    }
+
+    sub rect(uword x, uword y, uword width, uword height) {
+        cx16.GRAPH_draw_rect(x, y, width, height, 0, 0)
+    }
+
+    sub horizontal_line(uword x, uword y, uword length) {
+        if length
+            cx16.GRAPH_draw_line(x, y, x+length-1, y)
+    }
+
+    sub vertical_line(uword x, uword y, uword height) {
+        if height
+            cx16.GRAPH_draw_line(x, y, x, y+height-1)
+    }
+
+    sub circle(uword xcenter, ubyte ycenter, ubyte radius) {
+        ;cx16.r0 = xcenter - radius/2
+        ;cx16.r1 = ycenter - radius/2
+        ;cx16.r2 = radius*2
+        ;cx16.r3 = radius*2
+        ;cx16.GRAPH_draw_oval(false)          ; currently this call is not implemented on cx16, does a BRK
+
+        ; Midpoint algorithm
+        if radius==0
+            return
+        ubyte @zp xx = radius
+        ubyte @zp yy = 0
+        word @zp decisionOver2 = (1 as word)-xx
+
+        while xx>=yy {
+            cx16.r0 = xcenter + xx
+            cx16.r1 = ycenter + yy
+            cx16.FB_cursor_position2()
+            cx16.FB_set_pixel(1)
+            cx16.r0 = xcenter - xx
+            cx16.FB_cursor_position2()
+            cx16.FB_set_pixel(1)
+            cx16.r0 = xcenter + xx
+            cx16.r1 = ycenter - yy
+            cx16.FB_cursor_position2()
+            cx16.FB_set_pixel(1)
+            cx16.r0 = xcenter - xx
+            cx16.FB_cursor_position2()
+            cx16.FB_set_pixel(1)
+            cx16.r0 = xcenter + yy
+            cx16.r1 = ycenter + xx
+            cx16.FB_cursor_position2()
+            cx16.FB_set_pixel(1)
+            cx16.r0 = xcenter - yy
+            cx16.FB_cursor_position2()
+            cx16.FB_set_pixel(1)
+            cx16.r0 = xcenter + yy
+            cx16.r1 = ycenter - xx
+            cx16.FB_cursor_position2()
+            cx16.FB_set_pixel(1)
+            cx16.r0 = xcenter - yy
+            cx16.FB_cursor_position2()
+            cx16.FB_set_pixel(1)
+            yy++
+            if decisionOver2<=0 {
+                decisionOver2 += (yy as word)*2+1
+            } else {
+                xx--
+                decisionOver2 += (yy as word -xx)*2+1
+            }
+        }
+    }
+
+    sub disc(uword xcenter, ubyte ycenter, ubyte radius) {
+        if radius==0
+            return
+        ubyte @zp yy = 0
+        word decisionOver2 = (1 as word)-radius
+
+        while radius>=yy {
+            horizontal_line(xcenter-radius, ycenter+yy, radius*2+1)
+            horizontal_line(xcenter-radius, ycenter-yy, radius*2+1)
+            horizontal_line(xcenter-yy, ycenter+radius, yy*2+1)
+            horizontal_line(xcenter-yy, ycenter-radius, yy*2+1)
+            yy++
+            if decisionOver2<=0
+                decisionOver2 += (yy as word)*2+1
+            else {
+                radius--
+                decisionOver2 += (yy as word -radius)*2+1
+            }
+        }
+    }
+
+    inline asmsub  plot(uword plotx @R0, uword ploty @R1) clobbers(A, X, Y) {
+        %asm {{
+            jsr  cx16.FB_cursor_position
+            lda  #1
+            jsr  cx16.FB_set_pixel
+        }}
+    }
+}

compiler/res/prog8lib/cx16/palette.p8

@@ -0,0 +1,183 @@
+%target cx16
+
+; Manipulate the Commander X16's display color palette.
+; Should you want to restore the default palette, you have to reinitialize the Vera yourself.
+
+palette {
+
+    uword vera_palette_ptr
+    ubyte c
+
+    sub set_color(ubyte index, uword color) {
+        vera_palette_ptr = $fa00+index*2
+        cx16.vpoke(1, vera_palette_ptr, lsb(color))
+        vera_palette_ptr++
+        cx16.vpoke(1, vera_palette_ptr, msb(color))
+    }
+
+    sub set_rgb4(uword palette_bytes_ptr, uword num_colors) {
+        ; 2 bytes per color entry, the Vera uses this, but the R/GB bytes order is swapped
+        vera_palette_ptr = $fa00
+        repeat num_colors {
+            cx16.vpoke(1, vera_palette_ptr+1, @(palette_bytes_ptr))
+            palette_bytes_ptr++
+            cx16.vpoke(1, vera_palette_ptr, @(palette_bytes_ptr))
+            palette_bytes_ptr++
+            vera_palette_ptr+=2
+        }
+    }
+
+    sub set_rgb(uword palette_words_ptr, uword num_colors) {
+        ; 1 word per color entry (in little endian format so $gb0r)
+        vera_palette_ptr = $fa00
+        repeat num_colors*2 {
+            cx16.vpoke(1, vera_palette_ptr, @(palette_words_ptr))
+            palette_words_ptr++
+            vera_palette_ptr++
+        }
+    }
+
+    sub set_rgb8(uword palette_bytes_ptr, uword num_colors) {
+        ; 3 bytes per color entry, adjust color depth from 8 to 4 bits per channel.
+        vera_palette_ptr = $fa00
+        ubyte red
+        ubyte greenblue
+        repeat num_colors {
+            red = @(palette_bytes_ptr) >> 4
+            palette_bytes_ptr++
+            greenblue = @(palette_bytes_ptr) & %11110000
+            palette_bytes_ptr++
+            greenblue |= @(palette_bytes_ptr) >> 4    ; add Blue
+            palette_bytes_ptr++
+            cx16.vpoke(1, vera_palette_ptr, greenblue)
+            vera_palette_ptr++
+            cx16.vpoke(1, vera_palette_ptr, red)
+            vera_palette_ptr++
+        }
+    }
+
+    sub set_monochrome(uword screencolorRGB, uword drawcolorRGB) {
+        vera_palette_ptr = $fa00
+        cx16.vpoke(1, vera_palette_ptr, lsb(screencolorRGB))   ; G,B
+        vera_palette_ptr++
+        cx16.vpoke(1, vera_palette_ptr, msb(screencolorRGB))   ; R
+        vera_palette_ptr++
+        repeat 255 {
+            cx16.vpoke(1, vera_palette_ptr, lsb(drawcolorRGB)) ; G,B
+            vera_palette_ptr++
+            cx16.vpoke(1, vera_palette_ptr, msb(drawcolorRGB)) ; R
+            vera_palette_ptr++
+        }
+    }
+
+    sub set_grayscale() {
+        vera_palette_ptr = $fa00
+        repeat 16 {
+            c=0
+            repeat 16 {
+                cx16.vpoke(1, vera_palette_ptr, c)
+                vera_palette_ptr++
+                cx16.vpoke(1, vera_palette_ptr, c)
+                vera_palette_ptr++
+                c += $11
+            }
+        }
+    }
+
+    uword[] C64_colorpalette_dark = [   ; this is a darker palette with more contrast
+        $000,  ; 0 = black
+        $FFF,  ; 1 = white
+        $632,  ; 2 = red
+        $7AB,  ; 3 = cyan
+        $638,  ; 4 = purple
+        $584,  ; 5 = green
+        $327,  ; 6 = blue
+        $BC6,  ; 7 = yellow
+        $642,  ; 8 = orange
+        $430,  ; 9 = brown
+        $965,  ; 10 = light red
+        $444,  ; 11 = dark grey
+        $666,  ; 12 = medium grey
+        $9D8,  ; 13 = light green
+        $65B,  ; 14 = light blue
+        $999   ; 15 = light grey
+    ]
+
+    uword[] C64_colorpalette_pepto = [  ; # this is Pepto's Commodore-64 palette  http://www.pepto.de/projects/colorvic/
+        $000,  ; 0 = black
+        $FFF,  ; 1 = white
+        $833,  ; 2 = red
+        $7cc,  ; 3 = cyan
+        $839,  ; 4 = purple
+        $5a4,  ; 5 = green
+        $229,  ; 6 = blue
+        $ef7,  ; 7 = yellow
+        $852,  ; 8 = orange
+        $530,  ; 9 = brown
+        $c67,  ; 10 = light red
+        $444,  ; 11 = dark grey
+        $777,  ; 12 = medium grey
+        $af9,  ; 13 = light green
+        $76e,  ; 14 = light blue
+        $bbb   ; 15 = light grey
+    ]
+
+    uword[] C64_colorpalette_light = [  ; this is a lighter palette
+        $000,  ; 0 = black
+        $FFF,  ; 1 = white
+        $944,  ; 2 = red
+        $7CC,  ; 3 = cyan
+        $95A,  ; 4 = purple
+        $6A5,  ; 5 = green
+        $549,  ; 6 = blue
+        $CD8,  ; 7 = yellow
+        $963,  ; 8 = orange
+        $650,  ; 9 = brown
+        $C77,  ; 10 = light red
+        $666,  ; 11 = dark grey
+        $888,  ; 12 = medium grey
+        $AE9,  ; 13 = light green
+        $87C,  ; 14 = light blue
+        $AAA   ; 15 = light grey
+    ]
+
+    sub set_c64pepto() {
+        vera_palette_ptr = $fa00
+        repeat 16 {
+            for c in 0 to 15 {
+                uword cc = C64_colorpalette_pepto[c]
+                cx16.vpoke(1, vera_palette_ptr, lsb(cc))     ; G, B
+                vera_palette_ptr++
+                cx16.vpoke(1, vera_palette_ptr, msb(cc))     ; R
+                vera_palette_ptr++
+            }
+        }
+    }
+
+    sub set_c64light() {
+        vera_palette_ptr = $fa00
+        repeat 16 {
+            for c in 0 to 15 {
+                uword cc = C64_colorpalette_light[c]
+                cx16.vpoke(1, vera_palette_ptr, lsb(cc))     ; G, B
+                vera_palette_ptr++
+                cx16.vpoke(1, vera_palette_ptr, msb(cc))     ; R
+                vera_palette_ptr++
+            }
+        }
+    }
+
+    sub set_c64dark() {
+        vera_palette_ptr = $fa00
+        repeat 16 {
+            for c in 0 to 15 {
+                uword cc = C64_colorpalette_dark[c]
+                cx16.vpoke(1, vera_palette_ptr, lsb(cc))     ; G, B
+                vera_palette_ptr++
+                cx16.vpoke(1, vera_palette_ptr, msb(cc))     ; R
+                vera_palette_ptr++
+            }
+        }
+    }
+
+}

compiler/res/prog8lib/cx16/syslib.p8

@@ -0,0 +1,782 @@
+; Prog8 definitions for the CommanderX16
+; Including memory registers, I/O registers, Basic and Kernal subroutines.
+;
+; Written by Irmen de Jong (irmen@razorvine.net) - license: GNU GPL 3.0
+;
+; indent format: TABS, size=8
+
+%target cx16
+
+
+c64 {
+
+; ---- kernal routines, these are the same as on the Commodore-64 (hence the same block name) ----
+
+; STROUT --> use txt.print
+; CLEARSCR -> use txt.clear_screen
+; HOMECRSR -> use txt.plot
+
+romsub $FF81 = CINT() clobbers(A,X,Y)                           ; (alias: SCINIT) initialize screen editor and video chip
+romsub $FF84 = IOINIT() clobbers(A, X)                          ; initialize I/O devices (CIA, SID, IRQ)
+romsub $FF87 = RAMTAS() clobbers(A,X,Y)                         ; initialize RAM, tape buffer, screen
+romsub $FF8A = RESTOR() clobbers(A,X,Y)                         ; restore default I/O vectors
+romsub $FF8D = VECTOR(uword userptr @ XY, ubyte dir @ Pc) clobbers(A,Y)     ; read/set I/O vector table
+romsub $FF90 = SETMSG(ubyte value @ A)                          ; set Kernal message control flag
+romsub $FF93 = SECOND(ubyte address @ A) clobbers(A)            ; (alias: LSTNSA) send secondary address after LISTEN
+romsub $FF96 = TKSA(ubyte address @ A) clobbers(A)              ; (alias: TALKSA) send secondary address after TALK
+romsub $FF99 = MEMTOP(uword address @ XY, ubyte dir @ Pc) -> uword @ XY     ; read/set top of memory  pointer.   NOTE: as a Cx16 extension, also returns the number of RAM memory banks in register A !  See MEMTOP2
+romsub $FF9C = MEMBOT(uword address @ XY, ubyte dir @ Pc) -> uword @ XY     ; read/set bottom of memory  pointer
+romsub $FF9F = SCNKEY() clobbers(A,X,Y)                         ; scan the keyboard
+romsub $FFA2 = SETTMO(ubyte timeout @ A)                        ; set time-out flag for IEEE bus
+romsub $FFA5 = ACPTR() -> ubyte @ A                             ; (alias: IECIN) input byte from serial bus
+romsub $FFA8 = CIOUT(ubyte databyte @ A)                        ; (alias: IECOUT) output byte to serial bus
+romsub $FFAB = UNTLK() clobbers(A)                              ; command serial bus device to UNTALK
+romsub $FFAE = UNLSN() clobbers(A)                              ; command serial bus device to UNLISTEN
+romsub $FFB1 = LISTEN(ubyte device @ A) clobbers(A)             ; command serial bus device to LISTEN
+romsub $FFB4 = TALK(ubyte device @ A) clobbers(A)               ; command serial bus device to TALK
+romsub $FFB7 = READST() -> ubyte @ A                            ; read I/O status word
+romsub $FFBA = SETLFS(ubyte logical @ A, ubyte device @ X, ubyte address @ Y)   ; set logical file parameters
+romsub $FFBD = SETNAM(ubyte namelen @ A, str filename @ XY)     ; set filename parameters
+romsub $FFC0 = OPEN() clobbers(X,Y) -> ubyte @Pc, ubyte @A      ; (via 794 ($31A)) open a logical file
+romsub $FFC3 = CLOSE(ubyte logical @ A) clobbers(A,X,Y)         ; (via 796 ($31C)) close a logical file
+romsub $FFC6 = CHKIN(ubyte logical @ X) clobbers(A,X) -> ubyte @Pc    ; (via 798 ($31E)) define an input channel
+romsub $FFC9 = CHKOUT(ubyte logical @ X) clobbers(A,X)          ; (via 800 ($320)) define an output channel
+romsub $FFCC = CLRCHN() clobbers(A,X)                           ; (via 802 ($322)) restore default devices
+romsub $FFCF = CHRIN() clobbers(X, Y) -> ubyte @ A   ; (via 804 ($324)) input a character (for keyboard, read a whole line from the screen) A=byte read.
+romsub $FFD2 = CHROUT(ubyte char @ A)                           ; (via 806 ($326)) output a character
+romsub $FFD5 = LOAD(ubyte verify @ A, uword address @ XY) -> ubyte @Pc, ubyte @ A, ubyte @ X, ubyte @ Y     ; (via 816 ($330)) load from device
+romsub $FFD8 = SAVE(ubyte zp_startaddr @ A, uword endaddr @ XY) -> ubyte @ Pc, ubyte @ A                    ; (via 818 ($332)) save to a device
+romsub $FFDB = SETTIM(ubyte low @ A, ubyte middle @ X, ubyte high @ Y)      ; set the software clock
+romsub $FFDE = RDTIM() -> ubyte @ A, ubyte @ X, ubyte @ Y       ; read the software clock
+romsub $FFE1 = STOP() clobbers(X) -> ubyte @ Pz, ubyte @ A      ; (via 808 ($328)) check the STOP key (and some others in A)
+romsub $FFE4 = GETIN() clobbers(X,Y) -> ubyte @Pc, ubyte @ A    ; (via 810 ($32A)) get a character
+romsub $FFE7 = CLALL() clobbers(A,X)                            ; (via 812 ($32C)) close all files
+romsub $FFEA = UDTIM() clobbers(A,X)                            ; update the software clock
+romsub $FFED = SCREEN() -> ubyte @ X, ubyte @ Y                 ; read number of screen rows and columns
+romsub $FFF0 = PLOT(ubyte col @ Y, ubyte row @ X, ubyte dir @ Pc) -> ubyte @ X, ubyte @ Y       ; read/set position of cursor on screen.  Use txt.plot for a 'safe' wrapper that preserves X.
+romsub $FFF3 = IOBASE() -> uword @ XY                           ; read base address of I/O devices
+
+; ---- utility
+
+asmsub STOP2() -> ubyte @A  {
+    ; -- check if STOP key was pressed, returns true if so.  More convenient to use than STOP() because that only sets the carry status flag.
+    %asm {{
+        phx
+        jsr  c64.STOP
+        beq  +
+        plx
+        lda  #0
+        rts
++       plx
+        lda  #1
+        rts
+    }}
+}
+
+asmsub RDTIM16() -> uword @AY {
+    ; --  like RDTIM() but only returning the lower 16 bits for convenience
+    %asm {{
+        phx
+        jsr  c64.RDTIM
+        pha
+        txa
+        tay
+        pla
+        plx
+        rts
+    }}
+}
+
+asmsub MEMTOP2() -> ubyte @A {
+    ; -- uses MEMTOP's cx16 extension to query the number of available RAM banks.
+    %asm {{
+        phx
+        sec
+        jsr  c64.MEMTOP
+        plx
+        rts
+    }}
+}
+
+}
+
+cx16 {
+
+; irq and hardware vectors:
+    &uword  CINV            = $0314     ; IRQ vector (in ram)
+    &uword  NMI_VEC         = $FFFA     ; 65c02 nmi vector, determined by the kernal if banked in
+    &uword  RESET_VEC       = $FFFC     ; 65c02 reset vector, determined by the kernal if banked in
+    &uword  IRQ_VEC         = $FFFE     ; 65c02 interrupt vector, determined by the kernal if banked in
+
+
+; the sixteen virtual 16-bit registers
+    &uword r0  = $0002
+    &uword r1  = $0004
+    &uword r2  = $0006
+    &uword r3  = $0008
+    &uword r4  = $000a
+    &uword r5  = $000c
+    &uword r6  = $000e
+    &uword r7  = $0010
+    &uword r8  = $0012
+    &uword r9  = $0014
+    &uword r10 = $0016
+    &uword r11 = $0018
+    &uword r12 = $001a
+    &uword r13 = $001c
+    &uword r14 = $001e
+    &uword r15 = $0020
+
+; VERA registers
+
+    const uword VERA_BASE       = $9F20
+    &ubyte  VERA_ADDR_L         = VERA_BASE + $0000
+    &ubyte  VERA_ADDR_M         = VERA_BASE + $0001
+    &ubyte  VERA_ADDR_H         = VERA_BASE + $0002
+    &ubyte  VERA_DATA0          = VERA_BASE + $0003
+    &ubyte  VERA_DATA1          = VERA_BASE + $0004
+    &ubyte  VERA_CTRL           = VERA_BASE + $0005
+    &ubyte  VERA_IEN            = VERA_BASE + $0006
+    &ubyte  VERA_ISR            = VERA_BASE + $0007
+    &ubyte  VERA_IRQ_LINE_L     = VERA_BASE + $0008
+    &ubyte  VERA_DC_VIDEO       = VERA_BASE + $0009
+    &ubyte  VERA_DC_HSCALE      = VERA_BASE + $000A
+    &ubyte  VERA_DC_VSCALE      = VERA_BASE + $000B
+    &ubyte  VERA_DC_BORDER      = VERA_BASE + $000C
+    &ubyte  VERA_DC_HSTART      = VERA_BASE + $0009
+    &ubyte  VERA_DC_HSTOP       = VERA_BASE + $000A
+    &ubyte  VERA_DC_VSTART      = VERA_BASE + $000B
+    &ubyte  VERA_DC_VSTOP       = VERA_BASE + $000C
+    &ubyte  VERA_L0_CONFIG      = VERA_BASE + $000D
+    &ubyte  VERA_L0_MAPBASE     = VERA_BASE + $000E
+    &ubyte  VERA_L0_TILEBASE    = VERA_BASE + $000F
+    &ubyte  VERA_L0_HSCROLL_L   = VERA_BASE + $0010
+    &ubyte  VERA_L0_HSCROLL_H   = VERA_BASE + $0011
+    &ubyte  VERA_L0_VSCROLL_L   = VERA_BASE + $0012
+    &ubyte  VERA_L0_VSCROLL_H   = VERA_BASE + $0013
+    &ubyte  VERA_L1_CONFIG      = VERA_BASE + $0014
+    &ubyte  VERA_L1_MAPBASE     = VERA_BASE + $0015
+    &ubyte  VERA_L1_TILEBASE    = VERA_BASE + $0016
+    &ubyte  VERA_L1_HSCROLL_L   = VERA_BASE + $0017
+    &ubyte  VERA_L1_HSCROLL_H   = VERA_BASE + $0018
+    &ubyte  VERA_L1_VSCROLL_L   = VERA_BASE + $0019
+    &ubyte  VERA_L1_VSCROLL_H   = VERA_BASE + $001A
+    &ubyte  VERA_AUDIO_CTRL     = VERA_BASE + $001B
+    &ubyte  VERA_AUDIO_RATE     = VERA_BASE + $001C
+    &ubyte  VERA_AUDIO_DATA     = VERA_BASE + $001D
+    &ubyte  VERA_SPI_DATA       = VERA_BASE + $001E
+    &ubyte  VERA_SPI_CTRL       = VERA_BASE + $001F
+; VERA_PSG_BASE     = $1F9C0
+; VERA_PALETTE_BASE = $1FA00
+; VERA_SPRITES_BASE = $1FC00
+
+; I/O
+
+    const uword  via1   = $9f60                  ;VIA 6522 #1
+    &ubyte  d1prb	= via1+0
+    &ubyte  d1pra	= via1+1
+    &ubyte  d1ddrb	= via1+2
+    &ubyte  d1ddra	= via1+3
+    &ubyte  d1t1l	= via1+4
+    &ubyte  d1t1h	= via1+5
+    &ubyte  d1t1ll	= via1+6
+    &ubyte  d1t1lh	= via1+7
+    &ubyte  d1t2l	= via1+8
+    &ubyte  d1t2h	= via1+9
+    &ubyte  d1sr	= via1+10
+    &ubyte  d1acr	= via1+11
+    &ubyte  d1pcr	= via1+12
+    &ubyte  d1ifr	= via1+13
+    &ubyte  d1ier	= via1+14
+    &ubyte  d1ora	= via1+15
+
+    const uword  via2   = $9f70                  ;VIA 6522 #2
+    &ubyte  d2prb	= via2+0
+    &ubyte  d2pra	= via2+1
+    &ubyte  d2ddrb	= via2+2
+    &ubyte  d2ddra	= via2+3
+    &ubyte  d2t1l	= via2+4
+    &ubyte  d2t1h	= via2+5
+    &ubyte  d2t1ll	= via2+6
+    &ubyte  d2t1lh	= via2+7
+    &ubyte  d2t2l	= via2+8
+    &ubyte  d2t2h	= via2+9
+    &ubyte  d2sr	= via2+10
+    &ubyte  d2acr	= via2+11
+    &ubyte  d2pcr	= via2+12
+    &ubyte  d2ifr	= via2+13
+    &ubyte  d2ier	= via2+14
+    &ubyte  d2ora	= via2+15
+
+
+; ---- Commander X-16 additions on top of C64 kernal routines ----
+; spelling of the names is taken from the Commander X-16 rom sources
+
+; supported C128 additions
+romsub $ff4a = close_all(ubyte device @A)  clobbers(A,X,Y)
+romsub $ff59 = lkupla(ubyte la @A)  clobbers(A,X,Y)
+romsub $ff5c = lkupsa(ubyte sa @Y)  clobbers(A,X,Y)
+romsub $ff5f = screen_set_mode(ubyte mode @A)  clobbers(A, X, Y) -> ubyte @Pc
+romsub $ff62 = screen_set_charset(ubyte charset @A, uword charsetptr @XY)  clobbers(A,X,Y)      ; incompatible with C128  dlchr()
+; not yet supported: romsub $ff65 = pfkey()  clobbers(A,X,Y)
+romsub $ff6e = jsrfar()
+romsub $ff74 = fetch(ubyte bank @X, ubyte index @Y)  clobbers(X) -> ubyte @A
+romsub $ff77 = stash(ubyte data @A, ubyte bank @X, ubyte index @Y)  clobbers(X)
+romsub $ff7a = cmpare(ubyte data @A, ubyte bank @X, ubyte index @Y)  clobbers(X)
+romsub $ff7d = primm()
+
+; X16 additions
+romsub $ff44 = macptr()  clobbers(A,X,Y)
+romsub $ff47 = enter_basic(ubyte cold_or_warm @Pc)  clobbers(A,X,Y)
+romsub $ff68 = mouse_config(ubyte shape @A, ubyte scale @X)  clobbers (A, X, Y)
+romsub $ff6b = mouse_get(ubyte zpdataptr @X)  clobbers(A)
+romsub $ff71 = mouse_scan()  clobbers(A, X, Y)
+romsub $ff53 = joystick_scan()  clobbers(A, X, Y)
+romsub $ff56 = joystick_get(ubyte joynr @A) -> ubyte @A, ubyte @X, ubyte @Y
+romsub $ff4d = clock_set_date_time(uword yearmonth @R0, uword dayhours @R1, uword minsecs @R2, ubyte jiffies @R3)  clobbers(A, X, Y)
+romsub $ff50 = clock_get_date_time()  clobbers(A, X, Y)  -> uword @R0, uword @R1, uword @R2, ubyte @R3   ; result registers see clock_set_date_time()
+
+; TODO specify the correct clobbers for alle these functions below, we now assume all 3 regs are clobbered
+
+; high level graphics & fonts
+romsub $ff20 = GRAPH_init(uword vectors @R0)  clobbers(A,X,Y)
+romsub $ff23 = GRAPH_clear()  clobbers(A,X,Y)
+romsub $ff26 = GRAPH_set_window(uword x @R0, uword y @R1, uword width @R2, uword height @R3)  clobbers(A,X,Y)
+romsub $ff29 = GRAPH_set_colors(ubyte stroke @A, ubyte fill @X, ubyte background @Y)  clobbers (A,X,Y)
+romsub $ff2c = GRAPH_draw_line(uword x1 @R0, uword y1 @R1, uword x2 @R2, uword y2 @R3)  clobbers(A,X,Y)
+romsub $ff2f = GRAPH_draw_rect(uword x @R0, uword y @R1, uword width @R2, uword height @R3, uword cornerradius @R4, ubyte fill @Pc)  clobbers(A,X,Y)
+romsub $ff32 = GRAPH_move_rect(uword sx @R0, uword sy @R1, uword tx @R2, uword ty @R3, uword width @R4, uword height @R5)  clobbers(A,X,Y)
+romsub $ff35 = GRAPH_draw_oval(uword x @R0, uword y @R1, uword width @R2, uword height @R3, ubyte fill @Pc)  clobbers(A,X,Y)
+romsub $ff38 = GRAPH_draw_image(uword x @R0, uword y @R1, uword ptr @R2, uword width @R3, uword height @R4)  clobbers(A,X,Y)
+romsub $ff3b = GRAPH_set_font(uword fontptr @R0)  clobbers(A,X,Y)
+romsub $ff3e = GRAPH_get_char_size(ubyte baseline @A, ubyte width @X, ubyte height_or_style @Y, ubyte is_control @Pc)  clobbers(A,X,Y)
+romsub $ff41 = GRAPH_put_char(uword x @R0, uword y @R1, ubyte char @A)  clobbers(A,X,Y)
+romsub $ff41 = GRAPH_put_next_char(ubyte char @A)  clobbers(A,X,Y)     ; alias for the routine above that doesn't reset the position of the initial character
+
+; framebuffer
+romsub $fef6 = FB_init()  clobbers(A,X,Y)
+romsub $fef9 = FB_get_info()  clobbers(X,Y) -> byte @A, uword @R0, uword @R1    ; width=r0, height=r1
+romsub $fefc = FB_set_palette(uword pointer @R0, ubyte index @A, ubyte bytecount @X)  clobbers(A,X,Y)
+romsub $feff = FB_cursor_position(uword x @R0, uword y @R1)  clobbers(A,X,Y)
+romsub $feff = FB_cursor_position2()  clobbers(A,X,Y)           ;  alias for the previous routine, but avoiding having to respecify both x and y every time
+romsub $ff02 = FB_cursor_next_line(uword x @R0)  clobbers(A,X,Y)
+romsub $ff05 = FB_get_pixel()  clobbers(X,Y) -> ubyte @A
+romsub $ff08 = FB_get_pixels(uword pointer @R0, uword count @R1)  clobbers(A,X,Y)
+romsub $ff0b = FB_set_pixel(ubyte color @A)  clobbers(A,X,Y)
+romsub $ff0e = FB_set_pixels(uword pointer @R0, uword count @R1)  clobbers(A,X,Y)
+romsub $ff11 = FB_set_8_pixels(ubyte pattern @A, ubyte color @X)  clobbers(A,X,Y)
+romsub $ff14 = FB_set_8_pixels_opaque(ubyte pattern @R0, ubyte mask @A, ubyte color1 @X, ubyte color2 @Y)  clobbers(A,X,Y)
+romsub $ff17 = FB_fill_pixels(uword count @R0, uword pstep @R1, ubyte color @A)  clobbers(A,X,Y)
+romsub $ff1a = FB_filter_pixels(uword pointer @ R0, uword count @R1)  clobbers(A,X,Y)
+romsub $ff1d = FB_move_pixels(uword sx @R0, uword sy @R1, uword tx @R2, uword ty @R3, uword count @R4)  clobbers(A,X,Y)
+
+; misc
+romsub $fef0 = sprite_set_image(uword pixels @R0, uword mask @R1, ubyte bpp @R2, ubyte number @A, ubyte width @X, ubyte height @Y, ubyte apply_mask @Pc)  clobbers(A,X,Y) -> ubyte @Pc
+romsub $fef3 = sprite_set_position(uword x @R0, uword y @R1, ubyte number @A)  clobbers(A,X,Y)
+romsub $fee4 = memory_fill(uword address @R0, uword num_bytes @R1, ubyte value @A)  clobbers(A,X,Y)
+romsub $fee7 = memory_copy(uword source @R0, uword target @R1, uword num_bytes @R2)  clobbers(A,X,Y)
+romsub $feea = memory_crc(uword address @R0, uword num_bytes @R1)  clobbers(A,X,Y) -> uword @R2
+romsub $feed = memory_decompress(uword input @R0, uword output @R1)  clobbers(A,X,Y) -> uword @R1       ; last address +1 is result in R1
+romsub $fedb = console_init(uword x @R0, uword y @R1, uword width @R2, uword height @R3)  clobbers(A,X,Y)
+romsub $fede = console_put_char(ubyte char @A, ubyte wrapping @Pc)  clobbers(A,X,Y)
+romsub $fee1 = console_get_char()  clobbers(X,Y) -> ubyte @A
+romsub $fed8 = console_put_image(uword pointer @R0, uword width @R1, uword height @R2)  clobbers(A,X,Y)
+romsub $fed5 = console_set_paging_message(uword msgptr @R0)  clobbers(A,X,Y)
+romsub $fed2 = kbdbuf_put(ubyte key @A)  clobbers(A,X,Y)
+romsub $fecf = entropy_get() -> ubyte @A, ubyte @X, ubyte @Y
+romsub $fecc = monitor()  clobbers(A,X,Y)
+
+; ---- end of kernal routines ----
+
+
+; ---- utilities -----
+
+inline asmsub rombank(ubyte rombank @A) {
+    ; -- set the rom banks
+    %asm {{
+        sta  $01            ; rom bank register (new)
+        sta  cx16.d1prb     ; rom bank register (old)
+    }}
+}
+
+inline asmsub rambank(ubyte rambank @A) {
+    ; -- set the ram bank
+    %asm {{
+        sta  $00            ; ram bank register (new)
+        sta  cx16.d1pra     ; ram bank register (old)
+    }}
+}
+
+asmsub vpeek(ubyte bank @A, uword address @XY) -> ubyte @A {
+        ; -- get a byte from VERA's video memory
+        ;    note: inefficient when reading multiple sequential bytes!
+        %asm {{
+                pha
+                lda  #1
+                sta  cx16.VERA_CTRL
+                pla
+                and  #1
+                sta  cx16.VERA_ADDR_H
+                sty  cx16.VERA_ADDR_M
+                stx  cx16.VERA_ADDR_L
+                lda  cx16.VERA_DATA1
+                rts
+            }}
+}
+
+asmsub vaddr(ubyte bank @A, uword address @R0, ubyte addrsel @R1, byte autoIncrOrDecrByOne @Y) clobbers(A) {
+        ; -- setup the VERA's data address register 0 or 1
+        %asm {{
+            and  #1
+            pha
+            lda  cx16.r1
+            and  #1
+            sta  cx16.VERA_CTRL
+            lda  cx16.r0
+            sta  cx16.VERA_ADDR_L
+            lda  cx16.r0+1
+            sta  cx16.VERA_ADDR_M
+            pla
+            cpy  #0
+            bmi  ++
+            beq  +
+            ora  #%00010000
++           sta  cx16.VERA_ADDR_H
+            rts
++           ora  #%00011000
+            sta  cx16.VERA_ADDR_H
+            rts
+        }}
+}
+
+asmsub vpoke(ubyte bank @A, uword address @R0, ubyte value @Y) clobbers(A) {
+        ; -- write a single byte to VERA's video memory
+        ;    note: inefficient when writing multiple sequential bytes!
+        %asm {{
+            stz  cx16.VERA_CTRL
+            and  #1
+            sta  cx16.VERA_ADDR_H
+            lda  cx16.r0
+            sta  cx16.VERA_ADDR_L
+            lda  cx16.r0+1
+            sta  cx16.VERA_ADDR_M
+            sty  cx16.VERA_DATA0
+            rts
+        }}
+}
+
+asmsub vpoke_or(ubyte bank @A, uword address @R0, ubyte value @Y) clobbers (A) {
+        ; -- or a single byte to the value already in the VERA's video memory at that location
+        ;    note: inefficient when writing multiple sequential bytes!
+        %asm {{
+            stz  cx16.VERA_CTRL
+            and  #1
+            sta  cx16.VERA_ADDR_H
+            lda  cx16.r0
+            sta  cx16.VERA_ADDR_L
+            lda  cx16.r0+1
+            sta  cx16.VERA_ADDR_M
+            tya
+            ora  cx16.VERA_DATA0
+            sta  cx16.VERA_DATA0
+            rts
+        }}
+}
+
+asmsub vpoke_and(ubyte bank @A, uword address @R0, ubyte value @Y) clobbers(A) {
+        ; -- and a single byte to the value already in the VERA's video memory at that location
+        ;    note: inefficient when writing multiple sequential bytes!
+        %asm {{
+            stz  cx16.VERA_CTRL
+            and  #1
+            sta  cx16.VERA_ADDR_H
+            lda  cx16.r0
+            sta  cx16.VERA_ADDR_L
+            lda  cx16.r0+1
+            sta  cx16.VERA_ADDR_M
+            tya
+            and  cx16.VERA_DATA0
+            sta  cx16.VERA_DATA0
+            rts
+        }}
+}
+
+asmsub vpoke_xor(ubyte bank @A, uword address @R0, ubyte value @Y) clobbers (A) {
+        ; -- xor a single byte to the value already in the VERA's video memory at that location
+        ;    note: inefficient when writing multiple sequential bytes!
+        %asm {{
+            stz  cx16.VERA_CTRL
+            and  #1
+            sta  cx16.VERA_ADDR_H
+            lda  cx16.r0
+            sta  cx16.VERA_ADDR_L
+            lda  cx16.r0+1
+            sta  cx16.VERA_ADDR_M
+            tya
+            eor  cx16.VERA_DATA0
+            sta  cx16.VERA_DATA0
+            rts
+        }}
+}
+
+sub FB_set_pixels_from_buf(uword buffer, uword count) {
+    %asm {{
+            ; -- This is replacement code for the normal FB_set_pixels subroutine in ROM
+            ;    However that routine contains a bug in the current v38 ROM that makes it crash when count > 255.
+            ;    So the code below replaces that. Once the ROM is patched this routine is no longer necessary.
+            ;    See https://github.com/commanderx16/x16-rom/issues/179
+            phx
+            lda  buffer
+            ldy  buffer+1
+            sta  P8ZP_SCRATCH_W1
+            sty  P8ZP_SCRATCH_W1+1
+            jsr  _pixels
+            plx
+            rts
+
+_pixels     lda  count+1
+            beq  +
+            ldx  #0
+-           jsr  _loop
+            inc  P8ZP_SCRATCH_W1+1
+            dec  count+1
+            bne  -
+
++           ldx  count
+_loop       ldy  #0
+-           lda  (P8ZP_SCRATCH_W1),y
+            sta  cx16.VERA_DATA0
+            iny
+            dex
+            bne  -
+            rts
+        }}
+}
+
+; ---- system stuff -----
+asmsub  init_system()  {
+    ; Initializes the machine to a sane starting state.
+    ; Called automatically by the loader program logic.
+    %asm {{
+        sei
+        cld
+        ;stz  $00
+        ;stz  $01
+        ;stz  d1prb      ; select rom bank 0 (enable kernal)
+        lda  #$80
+        sta  VERA_CTRL
+        jsr  c64.IOINIT
+        jsr  c64.RESTOR
+        jsr  c64.CINT
+        lda  #$90       ; black
+        jsr  c64.CHROUT
+        lda  #1         ; swap fg/bg
+        jsr  c64.CHROUT
+        lda  #$9e       ; yellow
+        jsr  c64.CHROUT
+        lda  #147       ; clear screen
+        jsr  c64.CHROUT
+        lda  #0
+        tax
+        tay
+        clc
+        clv
+        cli
+        rts
+    }}
+}
+
+asmsub  init_system_phase2()  {
+    %asm {{
+        sei
+        lda  cx16.CINV
+        sta  restore_irq._orig_irqvec
+        lda  cx16.CINV+1
+        sta  restore_irq._orig_irqvec+1
+        cli
+        rts
+    }}
+}
+
+asmsub  set_irq(uword handler @AY, ubyte useKernal @Pc) clobbers(A)  {
+	%asm {{
+	        sta  _modified+1
+	        sty  _modified+2
+	        lda  #0
+	        adc  #0
+	        sta  _use_kernal
+		sei
+		lda  #<_irq_handler
+		sta  cx16.CINV
+		lda  #>_irq_handler
+		sta  cx16.CINV+1
+                lda  cx16.VERA_IEN
+                ora  #%00000001     ; enable the vsync irq
+                sta  cx16.VERA_IEN
+		cli
+		rts
+
+_irq_handler    jsr  _irq_handler_init
+_modified	jsr  $ffff                      ; modified
+		jsr  _irq_handler_end
+		lda  _use_kernal
+		bne  +
+		; end irq processing - don't use kernal's irq handling
+		lda  cx16.VERA_ISR
+		ora  #1
+		sta  cx16.VERA_ISR      ; clear Vera Vsync irq status
+		ply
+		plx
+		pla
+		rti
++		jmp  (restore_irq._orig_irqvec)   ; continue with normal kernal irq routine
+
+_use_kernal     .byte  0
+
+_irq_handler_init
+		; save all zp scratch registers and the X register as these might be clobbered by the irq routine
+		stx  IRQ_X_REG
+		lda  P8ZP_SCRATCH_B1
+		sta  IRQ_SCRATCH_ZPB1
+		lda  P8ZP_SCRATCH_REG
+		sta  IRQ_SCRATCH_ZPREG
+		lda  P8ZP_SCRATCH_W1
+		sta  IRQ_SCRATCH_ZPWORD1
+		lda  P8ZP_SCRATCH_W1+1
+		sta  IRQ_SCRATCH_ZPWORD1+1
+		lda  P8ZP_SCRATCH_W2
+		sta  IRQ_SCRATCH_ZPWORD2
+		lda  P8ZP_SCRATCH_W2+1
+		sta  IRQ_SCRATCH_ZPWORD2+1
+		; stack protector; make sure we don't clobber the top of the evaluation stack
+		dex
+		dex
+		dex
+		dex
+		dex
+		dex
+		cld
+		rts
+
+_irq_handler_end
+		; restore all zp scratch registers and the X register
+		lda  IRQ_SCRATCH_ZPB1
+		sta  P8ZP_SCRATCH_B1
+		lda  IRQ_SCRATCH_ZPREG
+		sta  P8ZP_SCRATCH_REG
+		lda  IRQ_SCRATCH_ZPWORD1
+		sta  P8ZP_SCRATCH_W1
+		lda  IRQ_SCRATCH_ZPWORD1+1
+		sta  P8ZP_SCRATCH_W1+1
+		lda  IRQ_SCRATCH_ZPWORD2
+		sta  P8ZP_SCRATCH_W2
+		lda  IRQ_SCRATCH_ZPWORD2+1
+		sta  P8ZP_SCRATCH_W2+1
+		ldx  IRQ_X_REG
+		rts
+
+IRQ_X_REG		.byte  0
+IRQ_SCRATCH_ZPB1	.byte  0
+IRQ_SCRATCH_ZPREG	.byte  0
+IRQ_SCRATCH_ZPWORD1	.word  0
+IRQ_SCRATCH_ZPWORD2	.word  0
+
+		}}
+}
+
+
+asmsub  restore_irq() clobbers(A) {
+	%asm {{
+	    sei
+	    lda  _orig_irqvec
+	    sta  cx16.CINV
+	    lda  _orig_irqvec+1
+	    sta  cx16.CINV+1
+	    lda  cx16.VERA_IEN
+	    and  #%11110000     ; disable all Vera IRQs
+	    ora  #%00000001     ; enable only the vsync Irq
+	    sta  cx16.VERA_IEN
+	    cli
+	    rts
+_orig_irqvec    .word  0
+        }}
+}
+
+asmsub  set_rasterirq(uword handler @AY, uword rasterpos @R0) clobbers(A) {
+	%asm {{
+            sta  _modified+1
+            sty  _modified+2
+            lda  cx16.r0
+            ldy  cx16.r0+1
+            sei
+            lda  cx16.VERA_IEN
+            and  #%11110000     ; clear other IRQs
+            ora  #%00000010     ; enable the line (raster) irq
+            sta  cx16.VERA_IEN
+            lda  cx16.r0
+            ldy  cx16.r0+1
+            jsr  set_rasterline
+            lda  #<_raster_irq_handler
+            sta  cx16.CINV
+            lda  #>_raster_irq_handler
+            sta  cx16.CINV+1
+            cli
+            rts
+
+_raster_irq_handler
+            jsr  set_irq._irq_handler_init
+_modified   jsr  $ffff                      ; modified
+            jsr  set_irq._irq_handler_end
+            ; end irq processing - don't use kernal's irq handling
+            lda  cx16.VERA_ISR
+            ora  #%00000010
+            sta  cx16.VERA_ISR      ; clear Vera line irq status
+            ply
+            plx
+            pla
+            rti
+        }}
+}
+
+asmsub  set_rasterline(uword line @AY) {
+    %asm {{
+        sta  cx16.VERA_IRQ_LINE_L
+        lda  cx16.VERA_IEN
+        and  #%01111111
+        sta  cx16.VERA_IEN
+        tya
+        lsr  a
+        ror  a
+        and  #%10000000
+        ora  cx16.VERA_IEN
+        sta  cx16.VERA_IEN
+        rts
+    }}
+}
+
+}
+
+
+sys {
+    ; ------- lowlevel system routines --------
+
+    const ubyte target = 16         ;  compilation target specifier.  64 = C64,  16 = CommanderX16.
+
+
+    asmsub reset_system() {
+        ; Soft-reset the system back to Basic prompt.
+        %asm {{
+            sei
+            stz  $01            ; bank the kernal in (new rom bank register)
+            stz  cx16.d1prb     ; bank the kernal in (old rom bank register)
+            jmp  (cx16.RESET_VEC)
+        }}
+    }
+
+    sub wait(uword jiffies) {
+        ; --- wait approximately the given number of jiffies (1/60th seconds)
+        repeat jiffies {
+            ubyte jiff = lsb(c64.RDTIM16())
+            while jiff==lsb(c64.RDTIM16()) {
+                ; wait until 1 jiffy has passed
+            }
+        }
+    }
+
+    inline asmsub memcopy(uword source @R0, uword target @R1, uword count @AY) clobbers(A,X,Y) {
+        %asm {{
+            sta  cx16.r2
+            sty  cx16.r2+1
+            jsr  cx16.memory_copy
+        }}
+    }
+
+    inline asmsub memset(uword mem @R0, uword numbytes @R1, ubyte value @A) clobbers(A,X,Y) {
+        %asm {{
+            jsr  cx16.memory_fill
+        }}
+    }
+
+    asmsub memsetw(uword mem @R0, uword numwords @R1, uword value @AY) clobbers (A,X,Y) {
+        %asm {{
+            ldx  cx16.r0
+            stx  P8ZP_SCRATCH_W1
+            ldx  cx16.r0+1
+            stx  P8ZP_SCRATCH_W1+1
+            ldx  cx16.r1
+            stx  P8ZP_SCRATCH_W2
+            ldx  cx16.r1+1
+            stx  P8ZP_SCRATCH_W2+1
+            jmp  prog8_lib.memsetw
+        }}
+    }
+
+    inline asmsub rsave() {
+        ; save cpu status flag and all registers A, X, Y.
+        ; see http://6502.org/tutorials/register_preservation.html
+        %asm {{
+            php
+            pha
+            phy
+            phx
+        }}
+    }
+
+    inline asmsub rrestore() {
+        ; restore all registers and cpu status flag
+        %asm {{
+            plx
+            ply
+            pla
+            plp
+        }}
+    }
+
+    inline asmsub read_flags() -> ubyte @A {
+        %asm {{
+            php
+            pla
+        }}
+    }
+
+    inline asmsub clear_carry() {
+        %asm {{
+            clc
+        }}
+    }
+
+    inline asmsub set_carry() {
+        %asm {{
+            sec
+        }}
+    }
+
+    inline asmsub clear_irqd() {
+        %asm {{
+            cli
+        }}
+    }
+
+    inline asmsub set_irqd() {
+        %asm {{
+            sei
+        }}
+    }
+
+    inline asmsub exit(ubyte returnvalue @A) {
+        ; -- immediately exit the program with a return code in the A register
+        %asm {{
+            jsr  c64.CLRCHN		; reset i/o channels
+            ldx  prog8_lib.orig_stackpointer
+            txs
+            rts		; return to original caller
+        }}
+    }
+
+    inline asmsub progend() -> uword @AY {
+        %asm {{
+            lda  #<prog8_program_end
+            ldy  #>prog8_program_end
+        }}
+    }
+}

compiler/res/prog8lib/cx16/textio.p8

@@ -0,0 +1,749 @@
+; Prog8 definitions for the Text I/O and Screen routines for the CommanderX16
+;
+; Written by Irmen de Jong (irmen@razorvine.net) - license: GNU GPL 3.0
+;
+; indent format: TABS, size=8
+
+%target cx16
+%import syslib
+%import conv
+
+
+txt {
+
+const ubyte DEFAULT_WIDTH = 80
+const ubyte DEFAULT_HEIGHT = 60
+
+
+sub clear_screen() {
+    txt.chrout(147)
+}
+
+sub home() {
+    txt.chrout(19)
+}
+
+sub nl() {
+    txt.chrout('\n')
+}
+
+sub spc() {
+    txt.chrout(' ')
+}
+
+asmsub column(ubyte col @A) clobbers(A, X, Y) {
+    ; ---- set the cursor on the given column (starting with 0) on the current line
+    %asm {{
+        sec
+        jsr  c64.PLOT
+        tay
+        clc
+        jmp  c64.PLOT
+    }}
+}
+
+asmsub  fill_screen (ubyte char @ A, ubyte color @ Y) clobbers(A)  {
+	; ---- fill the character screen with the given fill character and character color.
+	%asm {{
+	    sty  _ly+1
+        phx
+        pha
+        jsr  c64.SCREEN             ; get dimensions in X/Y
+        txa
+        lsr  a
+        lsr  a
+        sta  _lx+1
+        stz  cx16.VERA_CTRL
+        lda  #%00010000
+        sta  cx16.VERA_ADDR_H       ; enable auto increment by 1, bank 0.
+        stz  cx16.VERA_ADDR_L       ; start at (0,0)
+        stz  cx16.VERA_ADDR_M
+        pla
+_lx     ldx  #0                     ; modified
+        phy
+_ly     ldy  #1                     ; modified
+-       sta  cx16.VERA_DATA0
+        sty  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA0
+        sty  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA0
+        sty  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA0
+        sty  cx16.VERA_DATA0
+        dex
+        bne  -
+        ply
+        dey
+        beq  +
+        stz  cx16.VERA_ADDR_L
+        inc  cx16.VERA_ADDR_M       ; next line
+        bra  _lx
++       plx
+        rts
+        }}
+}
+
+asmsub  clear_screenchars (ubyte char @ A) clobbers(Y)  {
+	; ---- clear the character screen with the given fill character (leaves colors)
+	;      (assumes screen matrix is at the default address)
+	%asm {{
+        phx
+        pha
+        jsr  c64.SCREEN             ; get dimensions in X/Y
+        txa
+        lsr  a
+        lsr  a
+        sta  _lx+1
+        stz  cx16.VERA_CTRL
+        lda  #%00100000
+        sta  cx16.VERA_ADDR_H       ; enable auto increment by 2, bank 0.
+        stz  cx16.VERA_ADDR_L       ; start at (0,0)
+        stz  cx16.VERA_ADDR_M
+        pla
+_lx     ldx  #0                     ; modified
+-       sta  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA0
+        dex
+        bne  -
+        dey
+        beq  +
+        stz  cx16.VERA_ADDR_L
+        inc  cx16.VERA_ADDR_M       ; next line
+        bra  _lx
++       plx
+        rts
+        }}
+}
+
+asmsub  clear_screencolors (ubyte color @ A) clobbers(Y)  {
+	; ---- clear the character screen colors with the given color (leaves characters).
+	;      (assumes color matrix is at the default address)
+	%asm {{
+        phx
+        sta  _la+1
+        jsr  c64.SCREEN             ; get dimensions in X/Y
+        txa
+        lsr  a
+        lsr  a
+        sta  _lx+1
+        stz  cx16.VERA_CTRL
+        lda  #%00100000
+        sta  cx16.VERA_ADDR_H       ; enable auto increment by 2, bank 0.
+        lda  #1
+        sta  cx16.VERA_ADDR_L       ; start at (1,0)
+        stz  cx16.VERA_ADDR_M
+_lx     ldx  #0                     ; modified
+_la     lda  #0                     ; modified
+-       sta  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA0
+        dex
+        bne  -
+        dey
+        beq  +
+        lda  #1
+        sta  cx16.VERA_ADDR_L
+        inc  cx16.VERA_ADDR_M       ; next line
+        bra  _lx
++       plx
+        rts
+        }}
+}
+
+
+ubyte[16] color_to_charcode = [$90,$05,$1c,$9f,$9c,$1e,$1f,$9e,$81,$95,$96,$97,$98,$99,$9a,$9b]
+
+sub color (ubyte txtcol) {
+    txtcol &= 15
+    c64.CHROUT(color_to_charcode[txtcol])
+}
+
+sub color2 (ubyte txtcol, ubyte bgcol) {
+    txtcol &= 15
+    bgcol &= 15
+    c64.CHROUT(color_to_charcode[bgcol])
+    c64.CHROUT(1)       ; switch fg and bg colors
+    c64.CHROUT(color_to_charcode[txtcol])
+}
+
+sub lowercase() {
+    cx16.screen_set_charset(3, 0)  ; lowercase charset
+}
+
+sub uppercase() {
+    cx16.screen_set_charset(2, 0)  ; uppercase charset
+}
+
+asmsub  scroll_left() clobbers(A, Y)  {
+	; ---- scroll the whole screen 1 character to the left
+	;      contents of the rightmost column are unchanged, you should clear/refill this yourself
+	%asm {{
+	    phx
+	    jsr  c64.SCREEN
+	    dex
+	    stx  _lx+1
+        dey
+        sty  P8ZP_SCRATCH_B1    ; number of rows to scroll
+
+_nextline
+        stz  cx16.VERA_CTRL     ; data port 0: source column
+        lda  #%00010000         ; auto increment 1
+        sta  cx16.VERA_ADDR_H
+        lda  #2
+        sta  cx16.VERA_ADDR_L   ; begin in column 1
+        ldy  P8ZP_SCRATCH_B1
+        sty  cx16.VERA_ADDR_M
+        lda  #1
+        sta  cx16.VERA_CTRL     ; data port 1: destination column
+        lda  #%00010000         ; auto increment 1
+        sta  cx16.VERA_ADDR_H
+        stz  cx16.VERA_ADDR_L
+        sty  cx16.VERA_ADDR_M
+
+_lx     ldx  #0                ; modified
+-       lda  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA1    ; copy char
+        lda  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA1    ; copy color
+        dex
+        bne  -
+        dec  P8ZP_SCRATCH_B1
+        bpl  _nextline
+
+        lda  #0
+        sta  cx16.VERA_CTRL
+	    plx
+	    rts
+	}}
+}
+
+asmsub  scroll_right() clobbers(A)  {
+	; ---- scroll the whole screen 1 character to the right
+	;      contents of the leftmost column are unchanged, you should clear/refill this yourself
+	%asm {{
+	    phx
+	    jsr  c64.SCREEN
+	    dex
+	    stx  _lx+1
+	    txa
+	    asl  a
+	    dea
+	    sta  _rcol+1
+	    ina
+	    ina
+	    sta  _rcol2+1
+        dey
+        sty  P8ZP_SCRATCH_B1    ; number of rows to scroll
+
+_nextline
+        stz  cx16.VERA_CTRL     ; data port 0: source column
+        lda  #%00011000         ; auto decrement 1
+        sta  cx16.VERA_ADDR_H
+_rcol   lda  #79*2-1             ; modified
+        sta  cx16.VERA_ADDR_L   ; begin in rightmost column minus one
+        ldy  P8ZP_SCRATCH_B1
+        sty  cx16.VERA_ADDR_M
+        lda  #1
+        sta  cx16.VERA_CTRL     ; data port 1: destination column
+        lda  #%00011000         ; auto decrement 1
+        sta  cx16.VERA_ADDR_H
+_rcol2  lda  #79*2+1            ; modified
+        sta  cx16.VERA_ADDR_L
+        sty  cx16.VERA_ADDR_M
+
+_lx     ldx  #0                 ; modified
+-       lda  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA1    ; copy char
+        lda  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA1    ; copy color
+        dex
+        bne  -
+        dec  P8ZP_SCRATCH_B1
+        bpl  _nextline
+
+        lda  #0
+        sta  cx16.VERA_CTRL
+	    plx
+	    rts
+	}}
+}
+
+asmsub  scroll_up() clobbers(A, Y)  {
+	; ---- scroll the whole screen 1 character up
+	;      contents of the bottom row are unchanged, you should refill/clear this yourself
+	%asm {{
+	    phx
+	    jsr  c64.SCREEN
+	    stx  _nextline+1
+	    dey
+        sty  P8ZP_SCRATCH_B1
+        stz  cx16.VERA_CTRL         ; data port 0 is source
+        lda  #1
+        sta  cx16.VERA_ADDR_M       ; start at second line
+        stz  cx16.VERA_ADDR_L
+        lda  #%00010000
+        sta  cx16.VERA_ADDR_H       ; enable auto increment by 1, bank 0.
+
+        lda  #1
+        sta  cx16.VERA_CTRL         ; data port 1 is destination
+        stz  cx16.VERA_ADDR_M       ; start at top line
+        stz  cx16.VERA_ADDR_L
+        lda  #%00010000
+        sta  cx16.VERA_ADDR_H       ; enable auto increment by 1, bank 0.
+
+_nextline
+        ldx  #80        ; modified
+-       lda  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA1        ; copy char
+        lda  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA1        ; copy color
+        dex
+        bne  -
+        dec  P8ZP_SCRATCH_B1
+        beq  +
+        stz  cx16.VERA_CTRL         ; data port 0
+        stz  cx16.VERA_ADDR_L
+        inc  cx16.VERA_ADDR_M
+        lda  #1
+        sta  cx16.VERA_CTRL         ; data port 1
+        stz  cx16.VERA_ADDR_L
+        inc  cx16.VERA_ADDR_M
+        bra  _nextline
+
++       lda  #0
+        sta  cx16.VERA_CTRL
+	    plx
+	    rts
+	}}
+}
+
+asmsub  scroll_down() clobbers(A, Y)  {
+	; ---- scroll the whole screen 1 character down
+	;      contents of the top row are unchanged, you should refill/clear this yourself
+	%asm {{
+	    phx
+	    jsr  c64.SCREEN
+	    stx  _nextline+1
+	    dey
+        sty  P8ZP_SCRATCH_B1
+        stz  cx16.VERA_CTRL         ; data port 0 is source
+        dey
+        sty  cx16.VERA_ADDR_M       ; start at line before bottom line
+        stz  cx16.VERA_ADDR_L
+        lda  #%00010000
+        sta  cx16.VERA_ADDR_H       ; enable auto increment by 1, bank 0.
+
+        lda  #1
+        sta  cx16.VERA_CTRL         ; data port 1 is destination
+        iny
+        sty  cx16.VERA_ADDR_M       ; start at bottom line
+        stz  cx16.VERA_ADDR_L
+        lda  #%00010000
+        sta  cx16.VERA_ADDR_H       ; enable auto increment by 1, bank 0.
+
+_nextline
+        ldx  #80        ; modified
+-       lda  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA1        ; copy char
+        lda  cx16.VERA_DATA0
+        sta  cx16.VERA_DATA1        ; copy color
+        dex
+        bne  -
+        dec  P8ZP_SCRATCH_B1
+        beq  +
+        stz  cx16.VERA_CTRL         ; data port 0
+        stz  cx16.VERA_ADDR_L
+        dec  cx16.VERA_ADDR_M
+        lda  #1
+        sta  cx16.VERA_CTRL         ; data port 1
+        stz  cx16.VERA_ADDR_L
+        dec  cx16.VERA_ADDR_M
+        bra  _nextline
+
++       lda  #0
+        sta  cx16.VERA_CTRL
+	    plx
+	    rts
+	}}
+}
+
+romsub $FFD2 = chrout(ubyte char @ A)    ; for consistency. You can also use c64.CHROUT directly ofcourse.
+
+asmsub  print (str text @ AY) clobbers(A,Y)  {
+	; ---- print null terminated string from A/Y
+	; note: the compiler contains an optimization that will replace
+	;       a call to this subroutine with a string argument of just one char,
+	;       by just one call to c64.CHROUT of that single char.
+	%asm {{
+		sta  P8ZP_SCRATCH_B1
+		sty  P8ZP_SCRATCH_REG
+		ldy  #0
+-		lda  (P8ZP_SCRATCH_B1),y
+		beq  +
+		jsr  c64.CHROUT
+		iny
+		bne  -
++		rts
+	}}
+}
+
+asmsub  print_ub0  (ubyte value @ A) clobbers(A,Y)  {
+	; ---- print the ubyte in A in decimal form, with left padding 0s (3 positions total)
+	%asm {{
+		phx
+		jsr  conv.ubyte2decimal
+		pha
+		tya
+		jsr  c64.CHROUT
+		pla
+		jsr  c64.CHROUT
+		txa
+		jsr  c64.CHROUT
+		plx
+		rts
+	}}
+}
+
+asmsub  print_ub  (ubyte value @ A) clobbers(A,Y)  {
+	; ---- print the ubyte in A in decimal form, without left padding 0s
+	%asm {{
+		phx
+		jsr  conv.ubyte2decimal
+_print_byte_digits
+		pha
+		cpy  #'0'
+		beq  +
+		tya
+		jsr  c64.CHROUT
+		pla
+		jsr  c64.CHROUT
+		jmp  _ones
++       pla
+        cmp  #'0'
+        beq  _ones
+        jsr  c64.CHROUT
+_ones   txa
+		jsr  c64.CHROUT
+		plx
+		rts
+	}}
+}
+
+asmsub  print_b  (byte value @ A) clobbers(A,Y)  {
+	; ---- print the byte in A in decimal form, without left padding 0s
+	%asm {{
+		phx
+		pha
+		cmp  #0
+		bpl  +
+		lda  #'-'
+		jsr  c64.CHROUT
++		pla
+		jsr  conv.byte2decimal
+		jmp  print_ub._print_byte_digits
+	}}
+}
+
+asmsub  print_ubhex  (ubyte value @ A, ubyte prefix @ Pc) clobbers(A,Y)  {
+	; ---- print the ubyte in A in hex form (if Carry is set, a radix prefix '$' is printed as well)
+	%asm {{
+		phx
+		bcc  +
+		pha
+		lda  #'$'
+		jsr  c64.CHROUT
+		pla
++		jsr  conv.ubyte2hex
+		jsr  c64.CHROUT
+		tya
+		jsr  c64.CHROUT
+		plx
+		rts
+	}}
+}
+
+asmsub  print_ubbin  (ubyte value @ A, ubyte prefix @ Pc) clobbers(A,Y)  {
+	; ---- print the ubyte in A in binary form (if Carry is set, a radix prefix '%' is printed as well)
+	%asm {{
+		phx
+		sta  P8ZP_SCRATCH_B1
+		bcc  +
+		lda  #'%'
+		jsr  c64.CHROUT
++		ldy  #8
+-		lda  #'0'
+		asl  P8ZP_SCRATCH_B1
+		bcc  +
+		lda  #'1'
++		jsr  c64.CHROUT
+		dey
+		bne  -
+		plx
+		rts
+	}}
+}
+
+asmsub  print_uwbin  (uword value @ AY, ubyte prefix @ Pc) clobbers(A,Y)  {
+	; ---- print the uword in A/Y in binary form (if Carry is set, a radix prefix '%' is printed as well)
+	%asm {{
+		pha
+		tya
+		jsr  print_ubbin
+		pla
+		clc
+		jmp  print_ubbin
+	}}
+}
+
+asmsub  print_uwhex  (uword value @ AY, ubyte prefix @ Pc) clobbers(A,Y)  {
+	; ---- print the uword in A/Y in hexadecimal form (4 digits)
+	;      (if Carry is set, a radix prefix '$' is printed as well)
+	%asm {{
+		pha
+		tya
+		jsr  print_ubhex
+		pla
+		clc
+		jmp  print_ubhex
+	}}
+}
+
+asmsub  print_uw0  (uword value @ AY) clobbers(A,Y)  {
+	; ---- print the uword in A/Y in decimal form, with left padding 0s (5 positions total)
+	%asm {{
+	    phx
+		jsr  conv.uword2decimal
+		ldy  #0
+-		lda  conv.uword2decimal.decTenThousands,y
+        beq  +
+		jsr  c64.CHROUT
+		iny
+		bne  -
++		plx
+		rts
+	}}
+}
+
+asmsub  print_uw  (uword value @ AY) clobbers(A,Y)  {
+	; ---- print the uword in A/Y in decimal form, without left padding 0s
+	%asm {{
+	    phx
+		jsr  conv.uword2decimal
+		plx
+		ldy  #0
+-		lda  conv.uword2decimal.decTenThousands,y
+		beq  _allzero
+		cmp  #'0'
+		bne  _gotdigit
+		iny
+		bne  -
+
+_gotdigit
+		jsr  c64.CHROUT
+		iny
+		lda  conv.uword2decimal.decTenThousands,y
+		bne  _gotdigit
+		rts
+_allzero
+        lda  #'0'
+        jmp  c64.CHROUT
+	}}
+}
+
+asmsub  print_w  (word value @ AY) clobbers(A,Y)  {
+	; ---- print the (signed) word in A/Y in decimal form, without left padding 0's
+	%asm {{
+		cpy  #0
+		bpl  +
+		pha
+		lda  #'-'
+		jsr  c64.CHROUT
+		tya
+		eor  #255
+		tay
+		pla
+		eor  #255
+		clc
+		adc  #1
+		bcc  +
+		iny
++		jmp  print_uw
+	}}
+}
+
+asmsub  input_chars  (uword buffer @ AY) clobbers(A) -> ubyte @ Y  {
+	; ---- Input a string (max. 80 chars) from the keyboard. Returns length in Y. (string is terminated with a 0 byte as well)
+	;      It assumes the keyboard is selected as I/O channel!
+
+	%asm {{
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		ldy  #0				; char counter = 0
+-		jsr  c64.CHRIN
+		cmp  #$0d			; return (ascii 13) pressed?
+		beq  +				; yes, end.
+		sta  (P8ZP_SCRATCH_W1),y	; else store char in buffer
+		iny
+		bne  -
++		lda  #0
+		sta  (P8ZP_SCRATCH_W1),y	; finish string with 0 byte
+		rts
+
+	}}
+}
+
+asmsub  setchr  (ubyte col @X, ubyte row @Y, ubyte character @A) clobbers(A)  {
+	; ---- sets the character in the screen matrix at the given position
+	%asm {{
+            pha
+            txa
+            asl  a
+            stz  cx16.VERA_CTRL
+            stz  cx16.VERA_ADDR_H
+            sta  cx16.VERA_ADDR_L
+            sty  cx16.VERA_ADDR_M
+            pla
+            sta  cx16.VERA_DATA0
+            rts
+	}}
+}
+
+asmsub  getchr  (ubyte col @A, ubyte row @Y) -> ubyte @ A {
+	; ---- get the character in the screen matrix at the given location
+	%asm  {{
+            asl  a
+            stz  cx16.VERA_CTRL
+            stz  cx16.VERA_ADDR_H
+            sta  cx16.VERA_ADDR_L
+            sty  cx16.VERA_ADDR_M
+            lda  cx16.VERA_DATA0
+            rts
+	}}
+}
+
+asmsub  setclr  (ubyte col @X, ubyte row @Y, ubyte color @A) clobbers(A)  {
+	; ---- set the color in A on the screen matrix at the given position
+	;      note: on the CommanderX16 this allows you to set both Fg and Bg colors;
+	;            use the high nybble in A to set the Bg color!
+	%asm {{
+            pha
+            txa
+            asl  a
+            ina
+            stz  cx16.VERA_CTRL
+            stz  cx16.VERA_ADDR_H
+            sta  cx16.VERA_ADDR_L
+            sty  cx16.VERA_ADDR_M
+            pla
+            sta  cx16.VERA_DATA0
+            rts
+	}}
+}
+
+asmsub  getclr  (ubyte col @A, ubyte row @Y) -> ubyte @ A {
+	; ---- get the color in the screen color matrix at the given location
+	%asm  {{
+            asl  a
+            ina
+            stz  cx16.VERA_CTRL
+            stz  cx16.VERA_ADDR_H
+            sta  cx16.VERA_ADDR_L
+            sty  cx16.VERA_ADDR_M
+            lda  cx16.VERA_DATA0
+            rts
+	}}
+}
+
+sub  setcc  (ubyte column, ubyte row, ubyte char, ubyte charcolor)  {
+	; ---- set char+color at the given position on the screen
+	;      note: color handling is the same as on the C64: it only sets the foreground color.
+	;            use setcc2 if you want Cx-16 specific feature of setting both Bg+Fg colors.
+	%asm {{
+            phx
+            lda  column
+            asl  a
+            tax
+            ldy  row
+            lda  charcolor
+            and  #$0f
+            sta  P8ZP_SCRATCH_B1
+            stz  cx16.VERA_CTRL
+            stz  cx16.VERA_ADDR_H
+            stx  cx16.VERA_ADDR_L
+            sty  cx16.VERA_ADDR_M
+            lda  char
+            sta  cx16.VERA_DATA0
+            inx
+            stz  cx16.VERA_ADDR_H
+            stx  cx16.VERA_ADDR_L
+            sty  cx16.VERA_ADDR_M
+            lda  cx16.VERA_DATA0
+            and  #$f0
+            ora  P8ZP_SCRATCH_B1
+            sta  cx16.VERA_DATA0
+            plx
+            rts
+    }}
+}
+
+sub  setcc2  (ubyte column, ubyte row, ubyte char, ubyte colors)  {
+	; ---- set char+color at the given position on the screen
+	;      note: on the CommanderX16 this allows you to set both Fg and Bg colors;
+	;            use the high nybble in A to set the Bg color!
+	%asm {{
+            phx
+            lda  column
+            asl  a
+            tax
+            ldy  row
+            stz  cx16.VERA_CTRL
+            stz  cx16.VERA_ADDR_H
+            stx  cx16.VERA_ADDR_L
+            sty  cx16.VERA_ADDR_M
+            lda  char
+            sta  cx16.VERA_DATA0
+            inx
+            stz  cx16.VERA_ADDR_H
+            stx  cx16.VERA_ADDR_L
+            sty  cx16.VERA_ADDR_M
+            lda  colors
+            sta  cx16.VERA_DATA0
+            plx
+            rts
+    }}
+}
+
+asmsub  plot  (ubyte col @ Y, ubyte row @ A) clobbers(A) {
+	; ---- safe wrapper around PLOT kernal routine, to save the X register.
+	%asm  {{
+		phx
+		tax
+		clc
+		jsr  c64.PLOT
+		plx
+		rts
+	}}
+}
+
+asmsub width() clobbers(X,Y) -> ubyte @A {
+    ; -- returns the text screen width (number of columns)
+    %asm {{
+        jsr  c64.SCREEN
+        txa
+        rts
+    }}
+}
+
+asmsub height() clobbers(X, Y) -> ubyte @A {
+    ; -- returns the text screen height (number of rows)
+    %asm {{
+        jsr  c64.SCREEN
+        tya
+        rts
+    }}
+}
+
+}

compiler/res/prog8lib/cx16logo.p8

@@ -0,0 +1,31 @@
+; routine to draw the Commander X16's log in petscii.
+
+%import textio
+
+cx16logo {
+    sub logo_at(ubyte column, ubyte row) {
+        uword strptr
+        for strptr in logo_lines {
+            txt.plot(column, row)
+            txt.print(strptr)
+            row++
+        }
+    }
+
+    sub logo() {
+        uword strptr
+        for strptr in logo_lines
+            txt.print(strptr)
+        txt.nl()
+    }
+
+    str[] logo_lines = [
+            "\uf10d\uf11a\uf139\uf11b     \uf11a\uf13a\uf11b\n",
+            "\uf10b\uf11a▎\uf139\uf11b   \uf11a\uf13a\uf130\uf11b\n",
+            "\uf10f\uf11a▌ \uf139\uf11b \uf11a\uf13a \uf11b▌\n",
+            "\uf102 \uf132\uf11a▖\uf11b \uf11a▗\uf11b\uf132\n",
+            "\uf10e ▂\uf11a▘\uf11b \uf11a▝\uf11b▂\n",
+            "\uf104 \uf11a \uf11b\uf13a\uf11b \uf139\uf11a \uf11b\n",
+            "\uf101\uf130\uf13a   \uf139▎\uf100"
+        ]
+}

compiler/res/prog8lib/diskio.p8

@@ -0,0 +1,450 @@
+; C64 and Cx16 disk drive I/O routines.
+;
+; Written by Irmen de Jong (irmen@razorvine.net) - license: GNU GPL 3.0
+
+%import textio
+%import string
+%import syslib
+
+diskio {
+
+    sub directory(ubyte drivenumber) -> ubyte {
+        ; -- Prints the directory contents of disk drive 8-11 to the screen. Returns success.
+
+        c64.SETNAM(1, "$")
+        c64.SETLFS(13, drivenumber, 0)
+        void c64.OPEN()          ; open 13,8,0,"$"
+        if_cs
+            goto io_error
+        void c64.CHKIN(13)        ; use #13 as input channel
+        if_cs
+            goto io_error
+
+        repeat 4 {
+            void c64.CHRIN()     ; skip the 4 prologue bytes
+        }
+
+        ; while not key pressed / EOF encountered, read data.
+        ubyte status = c64.READST()
+        while not status {
+            ubyte low = c64.CHRIN()
+            ubyte high = c64.CHRIN()
+            txt.print_uw(mkword(high, low))
+            txt.spc()
+            ubyte @zp char
+            repeat {
+                char = c64.CHRIN()
+                if char==0
+                    break
+                txt.chrout(char)
+            }
+            txt.nl()
+            void c64.CHRIN()     ; skip 2 bytes
+            void c64.CHRIN()
+            status = c64.READST()
+            if c64.STOP2()
+                break
+        }
+
+io_error:
+        status = c64.READST()
+        c64.CLRCHN()        ; restore default i/o devices
+        c64.CLOSE(13)
+
+        if status and status & $40 == 0 {            ; bit 6=end of file
+            txt.print("\ni/o error, status: ")
+            txt.print_ub(status)
+            txt.nl()
+            return false
+        }
+
+        return true
+    }
+
+
+    ; internal variables for the iterative file lister / loader
+    ubyte list_skip_disk_name
+    uword list_pattern
+    uword list_blocks
+    ubyte iteration_in_progress = false
+    ubyte @zp first_byte
+    ubyte have_first_byte
+    str   list_filename = "?" * 32
+
+
+    ; ----- get a list of files (uses iteration functions internally) -----
+
+    sub list_files(ubyte drivenumber, uword pattern_ptr, uword name_ptrs, ubyte max_names) -> ubyte {
+        ; -- fill the array 'name_ptrs' with (pointers to) the names of the files requested.
+        uword names_buffer = memory("filenames", 512)
+        uword buffer_start = names_buffer
+        ubyte files_found = 0
+        if lf_start_list(drivenumber, pattern_ptr) {
+            while lf_next_entry() {
+                @(name_ptrs) = lsb(names_buffer)
+                name_ptrs++
+                @(name_ptrs) = msb(names_buffer)
+                name_ptrs++
+                names_buffer += string.copy(diskio.list_filename, names_buffer) + 1
+                files_found++
+                if names_buffer - buffer_start > 512-18
+                    break
+                if files_found == max_names
+                    break
+            }
+            lf_end_list()
+        }
+        return files_found
+    }
+
+    ; ----- iterative file lister functions (uses io channel 12) -----
+
+    sub lf_start_list(ubyte drivenumber, uword pattern_ptr) -> ubyte {
+        ; -- start an iterative file listing with optional pattern matching.
+        ;    note: only a single iteration loop can be active at a time!
+        lf_end_list()
+        list_pattern = pattern_ptr
+        list_skip_disk_name = true
+        iteration_in_progress = true
+
+        c64.SETNAM(1, "$")
+        c64.SETLFS(12, drivenumber, 0)
+        void c64.OPEN()          ; open 12,8,0,"$"
+        if_cs
+            goto io_error
+        void c64.CHKIN(12)        ; use #12 as input channel
+        if_cs
+            goto io_error
+
+        repeat 4 {
+            void c64.CHRIN()     ; skip the 4 prologue bytes
+        }
+
+        if c64.READST()==0
+            return true
+
+io_error:
+        lf_end_list()
+        return false
+    }
+
+    sub lf_next_entry() -> ubyte {
+        ; -- retrieve the next entry from an iterative file listing session.
+        ;    results will be found in list_blocks and list_filename.
+        ;    if it returns false though, there are no more entries (or an error occurred).
+
+        if not iteration_in_progress
+            return false
+
+        repeat {
+            void c64.CHKIN(12)        ; use #12 as input channel again
+
+            uword nameptr = &list_filename
+            ubyte blocks_lsb = c64.CHRIN()
+            ubyte blocks_msb = c64.CHRIN()
+
+            if c64.READST()
+                goto close_end
+
+            list_blocks = mkword(blocks_msb, blocks_lsb)
+
+            ; read until the filename starts after the first "
+            while c64.CHRIN()!='\"'  {
+                if c64.READST()
+                    goto close_end
+            }
+
+            ; read the filename
+            repeat {
+                ubyte char = c64.CHRIN()
+                if char==0
+                    break
+                if char=='\"'
+                    break
+                @(nameptr) = char
+                nameptr++
+            }
+
+            @(nameptr) = 0
+
+            while c64.CHRIN() {
+                ; read the rest of the entry until the end
+            }
+
+            void c64.CHRIN()     ; skip 2 bytes
+            void c64.CHRIN()
+
+            if not list_skip_disk_name {
+                if not list_pattern
+                    return true
+                if prog8_lib.pattern_match(list_filename, list_pattern)
+                    return true
+            }
+            list_skip_disk_name = false
+        }
+
+close_end:
+        lf_end_list()
+        return false
+    }
+
+    sub lf_end_list() {
+        ; -- end an iterative file listing session (close channels).
+        if iteration_in_progress {
+            c64.CLRCHN()
+            c64.CLOSE(12)
+            iteration_in_progress = false
+        }
+    }
+
+
+    ; ----- iterative file loader functions (uses io channel 11) -----
+
+    sub f_open(ubyte drivenumber, uword filenameptr) -> ubyte {
+        ; -- open a file for iterative reading with f_read
+        ;    note: only a single iteration loop can be active at a time!
+        f_close()
+
+        c64.SETNAM(string.length(filenameptr), filenameptr)
+        c64.SETLFS(11, drivenumber, 0)
+        void c64.OPEN()          ; open 11,8,0,"filename"
+        if_cc {
+            iteration_in_progress = true
+            have_first_byte = false
+            void c64.CHKIN(11)        ; use #11 as input channel
+            if_cc {
+                first_byte = c64.CHRIN()   ; read first byte to test for file not found
+                if not c64.READST() {
+                    have_first_byte = true
+                    return true
+                }
+            }
+        }
+        f_close()
+        return false
+    }
+
+    sub f_read(uword bufferpointer, uword num_bytes) -> uword {
+        ; -- read from the currently open file, up to the given number of bytes.
+        ;    returns the actual number of bytes read.  (checks for End-of-file and error conditions)
+        if not iteration_in_progress or not num_bytes
+            return 0
+
+        list_blocks = 0     ; we reuse this variable for the total number of bytes read
+        if have_first_byte {
+            have_first_byte=false
+            @(bufferpointer) = first_byte
+            bufferpointer++
+            list_blocks++
+            num_bytes--
+        }
+
+        void c64.CHKIN(11)        ; use #11 as input channel again
+        %asm {{
+            lda  bufferpointer
+            sta  _in_buffer+1
+            lda  bufferpointer+1
+            sta  _in_buffer+2
+        }}
+        repeat num_bytes {
+            %asm {{
+                jsr  c64.CHRIN
+                sta  cx16.r5
+_in_buffer      sta  $ffff
+                inc  _in_buffer+1
+                bne  +
+                inc  _in_buffer+2
++               inc  list_blocks
+                bne  +
+                inc  list_blocks+1
++
+            }}
+
+            if cx16.r5==$0d {   ; chance on I/o error status?
+                first_byte = c64.READST()
+                if first_byte & $40
+                    f_close()       ; end of file, close it
+                if first_byte
+                    return list_blocks
+            }
+        }
+        return list_blocks
+    }
+
+    sub f_read_all(uword bufferpointer) -> uword {
+        ; -- read the full contents of the file, returns number of bytes read.
+        if not iteration_in_progress
+            return 0
+
+        list_blocks = 0     ; we reuse this variable for the total number of bytes read
+        if have_first_byte {
+            have_first_byte=false
+            @(bufferpointer) = first_byte
+            bufferpointer++
+            list_blocks++
+        }
+
+        while not c64.READST() {
+            list_blocks += f_read(bufferpointer, 256)
+            bufferpointer += 256
+        }
+        return list_blocks
+    }
+
+    asmsub f_readline(uword bufptr @AY) clobbers(X) -> ubyte @Y {
+        ; Routine to read text lines from a text file. Lines must be less than 255 characters.
+        ; Reads characters from the input file UNTIL a newline or return character (or EOF).
+        ; The line read will be 0-terminated in the buffer (and not contain the end of line character).
+        ; The length of the line is returned in Y. Note that an empty line is okay and is length 0!
+        ; I/O error status should be checked by the caller itself via READST() routine.
+        %asm {{
+            sta  P8ZP_SCRATCH_W1
+            sty  P8ZP_SCRATCH_W1+1
+            ldx  #11
+            jsr  c64.CHKIN              ; use channel 11 again for input
+            ldy  #0
+            lda  have_first_byte
+            beq  _loop
+            lda  #0
+            sta  have_first_byte
+            lda  first_byte
+            sta  (P8ZP_SCRATCH_W1),y
+            iny
+_loop       jsr  c64.CHRIN
+            sta  (P8ZP_SCRATCH_W1),y
+            beq  _end
+            iny
+            cmp  #$0a
+            beq  _line_end
+            cmp  #$0d
+            bne  _loop
+_line_end   dey     ; get rid of the trailing end-of-line char
+            lda  #0
+            sta  (P8ZP_SCRATCH_W1),y
+_end        rts
+        }}
+    }
+
+
+    sub f_close() {
+        ; -- end an iterative file loading session (close channels).
+        if iteration_in_progress {
+            c64.CLRCHN()
+            c64.CLOSE(11)
+            iteration_in_progress = false
+        }
+    }
+
+
+    sub status(ubyte drivenumber) -> uword {
+        ; -- retrieve the disk drive's current status message
+        uword messageptr = &filename
+        c64.SETNAM(0, filename)
+        c64.SETLFS(15, drivenumber, 15)
+        void c64.OPEN()          ; open 15,8,15
+        if_cs
+            goto io_error
+        void c64.CHKIN(15)        ; use #15 as input channel
+        if_cs
+            goto io_error
+
+        while not c64.READST() {
+            @(messageptr) = c64.CHRIN()
+            messageptr++
+        }
+
+io_error:
+        @(messageptr) = 0
+        c64.CLRCHN()        ; restore default i/o devices
+        c64.CLOSE(15)
+        return filename
+    }
+
+
+    sub save(ubyte drivenumber, uword filenameptr, uword address, uword size) -> ubyte {
+        c64.SETNAM(string.length(filenameptr), filenameptr)
+        c64.SETLFS(1, drivenumber, 0)
+        uword end_address = address + size
+
+        %asm {{
+            lda  address
+            sta  P8ZP_SCRATCH_W1
+            lda  address+1
+            sta  P8ZP_SCRATCH_W1+1
+            stx  P8ZP_SCRATCH_REG
+            lda  #<P8ZP_SCRATCH_W1
+            ldx  end_address
+            ldy  end_address+1
+            jsr  c64.SAVE
+            php
+            ldx  P8ZP_SCRATCH_REG
+            plp
+        }}
+
+        first_byte = 0      ; result var reuse
+        if_cc
+            first_byte = c64.READST()==0
+
+        c64.CLRCHN()
+        c64.CLOSE(1)
+
+        return first_byte
+    }
+
+    sub load(ubyte drivenumber, uword filenameptr, uword address_override) -> uword {
+        c64.SETNAM(string.length(filenameptr), filenameptr)
+        ubyte secondary = 1
+        uword end_of_load = 0
+        if address_override
+            secondary = 0
+        c64.SETLFS(1, drivenumber, secondary)
+        %asm {{
+            stx  P8ZP_SCRATCH_REG
+            lda  #0
+            ldx  address_override
+            ldy  address_override+1
+            jsr  c64.LOAD
+            bcs  +
+            stx  end_of_load
+            sty  end_of_load+1
++           ldx  P8ZP_SCRATCH_REG
+        }}
+
+        c64.CLRCHN()
+        c64.CLOSE(1)
+
+        if end_of_load
+            return end_of_load - address_override
+
+        return 0
+    }
+
+
+    str filename = "0:??????????????????????????????????????"
+
+    sub delete(ubyte drivenumber, uword filenameptr) {
+        ; -- delete a file on the drive
+        filename[0] = 's'
+        filename[1] = ':'
+        ubyte flen = string.copy(filenameptr, &filename+2)
+        c64.SETNAM(flen+2, filename)
+        c64.SETLFS(1, drivenumber, 15)
+        void c64.OPEN()
+        c64.CLRCHN()
+        c64.CLOSE(1)
+    }
+
+    sub rename(ubyte drivenumber, uword oldfileptr, uword newfileptr) {
+        ; -- rename a file on the drive
+        filename[0] = 'r'
+        filename[1] = ':'
+        ubyte flen_new = string.copy(newfileptr, &filename+2)
+        filename[flen_new+2] = '='
+        ubyte flen_old = string.copy(oldfileptr, &filename+3+flen_new)
+        c64.SETNAM(3+flen_new+flen_old, filename)
+        c64.SETLFS(1, drivenumber, 15)
+        void c64.OPEN()
+        c64.CLRCHN()
+        c64.CLOSE(1)
+    }
+}

compiler/res/prog8lib/math.p8

@@ -1,10 +1,6 @@
-; Prog8 internal Math library routines - always included by the compiler
+; Internal Math library routines - always included by the compiler
 ;
 ; Written by Irmen de Jong (irmen@razorvine.net) - license: GNU GPL 3.0
-;
-; indent format: TABS, size=8
-
-%import c64lib
 
 math {
 	%asminclude "library:math.asm", ""

compiler/res/prog8lib/prog8_lib.p8

@@ -0,0 +1,84 @@
+; Internal library routines - always included by the compiler
+;
+; Written by Irmen de Jong (irmen@razorvine.net) - license: GNU GPL 3.0
+
+prog8_lib {
+	%asminclude "library:prog8_lib.asm", ""
+	%asminclude "library:prog8_funcs.asm", ""
+
+	uword @zp retval_interm_uw      ; to store intermediary expression results for return values (hopefully allocated on ZP to reduce code size)
+	word @zp retval_interm_w        ; to store intermediary expression results for return values (hopefully allocated on ZP to reduce code size)
+	ubyte @zp retval_interm_ub      ; to store intermediary expression results for return values (hopefully allocated on ZP to reduce code size)
+	byte @zp retval_interm_b        ; to store intermediary expression results for return values (hopefully allocated on ZP to reduce code size)
+
+	asmsub pattern_match(str string @AY, str pattern @R0) clobbers(Y) -> ubyte @A {
+		%asm {{
+; pattern matching of a string.
+; Input:  cx16.r0:  A NUL-terminated, <255-length pattern
+;              AY:  A NUL-terminated, <255-length string
+;
+; Output: A = 1 if the string matches the pattern, A = 0 if not.
+;
+; Notes:  Clobbers A, X, Y. Each * in the pattern uses 4 bytes of stack.
+;
+; see http://6502.org/source/strings/patmatch.htm
+
+str = P8ZP_SCRATCH_W1
+
+	stx  P8ZP_SCRATCH_REG
+	sta  str
+	sty  str+1
+	lda  cx16.r0
+	sta  modify_pattern1+1
+	sta  modify_pattern2+1
+	lda  cx16.r0+1
+	sta  modify_pattern1+2
+	sta  modify_pattern2+2
+	jsr  _match
+	lda  #0
+	adc  #0
+	ldx  P8ZP_SCRATCH_REG
+	rts
+
+
+_match
+	ldx #$00        ; x is an index in the pattern
+	ldy #$ff        ; y is an index in the string
+modify_pattern1
+next    lda $ffff,x   ; look at next pattern character    MODIFIED
+	cmp #'*'     ; is it a star?
+	beq star        ; yes, do the complicated stuff
+	iny             ; no, let's look at the string
+	cmp #'?'     ; is the pattern caracter a ques?
+	bne reg         ; no, it's a regular character
+	lda (str),y     ; yes, so it will match anything
+	beq fail        ;  except the end of string
+reg     cmp (str),y     ; are both characters the same?
+	bne fail        ; no, so no match
+	inx             ; yes, keep checking
+	cmp #0          ; are we at end of string?
+	bne next        ; not yet, loop
+found   rts             ; success, return with c=1
+
+star    inx             ; skip star in pattern
+modify_pattern2
+	cmp $ffff,x   	; string of stars equals one star	MODIFIED
+	beq star        ;  so skip them also
+stloop  txa             ; we first try to match with * = ""
+	pha             ;  and grow it by 1 character every
+	tya             ;  time we loop
+	pha             ; save x and y on stack
+	jsr next        ; recursive call
+	pla             ; restore x and y
+	tay
+	pla
+	tax
+	bcs found       ; we found a match, return with c=1
+	iny             ; no match yet, try to grow * string
+	lda (str),y     ; are we at the end of string?
+	bne stloop      ; not yet, add a character
+fail    clc             ; yes, no match found, return with c=0
+	rts
+		}}
+	}
+}

compiler/res/prog8lib/prog8lib.p8

@@ -1,11 +0,0 @@
-; Prog8 internal library routines - always included by the compiler
-;
-; Written by Irmen de Jong (irmen@razorvine.net) - license: GNU GPL 3.0
-;
-; indent format: TABS, size=8
-
-%import c64lib
-
-prog8_lib {
-	%asminclude "library:prog8lib.asm", ""
-}

compiler/res/prog8lib/string.p8

@@ -0,0 +1,235 @@
+; 0-terminated string manipulation routines.
+;
+; Written by Irmen de Jong (irmen@razorvine.net) - license: GNU GPL 3.0
+
+
+string {
+
+    asmsub length(uword string @AY) clobbers(A) -> ubyte @Y {
+        ; Returns the number of bytes in the string.
+        ; This value is determined during runtime and counts upto the first terminating 0 byte in the string,
+        ; regardless of the size of the string during compilation time. Don’t confuse this with len and sizeof!
+
+        %asm {{
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		ldy  #0
+-		lda  (P8ZP_SCRATCH_W1),y
+		beq  +
+		iny
+		bne  -
++		rts
+        }}
+    }
+
+    asmsub left(uword source @R0, ubyte length @A, uword target @R1) clobbers(A, Y) {
+        ; Copies the left side of the source string of the given length to target string.
+        ; It is assumed the target string buffer is large enough to contain the result.
+        ; Also, you have to make sure yourself that length is smaller or equal to the length of the source string.
+        ; Modifies in-place, doesn’t return a value (so can’t be used in an expression).
+        %asm {{
+                ; need to copy the the cx16 virtual registers to zeropage to be compatible with C64...
+		ldy  cx16.r0
+		sty  P8ZP_SCRATCH_W1
+		ldy  cx16.r0+1
+		sty  P8ZP_SCRATCH_W1+1
+		ldy  cx16.r1
+		sty  P8ZP_SCRATCH_W2
+		ldy  cx16.r1+1
+		sty  P8ZP_SCRATCH_W2+1
+		tay
+		lda  #0
+		sta  (P8ZP_SCRATCH_W2),y
+		cpy  #0
+		bne  _loop
+		rts
+_loop		dey
+		lda  (P8ZP_SCRATCH_W1),y
+		sta  (P8ZP_SCRATCH_W2),y
+		cpy  #0
+		bne  _loop
++		rts
+        }}
+;                asmgen.out("  jsr  prog8_lib.func_leftstr")
+    }
+
+    asmsub right(uword source @R0, ubyte length @A, uword target @R1) clobbers(A,Y) {
+        ; Copies the right side of the source string of the given length to target string.
+        ; It is assumed the target string buffer is large enough to contain the result.
+        ; Also, you have to make sure yourself that length is smaller or equal to the length of the source string.
+        ; Modifies in-place, doesn’t return a value (so can’t be used in an expression).
+        %asm {{
+                ; need to copy the the cx16 virtual registers to zeropage to be compatible with C64...
+                sta  P8ZP_SCRATCH_B1
+                lda  cx16.r0
+                ldy  cx16.r0+1
+                jsr  string.length
+                tya
+                sec
+                sbc  P8ZP_SCRATCH_B1
+                clc
+                adc  cx16.r0
+		sta  P8ZP_SCRATCH_W1
+		lda  cx16.r0+1
+		adc  #0
+		sta  P8ZP_SCRATCH_W1+1
+		ldy  cx16.r1
+		sty  P8ZP_SCRATCH_W2
+		ldy  cx16.r1+1
+		sty  P8ZP_SCRATCH_W2+1
+		ldy  P8ZP_SCRATCH_B1
+		lda  #0
+		sta  (P8ZP_SCRATCH_W2),y
+		cpy  #0
+		bne  _loop
+		rts
+_loop		dey
+		lda  (P8ZP_SCRATCH_W1),y
+		sta  (P8ZP_SCRATCH_W2),y
+		cpy  #0
+		bne  _loop
++		rts
+        }}
+    }
+
+    asmsub slice(uword source @R0, ubyte start @A, ubyte length @Y, uword target @R1) clobbers(A, Y) {
+        ; Copies a segment from the source string, starting at the given index,
+        ;  and of the given length to target string.
+        ; It is assumed the target string buffer is large enough to contain the result.
+        ; Also, you have to make sure yourself that start and length are within bounds of the strings.
+        ; Modifies in-place, doesn’t return a value (so can’t be used in an expression).
+        %asm {{
+                ; need to copy the the cx16 virtual registers to zeropage to be compatible with C64...
+		; substr(source, target, start, length)
+		sta  P8ZP_SCRATCH_B1
+		lda  cx16.r0
+		sta  P8ZP_SCRATCH_W1
+		lda  cx16.r0+1
+		sta  P8ZP_SCRATCH_W1+1
+		lda  cx16.r1
+		sta  P8ZP_SCRATCH_W2
+		lda  cx16.r1+1
+		sta  P8ZP_SCRATCH_W2+1
+
+		; adjust src location
+		clc
+		lda  P8ZP_SCRATCH_W1
+		adc  P8ZP_SCRATCH_B1
+		sta  P8ZP_SCRATCH_W1
+		bcc  +
+		inc  P8ZP_SCRATCH_W1+1
++		lda  #0
+		sta  (P8ZP_SCRATCH_W2),y
+		beq  _startloop
+-		lda  (P8ZP_SCRATCH_W1),y
+		sta  (P8ZP_SCRATCH_W2),y
+_startloop	dey
+		cpy  #$ff
+		bne  -
+		rts
+        }}
+    }
+
+    asmsub find(uword string @R0, ubyte character @A) -> uword @AY {
+        ; Locates the first position of the given character in the string,
+        ;  returns the string starting with this character or $0000 if the character is not found.
+        %asm {{
+                ; need to copy the the cx16 virtual registers to zeropage to be compatible with C64...
+                sta  P8ZP_SCRATCH_B1
+		lda  cx16.r0
+		ldy  cx16.r0+1
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		ldy  #0
+-		lda  (P8ZP_SCRATCH_W1),y
+		beq  _notfound
+		cmp  P8ZP_SCRATCH_B1
+		beq  _found
+		iny
+		bne  -
+_notfound	lda  #0
+		ldy  #0
+		rts
+_found		sty  P8ZP_SCRATCH_B1
+		ldy  P8ZP_SCRATCH_W1+1
+		lda  P8ZP_SCRATCH_W1
+		clc
+		adc  P8ZP_SCRATCH_B1
+		bcc  +
+		iny
++		rts
+
+        }}
+    }
+
+    asmsub copy(uword source @R0, uword target @AY) clobbers(A) -> ubyte @Y {
+        ; Copy a string to another, overwriting that one.
+        ; Returns the length of the string that was copied.
+        ; Often you don’t have to call this explicitly and can just write string1 = string2
+        ; but this function is useful if you’re dealing with addresses for instance.
+        %asm {{
+		sta  P8ZP_SCRATCH_W1
+		sty  P8ZP_SCRATCH_W1+1
+		lda  cx16.r0
+		ldy  cx16.r0+1
+		jmp  prog8_lib.strcpy
+        }}
+    }
+
+    asmsub compare(uword string1 @R0, uword string2 @AY) clobbers(Y) -> byte @A {
+        ; Compares two strings for sorting.
+        ; Returns -1 (255), 0 or 1 depeding on wether string1 sorts before, equal or after string2.
+        ; Note that you can also directly compare strings and string values with eachother using
+        ; comparison operators ==, < etcetera (it will use strcmp for you under water automatically).
+        %asm {{
+		sta  P8ZP_SCRATCH_W2
+		sty  P8ZP_SCRATCH_W2+1
+		lda  cx16.r0
+		ldy  cx16.r0+1
+		jmp  prog8_lib.strcmp_mem
+        }}
+    }
+
+    asmsub lower(uword st @AY) {
+        ; Lowercases the petscii string in-place.
+        ; (for efficiency, non-letter characters > 128 will also not be left intact,
+        ;  but regular text doesn't usually contain those characters anyway.)
+        %asm {{
+            sta  P8ZP_SCRATCH_W1
+            sty  P8ZP_SCRATCH_W1+1
+            ldy  #0
+-           lda  (P8ZP_SCRATCH_W1),y
+            beq  _done
+            and  #$7f
+            cmp  #97
+            bcc  +
+            cmp  #123
+            bcs  +
+            and  #%11011111
++           sta  (P8ZP_SCRATCH_W1),y
+            iny
+            bne  -
+_done       rts
+        }}
+    }
+
+    asmsub upper(uword st @AY) {
+        ; Uppercases the petscii string in-place.
+        %asm {{
+            sta  P8ZP_SCRATCH_W1
+            sty  P8ZP_SCRATCH_W1+1
+            ldy  #0
+-           lda  (P8ZP_SCRATCH_W1),y
+            beq  _done
+            cmp  #65
+            bcc  +
+            cmp  #91
+            bcs  +
+            ora  #%00100000
++           sta  (P8ZP_SCRATCH_W1),y
+            iny
+            bne  -
+_done       rts
+        }}
+    }
+}

compiler/res/prog8lib/test_stack.p8

@@ -0,0 +1,50 @@
+; utility debug code to print the X (evalstack) and SP (cpu stack) registers.
+
+%import textio
+
+test_stack {
+
+    asmsub test() {
+        %asm {{
+	stx  _saveX
+	lda  #13
+	jsr  txt.chrout
+	lda  #'-'
+	ldy  #12
+-	jsr  txt.chrout
+	dey
+	bne  -
+	lda  #13
+	jsr  txt.chrout
+	lda  #'x'
+	jsr  txt.chrout
+	lda  #'='
+	jsr  txt.chrout
+	lda  _saveX
+	jsr  txt.print_ub
+	lda  #' '
+	jsr  txt.chrout
+	lda  #'s'
+	jsr  txt.chrout
+	lda  #'p'
+	jsr  txt.chrout
+	lda  #'='
+	jsr  txt.chrout
+	tsx
+	txa
+	jsr  txt.print_ub
+	lda  #13
+	jsr  txt.chrout
+	lda  #'-'
+	ldy  #12
+-	jsr  txt.chrout
+	dey
+	bne  -
+	lda  #13
+	jsr  txt.chrout
+	ldx  _saveX
+	rts
+_saveX	.byte 0
+        }}
+    }
+}

compiler/res/version.txt

@@ -1 +1 @@
-2.2
+6.2

compiler/src/prog8/CompilerMain.kt

@@ -1,15 +1,15 @@
 package prog8
 
-import kotlinx.cli.*
+import kotlinx.cli.ArgParser
+import kotlinx.cli.ArgType
+import kotlinx.cli.default
+import kotlinx.cli.multiple
 import prog8.ast.base.AstException
 import prog8.compiler.CompilationResult
 import prog8.compiler.compileProgram
-import prog8.compiler.target.CompilationTarget
-import prog8.compiler.target.c64.C64MachineDefinition
-import prog8.compiler.target.c64.Petscii
-import prog8.compiler.target.c64.codegen.AsmGen
+import prog8.compiler.target.C64Target
+import prog8.compiler.target.Cx16Target
 import prog8.parser.ParsingFailedError
-import java.io.IOException
 import java.nio.file.FileSystems
 import java.nio.file.Path
 import java.nio.file.StandardWatchEventKinds
@@ -34,73 +34,68 @@ fun pathFrom(stringPath: String, vararg rest: String): Path  = FileSystems.getDe
 
 
 private fun compileMain(args: Array<String>) {
-    val cli = CommandLineInterface("prog8compiler")
-    val startEmulator by cli.flagArgument("-emu", "auto-start the Vice C-64 emulator after successful compilation")
-    val outputDir by cli.flagValueArgument("-out", "directory", "directory for output files instead of current directory", ".")
-    val dontWriteAssembly by cli.flagArgument("-noasm", "don't create assembly code")
-    val dontOptimize by cli.flagArgument("-noopt", "don't perform any optimizations")
-    val watchMode by cli.flagArgument("-watch", "continuous compilation mode (watches for file changes), greatly increases compilation speed")
-    val compilationTarget by cli.flagValueArgument("-target", "compilertarget", "target output of the compiler, currently only 'c64' (C64 6502 assembly) available", "c64")
-    val moduleFiles by cli.positionalArgumentsList("modules", "main module file(s) to compile", minArgs = 1)
+    val cli = ArgParser("prog8compiler", prefixStyle = ArgParser.OptionPrefixStyle.JVM)
+    val startEmulator by cli.option(ArgType.Boolean, fullName = "emu", description = "auto-start emulator after successful compilation")
+    val outputDir by cli.option(ArgType.String, fullName = "out", description = "directory for output files instead of current directory").default(".")
+    val dontWriteAssembly by cli.option(ArgType.Boolean, fullName = "noasm", description="don't create assembly code")
+    val dontOptimize by cli.option(ArgType.Boolean, fullName = "noopt", description = "don't perform any optimizations")
+    val watchMode by cli.option(ArgType.Boolean, fullName = "watch", description = "continuous compilation mode (watches for file changes), greatly increases compilation speed")
+    val slowCodegenWarnings by cli.option(ArgType.Boolean, fullName = "slowwarn", description="show debug warnings about slow/problematic assembly code generation")
+    val compilationTarget by cli.option(ArgType.String, fullName = "target", description = "target output of the compiler, currently '${C64Target.name}' and '${Cx16Target.name}' available").default(C64Target.name)
+    val moduleFiles by cli.argument(ArgType.String, fullName = "modules", description = "main module file(s) to compile").multiple(999)
 
     try {
         cli.parse(args)
-    } catch (e: Exception) {
+    } catch (e: IllegalStateException) {
+        System.err.println(e.message)
         exitProcess(1)
     }
 
-    when(compilationTarget) {
-        "c64" -> {
-            with(CompilationTarget) {
-                name = "c64"
-                machine = C64MachineDefinition
-                encodeString = { str, altEncoding ->
-                    if(altEncoding) Petscii.encodeScreencode(str, true) else Petscii.encodePetscii(str, true)
-                }
-                decodeString = { bytes, altEncoding ->
-                    if(altEncoding) Petscii.decodeScreencode(bytes, true) else Petscii.decodePetscii(bytes, true)
-                }
-                asmGenerator = ::AsmGen
-            }
-        }
-        else -> {
-            System.err.println("invalid compilation target")
-            exitProcess(1)
-        }
-    }
-
     val outputPath = pathFrom(outputDir)
     if(!outputPath.toFile().isDirectory) {
         System.err.println("Output path doesn't exist")
         exitProcess(1)
     }
 
-    if(watchMode && moduleFiles.size<=1) {
+    if(watchMode==true) {
         val watchservice = FileSystems.getDefault().newWatchService()
+        val allImportedFiles = mutableSetOf<Path>()
 
         while(true) {
-            val filepath = pathFrom(moduleFiles.single()).normalize()
-            println("Continuous watch mode active. Main module: $filepath")
+            println("Continuous watch mode active. Modules: $moduleFiles")
+            val results = mutableListOf<CompilationResult>()
+            for(filepathRaw in moduleFiles) {
+                val filepath = pathFrom(filepathRaw).normalize()
+                val compilationResult = compileProgram(filepath, dontOptimize!=true, dontWriteAssembly!=true, slowCodegenWarnings==true, compilationTarget, outputPath)
+                results.add(compilationResult)
+            }
 
-            try {
-                val compilationResult = compileProgram(filepath, !dontOptimize, !dontWriteAssembly, outputDir=outputPath)
-                println("Imported files (now watching:)")
-                for (importedFile in compilationResult.importedFiles) {
-                    print("  ")
-                    println(importedFile)
-                    importedFile.parent.register(watchservice, StandardWatchEventKinds.ENTRY_MODIFY)
-                }
-                println("[${LocalDateTime.now().withNano(0)}]  Waiting for file changes.")
+            val allNewlyImportedFiles = results.flatMap { it.importedFiles }
+            allImportedFiles.addAll(allNewlyImportedFiles)
+
+            println("Imported files (now watching:)")
+            for (importedFile in allImportedFiles) {
+                print("  ")
+                println(importedFile)
+                val watchDir = importedFile.parent ?: Path.of(".")
+                watchDir.register(watchservice, StandardWatchEventKinds.ENTRY_MODIFY)
+            }
+            println("[${LocalDateTime.now().withNano(0)}]  Waiting for file changes.")
+
+            var recompile=false
+            while(!recompile) {
                 val event = watchservice.take()
-                for(changed in event.pollEvents()) {
+                for (changed in event.pollEvents()) {
                     val changedPath = changed.context() as Path
-                    println("  change detected: $changedPath")
+                    if(allImportedFiles.any { it.fileName == changedPath.fileName }) {
+                        println("  change detected: $changedPath")
+                        recompile = true
+                    }
                 }
                 event.reset()
-                println("\u001b[H\u001b[2J")      // clear the screen
-            } catch (x: Exception) {
-                throw x
             }
+
+            println("\u001b[H\u001b[2J")      // clear the screen
         }
 
     } else {
@@ -108,7 +103,7 @@ private fun compileMain(args: Array<String>) {
             val filepath = pathFrom(filepathRaw).normalize()
             val compilationResult: CompilationResult
             try {
-                compilationResult = compileProgram(filepath, !dontOptimize, !dontWriteAssembly, outputDir=outputPath)
+                compilationResult = compileProgram(filepath, dontOptimize!=true, dontWriteAssembly!=true, slowCodegenWarnings==true, compilationTarget, outputPath)
                 if(!compilationResult.success)
                     exitProcess(1)
             } catch (x: ParsingFailedError) {
@@ -117,24 +112,11 @@ private fun compileMain(args: Array<String>) {
                 exitProcess(1)
             }
 
-            if (startEmulator) {
+            if (startEmulator==true) {
                 if (compilationResult.programName.isEmpty())
                     println("\nCan't start emulator because no program was assembled.")
-                else if(startEmulator) {
-                    for(emulator in listOf("x64sc", "x64")) {
-                        println("\nStarting C-64 emulator $emulator...")
-                        val cmdline = listOf(emulator, "-silent", "-moncommands", "${compilationResult.programName}.vice-mon-list",
-                                "-autostartprgmode", "1", "-autostart-warp", "-autostart", compilationResult.programName + ".prg")
-                        val processb = ProcessBuilder(cmdline).inheritIO()
-                        val process: Process
-                        try {
-                            process=processb.start()
-                        } catch(x: IOException) {
-                            continue  // try the next emulator executable
-                        }
-                        process.waitFor()
-                        break
-                    }
+                else {
+                    compilationResult.compTarget.machine.launchEmulator(compilationResult.programName)
                 }
             }
         }

compiler/src/prog8/ast/base/Extensions.kt

@@ -1,77 +0,0 @@
-package prog8.ast.base
-
-import prog8.ast.Module
-import prog8.ast.Program
-import prog8.ast.processing.*
-import prog8.compiler.CompilationOptions
-import prog8.compiler.BeforeAsmGenerationAstChanger
-import prog8.optimizer.AssignmentTransformer
-import prog8.optimizer.FlattenAnonymousScopesAndNopRemover
-
-
-internal fun Program.checkValid(compilerOptions: CompilationOptions, errors: ErrorReporter) {
-    val checker = AstChecker(this, compilerOptions, errors)
-    checker.visit(this)
-}
-
-internal fun Program.processAstBeforeAsmGeneration(errors: ErrorReporter) {
-    val fixer = BeforeAsmGenerationAstChanger(this, errors)
-    fixer.visit(this)
-    fixer.applyModifications()
-}
-
-internal fun Program.reorderStatements() {
-    val reorder = StatementReorderer(this)
-    reorder.visit(this)
-    reorder.applyModifications()
-}
-
-internal fun Program.addTypecasts(errors: ErrorReporter) {
-    val caster = TypecastsAdder(this, errors)
-    caster.visit(this)
-    caster.applyModifications()
-}
-
-internal fun Program.transformAssignments(errors: ErrorReporter) {
-    val transform = AssignmentTransformer(this, errors)
-    transform.visit(this)
-    while(transform.optimizationsDone>0 && errors.isEmpty()) {
-        transform.applyModifications()
-        transform.optimizationsDone = 0
-        transform.visit(this)
-    }
-    transform.applyModifications()
-}
-
-internal fun Module.checkImportedValid() {
-    val imr = ImportedModuleDirectiveRemover()
-    imr.visit(this, this.parent)
-    imr.applyModifications()
-}
-
-internal fun Program.checkRecursion(errors: ErrorReporter) {
-    val checker = AstRecursionChecker(namespace, errors)
-    checker.visit(this)
-    checker.processMessages(name)
-}
-
-internal fun Program.checkIdentifiers(errors: ErrorReporter) {
-
-    val checker2 = AstIdentifiersChecker(this, errors)
-    checker2.visit(this)
-
-    if(errors.isEmpty()) {
-        val transforms = AstVariousTransforms(this)
-        transforms.visit(this)
-        transforms.applyModifications()
-    }
-
-    if (modules.map { it.name }.toSet().size != modules.size) {
-        throw FatalAstException("modules should all be unique")
-    }
-}
-
-internal fun Program.removeNopsFlattenAnonScopes() {
-    val flattener = FlattenAnonymousScopesAndNopRemover()
-    flattener.visit(this)
-}

compiler/src/prog8/ast/processing/AstRecursionChecker.kt

@@ -1,118 +0,0 @@
-package prog8.ast.processing
-
-import prog8.ast.INameScope
-import prog8.ast.base.ErrorReporter
-import prog8.ast.base.Position
-import prog8.ast.expressions.FunctionCall
-import prog8.ast.statements.FunctionCallStatement
-import prog8.ast.statements.Subroutine
-
-
-internal class AstRecursionChecker(private val namespace: INameScope,
-                                   private val errors: ErrorReporter) : IAstVisitor {
-    private val callGraph = DirectedGraph<INameScope>()
-
-    fun processMessages(modulename: String) {
-        val cycle = callGraph.checkForCycle()
-        if(cycle.isEmpty())
-            return
-        val chain = cycle.joinToString(" <-- ") { "${it.name} at ${it.position}" }
-        errors.err("Program contains recursive subroutine calls, this is not supported. Recursive chain:\n (a subroutine call in) $chain", Position.DUMMY)
-    }
-
-    override fun visit(functionCallStatement: FunctionCallStatement) {
-        val scope = functionCallStatement.definingScope()
-        val targetStatement = functionCallStatement.target.targetStatement(namespace)
-        if(targetStatement!=null) {
-            val targetScope = when (targetStatement) {
-                is Subroutine -> targetStatement
-                else -> targetStatement.definingScope()
-            }
-            callGraph.add(scope, targetScope)
-        }
-        super.visit(functionCallStatement)
-    }
-
-    override fun visit(functionCall: FunctionCall) {
-        val scope = functionCall.definingScope()
-        val targetStatement = functionCall.target.targetStatement(namespace)
-        if(targetStatement!=null) {
-            val targetScope = when (targetStatement) {
-                is Subroutine -> targetStatement
-                else -> targetStatement.definingScope()
-            }
-            callGraph.add(scope, targetScope)
-        }
-        super.visit(functionCall)
-    }
-
-    private class DirectedGraph<VT> {
-        private val graph = mutableMapOf<VT, MutableSet<VT>>()
-        private var uniqueVertices = mutableSetOf<VT>()
-        val numVertices : Int
-            get() = uniqueVertices.size
-
-        fun add(from: VT, to: VT) {
-            var targets = graph[from]
-            if(targets==null) {
-                targets = mutableSetOf()
-                graph[from] = targets
-            }
-            targets.add(to)
-            uniqueVertices.add(from)
-            uniqueVertices.add(to)
-        }
-
-        fun print() {
-            println("#vertices: $numVertices")
-            graph.forEach { (from, to) ->
-                println("$from   CALLS:")
-                to.forEach { println("   $it") }
-            }
-            val cycle = checkForCycle()
-            if(cycle.isNotEmpty()) {
-                println("CYCLIC!  $cycle")
-            }
-        }
-
-        fun checkForCycle(): MutableList<VT> {
-            val visited = uniqueVertices.associateWith { false }.toMutableMap()
-            val recStack = uniqueVertices.associateWith { false }.toMutableMap()
-            val cycle = mutableListOf<VT>()
-            for(node in uniqueVertices) {
-                if(isCyclicUntil(node, visited, recStack, cycle))
-                    return cycle
-            }
-            return mutableListOf()
-        }
-
-        private fun isCyclicUntil(node: VT,
-                                  visited: MutableMap<VT, Boolean>,
-                                  recStack: MutableMap<VT, Boolean>,
-                                  cycleNodes: MutableList<VT>): Boolean {
-
-            if(recStack[node]==true) return true
-            if(visited[node]==true) return false
-
-            // mark current node as visited and add to recursion stack
-            visited[node] = true
-            recStack[node] = true
-
-            // recurse for all neighbours
-            val neighbors = graph[node]
-            if(neighbors!=null) {
-                for (neighbour in neighbors) {
-                    if (isCyclicUntil(neighbour, visited, recStack, cycleNodes)) {
-                        cycleNodes.add(node)
-                        return true
-                    }
-                }
-            }
-
-            // pop node from recursion stack
-            recStack[node] = false
-            return false
-        }
-    }
-
-}

compiler/src/prog8/ast/processing/AstVariousTransforms.kt

@@ -1,133 +0,0 @@
-package prog8.ast.processing
-
-import prog8.ast.Node
-import prog8.ast.Program
-import prog8.ast.base.*
-import prog8.ast.expressions.*
-import prog8.ast.statements.*
-
-
-internal class AstVariousTransforms(private val program: Program) : AstWalker() {
-    private val noModifications = emptyList<IAstModification>()
-
-    override fun before(functionCall: FunctionCall, parent: Node): Iterable<IAstModification> {
-        if(functionCall.target.nameInSource.size==1 && functionCall.target.nameInSource[0]=="lsb") {
-            // lsb(...) is just an alias for type cast to ubyte, so replace with "... as ubyte"
-            val typecast = TypecastExpression(functionCall.args.single(), DataType.UBYTE, false, functionCall.position)
-            return listOf(IAstModification.ReplaceNode(
-                    functionCall, typecast, parent
-            ))
-        }
-
-        return noModifications
-    }
-
-    override fun before(decl: VarDecl, parent: Node): Iterable<IAstModification> {
-        // is it a struct variable? then define all its struct members as mangled names,
-        //    and include the original decl as well.
-        if(decl.datatype==DataType.STRUCT && !decl.structHasBeenFlattened) {
-            val decls = decl.flattenStructMembers()
-            decls.add(decl)
-            val result = AnonymousScope(decls, decl.position)
-            return listOf(IAstModification.ReplaceNode(
-                    decl, result, parent
-            ))
-        }
-
-        return noModifications
-    }
-
-    override fun after(subroutine: Subroutine, parent: Node): Iterable<IAstModification> {
-        // For non-kernel subroutines and non-asm parameters:
-        // inject subroutine params as local variables (if they're not there yet).
-        val symbolsInSub = subroutine.allDefinedSymbols()
-        val namesInSub = symbolsInSub.map{ it.first }.toSet()
-        if(subroutine.asmAddress==null) {
-            if(subroutine.asmParameterRegisters.isEmpty()) {
-                return subroutine.parameters
-                        .filter { it.name !in namesInSub }
-                        .map {
-                            val vardecl = ParameterVarDecl(it.name, it.type, subroutine.position)
-                            IAstModification.InsertFirst(vardecl, subroutine)
-                        }
-            }
-        }
-
-        return noModifications
-    }
-
-    override fun before(expr: BinaryExpression, parent: Node): Iterable<IAstModification> {
-        when {
-            expr.left is StringLiteralValue ->
-                return listOf(IAstModification.ReplaceNode(
-                        expr,
-                        processBinaryExprWithString(expr.left as StringLiteralValue, expr.right, expr),
-                        parent
-                ))
-            expr.right is StringLiteralValue ->
-                return listOf(IAstModification.ReplaceNode(
-                        expr,
-                        processBinaryExprWithString(expr.right as StringLiteralValue, expr.left, expr),
-                        parent
-                ))
-        }
-
-        return noModifications
-    }
-
-    override fun after(string: StringLiteralValue, parent: Node): Iterable<IAstModification> {
-        if(string.parent !is VarDecl) {
-            // replace the literal string by a identifier reference to a new local vardecl
-            val vardecl = VarDecl.createAuto(string)
-            val identifier = IdentifierReference(listOf(vardecl.name), vardecl.position)
-            return listOf(
-                    IAstModification.ReplaceNode(string, identifier, parent),
-                    IAstModification.InsertFirst(vardecl, string.definingScope() as Node)
-            )
-        }
-        return noModifications
-    }
-
-    override fun after(array: ArrayLiteralValue, parent: Node): Iterable<IAstModification> {
-        val vardecl = array.parent as? VarDecl
-        if(vardecl!=null) {
-            // adjust the datatype of the array (to an educated guess)
-            val arrayDt = array.type
-            if(!arrayDt.istype(vardecl.datatype)) {
-                val cast = array.cast(vardecl.datatype)
-                if (cast != null && cast!=array)
-                    return listOf(IAstModification.ReplaceNode(vardecl.value!!, cast, vardecl))
-            }
-        } else {
-            val arrayDt = array.guessDatatype(program)
-            if(arrayDt.isKnown) {
-                // this array literal is part of an expression, turn it into an identifier reference
-                val litval2 = array.cast(arrayDt.typeOrElse(DataType.STRUCT))
-                if(litval2!=null && litval2!=array) {
-                    val vardecl = VarDecl.createAuto(litval2)
-                    val identifier = IdentifierReference(listOf(vardecl.name), vardecl.position)
-                    return listOf(
-                            IAstModification.ReplaceNode(array, identifier, parent),
-                            IAstModification.InsertFirst(vardecl, array.definingScope() as Node)
-                    )
-                }
-            }
-        }
-        return noModifications
-    }
-
-    private fun processBinaryExprWithString(string: StringLiteralValue, operand: Expression, expr: BinaryExpression): Expression {
-        val constvalue = operand.constValue(program)
-        if(constvalue!=null) {
-            if (expr.operator == "*") {
-                // repeat a string a number of times
-                return StringLiteralValue(string.value.repeat(constvalue.number.toInt()), string.altEncoding, expr.position)
-            }
-        }
-        if(expr.operator == "+" && operand is StringLiteralValue) {
-            // concatenate two strings
-            return StringLiteralValue("${string.value}${operand.value}", string.altEncoding, expr.position)
-        }
-        return expr
-    }
-}

compiler/src/prog8/ast/processing/ImportedModuleDirectiveRemover.kt

@@ -1,21 +0,0 @@
-package prog8.ast.processing
-
-import prog8.ast.Node
-import prog8.ast.statements.Directive
-
-
-internal class ImportedModuleDirectiveRemover: AstWalker() {
-    /**
-     * Most global directives don't apply for imported modules, so remove them
-     */
-
-    private val moduleLevelDirectives = listOf("%output", "%launcher", "%zeropage", "%zpreserved", "%address")
-    private val noModifications = emptyList<IAstModification>()
-
-    override fun before(directive: Directive, parent: Node): Iterable<IAstModification> {
-        if(directive.directive in moduleLevelDirectives) {
-            return listOf(IAstModification.Remove(directive, parent))
-        }
-        return noModifications
-    }
-}

compiler/src/prog8/ast/processing/StatementReorderer.kt

@@ -1,184 +0,0 @@
-package prog8.ast.processing
-
-import prog8.ast.*
-import prog8.ast.base.*
-import prog8.ast.expressions.*
-import prog8.ast.statements.*
-
-
-internal class StatementReorderer(val program: Program) : AstWalker() {
-    // Reorders the statements in a way the compiler needs.
-    // - 'main' block must be the very first statement UNLESS it has an address set.
-    // - library blocks are put last.
-    // - blocks are ordered by address, where blocks without address are placed last.
-    // - in every scope, most directives and vardecls are moved to the top.
-    // - the 'start' subroutine is moved to the top.
-    // - (syntax desugaring) a vardecl with a non-const initializer value is split into a regular vardecl and an assignment statement.
-    // - (syntax desugaring) augmented assignment is turned into regular assignment.
-    // - (syntax desugaring) struct value assignment is expanded into several struct member assignments.
-    // - sorts the choices in when statement.
-    // - insert AddressOf (&) expression where required (string params to a UWORD function param etc).
-
-    private val noModifications = emptyList<IAstModification>()
-    private val directivesToMove = setOf("%output", "%launcher", "%zeropage", "%zpreserved", "%address", "%option")
-
-    override fun after(module: Module, parent: Node): Iterable<IAstModification> {
-        val (blocks, other) = module.statements.partition { it is Block }
-        module.statements = other.asSequence().plus(blocks.sortedBy { (it as Block).address ?: Int.MAX_VALUE }).toMutableList()
-
-        val mainBlock = module.statements.filterIsInstance<Block>().firstOrNull { it.name=="main" }
-        if(mainBlock!=null && mainBlock.address==null) {
-            module.statements.remove(mainBlock)
-            module.statements.add(0, mainBlock)
-        }
-
-        reorderVardeclsAndDirectives(module.statements)
-        return noModifications
-    }
-
-    private fun reorderVardeclsAndDirectives(statements: MutableList<Statement>) {
-        val varDecls = statements.filterIsInstance<VarDecl>()
-        statements.removeAll(varDecls)
-        statements.addAll(0, varDecls)
-
-        val directives = statements.filterIsInstance<Directive>().filter {it.directive in directivesToMove}
-        statements.removeAll(directives)
-        statements.addAll(0, directives)
-    }
-
-    override fun before(block: Block, parent: Node): Iterable<IAstModification> {
-        parent as Module
-        if(block.isInLibrary) {
-            return listOf(
-                    IAstModification.Remove(block, parent),
-                    IAstModification.InsertLast(block, parent)
-            )
-        }
-
-        reorderVardeclsAndDirectives(block.statements)
-        return noModifications
-    }
-
-    override fun before(subroutine: Subroutine, parent: Node): Iterable<IAstModification> {
-        if(subroutine.name=="start" && parent is Block) {
-            if(parent.statements.filterIsInstance<Subroutine>().first().name!="start") {
-                return listOf(
-                        IAstModification.Remove(subroutine, parent),
-                        IAstModification.InsertFirst(subroutine, parent)
-                )
-            }
-        }
-        return noModifications
-    }
-
-    override fun after(decl: VarDecl, parent: Node): Iterable<IAstModification> {
-        val declValue = decl.value
-        if(declValue!=null && decl.type== VarDeclType.VAR && decl.datatype in NumericDatatypes) {
-            val declConstValue = declValue.constValue(program)
-            if(declConstValue==null) {
-                // move the vardecl (without value) to the scope and replace this with a regular assignment
-                decl.value = null
-                val target = AssignTarget(null, IdentifierReference(listOf(decl.name), decl.position), null, null, decl.position)
-                val assign = Assignment(target, null, declValue, decl.position)
-                return listOf(
-                        IAstModification.ReplaceNode(decl, assign, parent),
-                        IAstModification.InsertFirst(decl, decl.definingScope() as Node)
-                )
-            }
-        }
-        return noModifications
-    }
-
-    override fun after(whenStatement: WhenStatement, parent: Node): Iterable<IAstModification> {
-        val choices = whenStatement.choiceValues(program).sortedBy {
-            it.first?.first() ?: Int.MAX_VALUE
-        }
-        whenStatement.choices.clear()
-        choices.mapTo(whenStatement.choices) { it.second }
-        return noModifications
-    }
-
-    override fun before(assignment: Assignment, parent: Node): Iterable<IAstModification> {
-        if(assignment.aug_op!=null) {
-            return listOf(IAstModification.ReplaceNode(assignment, assignment.asDesugaredNonaugmented(), parent))
-        }
-
-        val valueType = assignment.value.inferType(program)
-        val targetType = assignment.target.inferType(program, assignment)
-        if(valueType.istype(DataType.STRUCT) && targetType.istype(DataType.STRUCT)) {
-            val assignments = if (assignment.value is StructLiteralValue) {
-                flattenStructAssignmentFromStructLiteral(assignment, program)    //  'structvar = { ..... } '
-            } else {
-                flattenStructAssignmentFromIdentifier(assignment, program)    //   'structvar1 = structvar2'
-            }
-            if(assignments.isNotEmpty()) {
-                val modifications = mutableListOf<IAstModification>()
-                assignments.reversed().mapTo(modifications) { IAstModification.InsertAfter(assignment, it, parent) }
-                modifications.add(IAstModification.Remove(assignment, parent))
-                return modifications
-            }
-        }
-
-        return noModifications
-    }
-
-    private fun flattenStructAssignmentFromStructLiteral(structAssignment: Assignment, program: Program): List<Assignment> {
-        val identifier = structAssignment.target.identifier!!
-        val identifierName = identifier.nameInSource.single()
-        val targetVar = identifier.targetVarDecl(program.namespace)!!
-        val struct = targetVar.struct!!
-
-        val slv = structAssignment.value as? StructLiteralValue
-        if(slv==null || slv.values.size != struct.numberOfElements)
-            throw FatalAstException("element count mismatch")
-
-        return struct.statements.zip(slv.values).map { (targetDecl, sourceValue) ->
-            targetDecl as VarDecl
-            val mangled = mangledStructMemberName(identifierName, targetDecl.name)
-            val idref = IdentifierReference(listOf(mangled), structAssignment.position)
-            val assign = Assignment(AssignTarget(null, idref, null, null, structAssignment.position),
-                    null, sourceValue, sourceValue.position)
-            assign.linkParents(structAssignment)
-            assign
-        }
-    }
-
-    private fun flattenStructAssignmentFromIdentifier(structAssignment: Assignment, program: Program): List<Assignment> {
-        val identifier = structAssignment.target.identifier!!
-        val identifierName = identifier.nameInSource.single()
-        val targetVar = identifier.targetVarDecl(program.namespace)!!
-        val struct = targetVar.struct!!
-        when (structAssignment.value) {
-            is IdentifierReference -> {
-                val sourceVar = (structAssignment.value as IdentifierReference).targetVarDecl(program.namespace)!!
-                if (sourceVar.struct == null)
-                    throw FatalAstException("can only assign arrays or structs to structs")
-                // struct memberwise copy
-                val sourceStruct = sourceVar.struct!!
-                if(sourceStruct!==targetVar.struct) {
-                    // structs are not the same in assignment
-                    return listOf()     // error will be printed elsewhere
-                }
-                return struct.statements.zip(sourceStruct.statements).map { member ->
-                    val targetDecl = member.first as VarDecl
-                    val sourceDecl = member.second as VarDecl
-                    if(targetDecl.name != sourceDecl.name)
-                        throw FatalAstException("struct member mismatch")
-                    val mangled = mangledStructMemberName(identifierName, targetDecl.name)
-                    val idref = IdentifierReference(listOf(mangled), structAssignment.position)
-                    val sourcemangled = mangledStructMemberName(sourceVar.name, sourceDecl.name)
-                    val sourceIdref = IdentifierReference(listOf(sourcemangled), structAssignment.position)
-                    val assign = Assignment(AssignTarget(null, idref, null, null, structAssignment.position),
-                            null, sourceIdref, member.second.position)
-                    assign.linkParents(structAssignment)
-                    assign
-                }
-            }
-            is StructLiteralValue -> {
-                throw IllegalArgumentException("not going to flatten a structLv assignment here")
-            }
-            else -> throw FatalAstException("strange struct value")
-        }
-    }
-
-}

compiler/src/prog8/compiler/BeforeAsmGenerationAstChanger.kt

@@ -1,61 +1,118 @@
 package prog8.compiler
 
+import prog8.ast.IFunctionCall
 import prog8.ast.Node
 import prog8.ast.Program
 import prog8.ast.base.*
 import prog8.ast.expressions.*
-import prog8.ast.processing.AstWalker
-import prog8.ast.processing.IAstModification
 import prog8.ast.statements.*
+import prog8.ast.walk.AstWalker
+import prog8.ast.walk.IAstModification
+import prog8.compiler.target.ICompilationTarget
 
 
-internal class BeforeAsmGenerationAstChanger(val program: Program, val errors: ErrorReporter) : AstWalker() {
+internal class BeforeAsmGenerationAstChanger(val program: Program, val errors: IErrorReporter, private val compTarget: ICompilationTarget) : AstWalker() {
 
     private val noModifications = emptyList<IAstModification>()
 
     override fun after(decl: VarDecl, parent: Node): Iterable<IAstModification> {
-        if (decl.value == null && decl.type == VarDeclType.VAR && decl.datatype in NumericDatatypes) {
-            // a numeric vardecl without an initial value is initialized with zero.
-            decl.value = decl.zeroElementValue()
+        subroutineVariables.add(decl.name to decl)
+        if (decl.value == null && !decl.autogeneratedDontRemove && decl.type == VarDeclType.VAR && decl.datatype in NumericDatatypes) {
+            // a numeric vardecl without an initial value is initialized with zero,
+            // unless there's already an assignment below, that initializes the value
+            if(decl.allowInitializeWithZero)
+            {
+                val nextAssign = decl.definingScope().nextSibling(decl) as? Assignment
+                if (nextAssign != null && nextAssign.target.isSameAs(IdentifierReference(listOf(decl.name), Position.DUMMY)))
+                    decl.value = null
+                else
+                    decl.value = decl.zeroElementValue()
+            }
         }
         return noModifications
     }
 
+    override fun after(assignment: Assignment, parent: Node): Iterable<IAstModification> {
+        // Try to replace A = B <operator> Something  by A= B, A = A <operator> Something
+        // this triggers the more efficent augmented assignment code generation more often.
+        // But it can only be done if the target variable IS NOT OCCURRING AS AN OPERAND ITSELF.
+        if(!assignment.isAugmentable
+                && assignment.target.identifier != null
+                && compTarget.isInRegularRAM(assignment.target, program)) {
+            val binExpr = assignment.value as? BinaryExpression
+            if (binExpr != null && binExpr.operator !in comparisonOperators) {
+                if (binExpr.left !is BinaryExpression) {
+                    if (binExpr.right.referencesIdentifier(*assignment.target.identifier!!.nameInSource.toTypedArray())) {
+                        // the right part of the expression contains the target variable itself.
+                        // we can't 'split' it trivially because the variable will be changed halfway through.
+                        if(binExpr.operator in associativeOperators) {
+                            // A = <something-without-A>  <associativeoperator>  <otherthing-with-A>
+                            // use the other part of the expression to split.
+                            val assignRight = Assignment(assignment.target, binExpr.right, assignment.position)
+                            return listOf(
+                                    IAstModification.InsertBefore(assignment, assignRight, assignment.definingScope()),
+                                    IAstModification.ReplaceNode(binExpr.right, binExpr.left, binExpr),
+                                    IAstModification.ReplaceNode(binExpr.left, assignment.target.toExpression(), binExpr))
+                        }
+                    } else {
+                        val assignLeft = Assignment(assignment.target, binExpr.left, assignment.position)
+                        return listOf(
+                                IAstModification.InsertBefore(assignment, assignLeft, assignment.definingScope()),
+                                IAstModification.ReplaceNode(binExpr.left, assignment.target.toExpression(), binExpr))
+                    }
+                }
+            }
+        }
+        return noModifications
+    }
+
+    private val subroutineVariables = mutableListOf<Pair<String, VarDecl>>()
+
+    override fun before(subroutine: Subroutine, parent: Node): Iterable<IAstModification> {
+        subroutineVariables.clear()
+        return noModifications
+    }
+
     override fun after(scope: AnonymousScope, parent: Node): Iterable<IAstModification> {
         val decls = scope.statements.filterIsInstance<VarDecl>()
+        subroutineVariables.addAll(decls.map { it.name to it })
+
         val sub = scope.definingSubroutine()
         if (sub != null) {
-            val existingVariables = sub.statements.filterIsInstance<VarDecl>().associateBy { it.name }
-            var conflicts = false
-            decls.forEach {
-                val existing = existingVariables[it.name]
-                if (existing != null) {
-                    errors.err("variable ${it.name} already defined in subroutine ${sub.name} at ${existing.position}", it.position)
-                    conflicts = true
-                }
-            }
-            if (!conflicts) {
-                val numericVarsWithValue = decls.filter { it.value != null && it.datatype in NumericDatatypes }
-                return numericVarsWithValue.map {
-                    val initValue = it.value!!  // assume here that value has always been set by now
-                    it.value = null     // make sure no value init assignment for this vardecl will be created later (would be superfluous)
-                    val target = AssignTarget(null, IdentifierReference(listOf(it.name), it.position), null, null, it.position)
-                    val assign = Assignment(target, null, initValue, it.position)
-                    initValue.parent = assign
-                    IAstModification.InsertFirst(assign, scope)
-                } +  decls.map { IAstModification.ReplaceNode(it, NopStatement(it.position), scope) } +
-                     decls.map { IAstModification.InsertFirst(it, sub) }    // move it up to the subroutine
+            // move vardecls of the scope into the upper scope. Make sure the position remains the same!
+            val numericVarsWithValue = decls.filter { it.value != null && it.datatype in NumericDatatypes }
+            val replaceVardecls =numericVarsWithValue.map {
+                val initValue = it.value!!  // assume here that value has always been set by now
+                it.value = null     // make sure no value init assignment for this vardecl will be created later (would be superfluous)
+                val target = AssignTarget(IdentifierReference(listOf(it.name), it.position), null, null, it.position)
+                val assign = Assignment(target, initValue, it.position)
+                initValue.parent = assign
+                IAstModification.ReplaceNode(it, assign, scope)
             }
+            val moveVardeclsUp = decls.map { IAstModification.InsertFirst(it, sub) }
+            return replaceVardecls + moveVardeclsUp
         }
         return noModifications
     }
 
     override fun after(subroutine: Subroutine, parent: Node): Iterable<IAstModification> {
-        // add the implicit return statement at the end (if it's not there yet), but only if it's not a kernel routine.
+        val firstDeclarations = mutableMapOf<String, VarDecl>()
+        for(decl in subroutineVariables) {
+            val existing = firstDeclarations[decl.first]
+            if(existing!=null && existing !== decl.second) {
+                errors.err("variable ${decl.first} already defined in subroutine ${subroutine.name} at ${existing.position}", decl.second.position)
+            } else {
+                firstDeclarations[decl.first] = decl.second
+            }
+        }
+
+
+        // add the implicit return statement at the end (if it's not there yet), but only if it's not a kernal routine.
         // and if an assembly block doesn't contain a rts/rti, and some other situations.
         val mods = mutableListOf<IAstModification>()
         val returnStmt = Return(null, subroutine.position)
         if (subroutine.asmAddress == null
+                && !subroutine.inline
                 && subroutine.statements.isNotEmpty()
                 && subroutine.amountOfRtsInAsm() == 0
                 && subroutine.statements.lastOrNull { it !is VarDecl } !is Return
@@ -72,9 +129,8 @@ internal class BeforeAsmGenerationAstChanger(val program: Program, val errors: E
                 && outerStatements[subroutineStmtIdx - 1] !is Subroutine
                 && outerStatements[subroutineStmtIdx - 1] !is Return
                 && outerScope !is Block) {
-            mods += IAstModification.InsertAfter(outerStatements[subroutineStmtIdx - 1], returnStmt, outerScope as Node)
+            mods += IAstModification.InsertAfter(outerStatements[subroutineStmtIdx - 1], returnStmt, outerScope)
         }
-
         return mods
     }
 
@@ -89,13 +145,40 @@ internal class BeforeAsmGenerationAstChanger(val program: Program, val errors: E
                 return listOf(IAstModification.ReplaceNode(typecast, typecast.expression, parent))
             }
         }
-        else if(sourceDt in PassByReferenceDatatypes) {
+
+
+        // Note: for various reasons (most importantly, code simplicity), the code generator assumes/requires
+        // that the types of assignment values and their target are the same,
+        // and that the types of both operands of a binaryexpression node are the same.
+        // So, it is not easily possible to remove the typecasts that are there to make these conditions true.
+        // The only place for now where we can do this is for:
+        //    asmsub register pair parameter.
+
+        if(typecast.type in WordDatatypes) {
+            val fcall = typecast.parent as? IFunctionCall
+            if (fcall != null) {
+                val sub = fcall.target.targetStatement(program) as? Subroutine
+                if (sub != null && sub.isAsmSubroutine) {
+                    return listOf(IAstModification.ReplaceNode(typecast, typecast.expression, parent))
+                }
+            }
+        }
+
+        if(sourceDt in PassByReferenceDatatypes) {
             if(typecast.type==DataType.UWORD) {
-                return listOf(IAstModification.ReplaceNode(
-                        typecast,
-                        AddressOf(typecast.expression as IdentifierReference, typecast.position),
-                        parent
-                ))
+                if(typecast.expression is IdentifierReference) {
+                    return listOf(IAstModification.ReplaceNode(
+                            typecast,
+                            AddressOf(typecast.expression as IdentifierReference, typecast.position),
+                            parent
+                    ))
+                } else if(typecast.expression is IFunctionCall) {
+                    return listOf(IAstModification.ReplaceNode(
+                            typecast,
+                            typecast.expression,
+                            parent
+                    ))
+                }
             } else {
                 errors.err("cannot cast pass-by-reference value to type ${typecast.type} (only to UWORD)", typecast.position)
             }
@@ -103,4 +186,34 @@ internal class BeforeAsmGenerationAstChanger(val program: Program, val errors: E
 
         return noModifications
     }
+
+    override fun after(ifStatement: IfStatement, parent: Node): Iterable<IAstModification> {
+        val binExpr = ifStatement.condition as? BinaryExpression
+        if(binExpr==null || binExpr.operator !in comparisonOperators) {
+            // if x  ->  if x!=0,    if x+5  ->  if x+5 != 0
+            val booleanExpr = BinaryExpression(ifStatement.condition, "!=", NumericLiteralValue.optimalInteger(0, ifStatement.condition.position), ifStatement.condition.position)
+            return listOf(IAstModification.ReplaceNode(ifStatement.condition, booleanExpr, ifStatement))
+        }
+        return noModifications
+    }
+
+    override fun after(untilLoop: UntilLoop, parent: Node): Iterable<IAstModification> {
+        val binExpr = untilLoop.condition as? BinaryExpression
+        if(binExpr==null || binExpr.operator !in comparisonOperators) {
+            // until x  ->  until x!=0,    until x+5  ->  until x+5 != 0
+            val booleanExpr = BinaryExpression(untilLoop.condition, "!=", NumericLiteralValue.optimalInteger(0, untilLoop.condition.position), untilLoop.condition.position)
+            return listOf(IAstModification.ReplaceNode(untilLoop.condition, booleanExpr, untilLoop))
+        }
+        return noModifications
+    }
+
+    override fun after(whileLoop: WhileLoop, parent: Node): Iterable<IAstModification> {
+        val binExpr = whileLoop.condition as? BinaryExpression
+        if(binExpr==null || binExpr.operator !in comparisonOperators) {
+            // while x  ->  while x!=0,    while x+5  ->  while x+5 != 0
+            val booleanExpr = BinaryExpression(whileLoop.condition, "!=", NumericLiteralValue.optimalInteger(0, whileLoop.condition.position), whileLoop.condition.position)
+            return listOf(IAstModification.ReplaceNode(whileLoop.condition, booleanExpr, whileLoop))
+        }
+        return noModifications
+    }
 }

compiler/src/prog8/compiler/Compiler.kt

@@ -1,9 +1,30 @@
 package prog8.compiler
 
+import prog8.ast.AstToSourceCode
+import prog8.ast.IBuiltinFunctions
+import prog8.ast.IMemSizer
+import prog8.ast.Program
+import prog8.ast.base.AstException
+import prog8.ast.base.Position
+import prog8.ast.expressions.Expression
+import prog8.ast.expressions.NumericLiteralValue
+import prog8.ast.statements.Directive
+import prog8.compiler.astprocessing.*
+import prog8.compiler.functions.*
+import prog8.compiler.target.C64Target
+import prog8.compiler.target.Cx16Target
+import prog8.compiler.target.ICompilationTarget
+import prog8.compiler.target.asmGeneratorFor
+import prog8.optimizer.*
+import prog8.parser.ModuleImporter
+import prog8.parser.ParsingFailedError
+import prog8.parser.moduleName
 import java.io.File
 import java.io.InputStream
 import java.nio.file.Path
-import kotlin.math.abs
+import kotlin.system.exitProcess
+import kotlin.system.measureTimeMillis
+
 
 enum class OutputType {
     RAW,
@@ -27,31 +48,284 @@ data class CompilationOptions(val output: OutputType,
                               val launcher: LauncherType,
                               val zeropage: ZeropageType,
                               val zpReserved: List<IntRange>,
-                              val floats: Boolean)
+                              val floats: Boolean,
+                              val noSysInit: Boolean,
+                              val compTarget: ICompilationTarget) {
+    var slowCodegenWarnings = false
+    var optimize = false
+}
 
 
 class CompilerException(message: String?) : Exception(message)
 
-fun Number.toHex(): String {
-    //  0..15 -> "0".."15"
-    //  16..255 -> "$10".."$ff"
-    //  256..65536 -> "$0100".."$ffff"
-    // negative values are prefixed with '-'.
-    val integer = this.toInt()
-    if(integer<0)
-        return '-' + abs(integer).toHex()
-    return when (integer) {
-        in 0 until 16 -> integer.toString()
-        in 0 until 0x100 -> "$"+integer.toString(16).padStart(2,'0')
-        in 0 until 0x10000 -> "$"+integer.toString(16).padStart(4,'0')
-        else -> throw CompilerException("number too large for 16 bits $this")
+class CompilationResult(val success: Boolean,
+                        val programAst: Program,
+                        val programName: String,
+                        val compTarget: ICompilationTarget,
+                        val importedFiles: List<Path>)
+
+
+fun compileProgram(filepath: Path,
+                   optimize: Boolean,
+                   writeAssembly: Boolean,
+                   slowCodegenWarnings: Boolean,
+                   compilationTarget: String,
+                   outputDir: Path): CompilationResult {
+    var programName = ""
+    lateinit var programAst: Program
+    lateinit var importedFiles: List<Path>
+    val errors = ErrorReporter()
+
+    val compTarget =
+        when(compilationTarget) {
+            C64Target.name -> C64Target
+            Cx16Target.name -> Cx16Target
+            else -> {
+                System.err.println("invalid compilation target")
+                exitProcess(1)
+            }
+        }
+
+    try {
+        val totalTime = measureTimeMillis {
+            // import main module and everything it needs
+            val (ast, compilationOptions, imported) = parseImports(filepath, errors, compTarget)
+            compilationOptions.slowCodegenWarnings = slowCodegenWarnings
+            compilationOptions.optimize = optimize
+            programAst = ast
+            importedFiles = imported
+            processAst(programAst, errors, compilationOptions)
+            if (compilationOptions.optimize)
+                optimizeAst(programAst, errors, BuiltinFunctionsFacade(BuiltinFunctions), compTarget)
+            postprocessAst(programAst, errors, compilationOptions)
+
+            // printAst(programAst)
+
+            if(writeAssembly)
+                programName = writeAssembly(programAst, errors, outputDir, compilationOptions)
+        }
+        System.out.flush()
+        System.err.flush()
+        println("\nTotal compilation+assemble time: ${totalTime / 1000.0} sec.")
+        return CompilationResult(true, programAst, programName, compTarget, importedFiles)
+
+    } catch (px: ParsingFailedError) {
+        System.err.print("\u001b[91m")  // bright red
+        System.err.println(px.message)
+        System.err.print("\u001b[0m")  // reset
+    } catch (ax: AstException) {
+        System.err.print("\u001b[91m")  // bright red
+        System.err.println(ax.toString())
+        System.err.print("\u001b[0m")  // reset
+    } catch (x: Exception) {
+        print("\u001b[91m")  // bright red
+        println("\n* internal error *")
+        print("\u001b[0m")  // reset
+        System.out.flush()
+        throw x
+    } catch (x: NotImplementedError) {
+        print("\u001b[91m")  // bright red
+        println("\n* internal error: missing feature/code *")
+        print("\u001b[0m")  // reset
+        System.out.flush()
+        throw x
     }
+
+    val failedProgram = Program("failed", mutableListOf(), BuiltinFunctionsFacade(BuiltinFunctions), compTarget)
+    return CompilationResult(false, failedProgram, programName, compTarget, emptyList())
+}
+
+private class BuiltinFunctionsFacade(functions: Map<String, FSignature>): IBuiltinFunctions {
+    lateinit var program: Program
+
+    override val names = functions.keys
+    override val purefunctionNames = functions.filter { it.value.pure }.map { it.key }.toSet()
+
+    override fun constValue(name: String, args: List<Expression>, position: Position, memsizer: IMemSizer): NumericLiteralValue? {
+        val func = BuiltinFunctions[name]
+        if(func!=null) {
+            val exprfunc = func.constExpressionFunc
+            if(exprfunc!=null) {
+                return try {
+                    exprfunc(args, position, program, memsizer)
+                } catch(x: NotConstArgumentException) {
+                    // const-evaluating the builtin function call failed.
+                    null
+                } catch(x: CannotEvaluateException) {
+                    // const-evaluating the builtin function call failed.
+                    null
+                }
+            }
+            else if(func.known_returntype==null)
+                throw IllegalArgumentException("builtin function $name can't be used here because it doesn't return a value")
+        }
+        return null
+    }
+    override fun returnType(name: String, args: MutableList<Expression>) =
+        builtinFunctionReturnType(name, args, program)
+}
+
+private fun parseImports(filepath: Path, errors: IErrorReporter, compTarget: ICompilationTarget): Triple<Program, CompilationOptions, List<Path>> {
+    val compilationTargetName = compTarget.name
+    println("Compiler target: $compilationTargetName. Parsing...")
+    val importer = ModuleImporter()
+    val bf = BuiltinFunctionsFacade(BuiltinFunctions)
+    val programAst = Program(moduleName(filepath.fileName), mutableListOf(), bf, compTarget)
+    bf.program = programAst
+    importer.importModule(programAst, filepath, compTarget, compilationTargetName)
+    errors.report()
+
+    val importedFiles = programAst.modules.filter { !it.source.startsWith("@embedded@") }.map { it.source }
+    val compilerOptions = determineCompilationOptions(programAst, compTarget)
+    if (compilerOptions.launcher == LauncherType.BASIC && compilerOptions.output != OutputType.PRG)
+        throw ParsingFailedError("${programAst.modules.first().position} BASIC launcher requires output type PRG.")
+
+    // depending on the machine and compiler options we may have to include some libraries
+    for(lib in compTarget.machine.importLibs(compilerOptions, compilationTargetName))
+        importer.importLibraryModule(programAst, lib, compTarget, compilationTargetName)
+
+    // always import prog8_lib and math
+    importer.importLibraryModule(programAst, "math", compTarget, compilationTargetName)
+    importer.importLibraryModule(programAst, "prog8_lib", compTarget, compilationTargetName)
+    errors.report()
+    return Triple(programAst, compilerOptions, importedFiles)
+}
+
+private fun determineCompilationOptions(program: Program, compTarget: ICompilationTarget): CompilationOptions {
+    val mainModule = program.mainModule
+    val outputType = (mainModule.statements.singleOrNull { it is Directive && it.directive == "%output" }
+            as? Directive)?.args?.single()?.name?.toUpperCase()
+    val launcherType = (mainModule.statements.singleOrNull { it is Directive && it.directive == "%launcher" }
+            as? Directive)?.args?.single()?.name?.toUpperCase()
+    val zpoption: String? = (mainModule.statements.singleOrNull { it is Directive && it.directive == "%zeropage" }
+            as? Directive)?.args?.single()?.name?.toUpperCase()
+    val allOptions = program.modules.flatMap { it.statements }.filter { it is Directive && it.directive == "%option" }.flatMap { (it as Directive).args }.toSet()
+    val floatsEnabled = allOptions.any { it.name == "enable_floats" }
+    val noSysInit = allOptions.any { it.name == "no_sysinit" }
+    var zpType: ZeropageType =
+            if (zpoption == null)
+                if(floatsEnabled) ZeropageType.FLOATSAFE else ZeropageType.KERNALSAFE
+            else
+                try {
+                    ZeropageType.valueOf(zpoption)
+                } catch (x: IllegalArgumentException) {
+                    ZeropageType.KERNALSAFE
+                    // error will be printed by the astchecker
+                }
+
+    if (zpType==ZeropageType.FLOATSAFE && compTarget.name == Cx16Target.name) {
+        System.err.println("Warning: Cx16 target must use zp option basicsafe instead of floatsafe")
+        zpType = ZeropageType.BASICSAFE
+    }
+
+    val zpReserved = mainModule.statements
+            .asSequence()
+            .filter { it is Directive && it.directive == "%zpreserved" }
+            .map { (it as Directive).args }
+            .map { it[0].int!!..it[1].int!! }
+            .toList()
+
+    if(outputType!=null && !OutputType.values().any {it.name==outputType}) {
+        System.err.println("invalid output type $outputType")
+        exitProcess(1)
+    }
+    if(launcherType!=null && !LauncherType.values().any {it.name==launcherType}) {
+        System.err.println("invalid launcher type $launcherType")
+        exitProcess(1)
+    }
+
+    return CompilationOptions(
+            if (outputType == null) OutputType.PRG else OutputType.valueOf(outputType),
+            if (launcherType == null) LauncherType.BASIC else LauncherType.valueOf(launcherType),
+            zpType, zpReserved, floatsEnabled, noSysInit,
+            compTarget
+    )
+}
+
+private fun processAst(programAst: Program, errors: IErrorReporter, compilerOptions: CompilationOptions) {
+    // perform initial syntax checks and processings
+    println("Processing for target ${compilerOptions.compTarget.name}...")
+    programAst.checkIdentifiers(errors, compilerOptions.compTarget)
+    errors.report()
+    programAst.constantFold(errors, compilerOptions.compTarget)
+    errors.report()
+    programAst.reorderStatements(errors)
+    errors.report()
+    programAst.addTypecasts(errors)
+    errors.report()
+    programAst.variousCleanups()
+    programAst.checkValid(compilerOptions, errors, compilerOptions.compTarget)
+    errors.report()
+    programAst.checkIdentifiers(errors, compilerOptions.compTarget)
+    errors.report()
+}
+
+private fun optimizeAst(programAst: Program, errors: IErrorReporter, functions: IBuiltinFunctions, compTarget: ICompilationTarget) {
+    // optimize the parse tree
+    println("Optimizing...")
+    while (true) {
+        // keep optimizing expressions and statements until no more steps remain
+        val optsDone1 = programAst.simplifyExpressions()
+        val optsDone2 = programAst.splitBinaryExpressions(compTarget)
+        val optsDone3 = programAst.optimizeStatements(errors, functions, compTarget, ::loadAsmIncludeFile)
+        programAst.constantFold(errors, compTarget) // because simplified statements and expressions can result in more constants that can be folded away
+        errors.report()
+        if (optsDone1 + optsDone2 + optsDone3 == 0)
+            break
+    }
+
+    val remover = UnusedCodeRemover(programAst, errors, compTarget, ::loadAsmIncludeFile)
+    remover.visit(programAst)
+    remover.applyModifications()
+    errors.report()
+}
+
+private fun postprocessAst(programAst: Program, errors: IErrorReporter, compilerOptions: CompilationOptions) {
+    programAst.addTypecasts(errors)
+    errors.report()
+    programAst.variousCleanups()
+    programAst.checkValid(compilerOptions, errors, compilerOptions.compTarget)          // check if final tree is still valid
+    errors.report()
+    val callGraph = CallGraph(programAst, ::loadAsmIncludeFile)
+    callGraph.checkRecursiveCalls(errors)
+    errors.report()
+    programAst.verifyFunctionArgTypes()
+    programAst.moveMainAndStartToFirst()
+}
+
+private fun writeAssembly(programAst: Program,
+                          errors: IErrorReporter,
+                          outputDir: Path,
+                          compilerOptions: CompilationOptions): String {
+    // asm generation directly from the Ast,
+    programAst.processAstBeforeAsmGeneration(errors, compilerOptions.compTarget)
+    errors.report()
+
+    // printAst(programAst)
+
+    compilerOptions.compTarget.machine.initializeZeropage(compilerOptions)
+    val assembly = asmGeneratorFor(compilerOptions.compTarget,
+            programAst,
+            errors,
+            compilerOptions.compTarget.machine.zeropage,
+            compilerOptions,
+            outputDir).compileToAssembly()
+    assembly.assemble(compilerOptions)
+    errors.report()
+    return assembly.name
+}
+
+fun printAst(programAst: Program) {
+    println()
+    val printer = AstToSourceCode(::print, programAst)
+    printer.visit(programAst)
+    println()
 }
 
 fun loadAsmIncludeFile(filename: String, source: Path): String {
     return if (filename.startsWith("library:")) {
         val resource = tryGetEmbeddedResource(filename.substring(8))
-                ?: throw IllegalArgumentException("library file '$filename' not found")
+            ?: throw IllegalArgumentException("library file '$filename' not found")
         resource.bufferedReader().use { it.readText() }
     } else {
         // first try in the isSameAs folder as where the containing file was imported from

compiler/src/prog8/compiler/ErrorReporting.kt

@@ -1,9 +1,18 @@
-package prog8.ast.base
+package prog8.compiler
 
+import prog8.ast.base.Position
 import prog8.parser.ParsingFailedError
 
 
-class ErrorReporter {
+interface IErrorReporter {
+    fun err(msg: String, position: Position)
+    fun warn(msg: String, position: Position)
+    fun isEmpty(): Boolean
+    fun report()
+}
+
+
+internal class ErrorReporter: IErrorReporter {
     private enum class MessageSeverity {
         WARNING,
         ERROR
@@ -13,10 +22,14 @@ class ErrorReporter {
     private val messages = mutableListOf<CompilerMessage>()
     private val alreadyReportedMessages = mutableSetOf<String>()
 
-    fun err(msg: String, position: Position) = messages.add(CompilerMessage(MessageSeverity.ERROR, msg, position))
-    fun warn(msg: String, position: Position) = messages.add(CompilerMessage(MessageSeverity.WARNING, msg, position))
+    override fun err(msg: String, position: Position) {
+        messages.add(CompilerMessage(MessageSeverity.ERROR, msg, position))
+    }
+    override fun warn(msg: String, position: Position) {
+        messages.add(CompilerMessage(MessageSeverity.WARNING, msg, position))
+    }
 
-    fun handle() {
+    override fun report() {
         var numErrors = 0
         var numWarnings = 0
         messages.forEach {
@@ -24,7 +37,7 @@ class ErrorReporter {
                 MessageSeverity.ERROR -> System.err.print("\u001b[91m")  // bright red
                 MessageSeverity.WARNING -> System.err.print("\u001b[93m")  // bright yellow
             }
-            val msg = "${it.position} ${it.severity} ${it.message}".trim()
+            val msg = "${it.position.toClickableStr()} ${it.severity} ${it.message}".trim()
             if(msg !in alreadyReportedMessages) {
                 System.err.println(msg)
                 alreadyReportedMessages.add(msg)
@@ -40,5 +53,5 @@ class ErrorReporter {
             throw ParsingFailedError("There are $numErrors errors and $numWarnings warnings.")
     }
 
-    fun isEmpty() = messages.isEmpty()
+    override fun isEmpty() = messages.isEmpty()
 }

compiler/src/prog8/compiler/Main.kt

@@ -1,215 +0,0 @@
-package prog8.compiler
-
-import prog8.ast.AstToSourceCode
-import prog8.ast.Program
-import prog8.ast.base.*
-import prog8.ast.statements.Directive
-import prog8.compiler.target.CompilationTarget
-import prog8.optimizer.UnusedCodeRemover
-import prog8.optimizer.constantFold
-import prog8.optimizer.optimizeStatements
-import prog8.optimizer.simplifyExpressions
-import prog8.parser.ModuleImporter
-import prog8.parser.ParsingFailedError
-import prog8.parser.moduleName
-import java.nio.file.Path
-import kotlin.system.measureTimeMillis
-
-
-class CompilationResult(val success: Boolean,
-                        val programAst: Program,
-                        val programName: String,
-                        val importedFiles: List<Path>)
-
-
-fun compileProgram(filepath: Path,
-                   optimize: Boolean,
-                   writeAssembly: Boolean,
-                   outputDir: Path): CompilationResult {
-    var programName = ""
-    lateinit var programAst: Program
-    lateinit var importedFiles: List<Path>
-    val errors = ErrorReporter()
-
-    try {
-        val totalTime = measureTimeMillis {
-            // import main module and everything it needs
-            val (ast, compilationOptions, imported) = parseImports(filepath, errors)
-            programAst = ast
-            importedFiles = imported
-            processAst(programAst, errors, compilationOptions)
-            if (optimize)
-                optimizeAst(programAst, errors)
-            postprocessAst(programAst, errors, compilationOptions)
-
-            // printAst(programAst) // TODO
-
-            if(writeAssembly)
-                programName = writeAssembly(programAst, errors, outputDir, optimize, compilationOptions)
-        }
-        System.out.flush()
-        System.err.flush()
-        println("\nTotal compilation+assemble time: ${totalTime / 1000.0} sec.")
-        return CompilationResult(true, programAst, programName, importedFiles)
-
-    } catch (px: ParsingFailedError) {
-        System.err.print("\u001b[91m")  // bright red
-        System.err.println(px.message)
-        System.err.print("\u001b[0m")  // reset
-    } catch (ax: AstException) {
-        System.err.print("\u001b[91m")  // bright red
-        System.err.println(ax.toString())
-        System.err.print("\u001b[0m")  // reset
-    } catch (x: Exception) {
-        print("\u001b[91m")  // bright red
-        println("\n* internal error *")
-        print("\u001b[0m")  // reset
-        System.out.flush()
-        throw x
-    } catch (x: NotImplementedError) {
-        print("\u001b[91m")  // bright red
-        println("\n* internal error: missing feature/code *")
-        print("\u001b[0m")  // reset
-        System.out.flush()
-        throw x
-    }
-
-    return CompilationResult(false, Program("failed", mutableListOf()), programName, emptyList())
-}
-
-private fun parseImports(filepath: Path, errors: ErrorReporter): Triple<Program, CompilationOptions, List<Path>> {
-    println("Parsing...")
-    val importer = ModuleImporter(errors)
-    val programAst = Program(moduleName(filepath.fileName), mutableListOf())
-    importer.importModule(programAst, filepath)
-    errors.handle()
-
-    val importedFiles = programAst.modules.filter { !it.source.startsWith("@embedded@") }.map { it.source }
-
-    val compilerOptions = determineCompilationOptions(programAst)
-    if (compilerOptions.launcher == LauncherType.BASIC && compilerOptions.output != OutputType.PRG)
-        throw ParsingFailedError("${programAst.modules.first().position} BASIC launcher requires output type PRG.")
-
-    // if we're producing a PRG or BASIC program, include the c64utils and c64lib libraries
-    if (compilerOptions.launcher == LauncherType.BASIC || compilerOptions.output == OutputType.PRG) {
-        importer.importLibraryModule(programAst, "c64lib")
-        importer.importLibraryModule(programAst, "c64utils")
-    }
-
-    // always import prog8lib and math
-    importer.importLibraryModule(programAst, "math")
-    importer.importLibraryModule(programAst, "prog8lib")
-    errors.handle()
-    return Triple(programAst, compilerOptions, importedFiles)
-}
-
-private fun determineCompilationOptions(program: Program): CompilationOptions {
-    val mainModule = program.modules.first()
-    val outputType = (mainModule.statements.singleOrNull { it is Directive && it.directive == "%output" }
-            as? Directive)?.args?.single()?.name?.toUpperCase()
-    val launcherType = (mainModule.statements.singleOrNull { it is Directive && it.directive == "%launcher" }
-            as? Directive)?.args?.single()?.name?.toUpperCase()
-    mainModule.loadAddress = (mainModule.statements.singleOrNull { it is Directive && it.directive == "%address" }
-            as? Directive)?.args?.single()?.int ?: 0
-    val zpoption: String? = (mainModule.statements.singleOrNull { it is Directive && it.directive == "%zeropage" }
-            as? Directive)?.args?.single()?.name?.toUpperCase()
-    val allOptions = program.modules.flatMap { it.statements }.filter { it is Directive && it.directive == "%option" }.flatMap { (it as Directive).args }.toSet()
-    val floatsEnabled = allOptions.any { it.name == "enable_floats" }
-    val zpType: ZeropageType =
-            if (zpoption == null)
-                if(floatsEnabled) ZeropageType.FLOATSAFE else ZeropageType.KERNALSAFE
-            else
-                try {
-                    ZeropageType.valueOf(zpoption)
-                } catch (x: IllegalArgumentException) {
-                    ZeropageType.KERNALSAFE
-                    // error will be printed by the astchecker
-                }
-    val zpReserved = mainModule.statements
-            .asSequence()
-            .filter { it is Directive && it.directive == "%zpreserved" }
-            .map { (it as Directive).args }
-            .map { it[0].int!!..it[1].int!! }
-            .toList()
-
-    return CompilationOptions(
-            if (outputType == null) OutputType.PRG else OutputType.valueOf(outputType),
-            if (launcherType == null) LauncherType.BASIC else LauncherType.valueOf(launcherType),
-            zpType, zpReserved, floatsEnabled
-    )
-}
-
-private fun processAst(programAst: Program, errors: ErrorReporter, compilerOptions: CompilationOptions) {
-    // perform initial syntax checks and processings
-    println("Processing...")
-    programAst.checkIdentifiers(errors)
-    errors.handle()
-    programAst.constantFold(errors)
-    errors.handle()
-    programAst.removeNopsFlattenAnonScopes()
-    programAst.reorderStatements()
-    programAst.addTypecasts(errors)
-    errors.handle()
-    programAst.checkValid(compilerOptions, errors)
-    errors.handle()
-    programAst.checkIdentifiers(errors)
-    errors.handle()
-}
-
-private fun optimizeAst(programAst: Program, errors: ErrorReporter) {
-    // optimize the parse tree
-    println("Optimizing...")
-    while (true) {
-        // keep optimizing expressions and statements until no more steps remain
-        val optsDone1 = programAst.simplifyExpressions()
-        val optsDone2 = programAst.optimizeStatements(errors)
-        programAst.constantFold(errors) // because simplified statements and expressions could give rise to more constants that can be folded away:
-        errors.handle()
-        if (optsDone1 + optsDone2 == 0)
-            break
-    }
-
-    val remover = UnusedCodeRemover()
-    remover.visit(programAst)
-    remover.applyModifications()
-}
-
-private fun postprocessAst(programAst: Program, errors: ErrorReporter, compilerOptions: CompilationOptions) {
-    programAst.transformAssignments(errors)
-    errors.handle()
-    programAst.addTypecasts(errors)
-    errors.handle()
-    programAst.removeNopsFlattenAnonScopes()
-    programAst.checkValid(compilerOptions, errors)          // check if final tree is still valid
-    errors.handle()
-    programAst.checkRecursion(errors)         // check if there are recursive subroutine calls
-    errors.handle()
-}
-
-private fun writeAssembly(programAst: Program, errors: ErrorReporter, outputDir: Path,
-                          optimize: Boolean, compilerOptions: CompilationOptions): String {
-    // asm generation directly from the Ast,
-    val zeropage = CompilationTarget.machine.getZeropage(compilerOptions)
-    programAst.processAstBeforeAsmGeneration(errors)
-    errors.handle()
-
-    // printAst(programAst)
-
-    val assembly = CompilationTarget.asmGenerator(
-            programAst,
-            errors,
-            zeropage,
-            compilerOptions,
-            outputDir).compileToAssembly(optimize)
-    assembly.assemble(compilerOptions)
-    errors.handle()
-    return assembly.name
-}
-
-fun printAst(programAst: Program) {
-    println()
-    val printer = AstToSourceCode(::print, programAst)
-    printer.visit(programAst)
-    println()
-}
-

compiler/src/prog8/compiler/Zeropage.kt

@@ -8,6 +8,12 @@ class ZeropageDepletedError(message: String) : Exception(message)
 
 abstract class Zeropage(protected val options: CompilationOptions) {
 
+    abstract val SCRATCH_B1 : Int      // temp storage for a single byte
+    abstract val SCRATCH_REG : Int     // temp storage for a register
+    abstract val SCRATCH_W1 : Int      // temp storage 1 for a word  $fb+$fc
+    abstract val SCRATCH_W2 : Int      // temp storage 2 for a word  $fb+$fc
+
+
     private val allocations = mutableMapOf<Int, Pair<String, DataType>>()
     val free = mutableListOf<Int>()     // subclasses must set this to the appropriate free locations.
 
@@ -15,8 +21,8 @@ abstract class Zeropage(protected val options: CompilationOptions) {
 
     fun available() = if(options.zeropage==ZeropageType.DONTUSE) 0 else free.size
 
-    fun allocate(scopedname: String, datatype: DataType, position: Position?, errors: ErrorReporter): Int {
-        assert(scopedname.isEmpty() || !allocations.values.any { it.first==scopedname } ) {"isSameAs scopedname can't be allocated twice"}
+    fun allocate(scopedname: String, datatype: DataType, position: Position?, errors: IErrorReporter): Int {
+        assert(scopedname.isEmpty() || !allocations.values.any { it.first==scopedname } ) {"scopedname can't be allocated twice"}
 
         if(options.zeropage==ZeropageType.DONTUSE)
             throw CompilerException("zero page usage has been disabled")
@@ -39,13 +45,13 @@ abstract class Zeropage(protected val options: CompilationOptions) {
 
         if(free.size > 0) {
             if(size==1) {
-                for(candidate in free.min()!! .. free.max()!!+1) {
+                for(candidate in free.minOrNull()!! .. free.maxOrNull()!!+1) {
                     if(loneByte(candidate))
                         return makeAllocation(candidate, 1, datatype, scopedname)
                 }
                 return makeAllocation(free[0], 1, datatype, scopedname)
             }
-            for(candidate in free.min()!! .. free.max()!!+1) {
+            for(candidate in free.minOrNull()!! .. free.maxOrNull()!!+1) {
                 if (sequentialFree(candidate, size))
                     return makeAllocation(candidate, size, datatype, scopedname)
             }
@@ -58,18 +64,10 @@ abstract class Zeropage(protected val options: CompilationOptions) {
 
     private fun makeAllocation(address: Int, size: Int, datatype: DataType, name: String?): Int {
         free.removeAll(address until address+size)
-        allocations[address] = Pair(name ?: "<unnamed>", datatype)
+        allocations[address] = (name ?: "<unnamed>") to datatype
         return address
     }
 
     private fun loneByte(address: Int) = address in free && address-1 !in free && address+1 !in free
     private fun sequentialFree(address: Int, size: Int) = free.containsAll((address until address+size).toList())
-
-    enum class ExitProgramStrategy {
-        CLEAN_EXIT,
-        SYSTEM_RESET
-    }
-
-    abstract val exitProgramStrategy: ExitProgramStrategy
-
 }

compiler/src/prog8/compiler/astprocessing/AstExtensions.kt

@@ -0,0 +1,97 @@
+package prog8.compiler.astprocessing
+
+import prog8.ast.Program
+import prog8.ast.base.FatalAstException
+import prog8.ast.statements.Directive
+import prog8.compiler.BeforeAsmGenerationAstChanger
+import prog8.compiler.CompilationOptions
+import prog8.compiler.IErrorReporter
+import prog8.compiler.target.ICompilationTarget
+
+
+internal fun Program.checkValid(compilerOptions: CompilationOptions, errors: IErrorReporter, compTarget: ICompilationTarget) {
+    val checker = AstChecker(this, compilerOptions, errors, compTarget)
+    checker.visit(this)
+}
+
+internal fun Program.processAstBeforeAsmGeneration(errors: IErrorReporter, compTarget: ICompilationTarget) {
+    val fixer = BeforeAsmGenerationAstChanger(this, errors, compTarget)
+    fixer.visit(this)
+    fixer.applyModifications()
+}
+
+internal fun Program.reorderStatements(errors: IErrorReporter) {
+    val reorder = StatementReorderer(this, errors)
+    reorder.visit(this)
+    reorder.applyModifications()
+}
+
+internal fun Program.addTypecasts(errors: IErrorReporter) {
+    val caster = TypecastsAdder(this, errors)
+    caster.visit(this)
+    caster.applyModifications()
+}
+
+internal fun Program.verifyFunctionArgTypes() {
+    val fixer = VerifyFunctionArgTypes(this)
+    fixer.visit(this)
+}
+
+internal fun Program.checkIdentifiers(errors: IErrorReporter, compTarget: ICompilationTarget) {
+
+    val checker2 = AstIdentifiersChecker(this, errors, compTarget)
+    checker2.visit(this)
+
+    if(errors.isEmpty()) {
+        val transforms = AstVariousTransforms(this)
+        transforms.visit(this)
+        transforms.applyModifications()
+        val lit2decl = LiteralsToAutoVars(this)
+        lit2decl.visit(this)
+        lit2decl.applyModifications()
+    }
+
+    if (modules.map { it.name }.toSet().size != modules.size) {
+        throw FatalAstException("modules should all be unique")
+    }
+}
+
+internal fun Program.variousCleanups() {
+    val process = VariousCleanups()
+    process.visit(this)
+    process.applyModifications()
+}
+
+internal fun Program.moveMainAndStartToFirst() {
+    // the module containing the program entrypoint is moved to the first in the sequence.
+    // the "main" block containing the entrypoint is moved to the top in there,
+    // and finally the entrypoint subroutine "start" itself is moved to the top in that block.
+
+    val directives = modules[0].statements.filterIsInstance<Directive>()
+    val start = this.entrypoint()
+    if(start!=null) {
+        val mod = start.definingModule()
+        val block = start.definingBlock()
+        if(!modules.remove(mod))
+            throw FatalAstException("module wrong")
+        modules.add(0, mod)
+        mod.remove(block)
+        var afterDirective = mod.statements.indexOfFirst { it !is Directive }
+        if(afterDirective<0)
+            mod.statements.add(block)
+        else
+            mod.statements.add(afterDirective, block)
+        block.remove(start)
+        afterDirective = block.statements.indexOfFirst { it !is Directive }
+        if(afterDirective<0)
+            block.statements.add(start)
+        else
+            block.statements.add(afterDirective, start)
+
+        // overwrite the directives in the module containing the entrypoint
+        for(directive in directives) {
+            modules[0].statements.removeAll { it is Directive && it.directive == directive.directive }
+            modules[0].statements.add(0, directive)
+        }
+    }
+}

compiler/src/prog8/compiler/astprocessing/AstIdentifiersChecker.kt

@@ -1,14 +1,20 @@
-package prog8.ast.processing
+package prog8.compiler.astprocessing
 
 import prog8.ast.Module
 import prog8.ast.Program
-import prog8.ast.base.*
-import prog8.ast.expressions.*
+import prog8.ast.base.DataType
+import prog8.ast.base.NumericDatatypes
+import prog8.ast.base.Position
+import prog8.ast.expressions.ArrayLiteralValue
+import prog8.ast.expressions.NumericLiteralValue
+import prog8.ast.expressions.StringLiteralValue
 import prog8.ast.statements.*
-import prog8.compiler.target.CompilationTarget
-import prog8.functions.BuiltinFunctions
+import prog8.ast.walk.IAstVisitor
+import prog8.compiler.IErrorReporter
+import prog8.compiler.functions.BuiltinFunctions
+import prog8.compiler.target.ICompilationTarget
 
-internal class AstIdentifiersChecker(private val program: Program, private val errors: ErrorReporter) : IAstVisitor {
+internal class AstIdentifiersChecker(private val program: Program, private val errors: IErrorReporter, private val compTarget: ICompilationTarget) : IAstVisitor {
     private var blocks = mutableMapOf<String, Block>()
 
     private fun nameError(name: String, position: Position, existing: Statement) {
@@ -22,22 +28,42 @@ internal class AstIdentifiersChecker(private val program: Program, private val e
     }
 
     override fun visit(block: Block) {
+        if(block.name in compTarget.machine.opcodeNames)
+            errors.err("can't use a cpu opcode name as a symbol: '${block.name}'", block.position)
+
         val existing = blocks[block.name]
         if(existing!=null)
             nameError(block.name, block.position, existing)
         else
             blocks[block.name] = block
 
+        if(!block.isInLibrary) {
+            val libraries = program.modules.filter { it.isLibraryModule }
+            val libraryBlockNames = libraries.flatMap { it.statements.filterIsInstance<Block>().map { b -> b.name } }
+            if(block.name in libraryBlockNames)
+                errors.err("block is already defined in an included library module", block.position)
+        }
+
         super.visit(block)
     }
 
+    override fun visit(directive: Directive) {
+        if(directive.directive=="%target") {
+            val compatibleTarget = directive.args.single().name
+            if (compatibleTarget != compTarget.name)
+                errors.err("module's compilation target ($compatibleTarget) differs from active target (${compTarget.name})", directive.position)
+        }
+
+        super.visit(directive)
+    }
+
     override fun visit(decl: VarDecl) {
         decl.datatypeErrors.forEach { errors.err(it.message, it.position) }
 
         if(decl.name in BuiltinFunctions)
             errors.err("builtin function cannot be redefined", decl.position)
 
-        if(decl.name in CompilationTarget.machine.opcodeNames)
+        if(decl.name in compTarget.machine.opcodeNames)
             errors.err("can't use a cpu opcode name as a symbol: '${decl.name}'", decl.position)
 
         if(decl.datatype==DataType.STRUCT) {
@@ -57,8 +83,8 @@ internal class AstIdentifiersChecker(private val program: Program, private val e
                 return super.visit(decl)
             }
 
-            if (decl.value != null && decl.value !is StructLiteralValue) {
-                errors.err("initializing requires struct literal value", decl.value?.position ?: decl.position)
+            if (decl.value != null && decl.value !is ArrayLiteralValue) {
+                errors.err("initializing a struct requires array literal value", decl.value?.position ?: decl.position)
                 return super.visit(decl)
             }
         }
@@ -67,11 +93,16 @@ internal class AstIdentifiersChecker(private val program: Program, private val e
         if (existing != null && existing !== decl)
             nameError(decl.name, decl.position, existing)
 
+        if(decl.definingBlock().name==decl.name)
+            nameError(decl.name, decl.position, decl.definingBlock())
+        if(decl.definingSubroutine()?.name==decl.name)
+            nameError(decl.name, decl.position, decl.definingSubroutine()!!)
+
         super.visit(decl)
     }
 
     override fun visit(subroutine: Subroutine) {
-        if(subroutine.name in CompilationTarget.machine.opcodeNames) {
+        if(subroutine.name in compTarget.machine.opcodeNames) {
             errors.err("can't use a cpu opcode name as a symbol: '${subroutine.name}'", subroutine.position)
         } else if(subroutine.name in BuiltinFunctions) {
             // the builtin functions can't be redefined
@@ -85,14 +116,6 @@ internal class AstIdentifiersChecker(private val program: Program, private val e
             if (existing != null && existing !== subroutine)
                 nameError(subroutine.name, subroutine.position, existing)
 
-            // does the parameter redefine a variable declared elsewhere?
-            for(param in subroutine.parameters) {
-                val existingVar = subroutine.lookup(listOf(param.name), subroutine)
-                if (existingVar != null && existingVar.parent !== subroutine) {
-                    nameError(param.name, param.position, existingVar)
-                }
-            }
-
             // check that there are no local variables, labels, or other subs that redefine the subroutine's parameters
             val symbolsInSub = subroutine.allDefinedSymbols()
             val namesInSub = symbolsInSub.map{ it.first }.toSet()
@@ -102,9 +125,12 @@ internal class AstIdentifiersChecker(private val program: Program, private val e
                 val labelOrVar = subroutine.getLabelOrVariable(name)
                 if(labelOrVar!=null && labelOrVar.position != subroutine.position)
                     nameError(name, labelOrVar.position, subroutine)
-                val sub = subroutine.statements.singleOrNull { it is Subroutine && it.name==name}
+                val sub = subroutine.statements.firstOrNull { it is Subroutine && it.name==name}
                 if(sub!=null)
-                    nameError(name, sub.position, subroutine)
+                    nameError(name, subroutine.position, sub)
+                val block = program.allBlocks().firstOrNull { it.name==name }
+                if(block!=null)
+                    nameError(name, subroutine.position, block)
             }
 
             if(subroutine.isAsmSubroutine && subroutine.statements.any{it !is InlineAssembly}) {
@@ -116,7 +142,7 @@ internal class AstIdentifiersChecker(private val program: Program, private val e
     }
 
     override fun visit(label: Label) {
-        if(label.name in CompilationTarget.machine.opcodeNames)
+        if(label.name in compTarget.machine.opcodeNames)
             errors.err("can't use a cpu opcode name as a symbol: '${label.name}'", label.position)
 
         if(label.name in BuiltinFunctions) {
@@ -137,24 +163,9 @@ internal class AstIdentifiersChecker(private val program: Program, private val e
         super.visit(label)
     }
 
-    override fun visit(forLoop: ForLoop) {
-        if (forLoop.loopRegister != null) {
-            if (forLoop.loopRegister == Register.X)
-                errors.warn("writing to the X register is dangerous, because it's used as an internal pointer", forLoop.position)
-        }
-
-        super.visit(forLoop)
-    }
-
-    override fun visit(assignTarget: AssignTarget) {
-        if(assignTarget.register== Register.X)
-            errors.warn("writing to the X register is dangerous, because it's used as an internal pointer", assignTarget.position)
-        super.visit(assignTarget)
-    }
-
     override fun visit(string: StringLiteralValue) {
-        if (string.value.length !in 1..255)
-            errors.err("string literal length must be between 1 and 255", string.position)
+        if (string.value.length > 255)
+            errors.err("string literal length max is 255", string.position)
 
         super.visit(string)
     }

compiler/src/prog8/compiler/astprocessing/AstVariousTransforms.kt

@@ -0,0 +1,111 @@
+package prog8.compiler.astprocessing
+
+import prog8.ast.Node
+import prog8.ast.Program
+import prog8.ast.base.DataType
+import prog8.ast.expressions.BinaryExpression
+import prog8.ast.expressions.StringLiteralValue
+import prog8.ast.statements.AnonymousScope
+import prog8.ast.statements.ParameterVarDecl
+import prog8.ast.statements.Subroutine
+import prog8.ast.statements.VarDecl
+import prog8.ast.walk.AstWalker
+import prog8.ast.walk.IAstModification
+
+
+internal class AstVariousTransforms(private val program: Program) : AstWalker() {
+    private val noModifications = emptyList<IAstModification>()
+
+    override fun before(decl: VarDecl, parent: Node): Iterable<IAstModification> {
+        // is it a struct variable? then define all its struct members as mangled names,
+        //    and include the original decl as well.
+        if(decl.datatype==DataType.STRUCT && !decl.structHasBeenFlattened) {
+            val decls = decl.flattenStructMembers()
+            decls.add(decl)
+            val result = AnonymousScope(decls, decl.position)
+            return listOf(IAstModification.ReplaceNode(
+                    decl, result, parent
+            ))
+        }
+
+        return noModifications
+    }
+
+    override fun after(subroutine: Subroutine, parent: Node): Iterable<IAstModification> {
+        // For non-kernal subroutines and non-asm parameters:
+        // inject subroutine params as local variables (if they're not there yet).
+        val symbolsInSub = subroutine.allDefinedSymbols()
+        val namesInSub = symbolsInSub.map{ it.first }.toSet()
+        if(subroutine.asmAddress==null) {
+            if(subroutine.asmParameterRegisters.isEmpty() && subroutine.parameters.isNotEmpty()) {
+                val vars = subroutine.statements.filterIsInstance<VarDecl>().map { it.name }.toSet()
+                if(!vars.containsAll(subroutine.parameters.map{it.name})) {
+                    return subroutine.parameters
+                            .filter { it.name !in namesInSub }
+                            .map {
+                                val vardecl = ParameterVarDecl(it.name, it.type, subroutine.position)
+                                IAstModification.InsertFirst(vardecl, subroutine)
+                            }
+                }
+            }
+        }
+
+        return noModifications
+    }
+
+    override fun after(expr: BinaryExpression, parent: Node): Iterable<IAstModification> {
+        val leftStr = expr.left as? StringLiteralValue
+        val rightStr = expr.right as? StringLiteralValue
+        if(expr.operator == "+") {
+            val concatenatedString = concatString(expr)
+            if(concatenatedString!=null)
+                return listOf(IAstModification.ReplaceNode(expr, concatenatedString, parent))
+        }
+        else if(expr.operator == "*") {
+            if (leftStr!=null) {
+                val amount = expr.right.constValue(program)
+                if(amount!=null) {
+                    val string = leftStr.value.repeat(amount.number.toInt())
+                    val strval = StringLiteralValue(string, leftStr.altEncoding, expr.position)
+                    return listOf(IAstModification.ReplaceNode(expr, strval, parent))
+                }
+            }
+            else if (rightStr!=null) {
+                val amount = expr.right.constValue(program)
+                if(amount!=null) {
+                    val string = rightStr.value.repeat(amount.number.toInt())
+                    val strval = StringLiteralValue(string, rightStr.altEncoding, expr.position)
+                    return listOf(IAstModification.ReplaceNode(expr, strval, parent))
+                }
+            }
+        }
+
+        return noModifications
+    }
+
+    private fun concatString(expr: BinaryExpression): StringLiteralValue? {
+        val rightStrval = expr.right as? StringLiteralValue
+        val leftStrval = expr.left as? StringLiteralValue
+        return when {
+            expr.operator!="+" -> null
+            expr.left is BinaryExpression && rightStrval!=null -> {
+                val subStrVal = concatString(expr.left as BinaryExpression)
+                if(subStrVal==null)
+                    null
+                else
+                    StringLiteralValue("${subStrVal.value}${rightStrval.value}", subStrVal.altEncoding, rightStrval.position)
+            }
+            expr.right is BinaryExpression && leftStrval!=null -> {
+                val subStrVal = concatString(expr.right as BinaryExpression)
+                if(subStrVal==null)
+                    null
+                else
+                    StringLiteralValue("${leftStrval.value}${subStrVal.value}", subStrVal.altEncoding, leftStrval.position)
+            }
+            leftStrval!=null && rightStrval!=null -> {
+                StringLiteralValue("${leftStrval.value}${rightStrval.value}", leftStrval.altEncoding, leftStrval.position)
+            }
+            else -> null
+        }
+    }
+}

compiler/src/prog8/compiler/astprocessing/LiteralsToAutoVars.kt

@@ -0,0 +1,55 @@
+package prog8.compiler.astprocessing
+
+import prog8.ast.Node
+import prog8.ast.Program
+import prog8.ast.base.DataType
+import prog8.ast.expressions.ArrayLiteralValue
+import prog8.ast.expressions.IdentifierReference
+import prog8.ast.expressions.StringLiteralValue
+import prog8.ast.statements.VarDecl
+import prog8.ast.statements.WhenChoice
+import prog8.ast.walk.AstWalker
+import prog8.ast.walk.IAstModification
+
+
+internal class LiteralsToAutoVars(private val program: Program) : AstWalker() {
+    private val noModifications = emptyList<IAstModification>()
+
+    override fun after(string: StringLiteralValue, parent: Node): Iterable<IAstModification> {
+        if(string.parent !is VarDecl && string.parent !is WhenChoice) {
+            // replace the literal string by a identifier reference to the interned string
+            val scopedName = program.internString(string)
+            val identifier = IdentifierReference(scopedName, string.position)
+            return listOf(IAstModification.ReplaceNode(string, identifier, parent))
+        }
+        return noModifications
+    }
+
+    override fun after(array: ArrayLiteralValue, parent: Node): Iterable<IAstModification> {
+        val vardecl = array.parent as? VarDecl
+        if(vardecl!=null) {
+            // adjust the datatype of the array (to an educated guess)
+            val arrayDt = array.type
+            if(!arrayDt.istype(vardecl.datatype)) {
+                val cast = array.cast(vardecl.datatype)
+                if (cast != null && cast !== array)
+                    return listOf(IAstModification.ReplaceNode(vardecl.value!!, cast, vardecl))
+            }
+        } else {
+            val arrayDt = array.guessDatatype(program)
+            if(arrayDt.isKnown) {
+                // this array literal is part of an expression, turn it into an identifier reference
+                val litval2 = array.cast(arrayDt.typeOrElse(DataType.STRUCT))
+                if(litval2!=null) {
+                    val vardecl2 = VarDecl.createAuto(litval2)
+                    val identifier = IdentifierReference(listOf(vardecl2.name), vardecl2.position)
+                    return listOf(
+                            IAstModification.ReplaceNode(array, identifier, parent),
+                            IAstModification.InsertFirst(vardecl2, array.definingScope())
+                    )
+                }
+            }
+        }
+        return noModifications
+    }
+}

compiler/src/prog8/compiler/astprocessing/ReflectionAstWalker.kt

@@ -1,4 +1,4 @@
-package prog8.ast.processing
+package prog8.compiler.astprocessing
 
 
 /*

compiler/src/prog8/compiler/astprocessing/StatementReorderer.kt

@@ -0,0 +1,470 @@
+package prog8.compiler.astprocessing
+
+import prog8.ast.*
+import prog8.ast.base.*
+import prog8.ast.expressions.*
+import prog8.ast.statements.*
+import prog8.ast.walk.AstWalker
+import prog8.ast.walk.IAstModification
+import prog8.compiler.IErrorReporter
+import prog8.compiler.functions.BuiltinFunctions
+
+
+internal class StatementReorderer(val program: Program, val errors: IErrorReporter) : AstWalker() {
+    // Reorders the statements in a way the compiler needs.
+    // - 'main' block must be the very first statement UNLESS it has an address set.
+    // - library blocks are put last.
+    // - blocks are ordered by address, where blocks without address are placed last.
+    // - in every block and module, most directives and vardecls are moved to the top. (not in subroutines!)
+    // - the 'start' subroutine is moved to the top.
+    // - (syntax desugaring) a vardecl with a non-const initializer value is split into a regular vardecl and an assignment statement.
+    // - (syntax desugaring) struct value assignment is expanded into several struct member assignments.
+    // - in-place assignments are reordered a bit so that they are mostly of the form A = A <operator> <rest>
+    // - sorts the choices in when statement.
+    // - insert AddressOf (&) expression where required (string params to a UWORD function param etc).
+
+    private val noModifications = emptyList<IAstModification>()
+    private val directivesToMove = setOf("%output", "%launcher", "%zeropage", "%zpreserved", "%address", "%option")
+
+    override fun after(module: Module, parent: Node): Iterable<IAstModification> {
+        val (blocks, other) = module.statements.partition { it is Block }
+        module.statements = other.asSequence().plus(blocks.sortedBy { (it as Block).address ?: Int.MAX_VALUE }).toMutableList()
+
+        val mainBlock = module.statements.filterIsInstance<Block>().firstOrNull { it.name=="main" }
+        if(mainBlock!=null && mainBlock.address==null) {
+            module.statements.remove(mainBlock)
+            module.statements.add(0, mainBlock)
+        }
+
+        reorderVardeclsAndDirectives(module.statements)
+        return noModifications
+    }
+
+    private fun reorderVardeclsAndDirectives(statements: MutableList<Statement>) {
+        val varDecls = statements.filterIsInstance<VarDecl>()
+        statements.removeAll(varDecls)
+        statements.addAll(0, varDecls)
+
+        val directives = statements.filterIsInstance<Directive>().filter {it.directive in directivesToMove}
+        statements.removeAll(directives)
+        statements.addAll(0, directives)
+    }
+
+    override fun before(block: Block, parent: Node): Iterable<IAstModification> {
+        parent as Module
+        if(block.isInLibrary) {
+            return listOf(
+                    IAstModification.Remove(block, parent),
+                    IAstModification.InsertLast(block, parent)
+            )
+        }
+
+        reorderVardeclsAndDirectives(block.statements)
+        return noModifications
+    }
+
+    override fun before(subroutine: Subroutine, parent: Node): Iterable<IAstModification> {
+        if(subroutine.name=="start" && parent is Block) {
+            if(parent.statements.filterIsInstance<Subroutine>().first().name!="start") {
+                return listOf(
+                        IAstModification.Remove(subroutine, parent),
+                        IAstModification.InsertFirst(subroutine, parent)
+                )
+            }
+        }
+
+        val subs = subroutine.statements.filterIsInstance<Subroutine>()
+        if(subs.isNotEmpty()) {
+            // all subroutines defined within this subroutine are moved to the end
+            return subs.map { IAstModification.Remove(it, subroutine) } +
+                    subs.map { IAstModification.InsertLast(it, subroutine) }
+        }
+
+        return noModifications
+    }
+
+    override fun after(arrayIndexedExpression: ArrayIndexedExpression, parent: Node): Iterable<IAstModification> {
+
+        val arrayVar = arrayIndexedExpression.arrayvar.targetVarDecl(program)
+        if(arrayVar!=null && arrayVar.datatype ==  DataType.UWORD) {
+            // rewrite   pointervar[index]  into  @(pointervar+index)
+            val indexer = arrayIndexedExpression.indexer
+            val index = (indexer.indexNum ?: indexer.indexVar)!!
+            val add = BinaryExpression(arrayIndexedExpression.arrayvar, "+", index, arrayIndexedExpression.position)
+            return if(parent is AssignTarget) {
+                // we're part of the target of an assignment, we have to actually change the assign target itself
+                val memwrite = DirectMemoryWrite(add, arrayIndexedExpression.position)
+                val newtarget = AssignTarget(null, null, memwrite, arrayIndexedExpression.position)
+                listOf(IAstModification.ReplaceNode(parent, newtarget, parent.parent))
+            } else {
+                val memread = DirectMemoryRead(add, arrayIndexedExpression.position)
+                listOf(IAstModification.ReplaceNode(arrayIndexedExpression, memread, parent))
+            }
+        }
+
+        when (val expr2 = arrayIndexedExpression.indexer.origExpression) {
+            is NumericLiteralValue -> {
+                arrayIndexedExpression.indexer.indexNum = expr2
+                arrayIndexedExpression.indexer.origExpression = null
+                return noModifications
+            }
+            is IdentifierReference -> {
+                arrayIndexedExpression.indexer.indexVar = expr2
+                arrayIndexedExpression.indexer.origExpression = null
+                return noModifications
+            }
+            is Expression -> {
+                // replace complex indexing with a temp variable
+                return getAutoIndexerVarFor(arrayIndexedExpression)
+            }
+            else -> return noModifications
+        }
+    }
+
+    override fun after(expr: BinaryExpression, parent: Node): Iterable<IAstModification> {
+        // when using a simple bit shift and assigning it to a variable of a different type,
+        // try to make the bit shifting 'wide enough' to fall into the variable's type.
+        // with this, for instance, uword x = 1 << 10  will result in 1024 rather than 0 (the ubyte result).
+        if(expr.operator=="<<" || expr.operator==">>") {
+            val leftDt = expr.left.inferType(program)
+            when (parent) {
+                is Assignment -> {
+                    val targetDt = parent.target.inferType(program)
+                    if(leftDt != targetDt) {
+                        val cast = TypecastExpression(expr.left, targetDt.typeOrElse(DataType.STRUCT), true, parent.position)
+                        return listOf(IAstModification.ReplaceNode(expr.left, cast, expr))
+                    }
+                }
+                is VarDecl -> {
+                    if(!leftDt.istype(parent.datatype)) {
+                        val cast = TypecastExpression(expr.left, parent.datatype, true, parent.position)
+                        return listOf(IAstModification.ReplaceNode(expr.left, cast, expr))
+                    }
+                }
+                is IFunctionCall -> {
+                    val argnum = parent.args.indexOf(expr)
+                    when (val callee = parent.target.targetStatement(program)) {
+                        is Subroutine -> {
+                            val paramType = callee.parameters[argnum].type
+                            if(leftDt isAssignableTo paramType) {
+                                val cast = TypecastExpression(expr.left, paramType, true, parent.position)
+                                return listOf(IAstModification.ReplaceNode(expr.left, cast, expr))
+                            }
+                        }
+                        is BuiltinFunctionStatementPlaceholder -> {
+                            val func = BuiltinFunctions.getValue(callee.name)
+                            val paramTypes = func.parameters[argnum].possibleDatatypes
+                            for(type in paramTypes) {
+                                if(leftDt isAssignableTo type) {
+                                    val cast = TypecastExpression(expr.left, type, true, parent.position)
+                                    return listOf(IAstModification.ReplaceNode(expr.left, cast, expr))
+                                }
+                            }
+                        }
+                        else -> throw FatalAstException("weird callee")
+                    }
+                }
+                else -> return noModifications
+            }
+        }
+        else if(expr.operator in logicalOperators) {
+            // make sure that logical expressions like "var and other-logical-expression
+            // is rewritten as "var!=0 and other-logical-expression", to avoid bitwise boolean and
+            // generating the wrong results later
+
+            fun wrapped(expr: Expression): Expression =
+                BinaryExpression(expr, "!=", NumericLiteralValue(DataType.UBYTE, 0, expr.position), expr.position)
+
+            fun isLogicalExpr(expr: Expression?): Boolean {
+                if(expr is BinaryExpression && expr.operator in (logicalOperators + comparisonOperators))
+                    return true
+                if(expr is PrefixExpression && expr.operator in logicalOperators)
+                    return true
+                return false
+            }
+
+            return if(isLogicalExpr(expr.left)) {
+                if(isLogicalExpr(expr.right))
+                    noModifications
+                else
+                    listOf(IAstModification.ReplaceNode(expr.right, wrapped(expr.right), expr))
+            } else {
+                if(isLogicalExpr(expr.right))
+                    listOf(IAstModification.ReplaceNode(expr.left, wrapped(expr.left), expr))
+                else {
+                    listOf(
+                        IAstModification.ReplaceNode(expr.left, wrapped(expr.left), expr),
+                        IAstModification.ReplaceNode(expr.right, wrapped(expr.right), expr)
+                    )
+                }
+            }
+        }
+        return noModifications
+    }
+
+    private fun getAutoIndexerVarFor(expr: ArrayIndexedExpression): MutableList<IAstModification> {
+        val modifications = mutableListOf<IAstModification>()
+        val subroutine = expr.definingSubroutine()!!
+        val statement = expr.containingStatement()
+        val indexerVarPrefix = "prog8_autovar_index_"
+        val repo = subroutine.asmGenInfo.usedAutoArrayIndexerForStatements
+
+        // TODO make this a bit smarter so it can reuse indexer variables. BUT BEWARE of scoping+initialization problems then
+        // add another loop index var to be used for this expression
+        val indexerVarName = "$indexerVarPrefix${expr.indexer.hashCode()}"
+        val indexerVar = AsmGenInfo.ArrayIndexerInfo(indexerVarName, expr.indexer)
+        repo.add(indexerVar)
+        // create the indexer var at block level scope
+        val vardecl = VarDecl(VarDeclType.VAR, DataType.UBYTE, ZeropageWish.PREFER_ZEROPAGE,
+                null, indexerVarName, null, null, isArray = false, autogeneratedDontRemove = true, position = expr.position)
+        modifications.add(IAstModification.InsertFirst(vardecl, subroutine))
+
+        // replace the indexer with just the variable
+        // assign the indexing expression to the helper variable, but only if that hasn't been done already
+        val indexerExpression = expr.indexer.origExpression!!
+        val target = AssignTarget(IdentifierReference(listOf(indexerVar.name), indexerExpression.position), null, null, indexerExpression.position)
+        val assign = Assignment(target, indexerExpression, indexerExpression.position)
+        modifications.add(IAstModification.InsertBefore(statement, assign, statement.definingScope()))
+        modifications.add(IAstModification.SetExpression( {
+            expr.indexer.indexVar = it as IdentifierReference
+            expr.indexer.indexNum = null
+            expr.indexer.origExpression = null
+        }, target.identifier!!.copy(), expr.indexer))
+
+        return modifications
+    }
+
+    override fun after(whenStatement: WhenStatement, parent: Node): Iterable<IAstModification> {
+        val choices = whenStatement.choiceValues(program).sortedBy {
+            it.first?.first() ?: Int.MAX_VALUE
+        }
+        whenStatement.choices.clear()
+        choices.mapTo(whenStatement.choices) { it.second }
+        return noModifications
+    }
+
+    override fun after(decl: VarDecl, parent: Node): Iterable<IAstModification> {
+        val declValue = decl.value
+        if(declValue!=null && decl.type== VarDeclType.VAR && decl.datatype in NumericDatatypes) {
+            val declConstValue = declValue.constValue(program)
+            if(declConstValue==null) {
+                // move the vardecl (without value) to the scope and replace this with a regular assignment
+                // Unless we're dealing with a floating point variable because that will actually make things less efficient at the moment (because floats are mostly calcualated via the stack)
+                if(decl.datatype!=DataType.FLOAT) {
+                    decl.value = null
+                    decl.allowInitializeWithZero = false
+                    val target = AssignTarget(IdentifierReference(listOf(decl.name), decl.position), null, null, decl.position)
+                    val assign = Assignment(target, declValue, decl.position)
+                    return listOf(
+                            IAstModification.ReplaceNode(decl, assign, parent),
+                            IAstModification.InsertFirst(decl, decl.definingScope())
+                    )
+                }
+            }
+        }
+        return noModifications
+    }
+
+    override fun before(assignment: Assignment, parent: Node): Iterable<IAstModification> {
+        val valueType = assignment.value.inferType(program)
+        val targetType = assignment.target.inferType(program)
+        var assignments = emptyList<Assignment>()
+
+        if(targetType.istype(DataType.STRUCT) && (valueType.istype(DataType.STRUCT) || valueType.typeOrElse(DataType.STRUCT) in ArrayDatatypes )) {
+            assignments = if (assignment.value is ArrayLiteralValue) {
+                flattenStructAssignmentFromStructLiteral(assignment)    //  'structvar = [ ..... ] '
+            } else {
+                flattenStructAssignmentFromIdentifier(assignment)    //   'structvar1 = structvar2'
+            }
+        }
+
+        if(targetType.typeOrElse(DataType.STRUCT) in ArrayDatatypes && valueType.typeOrElse(DataType.STRUCT) in ArrayDatatypes ) {
+            assignments = if (assignment.value is ArrayLiteralValue) {
+                flattenArrayAssignmentFromArrayLiteral(assignment)    //  'arrayvar = [ ..... ] '
+            } else {
+                flattenArrayAssignmentFromIdentifier(assignment)    //   'arrayvar1 = arrayvar2'
+            }
+        }
+
+        if(assignments.isNotEmpty()) {
+            val modifications = mutableListOf<IAstModification>()
+            val scope = assignment.definingScope()
+            assignments.reversed().mapTo(modifications) { IAstModification.InsertAfter(assignment, it, scope) }
+            modifications.add(IAstModification.Remove(assignment, scope))
+            return modifications
+        }
+
+        return noModifications
+    }
+
+    override fun after(assignment: Assignment, parent: Node): Iterable<IAstModification> {
+        // rewrite in-place assignment expressions a bit so that the assignment target usually is the leftmost operand
+        val binExpr = assignment.value as? BinaryExpression
+        if(binExpr!=null) {
+            if(binExpr.left isSameAs assignment.target) {
+                // A = A <operator> 5, unchanged
+                return noModifications
+            }
+
+            if(binExpr.operator in associativeOperators) {
+                if (binExpr.right isSameAs assignment.target) {
+                    // A = v <associative-operator> A  ==>  A = A <associative-operator> v
+                    return listOf(IAstModification.SwapOperands(binExpr))
+                }
+
+                val leftBinExpr = binExpr.left as? BinaryExpression
+                if(leftBinExpr?.operator == binExpr.operator) {
+                    return if(leftBinExpr.left isSameAs assignment.target) {
+                        // A = (A <associative-operator> x) <same-operator> y ==> A = A <associative-operator> (x <same-operator> y)
+                        val newRight = BinaryExpression(leftBinExpr.right, binExpr.operator, binExpr.right, binExpr.position)
+                        val newValue = BinaryExpression(leftBinExpr.left, binExpr.operator, newRight, binExpr.position)
+                        listOf(IAstModification.ReplaceNode(binExpr, newValue, assignment))
+                    } else {
+                        // A = (x <associative-operator> A) <same-operator> y ==> A = A <associative-operator> (x <same-operator> y)
+                        val newRight = BinaryExpression(leftBinExpr.left, binExpr.operator, binExpr.right, binExpr.position)
+                        val newValue = BinaryExpression(leftBinExpr.right, binExpr.operator, newRight, binExpr.position)
+                        listOf(IAstModification.ReplaceNode(binExpr, newValue, assignment))
+                    }
+                }
+                val rightBinExpr = binExpr.right as? BinaryExpression
+                if(rightBinExpr?.operator == binExpr.operator) {
+                    return if(rightBinExpr.left isSameAs assignment.target) {
+                        // A = x <associative-operator> (A <same-operator> y) ==> A = A <associative-operator> (x <same-operator> y)
+                        val newRight = BinaryExpression(binExpr.left, binExpr.operator, rightBinExpr.right, binExpr.position)
+                        val newValue = BinaryExpression(rightBinExpr.left, binExpr.operator, newRight, binExpr.position)
+                        listOf(IAstModification.ReplaceNode(binExpr, newValue, assignment))
+                    } else {
+                        // A = x <associative-operator> (y <same-operator> A) ==> A = A <associative-operator> (x <same-operator> y)
+                        val newRight = BinaryExpression(binExpr.left, binExpr.operator, rightBinExpr.left, binExpr.position)
+                        val newValue = BinaryExpression(rightBinExpr.right, binExpr.operator, newRight, binExpr.position)
+                        listOf(IAstModification.ReplaceNode(binExpr, newValue, assignment))
+                    }
+                }
+            }
+        }
+
+        return noModifications
+    }
+
+    private fun flattenArrayAssignmentFromArrayLiteral(assign: Assignment): List<Assignment> {
+        val identifier = assign.target.identifier!!
+        val targetVar = identifier.targetVarDecl(program)!!
+        val alv = assign.value as? ArrayLiteralValue
+        return flattenArrayAssign(targetVar, alv, identifier, assign.position)
+    }
+
+    private fun flattenArrayAssignmentFromIdentifier(assign: Assignment): List<Assignment> {
+        val identifier = assign.target.identifier!!
+        val targetVar = identifier.targetVarDecl(program)!!
+        val sourceIdent = assign.value as IdentifierReference
+        val sourceVar = sourceIdent.targetVarDecl(program)!!
+        if(!sourceVar.isArray) {
+            errors.err("value must be an array",  sourceIdent.position)
+            return emptyList()
+        }
+        val alv = sourceVar.value as? ArrayLiteralValue
+        return flattenArrayAssign(targetVar, alv, identifier, assign.position)
+    }
+
+    private fun flattenArrayAssign(targetVar: VarDecl, alv: ArrayLiteralValue?, identifier: IdentifierReference, position: Position): List<Assignment> {
+        if(targetVar.arraysize==null) {
+            errors.err("array has no defined size", identifier.position)
+            return emptyList()
+        }
+
+        if(alv==null || alv.value.size != targetVar.arraysize!!.constIndex()) {
+            errors.err("element count mismatch", position)
+            return emptyList()
+        }
+
+        // TODO use memcopy instead of individual assignments after certain amount of elements
+        // TODO what does assigning a struct var use?
+        return alv.value.mapIndexed { index, value ->
+            val idx = ArrayIndexedExpression(identifier, ArrayIndex(NumericLiteralValue(DataType.UBYTE, index, position), position), position)
+            Assignment(AssignTarget(null, idx, null, position), value, value.position)
+        }
+    }
+
+    private fun flattenStructAssignmentFromStructLiteral(structAssignment: Assignment): List<Assignment> {
+        val identifier = structAssignment.target.identifier!!
+        val identifierName = identifier.nameInSource.single()
+        val targetVar = identifier.targetVarDecl(program)!!
+        val struct = targetVar.struct!!
+
+        val slv = structAssignment.value as? ArrayLiteralValue
+        if(slv==null || slv.value.size != struct.numberOfElements) {
+            errors.err("element count mismatch", structAssignment.position)
+            return emptyList()
+        }
+
+        return struct.statements.zip(slv.value).map { (targetDecl, sourceValue) ->
+            targetDecl as VarDecl
+            val mangled = mangledStructMemberName(identifierName, targetDecl.name)
+            val idref = IdentifierReference(listOf(mangled), structAssignment.position)
+            val assign = Assignment(AssignTarget(idref, null, null, structAssignment.position),
+                    sourceValue, sourceValue.position)
+            assign.linkParents(structAssignment)
+            assign
+        }
+    }
+
+    private fun flattenStructAssignmentFromIdentifier(structAssignment: Assignment): List<Assignment> {
+        // TODO use memcopy beyond a certain number of elements
+        val identifier = structAssignment.target.identifier!!
+        val identifierName = identifier.nameInSource.single()
+        val targetVar = identifier.targetVarDecl(program)!!
+        val struct = targetVar.struct!!
+        when (structAssignment.value) {
+            is IdentifierReference -> {
+                val sourceVar = (structAssignment.value as IdentifierReference).targetVarDecl(program)!!
+                when {
+                    sourceVar.struct!=null -> {
+                        // struct memberwise copy
+                        val sourceStruct = sourceVar.struct!!
+                        if(sourceStruct!==targetVar.struct) {
+                            // structs are not the same in assignment
+                            return listOf()     // error will be printed elsewhere
+                        }
+                        if(struct.statements.size!=sourceStruct.statements.size)
+                            return listOf()     // error will be printed elsewhere
+                        return struct.statements.zip(sourceStruct.statements).map { member ->
+                            val targetDecl = member.first as VarDecl
+                            val sourceDecl = member.second as VarDecl
+                            if(targetDecl.name != sourceDecl.name)
+                                throw FatalAstException("struct member mismatch")
+                            val mangled = mangledStructMemberName(identifierName, targetDecl.name)
+                            val idref = IdentifierReference(listOf(mangled), structAssignment.position)
+                            val sourcemangled = mangledStructMemberName(sourceVar.name, sourceDecl.name)
+                            val sourceIdref = IdentifierReference(listOf(sourcemangled), structAssignment.position)
+                            val assign = Assignment(AssignTarget(idref, null, null, structAssignment.position), sourceIdref, member.second.position)
+                            assign.linkParents(structAssignment)
+                            assign
+                        }
+                    }
+                    sourceVar.isArray -> {
+                        val array = (sourceVar.value as ArrayLiteralValue).value
+                        if(struct.statements.size!=array.size)
+                            return listOf()     // error will be printed elsewhere
+                        return struct.statements.zip(array).map {
+                            val decl = it.first as VarDecl
+                            val mangled = mangledStructMemberName(identifierName, decl.name)
+                            val targetName = IdentifierReference(listOf(mangled), structAssignment.position)
+                            val target = AssignTarget(targetName, null, null, structAssignment.position)
+                            val assign = Assignment(target, it.second, structAssignment.position)
+                            assign.linkParents(structAssignment)
+                            assign
+                        }
+                    }
+                    else -> {
+                        throw FatalAstException("can only assign arrays or structs to structs")
+                    }
+                }
+            }
+            is ArrayLiteralValue -> {
+                throw IllegalArgumentException("not going to flatten a structLv assignment here")
+            }
+            else -> throw FatalAstException("strange struct value")
+        }
+    }
+
+}

compiler/src/prog8/compiler/astprocessing/TypecastsAdder.kt

@@ -1,16 +1,18 @@
-package prog8.ast.processing
+package prog8.compiler.astprocessing
 
 import prog8.ast.IFunctionCall
-import prog8.ast.INameScope
 import prog8.ast.Node
 import prog8.ast.Program
 import prog8.ast.base.*
 import prog8.ast.expressions.*
 import prog8.ast.statements.*
-import prog8.functions.BuiltinFunctions
+import prog8.ast.walk.AstWalker
+import prog8.ast.walk.IAstModification
+import prog8.compiler.IErrorReporter
+import prog8.compiler.functions.BuiltinFunctions
 
 
-class TypecastsAdder(val program: Program, val errors: ErrorReporter) : AstWalker() {
+class TypecastsAdder(val program: Program, val errors: IErrorReporter) : AstWalker() {
     /*
      * Make sure any value assignments get the proper type casts if needed to cast them into the target variable's type.
      * (this includes function call arguments)
@@ -18,6 +20,26 @@ class TypecastsAdder(val program: Program, val errors: ErrorReporter) : AstWalke
 
     private val noModifications = emptyList<IAstModification>()
 
+    override fun after(decl: VarDecl, parent: Node): Iterable<IAstModification> {
+        val declValue = decl.value
+        if(decl.type==VarDeclType.VAR && declValue!=null && decl.struct==null) {
+            val valueDt = declValue.inferType(program)
+            if(!valueDt.istype(decl.datatype)) {
+
+                // don't add a typecast on an array initializer value
+                if(valueDt.typeOrElse(DataType.STRUCT) in IntegerDatatypes && decl.datatype in ArrayDatatypes)
+                    return noModifications
+
+                return listOf(IAstModification.ReplaceNode(
+                        declValue,
+                        TypecastExpression(declValue, decl.datatype, true, declValue.position),
+                        decl
+                ))
+            }
+        }
+        return noModifications
+    }
+
     override fun after(expr: BinaryExpression, parent: Node): Iterable<IAstModification> {
         val leftDt = expr.left.inferType(program)
         val rightDt = expr.right.inferType(program)
@@ -40,19 +62,27 @@ class TypecastsAdder(val program: Program, val errors: ErrorReporter) : AstWalke
     override fun after(assignment: Assignment, parent: Node): Iterable<IAstModification> {
         // see if a typecast is needed to convert the value's type into the proper target type
         val valueItype = assignment.value.inferType(program)
-        val targetItype = assignment.target.inferType(program, assignment)
+        val targetItype = assignment.target.inferType(program)
         if(targetItype.isKnown && valueItype.isKnown) {
             val targettype = targetItype.typeOrElse(DataType.STRUCT)
             val valuetype = valueItype.typeOrElse(DataType.STRUCT)
             if (valuetype != targettype) {
                 if (valuetype isAssignableTo targettype) {
+                    if(valuetype in IterableDatatypes && targettype==DataType.UWORD)
+                        // special case, don't typecast STR/arrays to UWORD, we support those assignments "directly"
+                        return noModifications
                     return listOf(IAstModification.ReplaceNode(
                             assignment.value,
                             TypecastExpression(assignment.value, targettype, true, assignment.value.position),
                             assignment))
                 } else {
-                    fun castLiteral(cvalue: NumericLiteralValue): List<IAstModification.ReplaceNode> =
-                        listOf(IAstModification.ReplaceNode(cvalue, cvalue.cast(targettype), cvalue.parent))
+                    fun castLiteral(cvalue: NumericLiteralValue): List<IAstModification.ReplaceNode> {
+                        val cast = cvalue.cast(targettype)
+                        return if(cast.isValid)
+                            listOf(IAstModification.ReplaceNode(cvalue, cast.valueOrZero(), cvalue.parent))
+                        else
+                            emptyList()
+                    }
                     val cvalue = assignment.value.constValue(program)
                     if(cvalue!=null) {
                         val number = cvalue.number.toDouble()
@@ -78,72 +108,80 @@ class TypecastsAdder(val program: Program, val errors: ErrorReporter) : AstWalke
     }
 
     override fun after(functionCallStatement: FunctionCallStatement, parent: Node): Iterable<IAstModification> {
-        return afterFunctionCallArgs(functionCallStatement, functionCallStatement.definingScope())
+        return afterFunctionCallArgs(functionCallStatement)
     }
 
     override fun after(functionCall: FunctionCall, parent: Node): Iterable<IAstModification> {
-        return afterFunctionCallArgs(functionCall, functionCall.definingScope())
+        return afterFunctionCallArgs(functionCall)
     }
 
-    private fun afterFunctionCallArgs(call: IFunctionCall, scope: INameScope): Iterable<IAstModification> {
+    private fun afterFunctionCallArgs(call: IFunctionCall): Iterable<IAstModification> {
         // see if a typecast is needed to convert the arguments into the required parameter's type
-        return when(val sub = call.target.targetStatement(scope)) {
+        val modifications = mutableListOf<IAstModification>()
+
+        when(val sub = call.target.targetStatement(program)) {
             is Subroutine -> {
-                for(arg in sub.parameters.zip(call.args.withIndex())) {
-                    val argItype = arg.second.value.inferType(program)
+                sub.parameters.zip(call.args).forEachIndexed { index, pair ->
+                    val argItype = pair.second.inferType(program)
                     if(argItype.isKnown) {
                         val argtype = argItype.typeOrElse(DataType.STRUCT)
-                        val requiredType = arg.first.type
+                        val requiredType = pair.first.type
                         if (requiredType != argtype) {
                             if (argtype isAssignableTo requiredType) {
-                                return listOf(IAstModification.ReplaceNode(
-                                        call.args[arg.second.index],
-                                        TypecastExpression(arg.second.value, requiredType, true, arg.second.value.position),
-                                        call as Node))
+                                modifications += IAstModification.ReplaceNode(
+                                        call.args[index],
+                                        TypecastExpression(pair.second, requiredType, true, pair.second.position),
+                                        call as Node)
                             } else if(requiredType == DataType.UWORD && argtype in PassByReferenceDatatypes) {
                                 // we allow STR/ARRAY values in place of UWORD parameters. Take their address instead.
-                                return listOf(IAstModification.ReplaceNode(
-                                        call.args[arg.second.index],
-                                        AddressOf(arg.second.value as IdentifierReference, arg.second.value.position),
-                                        call as Node))
+                                if(pair.second is IdentifierReference) {
+                                    modifications += IAstModification.ReplaceNode(
+                                            call.args[index],
+                                            AddressOf(pair.second as IdentifierReference, pair.second.position),
+                                            call as Node)
+                                }
+                            } else if(pair.second is NumericLiteralValue) {
+                                val cast = (pair.second as NumericLiteralValue).cast(requiredType)
+                                if(cast.isValid)
+                                    modifications += IAstModification.ReplaceNode(
+                                            call.args[index],
+                                            cast.valueOrZero(),
+                                            call as Node)
                             }
                         }
                     }
                 }
-                emptyList()
             }
             is BuiltinFunctionStatementPlaceholder -> {
                 val func = BuiltinFunctions.getValue(sub.name)
-                if(func.pure) {
-                    // non-pure functions don't get automatic typecasts because sometimes they act directly on their parameters
-                    for (arg in func.parameters.zip(call.args.withIndex())) {
-                        val argItype = arg.second.value.inferType(program)
-                        if (argItype.isKnown) {
-                            val argtype = argItype.typeOrElse(DataType.STRUCT)
-                            if (arg.first.possibleDatatypes.any { argtype == it })
-                                continue
-                            for (possibleType in arg.first.possibleDatatypes) {
+                func.parameters.zip(call.args).forEachIndexed { index, pair ->
+                    val argItype = pair.second.inferType(program)
+                    if (argItype.isKnown) {
+                        val argtype = argItype.typeOrElse(DataType.STRUCT)
+                        if (pair.first.possibleDatatypes.all { argtype != it }) {
+                            for (possibleType in pair.first.possibleDatatypes) {
                                 if (argtype isAssignableTo possibleType) {
-                                    return listOf(IAstModification.ReplaceNode(
-                                            call.args[arg.second.index],
-                                            TypecastExpression(arg.second.value, possibleType, true, arg.second.value.position),
-                                            call as Node))
+                                    modifications += IAstModification.ReplaceNode(
+                                            call.args[index],
+                                            TypecastExpression(pair.second, possibleType, true, pair.second.position),
+                                            call as Node)
+                                    break
                                 }
                             }
                         }
                     }
                 }
-                emptyList()
             }
-            null -> emptyList()
-            else -> throw FatalAstException("call to something weird $sub   ${call.target}")
+            else -> { }
         }
+
+        return modifications
     }
 
     override fun after(typecast: TypecastExpression, parent: Node): Iterable<IAstModification> {
         // warn about any implicit type casts to Float, because that may not be intended
         if(typecast.implicit && typecast.type in setOf(DataType.FLOAT, DataType.ARRAY_F)) {
-            errors.warn("byte or word value implicitly converted to float. Suggestion: use explicit cast as float, a float number, or revert to integer arithmetic", typecast.position)
+            errors.warn("integer implicitly converted to float. Suggestion: use float literals, add an explicit cast, or revert to integer arithmetic", typecast.position)
         }
         return noModifications
     }
@@ -152,7 +190,7 @@ class TypecastsAdder(val program: Program, val errors: ErrorReporter) : AstWalke
         // make sure the memory address is an uword
         val dt = memread.addressExpression.inferType(program)
         if(dt.isKnown && dt.typeOrElse(DataType.UWORD)!=DataType.UWORD) {
-            val typecast = (memread.addressExpression as? NumericLiteralValue)?.cast(DataType.UWORD)
+            val typecast = (memread.addressExpression as? NumericLiteralValue)?.cast(DataType.UWORD)?.valueOrZero()
                     ?: TypecastExpression(memread.addressExpression, DataType.UWORD, true, memread.addressExpression.position)
             return listOf(IAstModification.ReplaceNode(memread.addressExpression, typecast, memread))
         }
@@ -163,58 +201,13 @@ class TypecastsAdder(val program: Program, val errors: ErrorReporter) : AstWalke
         // make sure the memory address is an uword
         val dt = memwrite.addressExpression.inferType(program)
         if(dt.isKnown && dt.typeOrElse(DataType.UWORD)!=DataType.UWORD) {
-            val typecast = (memwrite.addressExpression as? NumericLiteralValue)?.cast(DataType.UWORD)
+            val typecast = (memwrite.addressExpression as? NumericLiteralValue)?.cast(DataType.UWORD)?.valueOrZero()
                     ?: TypecastExpression(memwrite.addressExpression, DataType.UWORD, true, memwrite.addressExpression.position)
             return listOf(IAstModification.ReplaceNode(memwrite.addressExpression, typecast, memwrite))
         }
         return noModifications
     }
 
-    override fun after(structLv: StructLiteralValue, parent: Node): Iterable<IAstModification> {
-        // assignment of a struct literal value, some member values may need proper typecast
-
-        fun addTypecastsIfNeeded(struct: StructDecl): Iterable<IAstModification> {
-            val newValues = struct.statements.zip(structLv.values).map { (structMemberDecl, memberValue) ->
-                val memberDt = (structMemberDecl as VarDecl).datatype
-                val valueDt = memberValue.inferType(program)
-                if (valueDt.typeOrElse(memberDt) != memberDt)
-                    TypecastExpression(memberValue, memberDt, true, memberValue.position)
-                else
-                    memberValue
-            }
-
-            class StructLvValueReplacer(val targetStructLv: StructLiteralValue, val typecastValues: List<Expression>) : IAstModification {
-                override fun perform() {
-                    targetStructLv.values = typecastValues
-                    typecastValues.forEach { it.linkParents(targetStructLv) }
-                }
-            }
-
-            return if(structLv.values.zip(newValues).any { (v1, v2) -> v1 !== v2})
-                listOf(StructLvValueReplacer(structLv, newValues))
-            else
-                emptyList()
-        }
-
-        val decl = structLv.parent as? VarDecl
-        if(decl != null) {
-            val struct = decl.struct
-            if(struct != null)
-                return addTypecastsIfNeeded(struct)
-        } else {
-            val assign = structLv.parent as? Assignment
-            if (assign != null) {
-                val decl2 = assign.target.identifier?.targetVarDecl(program.namespace)
-                if(decl2 != null) {
-                    val struct = decl2.struct
-                    if(struct != null)
-                        return addTypecastsIfNeeded(struct)
-                }
-            }
-        }
-        return noModifications
-    }
-
     override fun after(returnStmt: Return, parent: Node): Iterable<IAstModification> {
         // add a typecast to the return type if it doesn't match the subroutine's signature
         val returnValue = returnStmt.value
@@ -225,7 +218,9 @@ class TypecastsAdder(val program: Program, val errors: ErrorReporter) : AstWalke
                 if (returnValue.inferType(program).istype(subReturnType))
                     return noModifications
                 if (returnValue is NumericLiteralValue) {
-                    returnStmt.value = returnValue.cast(subroutine.returntypes.single())
+                    val cast = returnValue.cast(subroutine.returntypes.single())
+                    if(cast.isValid)
+                        returnStmt.value = cast.valueOrZero()
                 } else {
                     return listOf(IAstModification.ReplaceNode(
                             returnValue,

compiler/src/prog8/compiler/astprocessing/VariousCleanups.kt

@@ -0,0 +1,73 @@
+package prog8.compiler.astprocessing
+
+import prog8.ast.IFunctionCall
+import prog8.ast.INameScope
+import prog8.ast.Node
+import prog8.ast.base.Position
+import prog8.ast.expressions.DirectMemoryRead
+import prog8.ast.expressions.FunctionCall
+import prog8.ast.expressions.NumericLiteralValue
+import prog8.ast.expressions.TypecastExpression
+import prog8.ast.statements.*
+import prog8.ast.walk.AstWalker
+import prog8.ast.walk.IAstModification
+
+
+internal class VariousCleanups: AstWalker() {
+    private val noModifications = emptyList<IAstModification>()
+
+    override fun before(nopStatement: NopStatement, parent: Node): Iterable<IAstModification> {
+        return listOf(IAstModification.Remove(nopStatement, parent as INameScope))
+    }
+
+    override fun before(scope: AnonymousScope, parent: Node): Iterable<IAstModification> {
+        return if(parent is INameScope)
+            listOf(ScopeFlatten(scope, parent as INameScope))
+        else
+            noModifications
+    }
+
+    class ScopeFlatten(val scope: AnonymousScope, val into: INameScope) : IAstModification {
+        override fun perform() {
+            val idx = into.statements.indexOf(scope)
+            if(idx>=0) {
+                into.statements.addAll(idx+1, scope.statements)
+                into.statements.remove(scope)
+            }
+        }
+    }
+
+    override fun before(typecast: TypecastExpression, parent: Node): Iterable<IAstModification> {
+        if(typecast.expression is NumericLiteralValue) {
+            val value = (typecast.expression as NumericLiteralValue).cast(typecast.type)
+            if(value.isValid)
+                return listOf(IAstModification.ReplaceNode(typecast, value.valueOrZero(), parent))
+        }
+
+        return noModifications
+    }
+
+    override fun before(functionCallStatement: FunctionCallStatement, parent: Node): Iterable<IAstModification> {
+        return before(functionCallStatement as IFunctionCall, parent, functionCallStatement.position)
+    }
+
+    override fun before(functionCall: FunctionCall, parent: Node): Iterable<IAstModification> {
+        return before(functionCall as IFunctionCall, parent, functionCall.position)
+    }
+
+    private fun before(functionCall: IFunctionCall, parent: Node, position: Position): Iterable<IAstModification> {
+        if(functionCall.target.nameInSource==listOf("peek")) {
+            // peek(a) is synonymous with @(a)
+            val memread = DirectMemoryRead(functionCall.args.single(), position)
+            return listOf(IAstModification.ReplaceNode(functionCall as Node, memread, parent))
+        }
+        if(functionCall.target.nameInSource==listOf("poke")) {
+            // poke(a, v) is synonymous with @(a) = v
+            val tgt = AssignTarget(null, null, DirectMemoryWrite(functionCall.args[0], position), position)
+            val assign = Assignment(tgt, functionCall.args[1], position)
+            return listOf(IAstModification.ReplaceNode(functionCall as Node, assign, parent))
+        }
+        return noModifications
+    }
+
+}

compiler/src/prog8/compiler/astprocessing/VerifyFunctionArgTypes.kt

@@ -0,0 +1,95 @@
+package prog8.compiler.astprocessing
+
+import prog8.ast.IFunctionCall
+import prog8.ast.Program
+import prog8.ast.base.DataType
+import prog8.ast.expressions.Expression
+import prog8.ast.expressions.FunctionCall
+import prog8.ast.expressions.TypecastExpression
+import prog8.ast.statements.*
+import prog8.ast.walk.IAstVisitor
+import prog8.compiler.CompilerException
+import prog8.compiler.functions.BuiltinFunctions
+
+class VerifyFunctionArgTypes(val program: Program) : IAstVisitor {
+
+    override fun visit(functionCall: FunctionCall) {
+        val error = checkTypes(functionCall as IFunctionCall, program)
+        if(error!=null)
+            throw CompilerException(error)
+    }
+
+    override fun visit(functionCallStatement: FunctionCallStatement) {
+        val error = checkTypes(functionCallStatement as IFunctionCall, program)
+        if (error!=null)
+            throw CompilerException(error)
+    }
+
+    companion object {
+
+        private fun argTypeCompatible(argDt: DataType, paramDt: DataType): Boolean {
+            if(argDt==paramDt)
+                return true
+
+            // there are some exceptions that are considered compatible, such as STR <> UWORD
+            if(argDt==DataType.STR && paramDt==DataType.UWORD ||
+                    argDt==DataType.UWORD && paramDt==DataType.STR)
+                return true
+
+            return false
+        }
+
+        fun checkTypes(call: IFunctionCall, program: Program): String? {
+            val argITypes = call.args.map { it.inferType(program) }
+            val firstUnknownDt = argITypes.indexOfFirst { it.isUnknown }
+            if(firstUnknownDt>=0)
+                return "argument ${firstUnknownDt+1} invalid argument type"
+            val argtypes = argITypes.map { it.typeOrElse(DataType.STRUCT) }
+            val target = call.target.targetStatement(program)
+            if (target is Subroutine) {
+                if(call.args.size != target.parameters.size)
+                    return "invalid number of arguments"
+                val paramtypes = target.parameters.map { it.type }
+                val mismatch = argtypes.zip(paramtypes).indexOfFirst { !argTypeCompatible(it.first, it.second) }
+                if(mismatch>=0) {
+                    val actual = argtypes[mismatch].toString()
+                    val expected = paramtypes[mismatch].toString()
+                    return "argument ${mismatch + 1} type mismatch, was: $actual expected: $expected"
+                }
+                if(target.isAsmSubroutine) {
+                    if(target.asmReturnvaluesRegisters.size>1) {
+                        // multiple return values will NOT work inside an expression.
+                        // they MIGHT work in a regular assignment or just a function call statement.
+                        val parent = if(call is Statement) call.parent else if(call is Expression) call.parent else null
+                        if (call !is FunctionCallStatement) {
+                            val checkParent =
+                                if(parent is TypecastExpression)
+                                    parent.parent
+                                else
+                                    parent
+                            if (checkParent !is Assignment && checkParent !is VarDecl) {
+                                return "can't use subroutine call that returns multiple return values here"
+                            }
+                        }
+                    }
+                }
+            }
+            else if (target is BuiltinFunctionStatementPlaceholder) {
+                val func = BuiltinFunctions.getValue(target.name)
+                if(call.args.size != func.parameters.size)
+                    return "invalid number of arguments"
+                val paramtypes = func.parameters.map { it.possibleDatatypes }
+                argtypes.zip(paramtypes).forEachIndexed { index, pair ->
+                    val anyCompatible = pair.second.any { argTypeCompatible(pair.first, it) }
+                    if (!anyCompatible) {
+                        val actual = pair.first.toString()
+                        val expected = pair.second.toString()
+                        return "argument ${index + 1} type mismatch, was: $actual expected: $expected"
+                    }
+                }
+            }
+
+            return null
+        }
+    }
+}

compiler/src/prog8/compiler/functions/BuiltinFunctions.kt

@@ -0,0 +1,488 @@
+package prog8.compiler.functions
+
+import prog8.ast.IMemSizer
+import prog8.ast.Program
+import prog8.ast.base.*
+import prog8.ast.expressions.*
+import prog8.ast.statements.StructDecl
+import prog8.ast.statements.VarDecl
+import prog8.compiler.CompilerException
+import kotlin.math.*
+
+
+class FParam(val name: String, val possibleDatatypes: Set<DataType>)
+
+
+typealias ConstExpressionCaller = (args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer) -> NumericLiteralValue
+
+
+class ReturnConvention(val dt: DataType, val reg: RegisterOrPair?, val floatFac1: Boolean)
+class ParamConvention(val dt: DataType, val reg: RegisterOrPair?, val variable: Boolean)
+class CallConvention(val params: List<ParamConvention>, val returns: ReturnConvention) {
+    override fun toString(): String {
+        val paramConvs =  params.mapIndexed { index, it ->
+            when {
+                it.reg!=null -> "$index:${it.reg}"
+                it.variable -> "$index:variable"
+                else -> "$index:???"
+            }
+        }
+        val returnConv =
+                when {
+                    returns.reg!=null -> returns.reg.toString()
+                    returns.floatFac1 -> "floatFAC1"
+                    else -> "<no returnvalue>"
+                }
+        return "CallConvention[" + paramConvs.joinToString() + " ; returns: $returnConv]"
+    }
+}
+
+
+class FSignature(val name: String,
+                 val pure: Boolean,      // does it have side effects?
+                 val parameters: List<FParam>,
+                 val known_returntype: DataType?,     // specify return type if fixed, otherwise null if it depends on the arguments
+                 val constExpressionFunc: ConstExpressionCaller? = null) {
+
+    fun callConvention(actualParamTypes: List<DataType>): CallConvention {
+        val returns = when(known_returntype) {
+            DataType.UBYTE, DataType.BYTE -> ReturnConvention(known_returntype, RegisterOrPair.A, false)
+            DataType.UWORD, DataType.WORD -> ReturnConvention(known_returntype, RegisterOrPair.AY, false)
+            DataType.FLOAT -> ReturnConvention(known_returntype, null, true)
+            in PassByReferenceDatatypes -> ReturnConvention(known_returntype!!, RegisterOrPair.AY, false)
+            else -> {
+                val paramType = actualParamTypes.first()
+                if(pure)
+                    // return type depends on arg type
+                    when(paramType) {
+                        DataType.UBYTE, DataType.BYTE -> ReturnConvention(paramType, RegisterOrPair.A, false)
+                        DataType.UWORD, DataType.WORD -> ReturnConvention(paramType, RegisterOrPair.AY, false)
+                        DataType.FLOAT -> ReturnConvention(paramType, null, true)
+                        in PassByReferenceDatatypes -> ReturnConvention(paramType, RegisterOrPair.AY, false)
+                        else -> ReturnConvention(paramType, null, false)
+                    }
+                else {
+                    ReturnConvention(paramType, null, false)
+                }
+            }
+        }
+        return when {
+            actualParamTypes.isEmpty() -> CallConvention(emptyList(), returns)
+            actualParamTypes.size==1 -> {
+                // one parameter? via register/registerpair
+                val paramConv = when(val paramType = actualParamTypes[0]) {
+                    DataType.UBYTE, DataType.BYTE -> ParamConvention(paramType, RegisterOrPair.A, false)
+                    DataType.UWORD, DataType.WORD -> ParamConvention(paramType, RegisterOrPair.AY, false)
+                    DataType.FLOAT -> ParamConvention(paramType, RegisterOrPair.AY, false)
+                    in PassByReferenceDatatypes -> ParamConvention(paramType, RegisterOrPair.AY, false)
+                    else -> ParamConvention(paramType, null, false)
+                }
+                CallConvention(listOf(paramConv), returns)
+            }
+            else -> {
+                // multiple parameters? via variables
+                val paramConvs = actualParamTypes.map { ParamConvention(it, null, true) }
+                CallConvention(paramConvs, returns)
+            }
+        }
+    }
+}
+
+@Suppress("UNUSED_ANONYMOUS_PARAMETER")
+private val functionSignatures: List<FSignature> = listOf(
+        // this set of function have no return value and operate in-place:
+    FSignature("rol"         , false, listOf(FParam("item", setOf(DataType.UBYTE, DataType.UWORD))), null),
+    FSignature("ror"         , false, listOf(FParam("item", setOf(DataType.UBYTE, DataType.UWORD))), null),
+    FSignature("rol2"        , false, listOf(FParam("item", setOf(DataType.UBYTE, DataType.UWORD))), null),
+    FSignature("ror2"        , false, listOf(FParam("item", setOf(DataType.UBYTE, DataType.UWORD))), null),
+    FSignature("sort"        , false, listOf(FParam("array", ArrayDatatypes)), null),
+    FSignature("reverse"     , false, listOf(FParam("array", ArrayDatatypes)), null),
+        // these few have a return value depending on the argument(s):
+    FSignature("max"         , true, listOf(FParam("values", ArrayDatatypes)), null) { a, p, prg, ct -> collectionArg(a, p, prg, ::builtinMax) },    // type depends on args
+    FSignature("min"         , true, listOf(FParam("values", ArrayDatatypes)), null) { a, p, prg, ct -> collectionArg(a, p, prg, ::builtinMin) },    // type depends on args
+    FSignature("sum"         , true, listOf(FParam("values", ArrayDatatypes)), null) { a, p, prg, ct -> collectionArg(a, p, prg, ::builtinSum) },      // type depends on args
+    FSignature("abs"         , true, listOf(FParam("value", NumericDatatypes)), null, ::builtinAbs),      // type depends on argument
+    FSignature("len"         , true, listOf(FParam("values", IterableDatatypes)), null, ::builtinLen),    // type is UBYTE or UWORD depending on actual length
+    FSignature("sizeof"      , true, listOf(FParam("object", DataType.values().toSet())), DataType.UBYTE, ::builtinSizeof),
+    FSignature("offsetof"    , true, listOf(FParam("object", DataType.values().toSet())), DataType.UBYTE, ::builtinOffsetof),
+        // normal functions follow:
+    FSignature("sgn"         , true, listOf(FParam("value", NumericDatatypes)), DataType.BYTE, ::builtinSgn ),
+    FSignature("sin"         , true, listOf(FParam("rads", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArg(a, p, prg, Math::sin) },
+    FSignature("sin8"        , true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.BYTE, ::builtinSin8 ),
+    FSignature("sin8u"       , true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.UBYTE, ::builtinSin8u ),
+    FSignature("sin16"       , true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.WORD, ::builtinSin16 ),
+    FSignature("sin16u"      , true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.UWORD, ::builtinSin16u ),
+    FSignature("cos"         , true, listOf(FParam("rads", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArg(a, p, prg, Math::cos) },
+    FSignature("cos8"        , true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.BYTE, ::builtinCos8 ),
+    FSignature("cos8u"       , true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.UBYTE, ::builtinCos8u ),
+    FSignature("cos16"       , true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.WORD, ::builtinCos16 ),
+    FSignature("cos16u"      , true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.UWORD, ::builtinCos16u ),
+    FSignature("tan"         , true, listOf(FParam("rads", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArg(a, p, prg, Math::tan) },
+    FSignature("atan"        , true, listOf(FParam("rads", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArg(a, p, prg, Math::atan) },
+    FSignature("ln"          , true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArg(a, p, prg, Math::log) },
+    FSignature("log2"        , true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArg(a, p, prg, ::log2) },
+    FSignature("sqrt16"      , true, listOf(FParam("value", setOf(DataType.UWORD))), DataType.UBYTE) { a, p, prg, ct -> oneIntArgOutputInt(a, p, prg) { sqrt(it.toDouble()).toInt() } },
+    FSignature("sqrt"        , true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArg(a, p, prg, Math::sqrt) },
+    FSignature("rad"         , true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArg(a, p, prg, Math::toRadians) },
+    FSignature("deg"         , true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArg(a, p, prg, Math::toDegrees) },
+    FSignature("round"       , true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArgOutputWord(a, p, prg, Math::round) },
+    FSignature("floor"       , true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArgOutputWord(a, p, prg, Math::floor) },
+    FSignature("ceil"        , true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg, ct -> oneDoubleArgOutputWord(a, p, prg, Math::ceil) },
+    FSignature("any"         , true, listOf(FParam("values", ArrayDatatypes)), DataType.UBYTE) { a, p, prg, ct -> collectionArg(a, p, prg, ::builtinAny) },
+    FSignature("all"         , true, listOf(FParam("values", ArrayDatatypes)), DataType.UBYTE) { a, p, prg, ct -> collectionArg(a, p, prg, ::builtinAll) },
+    FSignature("lsb"         , true, listOf(FParam("value", setOf(DataType.UWORD, DataType.WORD))), DataType.UBYTE) { a, p, prg, ct -> oneIntArgOutputInt(a, p, prg) { x: Int -> x and 255 } },
+    FSignature("msb"         , true, listOf(FParam("value", setOf(DataType.UWORD, DataType.WORD))), DataType.UBYTE) { a, p, prg, ct -> oneIntArgOutputInt(a, p, prg) { x: Int -> x ushr 8 and 255} },
+    FSignature("mkword"      , true, listOf(FParam("msb", setOf(DataType.UBYTE)), FParam("lsb", setOf(DataType.UBYTE))), DataType.UWORD, ::builtinMkword),
+    FSignature("peek"        , true, listOf(FParam("address", setOf(DataType.UWORD))), DataType.UBYTE),
+    FSignature("peekw"       , true, listOf(FParam("address", setOf(DataType.UWORD))), DataType.UWORD),
+    FSignature("poke"        , false, listOf(FParam("address", setOf(DataType.UWORD)), FParam("value", setOf(DataType.UBYTE))), null),
+    FSignature("pokew"       , false, listOf(FParam("address", setOf(DataType.UWORD)), FParam("value", setOf(DataType.UWORD))), null),
+    FSignature("rnd"         , false, emptyList(), DataType.UBYTE),
+    FSignature("rndw"        , false, emptyList(), DataType.UWORD),
+    FSignature("rndf"        , false, emptyList(), DataType.FLOAT),
+    FSignature("memory"      , true, listOf(FParam("name", setOf(DataType.STR)), FParam("size", setOf(DataType.UWORD))), DataType.UWORD),
+    FSignature("swap"        , false, listOf(FParam("first", NumericDatatypes), FParam("second", NumericDatatypes)), null),
+
+)
+
+val BuiltinFunctions = functionSignatures.associateBy { it.name }
+
+
+fun builtinMax(array: List<Number>): Number = array.maxByOrNull { it.toDouble() }!!
+
+fun builtinMin(array: List<Number>): Number = array.minByOrNull { it.toDouble() }!!
+
+fun builtinSum(array: List<Number>): Number = array.sumByDouble { it.toDouble() }
+
+fun builtinAny(array: List<Number>): Number = if(array.any { it.toDouble()!=0.0 }) 1 else 0
+
+fun builtinAll(array: List<Number>): Number = if(array.all { it.toDouble()!=0.0 }) 1 else 0
+
+
+fun builtinFunctionReturnType(function: String, args: List<Expression>, program: Program): InferredTypes.InferredType {
+
+    fun datatypeFromIterableArg(arglist: Expression): DataType {
+        if(arglist is ArrayLiteralValue) {
+            val dt = arglist.value.map {it.inferType(program).typeOrElse(DataType.STRUCT)}.toSet()
+            if(dt.any { it !in NumericDatatypes }) {
+                throw FatalAstException("fuction $function only accepts array of numeric values")
+            }
+            if(DataType.FLOAT in dt) return DataType.FLOAT
+            if(DataType.UWORD in dt) return DataType.UWORD
+            if(DataType.WORD in dt) return DataType.WORD
+            if(DataType.BYTE in dt) return DataType.BYTE
+            return DataType.UBYTE
+        }
+        if(arglist is IdentifierReference) {
+            val idt = arglist.inferType(program)
+            if(!idt.isKnown)
+                throw FatalAstException("couldn't determine type of iterable $arglist")
+            return when(val dt = idt.typeOrElse(DataType.STRUCT)) {
+                DataType.STR, in NumericDatatypes -> dt
+                in ArrayDatatypes -> ArrayElementTypes.getValue(dt)
+                else -> throw FatalAstException("function '$function' requires one argument which is an iterable")
+            }
+        }
+        throw FatalAstException("function '$function' requires one argument which is an iterable")
+    }
+
+    val func = BuiltinFunctions.getValue(function)
+    if(func.known_returntype!=null)
+        return InferredTypes.knownFor(func.known_returntype)
+
+    // function has return values, but the return type depends on the arguments
+    return when (function) {
+        "abs" -> {
+            val dt = args.single().inferType(program)
+            return if(dt.typeOrElse(DataType.STRUCT) in NumericDatatypes)
+                dt
+            else
+                InferredTypes.InferredType.unknown()
+        }
+        "max", "min" -> {
+            when(val dt = datatypeFromIterableArg(args.single())) {
+                DataType.STR -> InferredTypes.knownFor(DataType.UBYTE)
+                in NumericDatatypes -> InferredTypes.knownFor(dt)
+                in ArrayDatatypes -> InferredTypes.knownFor(ArrayElementTypes.getValue(dt))
+                else -> InferredTypes.unknown()
+            }
+        }
+        "sum" -> {
+            when(datatypeFromIterableArg(args.single())) {
+                DataType.UBYTE, DataType.UWORD -> InferredTypes.knownFor(DataType.UWORD)
+                DataType.BYTE, DataType.WORD -> InferredTypes.knownFor(DataType.WORD)
+                DataType.FLOAT -> InferredTypes.knownFor(DataType.FLOAT)
+                DataType.ARRAY_UB, DataType.ARRAY_UW -> InferredTypes.knownFor(DataType.UWORD)
+                DataType.ARRAY_B, DataType.ARRAY_W -> InferredTypes.knownFor(DataType.WORD)
+                DataType.ARRAY_F -> InferredTypes.knownFor(DataType.FLOAT)
+                DataType.STR -> InferredTypes.knownFor(DataType.UWORD)
+                else -> InferredTypes.unknown()
+            }
+        }
+        "len" -> {
+            // a length can be >255 so in that case, the result is an UWORD instead of an UBYTE
+            // but to avoid a lot of code duplication we simply assume UWORD in all cases for now
+            return InferredTypes.knownFor(DataType.UWORD)
+        }
+        else -> return InferredTypes.unknown()
+    }
+}
+
+
+class NotConstArgumentException: AstException("not a const argument to a built-in function")
+class CannotEvaluateException(func:String, msg: String): FatalAstException("cannot evaluate built-in function $func: $msg")
+
+
+private fun oneDoubleArg(args: List<Expression>, position: Position, program: Program, function: (arg: Double)->Number): NumericLiteralValue {
+    if(args.size!=1)
+        throw SyntaxError("built-in function requires one floating point argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val float = constval.number.toDouble()
+    return numericLiteral(function(float), args[0].position)
+}
+
+private fun oneDoubleArgOutputWord(args: List<Expression>, position: Position, program: Program, function: (arg: Double)->Number): NumericLiteralValue {
+    if(args.size!=1)
+        throw SyntaxError("built-in function requires one floating point argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val float = constval.number.toDouble()
+    return NumericLiteralValue(DataType.WORD, function(float).toInt(), args[0].position)
+}
+
+private fun oneIntArgOutputInt(args: List<Expression>, position: Position, program: Program, function: (arg: Int)->Number): NumericLiteralValue {
+    if(args.size!=1)
+        throw SyntaxError("built-in function requires one integer argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    if(constval.type != DataType.UBYTE && constval.type!= DataType.UWORD)
+        throw SyntaxError("built-in function requires one integer argument", position)
+
+    val integer = constval.number.toInt()
+    return numericLiteral(function(integer).toInt(), args[0].position)
+}
+
+private fun collectionArg(args: List<Expression>, position: Position, program: Program, function: (arg: List<Number>)->Number): NumericLiteralValue {
+    if(args.size!=1)
+        throw SyntaxError("builtin function requires one non-scalar argument", position)
+
+    val array= args[0] as? ArrayLiteralValue ?: throw NotConstArgumentException()
+    val constElements = array.value.map{it.constValue(program)?.number}
+    if(constElements.contains(null))
+        throw NotConstArgumentException()
+
+    return NumericLiteralValue.optimalNumeric(function(constElements.mapNotNull { it }), args[0].position)
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinAbs(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    // 1 arg, type = float or int, result type= isSameAs as argument type
+    if(args.size!=1)
+        throw SyntaxError("abs requires one numeric argument", position)
+
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    return when (constval.type) {
+        in IntegerDatatypes -> numericLiteral(abs(constval.number.toInt()), args[0].position)
+        DataType.FLOAT -> numericLiteral(abs(constval.number.toDouble()), args[0].position)
+        else -> throw SyntaxError("abs requires one numeric argument", position)
+    }
+}
+
+private fun builtinOffsetof(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    // 1 arg, type = anything, result type = ubyte
+    if(args.size!=1)
+        throw SyntaxError("offsetof requires one argument", position)
+    val idref = args[0] as? IdentifierReference
+        ?: throw SyntaxError("offsetof argument should be an identifier", position)
+
+    val vardecl = idref.targetVarDecl(program)!!
+    val struct = vardecl.struct
+    if (struct == null || vardecl.datatype == DataType.STRUCT)
+        throw SyntaxError("offsetof can only be used on struct members", position)
+
+    val membername = idref.nameInSource.last()
+    var offset = 0
+    for(member in struct.statements) {
+        if((member as VarDecl).name == membername)
+            return NumericLiteralValue(DataType.UBYTE, offset, position)
+        offset += memsizer.memorySize(member.datatype)
+    }
+    throw SyntaxError("undefined struct member", position)
+}
+
+private fun builtinSizeof(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    // 1 arg, type = anything, result type = ubyte
+    if(args.size!=1)
+        throw SyntaxError("sizeof requires one argument", position)
+    if(args[0] !is IdentifierReference)
+        throw SyntaxError("sizeof argument should be an identifier", position)
+
+    val dt = args[0].inferType(program)
+    if(dt.isKnown) {
+        val target = (args[0] as IdentifierReference).targetStatement(program)
+                ?: throw CannotEvaluateException("sizeof", "no target")
+
+        fun structSize(target: StructDecl) =
+                NumericLiteralValue(DataType.UBYTE, target.statements.map { memsizer.memorySize((it as VarDecl).datatype) }.sum(), position)
+
+        return when {
+            dt.typeOrElse(DataType.STRUCT) in ArrayDatatypes -> {
+                val length = (target as VarDecl).arraysize!!.constIndex() ?: throw CannotEvaluateException("sizeof", "unknown array size")
+                val elementDt = ArrayElementTypes.getValue(dt.typeOrElse(DataType.STRUCT))
+                numericLiteral(memsizer.memorySize(elementDt) * length, position)
+            }
+            dt.istype(DataType.STRUCT) -> {
+                when (target) {
+                    is VarDecl -> structSize(target.struct!!)
+                    is StructDecl -> structSize(target)
+                    else -> throw CompilerException("weird struct type $target")
+                }
+            }
+            dt.istype(DataType.STR) -> throw SyntaxError("sizeof str is undefined, did you mean len?", position)
+            else -> NumericLiteralValue(DataType.UBYTE, memsizer.memorySize(dt.typeOrElse(DataType.STRUCT)), position)
+        }
+    } else {
+        throw SyntaxError("sizeof invalid argument type", position)
+    }
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinLen(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    // note: in some cases the length is > 255 and then we have to return a UWORD type instead of a UBYTE.
+    if(args.size!=1)
+        throw SyntaxError("len requires one argument", position)
+
+    val directMemVar = ((args[0] as? DirectMemoryRead)?.addressExpression as? IdentifierReference)?.targetVarDecl(program)
+    var arraySize = directMemVar?.arraysize?.constIndex()
+    if(arraySize != null)
+        return NumericLiteralValue.optimalInteger(arraySize, position)
+    if(args[0] is ArrayLiteralValue)
+        return NumericLiteralValue.optimalInteger((args[0] as ArrayLiteralValue).value.size, position)
+    if(args[0] !is IdentifierReference)
+        throw SyntaxError("len argument should be an identifier", position)
+    val target = (args[0] as IdentifierReference).targetVarDecl(program)
+            ?: throw CannotEvaluateException("len", "no target vardecl")
+
+    return when(target.datatype) {
+        DataType.ARRAY_UB, DataType.ARRAY_B, DataType.ARRAY_UW, DataType.ARRAY_W, DataType.ARRAY_F -> {
+            arraySize = target.arraysize?.constIndex()
+            if(arraySize==null)
+                throw CannotEvaluateException("len", "arraysize unknown")
+            NumericLiteralValue.optimalInteger(arraySize, args[0].position)
+        }
+        DataType.STR -> {
+            val refLv = target.value as? StringLiteralValue ?: throw CannotEvaluateException("len", "stringsize unknown")
+            NumericLiteralValue.optimalInteger(refLv.value.length, args[0].position)
+        }
+        DataType.STRUCT -> throw SyntaxError("cannot use len on struct, did you mean sizeof?", args[0].position)
+        in NumericDatatypes -> throw SyntaxError("cannot use len on numeric value, did you mean sizeof?", args[0].position)
+        else -> throw CompilerException("weird datatype")
+    }
+}
+
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinMkword(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    if (args.size != 2)
+        throw SyntaxError("mkword requires msb and lsb arguments", position)
+    val constMsb = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val constLsb = args[1].constValue(program) ?: throw NotConstArgumentException()
+    val result = (constMsb.number.toInt() shl 8) or constLsb.number.toInt()
+    return NumericLiteralValue(DataType.UWORD, result, position)
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinSin8(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    if (args.size != 1)
+        throw SyntaxError("sin8 requires one argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
+    return NumericLiteralValue(DataType.BYTE, (127.0 * sin(rad)).toInt().toShort(), position)
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinSin8u(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    if (args.size != 1)
+        throw SyntaxError("sin8u requires one argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
+    return NumericLiteralValue(DataType.UBYTE, (128.0 + 127.5 * sin(rad)).toInt().toShort(), position)
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinCos8(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    if (args.size != 1)
+        throw SyntaxError("cos8 requires one argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
+    return NumericLiteralValue(DataType.BYTE, (127.0 * cos(rad)).toInt().toShort(), position)
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinCos8u(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    if (args.size != 1)
+        throw SyntaxError("cos8u requires one argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
+    return NumericLiteralValue(DataType.UBYTE, (128.0 + 127.5 * cos(rad)).toInt().toShort(), position)
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinSin16(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    if (args.size != 1)
+        throw SyntaxError("sin16 requires one argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
+    return NumericLiteralValue(DataType.WORD, (32767.0 * sin(rad)).toInt(), position)
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinSin16u(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    if (args.size != 1)
+        throw SyntaxError("sin16u requires one argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
+    return NumericLiteralValue(DataType.UWORD, (32768.0 + 32767.5 * sin(rad)).toInt(), position)
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinCos16(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    if (args.size != 1)
+        throw SyntaxError("cos16 requires one argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
+    return NumericLiteralValue(DataType.WORD, (32767.0 * cos(rad)).toInt(), position)
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinCos16u(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    if (args.size != 1)
+        throw SyntaxError("cos16u requires one argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
+    return NumericLiteralValue(DataType.UWORD, (32768.0 + 32767.5 * cos(rad)).toInt(), position)
+}
+
+@Suppress("UNUSED_PARAMETER")
+private fun builtinSgn(args: List<Expression>, position: Position, program: Program, memsizer: IMemSizer): NumericLiteralValue {
+    if (args.size != 1)
+        throw SyntaxError("sgn requires one argument", position)
+    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
+    return NumericLiteralValue(DataType.BYTE, constval.number.toDouble().sign.toInt().toShort(), position)
+}
+
+private fun numericLiteral(value: Number, position: Position): NumericLiteralValue {
+    val floatNum=value.toDouble()
+    val tweakedValue: Number =
+            if(floatNum== floor(floatNum) && (floatNum>=-32768 && floatNum<=65535))
+                floatNum.toInt()  // we have an integer disguised as a float.
+            else
+                floatNum
+
+    return when(tweakedValue) {
+        is Int -> NumericLiteralValue.optimalInteger(value.toInt(), position)
+        is Short -> NumericLiteralValue.optimalInteger(value.toInt(), position)
+        is Byte -> NumericLiteralValue(DataType.UBYTE, value.toShort(), position)
+        is Double -> NumericLiteralValue(DataType.FLOAT, value.toDouble(), position)
+        is Float -> NumericLiteralValue(DataType.FLOAT, value.toDouble(), position)
+        else -> throw FatalAstException("invalid number type ${value::class}")
+    }
+}

compiler/src/prog8/compiler/target/CompilationTarget.kt

@@ -1,18 +0,0 @@
-package prog8.compiler.target
-
-import prog8.ast.Program
-import prog8.ast.base.ErrorReporter
-import prog8.compiler.CompilationOptions
-import prog8.compiler.Zeropage
-import java.nio.file.Path
-
-
-internal interface CompilationTarget {
-    companion object {
-        lateinit var name: String
-        lateinit var machine: IMachineDefinition
-        lateinit var encodeString: (str: String, altEncoding: Boolean) -> List<Short>
-        lateinit var decodeString: (bytes: List<Short>, altEncoding: Boolean) -> String
-        lateinit var asmGenerator: (Program, ErrorReporter, Zeropage, CompilationOptions, Path) -> IAssemblyGenerator
-    }
-}

compiler/src/prog8/compiler/target/IAssemblyGenerator.kt

@@ -3,10 +3,12 @@ package prog8.compiler.target
 import prog8.compiler.CompilationOptions
 
 internal interface IAssemblyGenerator {
-    fun compileToAssembly(optimize: Boolean): IAssemblyProgram
+    fun compileToAssembly(): IAssemblyProgram
 }
 
 internal const val generatedLabelPrefix = "_prog8_label_"
+internal const val subroutineFloatEvalResultVar1 = "_prog8_float_eval_result1"
+internal const val subroutineFloatEvalResultVar2 = "_prog8_float_eval_result2"
 
 internal interface IAssemblyProgram {
     val name: String

compiler/src/prog8/compiler/target/ICompilationTarget.kt

@@ -0,0 +1,118 @@
+package prog8.compiler.target
+
+import prog8.ast.IMemSizer
+import prog8.ast.IStringEncoding
+import prog8.ast.Program
+import prog8.ast.base.*
+import prog8.ast.expressions.IdentifierReference
+import prog8.ast.expressions.NumericLiteralValue
+import prog8.ast.statements.AssignTarget
+import prog8.compiler.CompilationOptions
+import prog8.compiler.IErrorReporter
+import prog8.compiler.Zeropage
+import prog8.compiler.target.c64.C64MachineDefinition
+import prog8.compiler.target.c64.Petscii
+import prog8.compiler.target.c64.codegen.AsmGen
+import prog8.compiler.target.cx16.CX16MachineDefinition
+import java.nio.file.Path
+
+
+interface ICompilationTarget: IStringEncoding, IMemSizer {
+    val name: String
+    val machine: IMachineDefinition
+    override fun encodeString(str: String, altEncoding: Boolean): List<Short>
+    override fun decodeString(bytes: List<Short>, altEncoding: Boolean): String
+
+    fun isInRegularRAM(target: AssignTarget, program: Program): Boolean {
+        val memAddr = target.memoryAddress
+        val arrayIdx = target.arrayindexed
+        val ident = target.identifier
+        when {
+            memAddr != null -> {
+                return when (memAddr.addressExpression) {
+                    is NumericLiteralValue -> {
+                        machine.isRegularRAMaddress((memAddr.addressExpression as NumericLiteralValue).number.toInt())
+                    }
+                    is IdentifierReference -> {
+                        val decl = (memAddr.addressExpression as IdentifierReference).targetVarDecl(program)
+                        if ((decl?.type == VarDeclType.VAR || decl?.type == VarDeclType.CONST) && decl.value is NumericLiteralValue)
+                            machine.isRegularRAMaddress((decl.value as NumericLiteralValue).number.toInt())
+                        else
+                            false
+                    }
+                    else -> false
+                }
+            }
+            arrayIdx != null -> {
+                val targetStmt = arrayIdx.arrayvar.targetVarDecl(program)
+                return if (targetStmt?.type == VarDeclType.MEMORY) {
+                    val addr = targetStmt.value as? NumericLiteralValue
+                    if (addr != null)
+                        machine.isRegularRAMaddress(addr.number.toInt())
+                    else
+                        false
+                } else true
+            }
+            ident != null -> {
+                val decl = ident.targetVarDecl(program)!!
+                return if (decl.type == VarDeclType.MEMORY && decl.value is NumericLiteralValue)
+                    machine.isRegularRAMaddress((decl.value as NumericLiteralValue).number.toInt())
+                else
+                    true
+            }
+            else -> return true
+        }
+    }
+}
+
+
+internal object C64Target: ICompilationTarget {
+    override val name = "c64"
+    override val machine = C64MachineDefinition
+    override fun encodeString(str: String, altEncoding: Boolean) =
+            if(altEncoding) Petscii.encodeScreencode(str, true) else Petscii.encodePetscii(str, true)
+    override fun decodeString(bytes: List<Short>, altEncoding: Boolean) =
+            if(altEncoding) Petscii.decodeScreencode(bytes, true) else Petscii.decodePetscii(bytes, true)
+
+    override fun memorySize(dt: DataType): Int {
+        return when(dt) {
+            in ByteDatatypes -> 1
+            in WordDatatypes -> 2
+            DataType.FLOAT -> machine.FLOAT_MEM_SIZE
+            in PassByReferenceDatatypes -> machine.POINTER_MEM_SIZE
+            else -> -9999999
+        }
+    }
+}
+
+internal object Cx16Target: ICompilationTarget {
+    override val name = "cx16"
+    override val machine = CX16MachineDefinition
+    override fun encodeString(str: String, altEncoding: Boolean) =
+            if(altEncoding) Petscii.encodeScreencode(str, true) else Petscii.encodePetscii(str, true)
+    override fun decodeString(bytes: List<Short>, altEncoding: Boolean) =
+            if(altEncoding) Petscii.decodeScreencode(bytes, true) else Petscii.decodePetscii(bytes, true)
+
+    override fun memorySize(dt: DataType): Int {
+        return when(dt) {
+            in ByteDatatypes -> 1
+            in WordDatatypes -> 2
+            DataType.FLOAT -> machine.FLOAT_MEM_SIZE
+            in PassByReferenceDatatypes -> machine.POINTER_MEM_SIZE
+            else -> -9999999
+        }
+    }
+}
+
+
+internal fun asmGeneratorFor(
+    compTarget: ICompilationTarget,
+    program: Program,
+    errors: IErrorReporter,
+    zp: Zeropage,
+    options: CompilationOptions,
+    outputDir: Path
+): IAssemblyGenerator
+{
+    return AsmGen(program, errors, zp, options, compTarget, outputDir)
+}

compiler/src/prog8/compiler/target/IMachineDefinition.kt

@@ -4,12 +4,34 @@ import prog8.compiler.CompilationOptions
 import prog8.compiler.Zeropage
 
 
+interface IMachineFloat {
+    fun toDouble(): Double
+    fun makeFloatFillAsm(): String
+}
+
+enum class CpuType {
+    CPU6502,
+    CPU65c02
+}
+
 interface IMachineDefinition {
     val FLOAT_MAX_NEGATIVE: Double
     val FLOAT_MAX_POSITIVE: Double
     val FLOAT_MEM_SIZE: Int
+    val POINTER_MEM_SIZE: Int
+    val ESTACK_LO: Int
+    val ESTACK_HI: Int
+    val BASIC_LOAD_ADDRESS : Int
+    val RAW_LOAD_ADDRESS : Int
 
     val opcodeNames: Set<String>
+    var zeropage: Zeropage
+    val cpu: CpuType
 
-    fun getZeropage(compilerOptions: CompilationOptions): Zeropage
+    fun initializeZeropage(compilerOptions: CompilationOptions)
+    fun getFloat(num: Number): IMachineFloat
+
+    fun importLibs(compilerOptions: CompilationOptions, compilationTargetName: String): List<String>
+    fun launchEmulator(programName: String)
+    fun isRegularRAMaddress(address: Int): Boolean
 }

compiler/src/prog8/compiler/target/c64/AssemblyProgram.kt

@@ -7,27 +7,27 @@ import prog8.compiler.target.generatedLabelPrefix
 import java.nio.file.Path
 import kotlin.system.exitProcess
 
-class AssemblyProgram(override val name: String, outputDir: Path) : IAssemblyProgram {
+class AssemblyProgram(override val name: String, outputDir: Path, private val compTarget: String) : IAssemblyProgram {
     private val assemblyFile = outputDir.resolve("$name.asm")
     private val prgFile = outputDir.resolve("$name.prg")
     private val binFile = outputDir.resolve("$name.bin")
     private val viceMonListFile = outputDir.resolve("$name.vice-mon-list")
 
     override fun assemble(options: CompilationOptions) {
-        // add "-Wlong-branch"  to see warnings about conversion of branch instructions to jumps
+        // add "-Wlong-branch"  to see warnings about conversion of branch instructions to jumps (default = do this silently)
         val command = mutableListOf("64tass", "--ascii", "--case-sensitive", "--long-branch",
-                "-Wall", "-Wno-strict-bool", "-Wno-shadow", "-Werror", "-Wno-error=long-branch",
+                "-Wall", "-Wno-strict-bool", "-Wno-shadow", // "-Werror",
                 "--dump-labels", "--vice-labels", "-l", viceMonListFile.toString(), "--no-monitor")
 
         val outFile = when (options.output) {
             OutputType.PRG -> {
                 command.add("--cbm-prg")
-                println("\nCreating C-64 prg.")
+                println("\nCreating prg for target $compTarget.")
                 prgFile
             }
             OutputType.RAW -> {
                 command.add("--nostart")
-                println("\nCreating raw binary.")
+                println("\nCreating raw binary for target $compTarget.")
                 binFile
             }
         }

compiler/src/prog8/compiler/target/c64/C64MachineDefinition.kt

@@ -1,37 +1,62 @@
 package prog8.compiler.target.c64
 
-import prog8.compiler.CompilationOptions
-import prog8.compiler.CompilerException
-import prog8.compiler.Zeropage
-import prog8.compiler.ZeropageType
+import prog8.compiler.*
+import prog8.compiler.target.CpuType
 import prog8.compiler.target.IMachineDefinition
-import java.awt.Color
-import java.awt.image.BufferedImage
-import javax.imageio.ImageIO
+import prog8.compiler.target.IMachineFloat
+import java.io.IOException
 import kotlin.math.absoluteValue
 import kotlin.math.pow
 
-object C64MachineDefinition: IMachineDefinition {
+internal object C64MachineDefinition: IMachineDefinition {
+
+    override val cpu = CpuType.CPU6502
 
     // 5-byte cbm MFLPT format limitations:
     override val FLOAT_MAX_POSITIVE = 1.7014118345e+38         // bytes: 255,127,255,255,255
     override val FLOAT_MAX_NEGATIVE = -1.7014118345e+38        // bytes: 255,255,255,255,255
     override val FLOAT_MEM_SIZE = 5
-    const val BASIC_LOAD_ADDRESS = 0x0801
-    const val RAW_LOAD_ADDRESS = 0xc000
+    override val POINTER_MEM_SIZE = 2
+    override val BASIC_LOAD_ADDRESS = 0x0801
+    override val RAW_LOAD_ADDRESS = 0xc000
 
     // the 2*256 byte evaluation stack (on which bytes, words, and even floats are stored during calculations)
-    // and some heavily used string constants derived from the two values above
-    const val ESTACK_LO_VALUE       = 0xce00        //  $ce00-$ceff inclusive
-    const val ESTACK_HI_VALUE       = 0xcf00        //  $cf00-$cfff inclusive
-    const val ESTACK_LO_HEX         = "\$ce00"
-    const val ESTACK_LO_PLUS1_HEX   = "\$ce01"
-    const val ESTACK_LO_PLUS2_HEX   = "\$ce02"
-    const val ESTACK_HI_HEX         = "\$cf00"
-    const val ESTACK_HI_PLUS1_HEX   = "\$cf01"
-    const val ESTACK_HI_PLUS2_HEX   = "\$cf02"
+    override val ESTACK_LO = 0xce00        //  $ce00-$ceff inclusive
+    override val ESTACK_HI = 0xcf00        //  $ce00-$ceff inclusive
 
-    override fun getZeropage(compilerOptions: CompilationOptions) = C64Zeropage(compilerOptions)
+    override lateinit var zeropage: Zeropage
+
+    override fun getFloat(num: Number) = Mflpt5.fromNumber(num)
+
+    override fun importLibs(compilerOptions: CompilationOptions,compilationTargetName: String): List<String> {
+        return if (compilerOptions.launcher == LauncherType.BASIC || compilerOptions.output == OutputType.PRG)
+            listOf("syslib")
+        else
+            emptyList()
+    }
+
+    override fun launchEmulator(programName: String) {
+        for(emulator in listOf("x64sc", "x64")) {
+            println("\nStarting C-64 emulator $emulator...")
+            val cmdline = listOf(emulator, "-silent", "-moncommands", "$programName.vice-mon-list",
+                    "-autostartprgmode", "1", "-autostart-warp", "-autostart", programName + ".prg")
+            val processb = ProcessBuilder(cmdline).inheritIO()
+            val process: Process
+            try {
+                process=processb.start()
+            } catch(x: IOException) {
+                continue  // try the next emulator executable
+            }
+            process.waitFor()
+            break
+        }
+    }
+
+    override fun isRegularRAMaddress(address: Int): Boolean = address<0xa000 || address in 0xc000..0xcfff
+
+    override fun initializeZeropage(compilerOptions: CompilationOptions) {
+        zeropage = C64Zeropage(compilerOptions)
+    }
 
     // 6502 opcodes (including aliases and illegal opcodes), these cannot be used as variable or label names
     override val opcodeNames = setOf("adc", "ahx", "alr", "anc", "and", "ane", "arr", "asl", "asr", "axs", "bcc", "bcs",
@@ -45,20 +70,12 @@ object C64MachineDefinition: IMachineDefinition {
             "sta", "stx", "sty", "tas", "tax", "tay", "tsx", "txa", "txs", "tya", "xaa")
 
 
-    class C64Zeropage(options: CompilationOptions) : Zeropage(options) {
+    internal class C64Zeropage(options: CompilationOptions) : Zeropage(options) {
 
-        companion object {
-            const val SCRATCH_B1 = 0x02
-            const val SCRATCH_REG = 0x03    // temp storage for a register
-            const val SCRATCH_REG_X = 0xfa    // temp storage for register X (the evaluation stack pointer)
-            const val SCRATCH_W1 = 0xfb     // $fb+$fc
-            const val SCRATCH_W2 = 0xfd     // $fd+$fe
-        }
-
-        override val exitProgramStrategy: ExitProgramStrategy = when (options.zeropage) {
-            ZeropageType.BASICSAFE, ZeropageType.DONTUSE -> ExitProgramStrategy.CLEAN_EXIT
-            ZeropageType.FLOATSAFE, ZeropageType.KERNALSAFE, ZeropageType.FULL -> ExitProgramStrategy.SYSTEM_RESET
-        }
+        override val SCRATCH_B1 = 0x02      // temp storage for a single byte
+        override val SCRATCH_REG = 0x03     // temp storage for a register, must be B1+1
+        override val SCRATCH_W1 = 0xfb      // temp storage 1 for a word  $fb+$fc
+        override val SCRATCH_W2 = 0xfd      // temp storage 2 for a word  $fb+$fc
 
 
         init {
@@ -68,12 +85,14 @@ object C64MachineDefinition: IMachineDefinition {
             if (options.zeropage == ZeropageType.FULL) {
                 free.addAll(0x04..0xf9)
                 free.add(0xff)
-                free.removeAll(listOf(SCRATCH_B1, SCRATCH_REG, SCRATCH_REG_X, SCRATCH_W1, SCRATCH_W1 + 1, SCRATCH_W2, SCRATCH_W2 + 1))
+                free.removeAll(listOf(SCRATCH_B1, SCRATCH_REG, SCRATCH_W1, SCRATCH_W1 + 1, SCRATCH_W2, SCRATCH_W2 + 1))
                 free.removeAll(listOf(0xa0, 0xa1, 0xa2, 0x91, 0xc0, 0xc5, 0xcb, 0xf5, 0xf6))        // these are updated by IRQ
             } else {
                 if (options.zeropage == ZeropageType.KERNALSAFE || options.zeropage == ZeropageType.FLOATSAFE) {
-                    free.addAll(listOf(0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, 0x10, 0x11,
-                            0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x20, 0x21,
+                    free.addAll(listOf(
+                            0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, 0x10, 0x11,
+                            0x16, 0x17, 0x18, 0x19, 0x1a,
+                            0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x20, 0x21,
                             0x22, 0x23, 0x24, 0x25,
                             0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0x3e, 0x3f, 0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46,
                             0x47, 0x48, 0x4b, 0x4c, 0x4d, 0x4e, 0x4f, 0x51, 0x52, 0x53,
@@ -83,45 +102,43 @@ object C64MachineDefinition: IMachineDefinition {
                             0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, 0x7a, 0x7b, 0x7c,
                             0x7d, 0x7e, 0x7f, 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, 0x8a,
                             0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0xff
-                            // 0x90-0xfa is 'kernel work storage area'
+                            // 0x90-0xfa is 'kernal work storage area'
                     ))
                 }
 
                 if (options.zeropage == ZeropageType.FLOATSAFE) {
                     // remove the zero page locations used for floating point operations from the free list
                     free.removeAll(listOf(
-                            0x12, 0x26, 0x27, 0x28, 0x29, 0x2a,
+                            0x10, 0x11, 0x12, 0x26, 0x27, 0x28, 0x29, 0x2a,
                             0x57, 0x58, 0x59, 0x5a, 0x5b, 0x5c, 0x5d, 0x5e, 0x5f, 0x60,
                             0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68,
                             0x69, 0x6a, 0x6b, 0x6c, 0x6d, 0x6e, 0x6f, 0x70, 0x71, 0x72,
-                            0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0xf
+                            0x8b, 0x8c, 0x8d, 0x8e, 0x8f, 0xff
                     ))
                 }
 
                 if(options.zeropage!=ZeropageType.DONTUSE) {
-                    // add the other free Zp addresses,
-                    // these are valid for the C-64 (when no RS232 I/O is performed) but to keep BASIC running fully:
+                    // add the free Zp addresses
+                    // these are valid for the C-64 but allow BASIC to keep running fully *as long as you don't use tape I/O*
                     free.addAll(listOf(0x04, 0x05, 0x06, 0x0a, 0x0e,
-                            0x94, 0x95, 0xa7, 0xa8, 0xa9, 0xaa,
-                            0xb5, 0xb6, 0xf7, 0xf8, 0xf9))
+                            0x92, 0x96, 0x9b, 0x9c, 0x9e, 0x9f, 0xa5, 0xa6,
+                            0xb0, 0xb1, 0xbe, 0xbf, 0xf9))
                 } else {
                     // don't use the zeropage at all
                     free.clear()
                 }
             }
-            assert(SCRATCH_B1 !in free)
-            assert(SCRATCH_REG !in free)
-            assert(SCRATCH_REG_X !in free)
-            assert(SCRATCH_W1 !in free)
-            assert(SCRATCH_W2 !in free)
+            require(SCRATCH_B1 !in free)
+            require(SCRATCH_REG !in free)
+            require(SCRATCH_W1 !in free)
+            require(SCRATCH_W2 !in free)
 
             for (reserved in options.zpReserved)
                 reserve(reserved)
         }
     }
 
-
-    data class Mflpt5(val b0: Short, val b1: Short, val b2: Short, val b3: Short, val b4: Short) {
+    internal data class Mflpt5(val b0: Short, val b1: Short, val b2: Short, val b3: Short, val b4: Short): IMachineFloat {
 
         companion object {
             val zero = Mflpt5(0, 0, 0, 0, 0)
@@ -168,7 +185,7 @@ object C64MachineDefinition: IMachineDefinition {
             }
         }
 
-        fun toDouble(): Double {
+        override fun toDouble(): Double {
             if (this == zero) return 0.0
             val exp = b0 - 128
             val sign = (b1.toInt() and 0x80) > 0
@@ -176,91 +193,14 @@ object C64MachineDefinition: IMachineDefinition {
             val result = number.toDouble() * (2.0).pow(exp) / 0x100000000
             return if (sign) -result else result
         }
-    }
 
-    object Charset {
-        private val normalImg = ImageIO.read(javaClass.getResource("/charset/c64/charset-normal.png"))
-        private val shiftedImg = ImageIO.read(javaClass.getResource("/charset/c64/charset-shifted.png"))
-
-        private fun scanChars(img: BufferedImage): Array<BufferedImage> {
-
-            val transparent = BufferedImage(img.width, img.height, BufferedImage.TYPE_INT_ARGB)
-            transparent.createGraphics().drawImage(img, 0, 0, null)
-
-            val black = Color(0, 0, 0).rgb
-            val nopixel = Color(0, 0, 0, 0).rgb
-            for (y in 0 until transparent.height) {
-                for (x in 0 until transparent.width) {
-                    val col = transparent.getRGB(x, y)
-                    if (col == black)
-                        transparent.setRGB(x, y, nopixel)
-                }
-            }
-
-            val numColumns = transparent.width / 8
-            val charImages = (0..255).map {
-                val charX = it % numColumns
-                val charY = it / numColumns
-                transparent.getSubimage(charX * 8, charY * 8, 8, 8)
-            }
-            return charImages.toTypedArray()
+        override fun makeFloatFillAsm(): String {
+            val b0 = "$" + b0.toString(16).padStart(2, '0')
+            val b1 = "$" + b1.toString(16).padStart(2, '0')
+            val b2 = "$" + b2.toString(16).padStart(2, '0')
+            val b3 = "$" + b3.toString(16).padStart(2, '0')
+            val b4 = "$" + b4.toString(16).padStart(2, '0')
+            return "$b0, $b1, $b2, $b3, $b4"
         }
-
-        val normalChars = scanChars(normalImg)
-        val shiftedChars = scanChars(shiftedImg)
-
-        private val coloredNormalChars = mutableMapOf<Short, Array<BufferedImage>>()
-
-        fun getColoredChar(screenCode: Short, color: Short): BufferedImage {
-            val colorIdx = (color % colorPalette.size).toShort()
-            val chars = coloredNormalChars[colorIdx]
-            if (chars != null)
-                return chars[screenCode.toInt()]
-
-            val coloredChars = mutableListOf<BufferedImage>()
-            val transparent = Color(0, 0, 0, 0).rgb
-            val rgb = colorPalette[colorIdx.toInt()].rgb
-            for (c in normalChars) {
-                val colored = c.copy()
-                for (y in 0 until colored.height)
-                    for (x in 0 until colored.width) {
-                        if (colored.getRGB(x, y) != transparent) {
-                            colored.setRGB(x, y, rgb)
-                        }
-                    }
-                coloredChars.add(colored)
-            }
-            coloredNormalChars[colorIdx] = coloredChars.toTypedArray()
-            return coloredNormalChars.getValue(colorIdx)[screenCode.toInt()]
-        }
-
     }
-
-    private fun BufferedImage.copy(): BufferedImage {
-        val bcopy = BufferedImage(this.width, this.height, this.type)
-        val g = bcopy.graphics
-        g.drawImage(this, 0, 0, null)
-        g.dispose()
-        return bcopy
-    }
-
-    val colorPalette = listOf(         // this is Pepto's Commodore-64 palette  http://www.pepto.de/projects/colorvic/
-            Color(0x000000),  // 0 = black
-            Color(0xFFFFFF),  // 1 = white
-            Color(0x813338),  // 2 = red
-            Color(0x75cec8),  // 3 = cyan
-            Color(0x8e3c97),  // 4 = purple
-            Color(0x56ac4d),  // 5 = green
-            Color(0x2e2c9b),  // 6 = blue
-            Color(0xedf171),  // 7 = yellow
-            Color(0x8e5029),  // 8 = orange
-            Color(0x553800),  // 9 = brown
-            Color(0xc46c71),  // 10 = light red
-            Color(0x4a4a4a),  // 11 = dark grey
-            Color(0x7b7b7b),  // 12 = medium grey
-            Color(0xa9ff9f),  // 13 = light green
-            Color(0x706deb),  // 14 = light blue
-            Color(0xb2b2b2)   // 15 = light grey
-    )
-
 }

compiler/src/prog8/compiler/target/c64/Petscii.kt

@@ -172,7 +172,7 @@ object Petscii {
             '\u258c',    //  ▌    0xA1 -> LEFT HALF BLOCK
             '\u2584',    //  ▄    0xA2 -> LOWER HALF BLOCK
             '\u2594',    //  ▔    0xA3 -> UPPER ONE EIGHTH BLOCK
-            '\u2581',    //  ▁    0xA4 -> LOWER ONE EIGHTH BLOCK
+            '_',         //  ▁    0xA4 -> LOWER ONE EIGHTH BLOCK
             '\u258f',    //  ▏    0xA5 -> LEFT ONE EIGHTH BLOCK
             '\u2592',    //  ▒    0xA6 -> MEDIUM SHADE
             '\u2595',    //  ▕    0xA7 -> RIGHT ONE EIGHTH BLOCK
@@ -236,7 +236,7 @@ object Petscii {
             '\u258c',    //  ▌    0xE1 -> LEFT HALF BLOCK
             '\u2584',    //  ▄    0xE2 -> LOWER HALF BLOCK
             '\u2594',    //  ▔    0xE3 -> UPPER ONE EIGHTH BLOCK
-            '\u2581',    //  ▁    0xE4 -> LOWER ONE EIGHTH BLOCK
+            '_',         //  ▁    0xE4 -> LOWER ONE EIGHTH BLOCK
             '\u258f',    //  ▏    0xE5 -> LEFT ONE EIGHTH BLOCK
             '\u2592',    //  ▒    0xE6 -> MEDIUM SHADE
             '\u2595',    //  ▕    0xE7 -> RIGHT ONE EIGHTH BLOCK
@@ -431,7 +431,7 @@ object Petscii {
             '\u258c',    //  ▌    0xA1 -> LEFT HALF BLOCK
             '\u2584',    //  ▄    0xA2 -> LOWER HALF BLOCK
             '\u2594',    //  ▔    0xA3 -> UPPER ONE EIGHTH BLOCK
-            '\u2581',    //  ▁    0xA4 -> LOWER ONE EIGHTH BLOCK
+            '_',         //  ▁    0xA4 -> LOWER ONE EIGHTH BLOCK
             '\u258f',    //  ▏    0xA5 -> LEFT ONE EIGHTH BLOCK
             '\u2592',    //  ▒    0xA6 -> MEDIUM SHADE
             '\u2595',    //  ▕    0xA7 -> RIGHT ONE EIGHTH BLOCK
@@ -495,7 +495,7 @@ object Petscii {
             '\u258c',    //  ▌    0xE1 -> LEFT HALF BLOCK
             '\u2584',    //  ▄    0xE2 -> LOWER HALF BLOCK
             '\u2594',    //  ▔    0xE3 -> UPPER ONE EIGHTH BLOCK
-            '\u2581',    //  ▁    0xE4 -> LOWER ONE EIGHTH BLOCK
+            '_',         //  ▁    0xE4 -> LOWER ONE EIGHTH BLOCK
             '\u258f',    //  ▏    0xE5 -> LEFT ONE EIGHTH BLOCK
             '\u2592',    //  ▒    0xE6 -> MEDIUM SHADE
             '\u2595',    //  ▕    0xE7 -> RIGHT ONE EIGHTH BLOCK
@@ -626,7 +626,7 @@ object Petscii {
             '\u258c',    //  ▌    0x61 -> LEFT HALF BLOCK
             '\u2584',    //  ▄    0x62 -> LOWER HALF BLOCK
             '\u2594',    //  ▔    0x63 -> UPPER ONE EIGHTH BLOCK
-            '\u2581',    //  ▁    0x64 -> LOWER ONE EIGHTH BLOCK
+            '_',         //  ▁    0x64 -> LOWER ONE EIGHTH BLOCK
             '\u258f',    //  ▏    0x65 -> LEFT ONE EIGHTH BLOCK
             '\u2592',    //  ▒    0x66 -> MEDIUM SHADE
             '\u2595',    //  ▕    0x67 -> RIGHT ONE EIGHTH BLOCK
@@ -885,7 +885,7 @@ object Petscii {
             '\u258c',    //  ▌    0x61 -> LEFT HALF BLOCK
             '\u2584',    //  ▄    0x62 -> LOWER HALF BLOCK
             '\u2594',    //  ▔    0x63 -> UPPER ONE EIGHTH BLOCK
-            '\u2581',    //  ▁    0x64 -> LOWER ONE EIGHTH BLOCK
+            '_',         //  ▁    0x64 -> LOWER ONE EIGHTH BLOCK
             '\u258f',    //  ▏    0x65 -> LEFT ONE EIGHTH BLOCK
             '\u2592',    //  ▒    0x66 -> MEDIUM SHADE
             '\u2595',    //  ▕    0x67 -> RIGHT ONE EIGHTH BLOCK
@@ -1054,11 +1054,16 @@ object Petscii {
         val lookup = if(lowercase) encodingPetsciiLowercase else encodingPetsciiUppercase
         return text.map {
             val petscii = lookup[it]
-            petscii?.toShort() ?: if(it=='\u0000')
-                0.toShort()
-            else {
-                val case = if (lowercase) "lower" else "upper"
-                throw CharConversionException("no ${case}case Petscii character for '$it' (${it.toShort()})")
+            petscii?.toShort() ?: when (it) {
+                '\u0000' -> 0.toShort()
+                in '\u8000'..'\u80ff' -> {
+                    // special case: take the lower 8 bit hex value directly
+                    (it.toInt() - 0x8000).toShort()
+                }
+                else -> {
+                    val case = if (lowercase) "lower" else "upper"
+                    throw CharConversionException("no ${case}case Petscii character for '$it' (${it.toShort()})")
+                }
             }
         }
     }
@@ -1072,11 +1077,16 @@ object Petscii {
         val lookup = if(lowercase) encodingScreencodeLowercase else encodingScreencodeUppercase
         return text.map{
             val screencode = lookup[it]
-            screencode?.toShort() ?: if(it=='\u0000')
-                0.toShort()
-            else {
-                val case = if (lowercase) "lower" else "upper"
-                throw CharConversionException("no ${case}Screencode character for '$it' (${it.toShort()})")
+            screencode?.toShort() ?: when (it) {
+                '\u0000' -> 0.toShort()
+                in '\u8000'..'\u80ff' -> {
+                    // special case: take the lower 8 bit hex value directly
+                    (it.toInt() - 0x8000).toShort()
+                }
+                else -> {
+                    val case = if (lowercase) "lower" else "upper"
+                    throw CharConversionException("no ${case}Screencode character for '$it' (${it.toShort()})")
+                }
             }
         }
     }

compiler/src/prog8/compiler/target/c64/codegen/AsmOptimizer.kt

@@ -1,8 +1,5 @@
 package prog8.compiler.target.c64.codegen
 
-import prog8.compiler.target.c64.C64MachineDefinition.ESTACK_LO_HEX
-import prog8.compiler.target.c64.C64MachineDefinition.ESTACK_LO_PLUS1_HEX
-
 
 // note: see https://wiki.nesdev.com/w/index.php/6502_assembly_optimisations
 
@@ -78,19 +75,19 @@ private fun getLinesBy(lines: MutableList<String>, windowSize: Int) =
 
 private fun optimizeCmpSequence(linesByFour: List<List<IndexedValue<String>>>): List<Modification> {
     // the when statement (on bytes) generates a sequence of:
-    //	 lda $ce01,x
-    //	 cmp #$20
+    //	 lda  $ce01,x
+    //	 cmp  #$20
     //	 beq  check_prog8_s72choice_32
-    //	 lda $ce01,x
-    //	 cmp #$21
+    //	 lda  $ce01,x
+    //	 cmp  #$21
     //	 beq  check_prog8_s73choice_33
     // the repeated lda can be removed
     val mods = mutableListOf<Modification>()
     for(lines in linesByFour) {
-        if(lines[0].value.trim()=="lda  $ESTACK_LO_PLUS1_HEX,x" &&
+        if(lines[0].value.trim()=="lda  P8ESTACK_LO+1,x" &&
                 lines[1].value.trim().startsWith("cmp ") &&
                 lines[2].value.trim().startsWith("beq ") &&
-                lines[3].value.trim()=="lda  $ESTACK_LO_PLUS1_HEX,x") {
+                lines[3].value.trim()=="lda  P8ESTACK_LO+1,x") {
             mods.add(Modification(lines[3].index, true, null)) // remove the second lda
         }
     }
@@ -102,10 +99,10 @@ private fun optimizeUselessStackByteWrites(linesByFour: List<List<IndexedValue<S
     // this is a lot harder for word values because the instruction sequence varies.
     val mods = mutableListOf<Modification>()
     for(lines in linesByFour) {
-        if(lines[0].value.trim()=="sta  $ESTACK_LO_HEX,x" &&
+        if(lines[0].value.trim()=="sta  P8ESTACK_LO,x" &&
                 lines[1].value.trim()=="dex" &&
                 lines[2].value.trim()=="inx" &&
-                lines[3].value.trim()=="lda  $ESTACK_LO_HEX,x") {
+                lines[3].value.trim()=="lda  P8ESTACK_LO,x") {
             mods.add(Modification(lines[1].index, true, null))
             mods.add(Modification(lines[2].index, true, null))
             mods.add(Modification(lines[3].index, true, null))
@@ -154,7 +151,8 @@ private fun optimizeSameAssignments(linesByFourteen: List<List<IndexedValue<Stri
         }
 
         if(first.startsWith("lda") && second.startsWith("ldy") && third.startsWith("sta") && fourth.startsWith("sty") &&
-                fifth.startsWith("lda") && sixth.startsWith("ldy") && seventh.startsWith("jsr  c64flt.copy_float")) {
+                fifth.startsWith("lda") && sixth.startsWith("ldy") &&
+                (seventh.startsWith("jsr  floats.copy_float") || seventh.startsWith("jsr  cx16flt.copy_float"))) {
 
             val nineth = pair[8].value.trimStart()
             val tenth = pair[9].value.trimStart()
@@ -164,7 +162,8 @@ private fun optimizeSameAssignments(linesByFourteen: List<List<IndexedValue<Stri
             val fourteenth = pair[13].value.trimStart()
 
             if(eighth.startsWith("lda") && nineth.startsWith("ldy") && tenth.startsWith("sta") && eleventh.startsWith("sty") &&
-                    twelveth.startsWith("lda") && thirteenth.startsWith("ldy") && fourteenth.startsWith("jsr  c64flt.copy_float")) {
+                    twelveth.startsWith("lda") && thirteenth.startsWith("ldy") &&
+                    (fourteenth.startsWith("jsr  floats.copy_float") || fourteenth.startsWith("jsr  cx16flt.copy_float"))) {
 
                 if(first.substring(4) == eighth.substring(4) && second.substring(4)==nineth.substring(4)) {
                     // identical float init
@@ -180,7 +179,8 @@ private fun optimizeSameAssignments(linesByFourteen: List<List<IndexedValue<Stri
 }
 
 private fun optimizeStoreLoadSame(linesByFour: List<List<IndexedValue<String>>>): List<Modification> {
-    // sta X + lda X,  sty X + ldy X,   stx X + ldx X  -> the second instruction can be eliminated
+    // sta X + lda X,  sty X + ldy X,   stx X + ldx X  -> the second instruction can OFTEN be eliminated
+    // TODO this is not true if X is not a regular RAM memory address (but instead mapped I/O or ROM)
     val mods = mutableListOf<Modification>()
     for (pair in linesByFour) {
         val first = pair[0].value.trimStart()
@@ -196,10 +196,14 @@ private fun optimizeStoreLoadSame(linesByFour: List<List<IndexedValue<String>>>)
                 (first.startsWith("sty ") && second.startsWith("ldy ")) ||
                 (first.startsWith("stx ") && second.startsWith("ldx "))
         ) {
-            val firstLoc = first.substring(4)
-            val secondLoc = second.substring(4)
-            if (firstLoc == secondLoc) {
-                mods.add(Modification(pair[1].index, true, null))
+            val third = pair[2].value.trimStart()
+            if(!third.startsWith("b")) {
+                // no branch instruction follows, we can potentiall remove the load instruction
+                val firstLoc = first.substring(4).trimStart()
+                val secondLoc = second.substring(4).trimStart()
+                if (firstLoc == secondLoc) {
+                    mods.add(Modification(pair[1].index, true, null))
+                }
             }
         }
     }

compiler/src/prog8/compiler/target/c64/codegen/FunctionCallAsmGen.kt

@@ -1,238 +1,333 @@
 package prog8.compiler.target.c64.codegen
 
 import prog8.ast.IFunctionCall
+import prog8.ast.Node
 import prog8.ast.Program
 import prog8.ast.base.*
 import prog8.ast.expressions.*
-import prog8.ast.statements.AssignTarget
-import prog8.ast.statements.Subroutine
-import prog8.ast.statements.SubroutineParameter
+import prog8.ast.statements.*
 import prog8.compiler.AssemblyError
-import prog8.compiler.target.c64.C64MachineDefinition.ESTACK_HI_HEX
-import prog8.compiler.target.c64.C64MachineDefinition.ESTACK_LO_HEX
-import prog8.compiler.toHex
+import prog8.compiler.target.CpuType
+import prog8.compiler.target.c64.codegen.assignment.AsmAssignSource
+import prog8.compiler.target.c64.codegen.assignment.AsmAssignTarget
+import prog8.compiler.target.c64.codegen.assignment.AsmAssignment
+import prog8.compiler.target.c64.codegen.assignment.TargetStorageKind
 
 
 internal class FunctionCallAsmGen(private val program: Program, private val asmgen: AsmGen) {
 
+    internal fun translateFunctionCallStatement(stmt: IFunctionCall) {
+        saveXbeforeCall(stmt)
+        translateFunctionCall(stmt)
+        restoreXafterCall(stmt)
+        // just ignore any result values from the function call.
+    }
+
+    internal fun saveXbeforeCall(stmt: IFunctionCall) {
+        val sub = stmt.target.targetSubroutine(program) ?: throw AssemblyError("undefined subroutine ${stmt.target}")
+        if(sub.shouldSaveX()) {
+            val regSaveOnStack = sub.asmAddress==null       // rom-routines don't require registers to be saved on stack, normal subroutines do because they can contain nested calls
+            val (keepAonEntry: Boolean, keepAonReturn: Boolean) = sub.shouldKeepA()
+            if(regSaveOnStack)
+                asmgen.saveRegisterStack(CpuRegister.X, keepAonEntry)
+            else
+                asmgen.saveRegisterLocal(CpuRegister.X, (stmt as Node).definingSubroutine()!!)
+        }
+    }
+
+    internal fun restoreXafterCall(stmt: IFunctionCall) {
+        val sub = stmt.target.targetSubroutine(program) ?: throw AssemblyError("undefined subroutine ${stmt.target}")
+        if(sub.shouldSaveX()) {
+            val regSaveOnStack = sub.asmAddress==null       // rom-routines don't require registers to be saved on stack, normal subroutines do because they can contain nested calls
+            val (keepAonEntry: Boolean, keepAonReturn: Boolean) = sub.shouldKeepA()
+
+            if(regSaveOnStack)
+                asmgen.restoreRegisterStack(CpuRegister.X, keepAonReturn)
+            else
+                asmgen.restoreRegisterLocal(CpuRegister.X)
+        }
+    }
+
     internal fun translateFunctionCall(stmt: IFunctionCall) {
-        // output the code to setup the parameters and perform the actual call
-        // does NOT output the code to deal with the result values!
-        val sub = stmt.target.targetSubroutine(program.namespace) ?: throw AssemblyError("undefined subroutine ${stmt.target}")
-        val saveX = Register.X in sub.asmClobbers || sub.regXasResult()
-        if(saveX)
-            asmgen.out("  stx  c64.SCRATCH_ZPREGX")        // we only save X for now (required! is the eval stack pointer), screw A and Y...
+        // Output only the code to setup the parameters and perform the actual call
+        // NOTE: does NOT output the code to deal with the result values!
+        // NOTE: does NOT output code to save/restore the X register for this call! Every caller should deal with this in their own way!!
+        //       (you can use subroutine.shouldSaveX() and saveX()/restoreX() routines as a help for this)
 
-        val subName = asmgen.asmIdentifierName(stmt.target)
+        val sub = stmt.target.targetSubroutine(program) ?: throw AssemblyError("undefined subroutine ${stmt.target}")
+        val subName = asmgen.asmSymbolName(stmt.target)
         if(stmt.args.isNotEmpty()) {
-            for(arg in sub.parameters.withIndex().zip(stmt.args)) {
-                translateFuncArguments(arg.first, arg.second, sub)
-            }
-        }
-        asmgen.out("  jsr  $subName")
 
-        if(saveX)
-            asmgen.out("  ldx  c64.SCRATCH_ZPREGX")        // restore X again
-    }
+            if(sub.asmParameterRegisters.isEmpty()) {
+                // via variables
+                for(arg in sub.parameters.withIndex().zip(stmt.args)) {
+                    argumentViaVariable(sub, arg.first, arg.second)
+                }
+            } else {
+                // via registers
+                if(sub.parameters.size==1) {
+                    // just a single parameter, no risk of clobbering registers
+                    argumentViaRegister(sub, IndexedValue(0, sub.parameters.single()), stmt.args[0])
+                } else {
 
-    private fun translateFuncArguments(parameter: IndexedValue<SubroutineParameter>, value: Expression, sub: Subroutine) {
-        val sourceIDt = value.inferType(program)
-        if(!sourceIDt.isKnown)
-            throw AssemblyError("arg type unknown")
-        val sourceDt = sourceIDt.typeOrElse(DataType.STRUCT)
-        if(!argumentTypeCompatible(sourceDt, parameter.value.type))
-            throw AssemblyError("argument type incompatible")
-        if(sub.asmParameterRegisters.isEmpty()) {
-            // pass parameter via a regular variable (not via registers)
-            val paramVar = parameter.value
-            val scopedParamVar = (sub.scopedname+"."+paramVar.name).split(".")
-            val target = AssignTarget(null, IdentifierReference(scopedParamVar, sub.position), null, null, sub.position)
-            target.linkParents(value.parent)
-            when (value) {
-                is NumericLiteralValue -> {
-                    // optimize when the argument is a constant literal
-                    when(parameter.value.type) {
-                        in ByteDatatypes -> asmgen.assignFromByteConstant(target, value.number.toShort())
-                        in WordDatatypes -> asmgen.assignFromWordConstant(target, value.number.toInt())
-                        DataType.FLOAT -> asmgen.assignFromFloatConstant(target, value.number.toDouble())
-                        in PassByReferenceDatatypes -> throw AssemblyError("can't pass string/array as argument via a variable?")    // TODO huh
-                        else -> throw AssemblyError("weird parameter datatype")
-                    }
-                }
-                is IdentifierReference -> {
-                    // optimize when the argument is a variable
-                    when (parameter.value.type) {
-                        in ByteDatatypes -> asmgen.assignFromByteVariable(target, value)
-                        in WordDatatypes -> asmgen.assignFromWordVariable(target, value)
-                        DataType.FLOAT -> asmgen.assignFromFloatVariable(target, value)
-                        in PassByReferenceDatatypes -> throw AssemblyError("can't pass string/array as argument via a variable?")     // TODO huh
-                        else -> throw AssemblyError("weird parameter datatype")
-                    }
-                }
-                is RegisterExpr -> {
-                    asmgen.assignFromRegister(target, value.register)
-                }
-                is DirectMemoryRead -> {
-                    when(value.addressExpression) {
-                        is NumericLiteralValue -> {
-                            val address = (value.addressExpression as NumericLiteralValue).number.toInt()
-                            asmgen.assignFromMemoryByte(target, address, null)
+                    fun isNoClobberRisk(expr: Expression): Boolean {
+                        if(expr is AddressOf ||
+                                expr is NumericLiteralValue ||
+                                expr is StringLiteralValue ||
+                                expr is ArrayLiteralValue ||
+                                expr is IdentifierReference)
+                            return true
+
+                        if(expr is FunctionCall) {
+                            if(expr.target.nameInSource==listOf("lsb") || expr.target.nameInSource==listOf("msb"))
+                                return isNoClobberRisk(expr.args[0])
+                            if(expr.target.nameInSource==listOf("mkword"))
+                                return isNoClobberRisk(expr.args[0]) && isNoClobberRisk(expr.args[1])
                         }
-                        is IdentifierReference -> {
-                            asmgen.assignFromMemoryByte(target, null, value.addressExpression as IdentifierReference)
+
+                        return false
+                    }
+
+                    when {
+                        stmt.args.all {isNoClobberRisk(it)} -> {
+                            // There's no risk of clobbering for these simple argument types. Optimize the register loading directly from these values.
+                            // register assignment order: 1) cx16 virtual word registers, 2) actual CPU registers, 3) CPU Carry status flag.
+                            val argsInfo = sub.parameters.withIndex().zip(stmt.args).zip(sub.asmParameterRegisters)
+                            val (cx16virtualRegs, args2) = argsInfo.partition { it.second.registerOrPair in Cx16VirtualRegisters }
+                            val (cpuRegs, statusRegs) = args2.partition { it.second.registerOrPair!=null }
+                            for(arg in cx16virtualRegs)
+                                argumentViaRegister(sub, arg.first.first, arg.first.second)
+                            for(arg in cpuRegs)
+                                argumentViaRegister(sub, arg.first.first, arg.first.second)
+                            for(arg in statusRegs)
+                                argumentViaRegister(sub, arg.first.first, arg.first.second)
                         }
                         else -> {
-                            asmgen.translateExpression(value.addressExpression)
-                            asmgen.out("  jsr  prog8_lib.read_byte_from_address |  inx")
-                            asmgen.assignFromRegister(target, Register.A)
+                            // Risk of clobbering due to complex expression args. Evaluate first, then assign registers.
+                            registerArgsViaStackEvaluation(stmt, sub)
                         }
                     }
                 }
-                else -> {
-                    asmgen.translateExpression(value)
-                    asmgen.assignFromEvalResult(target)
-                }
             }
-        } else {
-            // pass parameter via a register parameter
-            val paramRegister = sub.asmParameterRegisters[parameter.index]
-            val statusflag = paramRegister.statusflag
-            val register = paramRegister.registerOrPair
-            val stack = paramRegister.stack
+        }
+
+        if(sub.inline && asmgen.options.optimize) {
+            if(sub.containsDefinedVariables())
+                throw AssemblyError("can't inline sub with vars")
+            if(!sub.isAsmSubroutine && sub.parameters.isNotEmpty())
+                throw AssemblyError("can't inline a non-asm subroutine with parameters")
+            asmgen.out("  \t; inlined routine follows: ${sub.name} from ${sub.position}")
+            val statements = sub.statements.filter { it !is ParameterVarDecl && it !is Directive }
+            statements.forEach { asmgen.translate(it) }
+        }
+        else {
+            asmgen.out("  jsr  $subName")
+        }
+
+        // remember: dealing with the X register and/or dealing with return values is the responsibility of the caller
+    }
+
+    private fun registerArgsViaStackEvaluation(stmt: IFunctionCall, sub: Subroutine) {
+        // this is called when one or more of the arguments are 'complex' and
+        // cannot be assigned to a register easily or risk clobbering other registers.
+
+        if(sub.parameters.isEmpty())
+            return
+
+        // 1. load all arguments reversed onto the stack: first arg goes last (is on top).
+        for (arg in stmt.args.reversed())
+            asmgen.translateExpression(arg)
+
+        var argForCarry: IndexedValue<Pair<Expression, RegisterOrStatusflag>>? = null
+        var argForXregister: IndexedValue<Pair<Expression, RegisterOrStatusflag>>? = null
+        var argForAregister: IndexedValue<Pair<Expression, RegisterOrStatusflag>>? = null
+
+        asmgen.out("  inx")     // align estack pointer
+
+        for(argi in stmt.args.zip(sub.asmParameterRegisters).withIndex()) {
+            val plusIdxStr = if(argi.index==0) "" else "+${argi.index}"
             when {
-                stack -> {
-                    // push arg onto the stack
-                    // note: argument order is reversed (first argument will be deepest on the stack)
-                    asmgen.translateExpression(value)
+                argi.value.second.statusflag == Statusflag.Pc -> {
+                    require(argForCarry == null)
+                    argForCarry = argi
                 }
-                statusflag!=null -> {
-                    if (statusflag == Statusflag.Pc) {
-                        // this param needs to be set last, right before the jsr
-                        // for now, this is already enforced on the subroutine definition by the Ast Checker
-                        when(value) {
-                            is NumericLiteralValue -> {
-                                val carrySet = value.number.toInt() != 0
-                                asmgen.out(if(carrySet) "  sec" else "  clc")
-                            }
-                            is IdentifierReference -> {
-                                val sourceName = asmgen.asmIdentifierName(value)
-                                asmgen.out("""
-            lda  $sourceName
-            beq  +
-            sec  
-            bcs  ++
-+           clc
-+
-""")
-                            }
-                            is RegisterExpr -> {
-                                when(value.register) {
-                                    Register.A -> asmgen.out("  cmp  #0")
-                                    Register.X -> asmgen.out("  txa")
-                                    Register.Y -> asmgen.out("  tya")
-                                }
-                                asmgen.out("""
-            beq  +
-            sec
-            bcs  ++
-+           clc
-+
-""")
-                            }
-                            else -> {
-                                asmgen.translateExpression(value)
-                                asmgen.out("""
-            inx                        
-            lda  $ESTACK_LO_HEX,x
-            beq  +
-            sec  
-            bcs  ++
-+           clc
-+
-""")
-                            }
-                        }
-                    }
-                    else throw AssemblyError("can only use Carry as status flag parameter")
+                argi.value.second.statusflag != null -> throw AssemblyError("can only use Carry as status flag parameter")
+                argi.value.second.registerOrPair in setOf(RegisterOrPair.X, RegisterOrPair.AX, RegisterOrPair.XY) -> {
+                    require(argForXregister==null)
+                    argForXregister = argi
                 }
-                register!=null && register.name.length==1 -> {
-                    when (value) {
-                        is NumericLiteralValue -> {
-                            val target = AssignTarget(Register.valueOf(register.name), null, null, null, sub.position)
-                            target.linkParents(value.parent)
-                            asmgen.assignFromByteConstant(target, value.number.toShort())
-                        }
-                        is IdentifierReference -> {
-                            val target = AssignTarget(Register.valueOf(register.name), null, null, null, sub.position)
-                            target.linkParents(value.parent)
-                            asmgen.assignFromByteVariable(target, value)
-                        }
-                        else -> {
-                            asmgen.translateExpression(value)
-                            when(register) {
-                                RegisterOrPair.A -> asmgen.out("  inx | lda  $ESTACK_LO_HEX,x")
-                                RegisterOrPair.X -> throw AssemblyError("can't pop into X register - use a variable instead")
-                                RegisterOrPair.Y -> asmgen.out("  inx | ldy  $ESTACK_LO_HEX,x")
-                                else -> throw AssemblyError("cannot assign to register pair")
-                            }
+                argi.value.second.registerOrPair in setOf(RegisterOrPair.A, RegisterOrPair.AY) -> {
+                    require(argForAregister == null)
+                    argForAregister = argi
+                }
+                argi.value.second.registerOrPair == RegisterOrPair.Y -> {
+                    asmgen.out("  ldy  P8ESTACK_LO$plusIdxStr,x")
+                }
+                argi.value.second.registerOrPair in Cx16VirtualRegisters -> {
+                    // immediately output code to load the virtual register, to avoid clobbering the A register later
+                    when (sub.parameters[argi.index].type) {
+                        in ByteDatatypes -> {
+                            // only load the lsb of the virtual register
+                            asmgen.out("""
+                                lda  P8ESTACK_LO$plusIdxStr,x
+                                sta  cx16.${argi.value.second.registerOrPair.toString().toLowerCase()}
+                            """)
+                            if (asmgen.isTargetCpu(CpuType.CPU65c02))
+                                asmgen.out("  stz  cx16.${argi.value.second.registerOrPair.toString().toLowerCase()}+1")
+                            else
+                                asmgen.out("  lda  #0 |  sta  cx16.${argi.value.second.registerOrPair.toString().toLowerCase()}+1")
                         }
+                        in WordDatatypes ->
+                            asmgen.out("""
+                                lda  P8ESTACK_LO$plusIdxStr,x
+                                sta  cx16.${argi.value.second.registerOrPair.toString().toLowerCase()}
+                                lda  P8ESTACK_HI$plusIdxStr,x
+                                sta  cx16.${argi.value.second.registerOrPair.toString().toLowerCase()}+1
+                            """)
+                        else -> throw AssemblyError("weird dt")
                     }
                 }
-                register!=null && register.name.length==2 -> {
-                    // register pair as a 16-bit value (only possible for subroutine parameters)
-                    when (value) {
-                        is NumericLiteralValue -> {
-                            // optimize when the argument is a constant literal
-                            val hex = value.number.toHex()
-                            when (register) {
-                                RegisterOrPair.AX -> asmgen.out("  lda  #<$hex  |  ldx  #>$hex")
-                                RegisterOrPair.AY -> asmgen.out("  lda  #<$hex  |  ldy  #>$hex")
-                                RegisterOrPair.XY -> asmgen.out("  ldx  #<$hex  |  ldy  #>$hex")
-                                else -> {}
-                            }
-                        }
-                        is AddressOf -> {
-                            // optimize when the argument is an address of something
-                            val sourceName = asmgen.asmIdentifierName(value.identifier)
-                            when (register) {
-                                RegisterOrPair.AX -> asmgen.out("  lda  #<$sourceName  |  ldx  #>$sourceName")
-                                RegisterOrPair.AY -> asmgen.out("  lda  #<$sourceName  |  ldy  #>$sourceName")
-                                RegisterOrPair.XY -> asmgen.out("  ldx  #<$sourceName  |  ldy  #>$sourceName")
-                                else -> {}
-                            }
-                        }
-                        is IdentifierReference -> {
-                            val sourceName = asmgen.asmIdentifierName(value)
-                            if(sourceDt in PassByReferenceDatatypes) {
-                                when (register) {
-                                    RegisterOrPair.AX -> asmgen.out("  lda  #<$sourceName  |  ldx  #>$sourceName")
-                                    RegisterOrPair.AY -> asmgen.out("  lda  #<$sourceName  |  ldy  #>$sourceName")
-                                    RegisterOrPair.XY -> asmgen.out("  ldx  #<$sourceName  |  ldy  #>$sourceName")
-                                    else -> {}
-                                }
-                            } else {
-                                when (register) {
-                                    RegisterOrPair.AX -> asmgen.out("  lda  $sourceName  |  ldx  $sourceName+1")
-                                    RegisterOrPair.AY -> asmgen.out("  lda  $sourceName  |  ldy  $sourceName+1")
-                                    RegisterOrPair.XY -> asmgen.out("  ldx  $sourceName  |  ldy  $sourceName+1")
-                                    else -> {}
-                                }
-                            }
-                        }
-                        else -> {
-                            asmgen.translateExpression(value)
-                            if (register == RegisterOrPair.AX || register == RegisterOrPair.XY)
-                                throw AssemblyError("can't use X register here - use a variable")
-                            else if (register == RegisterOrPair.AY)
-                                asmgen.out("  inx |  lda  $ESTACK_LO_HEX,x  |  ldy  $ESTACK_HI_HEX,x")
-                        }
+                else -> throw AssemblyError("weird argument")
+            }
+        }
+
+        if(argForCarry!=null) {
+            val plusIdxStr = if(argForCarry.index==0) "" else "+${argForCarry.index}"
+            asmgen.out("""
+                lda  P8ESTACK_LO$plusIdxStr,x
+                beq  +
+                sec
+                bcs  ++
++               clc
++               php""")             // push the status flags
+        }
+
+        if(argForAregister!=null) {
+            val plusIdxStr = if(argForAregister.index==0) "" else "+${argForAregister.index}"
+            when(argForAregister.value.second.registerOrPair) {
+                RegisterOrPair.A -> asmgen.out("  lda  P8ESTACK_LO$plusIdxStr,x")
+                RegisterOrPair.AY -> asmgen.out("  lda  P8ESTACK_LO$plusIdxStr,x |  ldy  P8ESTACK_HI$plusIdxStr,x")
+                else -> throw AssemblyError("weird arg")
+            }
+        }
+
+        if(argForXregister!=null) {
+            val plusIdxStr = if(argForXregister.index==0) "" else "+${argForXregister.index}"
+
+            if(argForAregister!=null)
+                asmgen.out("  pha")
+            when(argForXregister.value.second.registerOrPair) {
+                RegisterOrPair.X -> asmgen.out("  lda  P8ESTACK_LO$plusIdxStr,x |  tax")
+                RegisterOrPair.AX -> asmgen.out("  ldy  P8ESTACK_LO$plusIdxStr,x |  lda  P8ESTACK_HI$plusIdxStr,x |  tax |  tya")
+                RegisterOrPair.XY -> asmgen.out("  ldy  P8ESTACK_HI$plusIdxStr,x |  lda  P8ESTACK_LO$plusIdxStr,x |  tax")
+                else -> throw AssemblyError("weird arg")
+            }
+            if(argForAregister!=null)
+                asmgen.out("  pla")
+        } else {
+            repeat(sub.parameters.size - 1) { asmgen.out("  inx") }       // unwind stack
+        }
+
+        if(argForCarry!=null)
+            asmgen.out("  plp")       // set the carry flag back to correct value
+    }
+
+    private fun argumentViaVariable(sub: Subroutine, parameter: IndexedValue<SubroutineParameter>, value: Expression) {
+        // pass parameter via a regular variable (not via registers)
+        val valueIDt = value.inferType(program)
+        if(!valueIDt.isKnown)
+            throw AssemblyError("unknown dt")
+        val valueDt = valueIDt.typeOrElse(DataType.STRUCT)
+        if(!isArgumentTypeCompatible(valueDt, parameter.value.type))
+            throw AssemblyError("argument type incompatible")
+
+        val varName = asmgen.asmVariableName(sub.scopedname+"."+parameter.value.name)
+        asmgen.assignExpressionToVariable(value, varName, parameter.value.type, sub)
+    }
+
+    private fun argumentViaRegister(sub: Subroutine, parameter: IndexedValue<SubroutineParameter>, value: Expression) {
+        // pass argument via a register parameter
+        val valueIDt = value.inferType(program)
+        if(!valueIDt.isKnown)
+            throw AssemblyError("unknown dt")
+        val valueDt = valueIDt.typeOrElse(DataType.STRUCT)
+        if(!isArgumentTypeCompatible(valueDt, parameter.value.type))
+            throw AssemblyError("argument type incompatible")
+
+        val paramRegister = sub.asmParameterRegisters[parameter.index]
+        val statusflag = paramRegister.statusflag
+        val register = paramRegister.registerOrPair
+        val requiredDt = parameter.value.type
+        if(requiredDt!=valueDt) {
+            if(valueDt largerThan requiredDt)
+                throw AssemblyError("can only convert byte values to word param types")
+        }
+        if (statusflag!=null) {
+            if(requiredDt!=valueDt)
+                throw AssemblyError("for statusflag, byte value is required")
+            if (statusflag == Statusflag.Pc) {
+                // this param needs to be set last, right before the jsr
+                // for now, this is already enforced on the subroutine definition by the Ast Checker
+                when(value) {
+                    is NumericLiteralValue -> {
+                        val carrySet = value.number.toInt() != 0
+                        asmgen.out(if(carrySet) "  sec" else "  clc")
+                    }
+                    is IdentifierReference -> {
+                        val sourceName = asmgen.asmVariableName(value)
+                        asmgen.out("""
+                pha
+                lda  $sourceName
+                beq  +
+                sec  
+                bcs  ++
+    +           clc
+    +           pla
+    """)
+                    }
+                    else -> {
+                        asmgen.assignExpressionToRegister(value, RegisterOrPair.A)
+                        asmgen.out("""
+                                    beq  +
+                                    sec
+                                    bcs  ++
+    +                               clc
+    +""")
                     }
                 }
+            } else throw AssemblyError("can only use Carry as status flag parameter")
+        }
+        else {
+            // via register or register pair
+            register!!
+            if(requiredDt largerThan valueDt) {
+                // we need to sign extend the source, do this via temporary word variable
+                val scratchVar = asmgen.asmVariableName("P8ZP_SCRATCH_W1")
+                asmgen.assignExpressionToVariable(value, scratchVar, DataType.UBYTE, sub)
+                asmgen.signExtendVariableLsb(scratchVar, valueDt)
+                asmgen.assignVariableToRegister(scratchVar, register)
+            } else {
+                val target: AsmAssignTarget =
+                    if(parameter.value.type in ByteDatatypes && (register==RegisterOrPair.AX || register == RegisterOrPair.AY || register==RegisterOrPair.XY || register in Cx16VirtualRegisters))
+                        AsmAssignTarget(TargetStorageKind.REGISTER, program, asmgen, parameter.value.type, sub, register = register)
+                    else
+                        AsmAssignTarget.fromRegisters(register, sub, program, asmgen)
+                val src = if(valueDt in PassByReferenceDatatypes) {
+                    if(value is IdentifierReference) {
+                        val addr = AddressOf(value, Position.DUMMY)
+                        AsmAssignSource.fromAstSource(addr, program, asmgen).adjustSignedUnsigned(target)
+                    } else {
+                        AsmAssignSource.fromAstSource(value, program, asmgen).adjustSignedUnsigned(target)
+                    }
+                } else {
+                    AsmAssignSource.fromAstSource(value, program, asmgen).adjustSignedUnsigned(target)
+                }
+                asmgen.translateNormalAssignment(AsmAssignment(src, target, false, program.memsizer, Position.DUMMY))
             }
         }
     }
 
-    private fun argumentTypeCompatible(argType: DataType, paramType: DataType): Boolean {
+    private fun isArgumentTypeCompatible(argType: DataType, paramType: DataType): Boolean {
         if(argType isAssignableTo paramType)
             return true
         if(argType in ByteDatatypes && paramType in ByteDatatypes)

compiler/src/prog8/compiler/target/c64/codegen/PostIncrDecrAsmGen.kt

@@ -4,11 +4,9 @@ import prog8.ast.Program
 import prog8.ast.base.*
 import prog8.ast.expressions.IdentifierReference
 import prog8.ast.expressions.NumericLiteralValue
-import prog8.ast.expressions.RegisterExpr
 import prog8.ast.statements.PostIncrDecr
+import prog8.ast.toHex
 import prog8.compiler.AssemblyError
-import prog8.compiler.target.c64.C64MachineDefinition.C64Zeropage
-import prog8.compiler.toHex
 
 
 internal class PostIncrDecrAsmGen(private val program: Program, private val asmgen: AsmGen) {
@@ -17,28 +15,11 @@ internal class PostIncrDecrAsmGen(private val program: Program, private val asmg
         val targetIdent = stmt.target.identifier
         val targetMemory = stmt.target.memoryAddress
         val targetArrayIdx = stmt.target.arrayindexed
-        val targetRegister = stmt.target.register
+        val scope = stmt.definingSubroutine()
         when {
-            targetRegister!=null -> {
-                when(targetRegister) {
-                    Register.A -> {
-                        if(incr)
-                            asmgen.out("  clc |  adc  #1 ")
-                        else
-                            asmgen.out("  sec |  sbc  #1 ")
-                    }
-                    Register.X -> {
-                        if(incr) asmgen.out("  inx") else asmgen.out("  dex")
-                    }
-                    Register.Y -> {
-                        if(incr) asmgen.out("  iny") else asmgen.out("  dey")
-                    }
-                }
-            }
             targetIdent!=null -> {
-                val what = asmgen.asmIdentifierName(targetIdent)
-                val dt = stmt.target.inferType(program, stmt).typeOrElse(DataType.STRUCT)
-                when (dt) {
+                val what = asmgen.asmVariableName(targetIdent)
+                when (stmt.target.inferType(program).typeOrElse(DataType.STRUCT)) {
                     in ByteDatatypes -> asmgen.out(if (incr) "  inc  $what" else "  dec  $what")
                     in WordDatatypes -> {
                         if(incr)
@@ -53,7 +34,7 @@ internal class PostIncrDecrAsmGen(private val program: Program, private val asmg
                     }
                     DataType.FLOAT -> {
                         asmgen.out("  lda  #<$what |  ldy  #>$what")
-                        asmgen.out(if(incr) "  jsr  c64flt.inc_var_f" else "  jsr  c64flt.dec_var_f")
+                        asmgen.out(if(incr) "  jsr  floats.inc_var_f" else "  jsr  floats.dec_var_f")
                     }
                     else -> throw AssemblyError("need numeric type")
                 }
@@ -65,87 +46,82 @@ internal class PostIncrDecrAsmGen(private val program: Program, private val asmg
                         asmgen.out(if(incr) "  inc  $what" else "  dec  $what")
                     }
                     is IdentifierReference -> {
-                        val what = asmgen.asmIdentifierName(addressExpr)
+                        val what = asmgen.asmVariableName(addressExpr)
                         asmgen.out("  lda  $what |  sta  (+) +1 |  lda  $what+1 |  sta  (+) +2")
                         if(incr)
                             asmgen.out("+\tinc  ${'$'}ffff\t; modified")
                         else
                             asmgen.out("+\tdec  ${'$'}ffff\t; modified")
                     }
-                    else -> throw AssemblyError("weird target type $targetMemory")
+                    else -> {
+                        asmgen.assignExpressionToRegister(addressExpr, RegisterOrPair.AY)
+                        asmgen.out("  sta  (+) + 1 |  sty  (+) + 2")
+                        if(incr)
+                            asmgen.out("+\tinc  ${'$'}ffff\t; modified")
+                        else
+                            asmgen.out("+\tdec  ${'$'}ffff\t; modified")
+                    }
                 }
             }
             targetArrayIdx!=null -> {
-                val index = targetArrayIdx.arrayspec.index
-                val what = asmgen.asmIdentifierName(targetArrayIdx.identifier)
-                val arrayDt = targetArrayIdx.identifier.inferType(program).typeOrElse(DataType.STRUCT)
-                val elementDt = ArrayElementTypes.getValue(arrayDt)
-                when(index) {
-                    is NumericLiteralValue -> {
-                        val indexValue = index.number.toInt() * elementDt.memorySize()
-                        when(elementDt) {
-                            in ByteDatatypes -> asmgen.out(if (incr) "  inc  $what+$indexValue" else "  dec  $what+$indexValue")
-                            in WordDatatypes -> {
-                                if(incr)
-                                    asmgen.out(" inc  $what+$indexValue |  bne  + |  inc  $what+$indexValue+1 |+")
-                                else
-                                    asmgen.out("""
-        lda  $what+$indexValue
-        bne  +
-        dec  $what+$indexValue+1
-+       dec  $what+$indexValue 
+                val asmArrayvarname = asmgen.asmVariableName(targetArrayIdx.arrayvar)
+                val elementDt = targetArrayIdx.inferType(program).typeOrElse(DataType.STRUCT)
+                if(targetArrayIdx.indexer.indexNum!=null) {
+                    val indexValue = targetArrayIdx.indexer.constIndex()!! * program.memsizer.memorySize(elementDt)
+                    when(elementDt) {
+                        in ByteDatatypes -> asmgen.out(if (incr) "  inc  $asmArrayvarname+$indexValue" else "  dec  $asmArrayvarname+$indexValue")
+                        in WordDatatypes -> {
+                            if(incr)
+                                asmgen.out(" inc  $asmArrayvarname+$indexValue |  bne  + |  inc  $asmArrayvarname+$indexValue+1 |+")
+                            else
+                                asmgen.out("""
+    lda  $asmArrayvarname+$indexValue
+    bne  +
+    dec  $asmArrayvarname+$indexValue+1
++       dec  $asmArrayvarname+$indexValue 
 """)
-                            }
-                            DataType.FLOAT -> {
-                                asmgen.out("  lda  #<$what+$indexValue |  ldy  #>$what+$indexValue")
-                                asmgen.out(if(incr) "  jsr  c64flt.inc_var_f" else "  jsr  c64flt.dec_var_f")
-                            }
-                            else -> throw AssemblyError("need numeric type")
                         }
-                    }
-                    is RegisterExpr -> {
-                        asmgen.translateArrayIndexIntoA(targetArrayIdx)
-                        incrDecrArrayvalueWithIndexA(incr, arrayDt, what)
-                    }
-                    is IdentifierReference -> {
-                        asmgen.translateArrayIndexIntoA(targetArrayIdx)
-                        incrDecrArrayvalueWithIndexA(incr, arrayDt, what)
-                    }
-                    else -> {
-                        asmgen.translateArrayIndexIntoA(targetArrayIdx)
-                        incrDecrArrayvalueWithIndexA(incr, arrayDt, what)
+                        DataType.FLOAT -> {
+                            asmgen.out("  lda  #<$asmArrayvarname+$indexValue |  ldy  #>$asmArrayvarname+$indexValue")
+                            asmgen.out(if(incr) "  jsr  floats.inc_var_f" else "  jsr  floats.dec_var_f")
+                        }
+                        else -> throw AssemblyError("need numeric type")
                     }
                 }
-            }
-            else -> throw AssemblyError("weird target type ${stmt.target}")
-        }
-    }
-
-    private fun incrDecrArrayvalueWithIndexA(incr: Boolean, arrayDt: DataType, arrayVarName: String) {
-        asmgen.out("  stx  ${C64Zeropage.SCRATCH_REG_X} |  tax")
-        when(arrayDt) {
-            DataType.STR,
-            DataType.ARRAY_UB, DataType.ARRAY_B -> {
-                asmgen.out(if(incr) "  inc  $arrayVarName,x" else "  dec  $arrayVarName,x")
-            }
-            DataType.ARRAY_UW, DataType.ARRAY_W -> {
-                if(incr)
-                    asmgen.out(" inc  $arrayVarName,x |  bne  + |  inc  $arrayVarName+1,x |+")
                 else
-                    asmgen.out("""
-        lda  $arrayVarName,x
-        bne  +
-        dec  $arrayVarName+1,x
-+       dec  $arrayVarName 
+                {
+                    asmgen.loadScaledArrayIndexIntoRegister(targetArrayIdx, elementDt, CpuRegister.A)
+                    asmgen.saveRegisterLocal(CpuRegister.X, scope!!)
+                    asmgen.out("  tax")
+                    when(elementDt) {
+                        in ByteDatatypes -> {
+                            asmgen.out(if(incr) "  inc  $asmArrayvarname,x" else "  dec  $asmArrayvarname,x")
+                        }
+                        in WordDatatypes -> {
+                            if(incr)
+                                asmgen.out(" inc  $asmArrayvarname,x |  bne  + |  inc  $asmArrayvarname+1,x |+")
+                            else
+                                asmgen.out("""
+    lda  $asmArrayvarname,x
+    bne  +
+    dec  $asmArrayvarname+1,x
++       dec  $asmArrayvarname,x 
 """)
+                        }
+                        DataType.FLOAT -> {
+                            asmgen.out("""
+                    ldy  #>$asmArrayvarname
+                    clc
+                    adc  #<$asmArrayvarname
+                    bcc  +
+                    iny
++                       jsr  floats.inc_var_f""")
+                        }
+                        else -> throw AssemblyError("weird array elt dt")
+                    }
+                    asmgen.restoreRegisterLocal(CpuRegister.X)
+                }
             }
-            DataType.ARRAY_F -> {
-                asmgen.out("  lda  #<$arrayVarName |  ldy  #>$arrayVarName")
-                asmgen.out(if(incr) "  jsr  c64flt.inc_indexed_var_f" else "  jsr  c64flt.dec_indexed_var_f")
-            }
-            else -> throw AssemblyError("weird array dt")
         }
-        asmgen.out("  ldx  ${C64Zeropage.SCRATCH_REG_X}")
     }
-
 }

compiler/src/prog8/compiler/target/c64/codegen/assignment/AsmAssignment.kt

@@ -0,0 +1,220 @@
+package prog8.compiler.target.c64.codegen.assignment
+
+import prog8.ast.IMemSizer
+import prog8.ast.Program
+import prog8.ast.base.*
+import prog8.ast.expressions.*
+import prog8.ast.statements.*
+import prog8.compiler.AssemblyError
+import prog8.compiler.target.c64.codegen.AsmGen
+
+
+internal enum class TargetStorageKind {
+    VARIABLE,
+    ARRAY,
+    MEMORY,
+    REGISTER,
+    STACK
+}
+
+internal enum class SourceStorageKind {
+    LITERALNUMBER,
+    VARIABLE,
+    ARRAY,
+    MEMORY,
+    REGISTER,
+    STACK,              // value is already present on stack
+    EXPRESSION,         // expression in ast-form, still to be evaluated
+}
+
+internal class AsmAssignTarget(val kind: TargetStorageKind,
+                               private val program: Program,
+                               private val asmgen: AsmGen,
+                               val datatype: DataType,
+                               val scope: Subroutine?,
+                               private val variableAsmName: String? = null,
+                               val array: ArrayIndexedExpression? = null,
+                               val memory: DirectMemoryWrite? = null,
+                               val register: RegisterOrPair? = null,
+                               val origAstTarget: AssignTarget? = null
+                               )
+{
+    val constMemoryAddress by lazy { memory?.addressExpression?.constValue(program)?.number?.toInt() ?: 0}
+    val constArrayIndexValue by lazy { array?.indexer?.constIndex() }
+    val asmVarname: String
+        get() = if(array==null)
+            variableAsmName!!
+        else
+            asmgen.asmVariableName(array.arrayvar)
+
+    lateinit var origAssign: AsmAssignment
+
+    init {
+        if(register!=null && datatype !in NumericDatatypes)
+            throw AssemblyError("register must be integer or float type")
+    }
+
+    companion object {
+        fun fromAstAssignment(assign: Assignment, program: Program, asmgen: AsmGen): AsmAssignTarget = with(assign.target) {
+            val idt = inferType(program)
+            if(!idt.isKnown)
+                throw AssemblyError("unknown dt")
+            val dt = idt.typeOrElse(DataType.STRUCT)
+            when {
+                identifier != null -> AsmAssignTarget(TargetStorageKind.VARIABLE, program, asmgen, dt, assign.definingSubroutine(), variableAsmName = asmgen.asmVariableName(identifier!!), origAstTarget =  this)
+                arrayindexed != null -> AsmAssignTarget(TargetStorageKind.ARRAY, program, asmgen, dt, assign.definingSubroutine(), array = arrayindexed, origAstTarget =  this)
+                memoryAddress != null -> AsmAssignTarget(TargetStorageKind.MEMORY, program, asmgen, dt, assign.definingSubroutine(), memory =  memoryAddress, origAstTarget =  this)
+                else -> throw AssemblyError("weird target")
+            }
+        }
+
+        fun fromRegisters(registers: RegisterOrPair, scope: Subroutine?, program: Program, asmgen: AsmGen): AsmAssignTarget =
+                when(registers) {
+                    RegisterOrPair.A,
+                    RegisterOrPair.X,
+                    RegisterOrPair.Y -> AsmAssignTarget(TargetStorageKind.REGISTER, program, asmgen, DataType.UBYTE, scope, register = registers)
+                    RegisterOrPair.AX,
+                    RegisterOrPair.AY,
+                    RegisterOrPair.XY -> AsmAssignTarget(TargetStorageKind.REGISTER, program, asmgen, DataType.UWORD, scope, register = registers)
+                    RegisterOrPair.FAC1,
+                    RegisterOrPair.FAC2 -> AsmAssignTarget(TargetStorageKind.REGISTER, program, asmgen, DataType.FLOAT, scope, register = registers)
+                    RegisterOrPair.R0,
+                    RegisterOrPair.R1,
+                    RegisterOrPair.R2,
+                    RegisterOrPair.R3,
+                    RegisterOrPair.R4,
+                    RegisterOrPair.R5,
+                    RegisterOrPair.R6,
+                    RegisterOrPair.R7,
+                    RegisterOrPair.R8,
+                    RegisterOrPair.R9,
+                    RegisterOrPair.R10,
+                    RegisterOrPair.R11,
+                    RegisterOrPair.R12,
+                    RegisterOrPair.R13,
+                    RegisterOrPair.R14,
+                    RegisterOrPair.R15 -> AsmAssignTarget(TargetStorageKind.REGISTER, program, asmgen, DataType.UWORD, scope, register = registers)
+                }
+    }
+}
+
+internal class AsmAssignSource(val kind: SourceStorageKind,
+                               private val program: Program,
+                               private val asmgen: AsmGen,
+                               val datatype: DataType,
+                               private val variableAsmName: String? = null,
+                               val array: ArrayIndexedExpression? = null,
+                               val memory: DirectMemoryRead? = null,
+                               val register: RegisterOrPair? = null,
+                               val number: NumericLiteralValue? = null,
+                               val expression: Expression? = null
+)
+{
+    val constMemoryAddress by lazy { memory?.addressExpression?.constValue(program)?.number?.toInt() ?: 0}
+    val constArrayIndexValue by lazy { array?.indexer?.constIndex() }
+
+    val asmVarname: String
+        get() = if(array==null)
+            variableAsmName!!
+        else
+            asmgen.asmVariableName(array.arrayvar)
+
+    companion object {
+        fun fromAstSource(indexer: ArrayIndex, program: Program, asmgen: AsmGen): AsmAssignSource {
+            return when {
+                indexer.indexNum!=null -> fromAstSource(indexer.indexNum!!, program, asmgen)
+                indexer.indexVar!=null -> fromAstSource(indexer.indexVar!!, program, asmgen)
+                else -> throw AssemblyError("weird indexer")
+            }
+        }
+
+        fun fromAstSource(value: Expression, program: Program, asmgen: AsmGen): AsmAssignSource {
+            val cv = value.constValue(program)
+            if(cv!=null)
+                return AsmAssignSource(SourceStorageKind.LITERALNUMBER, program, asmgen, cv.type, number = cv)
+
+            return when(value) {
+                is NumericLiteralValue -> AsmAssignSource(SourceStorageKind.LITERALNUMBER, program, asmgen, value.type, number = cv)
+                is StringLiteralValue -> throw AssemblyError("string literal value should not occur anymore for asm generation")
+                is ArrayLiteralValue -> throw AssemblyError("array literal value should not occur anymore for asm generation")
+                is IdentifierReference -> {
+                    val dt = value.inferType(program).typeOrElse(DataType.STRUCT)
+                    val varName=asmgen.asmVariableName(value)
+                    // special case: "cx16.r[0-15]" are 16-bits virtual registers of the commander X16 system
+                    if(dt==DataType.UWORD && varName.toLowerCase().startsWith("cx16.r")) {
+                        val regStr = varName.toLowerCase().substring(5)
+                        val reg = RegisterOrPair.valueOf(regStr.toUpperCase())
+                        AsmAssignSource(SourceStorageKind.REGISTER, program, asmgen, dt, register = reg)
+                    } else {
+                        AsmAssignSource(SourceStorageKind.VARIABLE, program, asmgen, dt, variableAsmName = varName)
+                    }
+                }
+                is DirectMemoryRead -> {
+                    AsmAssignSource(SourceStorageKind.MEMORY, program, asmgen, DataType.UBYTE, memory = value)
+                }
+                is ArrayIndexedExpression -> {
+                    val dt = value.inferType(program).typeOrElse(DataType.STRUCT)
+                    AsmAssignSource(SourceStorageKind.ARRAY, program, asmgen, dt, array = value)
+                }
+                is FunctionCall -> {
+                    when (val sub = value.target.targetStatement(program)) {
+                        is Subroutine -> {
+                            val returnType = sub.returntypes.zip(sub.asmReturnvaluesRegisters).firstOrNull { rr -> rr.second.registerOrPair != null || rr.second.statusflag!=null }?.first
+                                    ?: throw AssemblyError("can't translate zero return values in assignment")
+
+                            AsmAssignSource(SourceStorageKind.EXPRESSION, program, asmgen, returnType, expression = value)
+                        }
+                        is BuiltinFunctionStatementPlaceholder -> {
+                            val returnType = value.inferType(program)
+                            if(!returnType.isKnown)
+                                throw AssemblyError("unknown dt")
+                            AsmAssignSource(SourceStorageKind.EXPRESSION, program, asmgen, returnType.typeOrElse(DataType.STRUCT), expression = value)
+                        }
+                        else -> {
+                            throw AssemblyError("weird call")
+                        }
+                    }
+                }
+                else -> {
+                    val dt = value.inferType(program)
+                    if(!dt.isKnown)
+                        throw AssemblyError("unknown dt")
+                    AsmAssignSource(SourceStorageKind.EXPRESSION, program, asmgen, dt.typeOrElse(DataType.STRUCT), expression = value)
+                }
+            }
+        }
+    }
+
+    fun adjustSignedUnsigned(target: AsmAssignTarget): AsmAssignSource {
+        // allow some signed/unsigned relaxations
+
+        fun withAdjustedDt(newType: DataType) =
+                AsmAssignSource(kind, program, asmgen, newType, variableAsmName, array, memory, register, number, expression)
+
+        if(target.datatype!=datatype) {
+            if(target.datatype in ByteDatatypes && datatype in ByteDatatypes) {
+                return withAdjustedDt(target.datatype)
+            } else if(target.datatype in WordDatatypes && datatype in WordDatatypes) {
+                return withAdjustedDt(target.datatype)
+            }
+        }
+        return this
+    }
+
+}
+
+
+internal class AsmAssignment(val source: AsmAssignSource,
+                             val target: AsmAssignTarget,
+                             val isAugmentable: Boolean,
+                             memsizer: IMemSizer,
+                             val position: Position) {
+
+    init {
+        if(target.register !in setOf(RegisterOrPair.XY, RegisterOrPair.AX, RegisterOrPair.AY))
+            require(source.datatype != DataType.STRUCT) { "must not be placeholder datatype" }
+            require(memsizer.memorySize(source.datatype) <= memsizer.memorySize(target.datatype)) {
+                "source storage size must be less or equal to target datatype storage size"
+            }
+    }
+}

compiler/src/prog8/compiler/target/cx16/CX16MachineDefinition.kt

@@ -0,0 +1,114 @@
+package prog8.compiler.target.cx16
+
+import prog8.compiler.*
+import prog8.compiler.target.CpuType
+import prog8.compiler.target.IMachineDefinition
+import prog8.compiler.target.c64.C64MachineDefinition
+import java.io.IOException
+
+internal object CX16MachineDefinition: IMachineDefinition {
+
+    override val cpu = CpuType.CPU65c02
+
+    // 5-byte cbm MFLPT format limitations:
+    override val FLOAT_MAX_POSITIVE = 1.7014118345e+38         // bytes: 255,127,255,255,255
+    override val FLOAT_MAX_NEGATIVE = -1.7014118345e+38        // bytes: 255,255,255,255,255
+    override val FLOAT_MEM_SIZE = 5
+    override val POINTER_MEM_SIZE = 2
+    override val BASIC_LOAD_ADDRESS = 0x0801
+    override val RAW_LOAD_ADDRESS = 0x8000
+
+    // the 2*256 byte evaluation stack (on which bytes, words, and even floats are stored during calculations)
+    override val ESTACK_LO = 0x0400        //  $0400-$04ff inclusive
+    override val ESTACK_HI = 0x0500        //  $0500-$05ff inclusive
+
+    override lateinit var zeropage: Zeropage
+
+    override fun getFloat(num: Number) = C64MachineDefinition.Mflpt5.fromNumber(num)
+    override fun importLibs(compilerOptions: CompilationOptions, compilationTargetName: String): List<String> {
+        return if (compilerOptions.launcher == LauncherType.BASIC || compilerOptions.output == OutputType.PRG)
+            listOf("syslib")
+        else
+            emptyList()
+    }
+
+    override fun launchEmulator(programName: String) {
+        for(emulator in listOf("x16emu")) {
+            println("\nStarting Commander X16 emulator $emulator...")
+            val cmdline = listOf(emulator, "-scale", "2", "-run", "-prg", "$programName.prg")
+            val processb = ProcessBuilder(cmdline).inheritIO()
+            val process: Process
+            try {
+                process=processb.start()
+            } catch(x: IOException) {
+                continue  // try the next emulator executable
+            }
+            process.waitFor()
+            break
+        }
+    }
+
+    override fun isRegularRAMaddress(address: Int): Boolean = address < 0x9f00 || address in 0xa000..0xbfff
+
+    override fun initializeZeropage(compilerOptions: CompilationOptions) {
+        zeropage = CX16Zeropage(compilerOptions)
+    }
+
+    // 6502 opcodes (including aliases and illegal opcodes), these cannot be used as variable or label names
+    override val opcodeNames = setOf("adc", "and", "asl", "bcc", "bcs",
+            "beq", "bge", "bit", "blt", "bmi", "bne", "bpl", "brk", "bvc", "bvs", "clc",
+            "cld", "cli", "clv", "cmp", "cpx", "cpy", "dec", "dex", "dey",
+            "eor", "gcc", "gcs", "geq", "gge", "glt", "gmi", "gne", "gpl", "gvc", "gvs",
+            "inc", "inx", "iny", "jmp", "jsr",
+            "lda", "ldx", "ldy", "lsr", "nop", "ora", "pha", "php",
+            "pla", "plp", "rol", "ror", "rti", "rts", "sbc",
+            "sec", "sed", "sei",
+            "sta", "stx", "sty", "tax", "tay", "tsx", "txa", "txs", "tya",
+            "bra", "phx", "phy", "plx", "ply", "stz", "trb", "tsb", "bbr", "bbs",
+            "rmb", "smb", "stp", "wai")
+
+
+    internal class CX16Zeropage(options: CompilationOptions) : Zeropage(options) {
+
+        override val SCRATCH_B1 = 0x7a      // temp storage for a single byte
+        override val SCRATCH_REG = 0x7b     // temp storage for a register, must be B1+1
+        override val SCRATCH_W1 = 0x7c      // temp storage 1 for a word  $7c+$7d
+        override val SCRATCH_W2 = 0x7e      // temp storage 2 for a word  $7e+$7f
+
+
+        init {
+            if (options.floats && options.zeropage !in setOf(ZeropageType.BASICSAFE, ZeropageType.DONTUSE ))
+                throw CompilerException("when floats are enabled, zero page type should be 'basicsafe' or 'dontuse'")
+
+            // the addresses 0x02 to 0x21 (inclusive) are taken for sixteen virtual 16-bit api registers.
+
+            when (options.zeropage) {
+                ZeropageType.FULL -> {
+                    free.addAll(0x22..0xff)
+                    free.removeAll(listOf(SCRATCH_B1, SCRATCH_REG, SCRATCH_W1, SCRATCH_W1 + 1, SCRATCH_W2, SCRATCH_W2 + 1))
+                }
+                ZeropageType.KERNALSAFE -> {
+                    free.addAll(0x22..0x7f)
+                    free.addAll(0xa9..0xff)
+                    free.removeAll(listOf(SCRATCH_B1, SCRATCH_REG, SCRATCH_W1, SCRATCH_W1 + 1, SCRATCH_W2, SCRATCH_W2 + 1))
+                }
+                ZeropageType.BASICSAFE -> {
+                    free.addAll(0x22..0x7f)
+                    free.removeAll(listOf(SCRATCH_B1, SCRATCH_REG, SCRATCH_W1, SCRATCH_W1 + 1, SCRATCH_W2, SCRATCH_W2 + 1))
+                }
+                ZeropageType.DONTUSE -> {
+                    free.clear() // don't use zeropage at all
+                }
+                else -> throw CompilerException("for this machine target, zero page type 'floatsafe' is not available. ${options.zeropage}")
+            }
+
+            require(SCRATCH_B1 !in free)
+            require(SCRATCH_REG !in free)
+            require(SCRATCH_W1 !in free)
+            require(SCRATCH_W2 !in free)
+
+            for (reserved in options.zpReserved)
+                reserve(reserved)
+        }
+    }
+}

compiler/src/prog8/functions/BuiltinFunctions.kt

@@ -1,397 +0,0 @@
-package prog8.functions
-
-import prog8.ast.Program
-import prog8.ast.base.*
-import prog8.ast.expressions.*
-import prog8.compiler.CompilerException
-import kotlin.math.*
-
-
-class FParam(val name: String, val possibleDatatypes: Set<DataType>)
-
-
-typealias ConstExpressionCaller = (args: List<Expression>, position: Position, program: Program) -> NumericLiteralValue
-
-
-class FSignature(val pure: Boolean,      // does it have side effects?
-                 val parameters: List<FParam>,
-                 val returntype: DataType?,
-                 val constExpressionFunc: ConstExpressionCaller? = null)
-
-
-val BuiltinFunctions = mapOf(
-        // this set of function have no return value and operate in-place:
-    "rol"         to FSignature(false, listOf(FParam("item", setOf(DataType.UBYTE, DataType.UWORD))), null),
-    "ror"         to FSignature(false, listOf(FParam("item", setOf(DataType.UBYTE, DataType.UWORD))), null),
-    "rol2"        to FSignature(false, listOf(FParam("item", setOf(DataType.UBYTE, DataType.UWORD))), null),
-    "ror2"        to FSignature(false, listOf(FParam("item", setOf(DataType.UBYTE, DataType.UWORD))), null),
-    "lsl"         to FSignature(false, listOf(FParam("item", IntegerDatatypes)), null),
-    "lsr"         to FSignature(false, listOf(FParam("item", IntegerDatatypes)), null),
-    "sort"        to FSignature(false, listOf(FParam("array", ArrayDatatypes)), null),
-    "reverse"     to FSignature(false, listOf(FParam("array", ArrayDatatypes)), null),
-        // these few have a return value depending on the argument(s):
-    "max"         to FSignature(true, listOf(FParam("values", ArrayDatatypes)), null) { a, p, prg -> collectionArg(a, p, prg, ::builtinMax) },    // type depends on args
-    "min"         to FSignature(true, listOf(FParam("values", ArrayDatatypes)), null) { a, p, prg -> collectionArg(a, p, prg, ::builtinMin) },    // type depends on args
-    "sum"         to FSignature(true, listOf(FParam("values", ArrayDatatypes)), null) { a, p, prg -> collectionArg(a, p, prg, ::builtinSum) },      // type depends on args
-    "abs"         to FSignature(true, listOf(FParam("value", NumericDatatypes)), null, ::builtinAbs),      // type depends on argument
-    "len"         to FSignature(true, listOf(FParam("values", IterableDatatypes)), null, ::builtinLen),    // type is UBYTE or UWORD depending on actual length
-        // normal functions follow:
-    "sgn"         to FSignature(true, listOf(FParam("value", NumericDatatypes)), DataType.BYTE, ::builtinSgn ),
-    "sin"         to FSignature(true, listOf(FParam("rads", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArg(a, p, prg, Math::sin) },
-    "sin8"        to FSignature(true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.BYTE, ::builtinSin8 ),
-    "sin8u"       to FSignature(true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.UBYTE, ::builtinSin8u ),
-    "sin16"       to FSignature(true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.WORD, ::builtinSin16 ),
-    "sin16u"      to FSignature(true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.UWORD, ::builtinSin16u ),
-    "cos"         to FSignature(true, listOf(FParam("rads", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArg(a, p, prg, Math::cos) },
-    "cos8"        to FSignature(true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.BYTE, ::builtinCos8 ),
-    "cos8u"       to FSignature(true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.UBYTE, ::builtinCos8u ),
-    "cos16"       to FSignature(true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.WORD, ::builtinCos16 ),
-    "cos16u"      to FSignature(true, listOf(FParam("angle8", setOf(DataType.UBYTE))), DataType.UWORD, ::builtinCos16u ),
-    "tan"         to FSignature(true, listOf(FParam("rads", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArg(a, p, prg, Math::tan) },
-    "atan"        to FSignature(true, listOf(FParam("rads", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArg(a, p, prg, Math::atan) },
-    "ln"          to FSignature(true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArg(a, p, prg, Math::log) },
-    "log2"        to FSignature(true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArg(a, p, prg, ::log2) },
-    "sqrt16"      to FSignature(true, listOf(FParam("value", setOf(DataType.UWORD))), DataType.UBYTE) { a, p, prg -> oneIntArgOutputInt(a, p, prg) { sqrt(it.toDouble()).toInt() } },
-    "sqrt"        to FSignature(true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArg(a, p, prg, Math::sqrt) },
-    "rad"         to FSignature(true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArg(a, p, prg, Math::toRadians) },
-    "deg"         to FSignature(true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArg(a, p, prg, Math::toDegrees) },
-    "round"       to FSignature(true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArgOutputWord(a, p, prg, Math::round) },
-    "floor"       to FSignature(true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArgOutputWord(a, p, prg, Math::floor) },
-    "ceil"        to FSignature(true, listOf(FParam("value", setOf(DataType.FLOAT))), DataType.FLOAT) { a, p, prg -> oneDoubleArgOutputWord(a, p, prg, Math::ceil) },
-    "any"         to FSignature(true, listOf(FParam("values", ArrayDatatypes)), DataType.UBYTE) { a, p, prg -> collectionArg(a, p, prg, ::builtinAny) },
-    "all"         to FSignature(true, listOf(FParam("values", ArrayDatatypes)), DataType.UBYTE) { a, p, prg -> collectionArg(a, p, prg, ::builtinAll) },
-    "lsb"         to FSignature(true, listOf(FParam("value", setOf(DataType.UWORD, DataType.WORD))), DataType.UBYTE) { a, p, prg -> oneIntArgOutputInt(a, p, prg) { x: Int -> x and 255 }},
-    "msb"         to FSignature(true, listOf(FParam("value", setOf(DataType.UWORD, DataType.WORD))), DataType.UBYTE) { a, p, prg -> oneIntArgOutputInt(a, p, prg) { x: Int -> x ushr 8 and 255}},
-    "mkword"      to FSignature(true, listOf(FParam("lsb", setOf(DataType.UBYTE)), FParam("msb", setOf(DataType.UBYTE))), DataType.UWORD, ::builtinMkword),
-    "rnd"         to FSignature(true, emptyList(), DataType.UBYTE),
-    "rndw"        to FSignature(true, emptyList(), DataType.UWORD),
-    "rndf"        to FSignature(true, emptyList(), DataType.FLOAT),
-    "exit"        to FSignature(false, listOf(FParam("returnvalue", setOf(DataType.UBYTE))), null),
-    "rsave"       to FSignature(false, emptyList(), null),
-    "rrestore"    to FSignature(false, emptyList(), null),
-    "set_carry"   to FSignature(false, emptyList(), null),
-    "clear_carry" to FSignature(false, emptyList(), null),
-    "set_irqd"    to FSignature(false, emptyList(), null),
-    "clear_irqd"  to FSignature(false, emptyList(), null),
-    "read_flags"  to FSignature(false, emptyList(), DataType.UBYTE),
-    "swap"        to FSignature(false, listOf(FParam("first", NumericDatatypes), FParam("second", NumericDatatypes)), null),
-    "memcopy"     to FSignature(false, listOf(
-                            FParam("from", IterableDatatypes + DataType.UWORD),
-                            FParam("to", IterableDatatypes + DataType.UWORD),
-                            FParam("numbytes", setOf(DataType.UBYTE))), null),
-    "memset"      to FSignature(false, listOf(
-                            FParam("address", IterableDatatypes + DataType.UWORD),
-                            FParam("numbytes", setOf(DataType.UWORD)),
-                            FParam("bytevalue", ByteDatatypes)), null),
-    "memsetw"     to FSignature(false, listOf(
-                            FParam("address", IterableDatatypes + DataType.UWORD),
-                            FParam("numwords", setOf(DataType.UWORD)),
-                            FParam("wordvalue", setOf(DataType.UWORD, DataType.WORD))), null),
-    "strlen"      to FSignature(true, listOf(FParam("string", setOf(DataType.STR))), DataType.UBYTE, ::builtinStrlen),
-    "substr"      to FSignature(false, listOf(
-            FParam("source", IterableDatatypes + DataType.UWORD),
-            FParam("target", IterableDatatypes + DataType.UWORD),
-            FParam("start", setOf(DataType.UBYTE)),
-            FParam("length", setOf(DataType.UBYTE))), null),
-    "leftstr"      to FSignature(false, listOf(
-            FParam("source", IterableDatatypes + DataType.UWORD),
-            FParam("target", IterableDatatypes + DataType.UWORD),
-            FParam("length", setOf(DataType.UBYTE))), null),
-    "rightstr"      to FSignature(false, listOf(
-            FParam("source", IterableDatatypes + DataType.UWORD),
-            FParam("target", IterableDatatypes + DataType.UWORD),
-            FParam("length", setOf(DataType.UBYTE))), null)
-)
-
-fun builtinMax(array: List<Number>): Number = array.maxBy { it.toDouble() }!!
-
-fun builtinMin(array: List<Number>): Number = array.minBy { it.toDouble() }!!
-
-fun builtinSum(array: List<Number>): Number = array.sumByDouble { it.toDouble() }
-
-fun builtinAny(array: List<Number>): Number = if(array.any { it.toDouble()!=0.0 }) 1 else 0
-
-fun builtinAll(array: List<Number>): Number = if(array.all { it.toDouble()!=0.0 }) 1 else 0
-
-
-fun builtinFunctionReturnType(function: String, args: List<Expression>, program: Program): InferredTypes.InferredType {
-
-    fun datatypeFromIterableArg(arglist: Expression): DataType {
-        if(arglist is ArrayLiteralValue) {
-            val dt = arglist.value.map {it.inferType(program).typeOrElse(DataType.STRUCT)}.toSet()
-            if(dt.any { it !in NumericDatatypes }) {
-                throw FatalAstException("fuction $function only accepts array of numeric values")
-            }
-            if(DataType.FLOAT in dt) return DataType.FLOAT
-            if(DataType.UWORD in dt) return DataType.UWORD
-            if(DataType.WORD in dt) return DataType.WORD
-            if(DataType.BYTE in dt) return DataType.BYTE
-            return DataType.UBYTE
-        }
-        if(arglist is IdentifierReference) {
-            val idt = arglist.inferType(program)
-            if(!idt.isKnown)
-                throw FatalAstException("couldn't determine type of iterable $arglist")
-            return when(val dt = idt.typeOrElse(DataType.STRUCT)) {
-                DataType.STR, in NumericDatatypes -> dt
-                in ArrayDatatypes -> ArrayElementTypes.getValue(dt)
-                else -> throw FatalAstException("function '$function' requires one argument which is an iterable")
-            }
-        }
-        throw FatalAstException("function '$function' requires one argument which is an iterable")
-    }
-
-    val func = BuiltinFunctions.getValue(function)
-    if(func.returntype!=null)
-        return InferredTypes.knownFor(func.returntype)
-    // function has return values, but the return type depends on the arguments
-
-    return when (function) {
-        "abs" -> {
-            val dt = args.single().inferType(program)
-            if(dt.typeOrElse(DataType.STRUCT) in NumericDatatypes)
-                return dt
-            else
-                throw FatalAstException("weird datatype passed to abs $dt")
-        }
-        "max", "min" -> {
-            when(val dt = datatypeFromIterableArg(args.single())) {
-                DataType.STR -> InferredTypes.knownFor(DataType.UBYTE)
-                in NumericDatatypes -> InferredTypes.knownFor(dt)
-                in ArrayDatatypes -> InferredTypes.knownFor(ArrayElementTypes.getValue(dt))
-                else -> InferredTypes.unknown()
-            }
-        }
-        "sum" -> {
-            when(datatypeFromIterableArg(args.single())) {
-                DataType.UBYTE, DataType.UWORD -> InferredTypes.knownFor(DataType.UWORD)
-                DataType.BYTE, DataType.WORD -> InferredTypes.knownFor(DataType.WORD)
-                DataType.FLOAT -> InferredTypes.knownFor(DataType.FLOAT)
-                DataType.ARRAY_UB, DataType.ARRAY_UW -> InferredTypes.knownFor(DataType.UWORD)
-                DataType.ARRAY_B, DataType.ARRAY_W -> InferredTypes.knownFor(DataType.WORD)
-                DataType.ARRAY_F -> InferredTypes.knownFor(DataType.FLOAT)
-                DataType.STR -> InferredTypes.knownFor(DataType.UWORD)
-                else -> InferredTypes.unknown()
-            }
-        }
-        "len" -> {
-            // a length can be >255 so in that case, the result is an UWORD instead of an UBYTE
-            // but to avoid a lot of code duplication we simply assume UWORD in all cases for now
-            return InferredTypes.knownFor(DataType.UWORD)
-        }
-        else -> return InferredTypes.unknown()
-    }
-}
-
-
-class NotConstArgumentException: AstException("not a const argument to a built-in function")
-class CannotEvaluateException(func:String, msg: String): FatalAstException("cannot evaluate built-in function $func: $msg")
-
-
-private fun oneDoubleArg(args: List<Expression>, position: Position, program: Program, function: (arg: Double)->Number): NumericLiteralValue {
-    if(args.size!=1)
-        throw SyntaxError("built-in function requires one floating point argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val float = constval.number.toDouble()
-    return numericLiteral(function(float), args[0].position)
-}
-
-private fun oneDoubleArgOutputWord(args: List<Expression>, position: Position, program: Program, function: (arg: Double)->Number): NumericLiteralValue {
-    if(args.size!=1)
-        throw SyntaxError("built-in function requires one floating point argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val float = constval.number.toDouble()
-    return NumericLiteralValue(DataType.WORD, function(float).toInt(), args[0].position)
-}
-
-private fun oneIntArgOutputInt(args: List<Expression>, position: Position, program: Program, function: (arg: Int)->Number): NumericLiteralValue {
-    if(args.size!=1)
-        throw SyntaxError("built-in function requires one integer argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    if(constval.type != DataType.UBYTE && constval.type!= DataType.UWORD)
-        throw SyntaxError("built-in function requires one integer argument", position)
-
-    val integer = constval.number.toInt()
-    return numericLiteral(function(integer).toInt(), args[0].position)
-}
-
-private fun collectionArg(args: List<Expression>, position: Position, program: Program, function: (arg: List<Number>)->Number): NumericLiteralValue {
-    if(args.size!=1)
-        throw SyntaxError("builtin function requires one non-scalar argument", position)
-
-    val array= args[0] as? ArrayLiteralValue ?: throw NotConstArgumentException()
-    val constElements = array.value.map{it.constValue(program)?.number}
-    if(constElements.contains(null))
-        throw NotConstArgumentException()
-
-    return NumericLiteralValue.optimalNumeric(function(constElements.mapNotNull { it }), args[0].position)
-}
-
-private fun builtinAbs(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    // 1 arg, type = float or int, result type= isSameAs as argument type
-    if(args.size!=1)
-        throw SyntaxError("abs requires one numeric argument", position)
-
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    return when (constval.type) {
-        in IntegerDatatypes -> numericLiteral(abs(constval.number.toInt()), args[0].position)
-        DataType.FLOAT -> numericLiteral(abs(constval.number.toDouble()), args[0].position)
-        else -> throw SyntaxError("abs requires one numeric argument", position)
-    }
-}
-
-private fun builtinStrlen(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 1)
-        throw SyntaxError("strlen requires one argument", position)
-    val argument = args[0].constValue(program) ?: throw NotConstArgumentException()
-    if(argument.type != DataType.STR)
-        throw SyntaxError("strlen must have string argument", position)
-
-    throw NotConstArgumentException()       // this function is not considering the string argument a constant
-}
-
-private fun builtinLen(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    // note: in some cases the length is > 255 and then we have to return a UWORD type instead of a UBYTE.
-    if(args.size!=1)
-        throw SyntaxError("len requires one argument", position)
-    val constArg = args[0].constValue(program)
-    if(constArg!=null)
-        throw SyntaxError("len of weird argument ${args[0]}", position)
-
-    val directMemVar = ((args[0] as? DirectMemoryRead)?.addressExpression as? IdentifierReference)?.targetVarDecl(program.namespace)
-    var arraySize = directMemVar?.arraysize?.size()
-    if(arraySize != null)
-        return NumericLiteralValue.optimalInteger(arraySize, position)
-    if(args[0] is ArrayLiteralValue)
-        return NumericLiteralValue.optimalInteger((args[0] as ArrayLiteralValue).value.size, position)
-    if(args[0] !is IdentifierReference)
-        throw SyntaxError("len argument should be an identifier, but is ${args[0]}", position)
-    val target = (args[0] as IdentifierReference).targetVarDecl(program.namespace)
-            ?: throw CannotEvaluateException("len", "no target vardecl")
-
-    return when(target.datatype) {
-        DataType.ARRAY_UB, DataType.ARRAY_B, DataType.ARRAY_UW, DataType.ARRAY_W -> {
-            arraySize = target.arraysize?.size()
-            if(arraySize==null)
-                throw CannotEvaluateException("len", "arraysize unknown")
-            if(arraySize>256)
-                throw CompilerException("array length exceeds byte limit ${target.position}")
-            NumericLiteralValue.optimalInteger(arraySize, args[0].position)
-        }
-        DataType.ARRAY_F -> {
-            arraySize = target.arraysize?.size()
-            if(arraySize==null)
-                throw CannotEvaluateException("len", "arraysize unknown")
-            if(arraySize>256)
-                throw CompilerException("array length exceeds byte limit ${target.position}")
-            NumericLiteralValue.optimalInteger(arraySize, args[0].position)
-        }
-        DataType.STR -> {
-            val refLv = target.value as StringLiteralValue
-            if(refLv.value.length>255)
-                throw CompilerException("string length exceeds byte limit ${refLv.position}")
-            NumericLiteralValue.optimalInteger(refLv.value.length, args[0].position)
-        }
-        in NumericDatatypes -> throw SyntaxError("len of weird argument ${args[0]}", position)
-        else -> throw CompilerException("weird datatype")
-    }
-}
-
-
-private fun builtinMkword(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 2)
-        throw SyntaxError("mkword requires lsb and msb arguments", position)
-    val constLsb = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val constMsb = args[1].constValue(program) ?: throw NotConstArgumentException()
-    val result = (constMsb.number.toInt() shl 8) or constLsb.number.toInt()
-    return NumericLiteralValue(DataType.UWORD, result, position)
-}
-
-private fun builtinSin8(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 1)
-        throw SyntaxError("sin8 requires one argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
-    return NumericLiteralValue(DataType.BYTE, (127.0 * sin(rad)).toShort(), position)
-}
-
-private fun builtinSin8u(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 1)
-        throw SyntaxError("sin8u requires one argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
-    return NumericLiteralValue(DataType.UBYTE, (128.0 + 127.5 * sin(rad)).toShort(), position)
-}
-
-private fun builtinCos8(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 1)
-        throw SyntaxError("cos8 requires one argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
-    return NumericLiteralValue(DataType.BYTE, (127.0 * cos(rad)).toShort(), position)
-}
-
-private fun builtinCos8u(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 1)
-        throw SyntaxError("cos8u requires one argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
-    return NumericLiteralValue(DataType.UBYTE, (128.0 + 127.5 * cos(rad)).toShort(), position)
-}
-
-private fun builtinSin16(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 1)
-        throw SyntaxError("sin16 requires one argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
-    return NumericLiteralValue(DataType.WORD, (32767.0 * sin(rad)).toInt(), position)
-}
-
-private fun builtinSin16u(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 1)
-        throw SyntaxError("sin16u requires one argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
-    return NumericLiteralValue(DataType.UWORD, (32768.0 + 32767.5 * sin(rad)).toInt(), position)
-}
-
-private fun builtinCos16(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 1)
-        throw SyntaxError("cos16 requires one argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
-    return NumericLiteralValue(DataType.WORD, (32767.0 * cos(rad)).toInt(), position)
-}
-
-private fun builtinCos16u(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 1)
-        throw SyntaxError("cos16u requires one argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    val rad = constval.number.toDouble() /256.0 * 2.0 * PI
-    return NumericLiteralValue(DataType.UWORD, (32768.0 + 32767.5 * cos(rad)).toInt(), position)
-}
-
-private fun builtinSgn(args: List<Expression>, position: Position, program: Program): NumericLiteralValue {
-    if (args.size != 1)
-        throw SyntaxError("sgn requires one argument", position)
-    val constval = args[0].constValue(program) ?: throw NotConstArgumentException()
-    return NumericLiteralValue(DataType.BYTE, constval.number.toDouble().sign.toShort(), position)
-}
-
-private fun numericLiteral(value: Number, position: Position): NumericLiteralValue {
-    val floatNum=value.toDouble()
-    val tweakedValue: Number =
-            if(floatNum== floor(floatNum) && (floatNum>=-32768 && floatNum<=65535))
-                floatNum.toInt()  // we have an integer disguised as a float.
-            else
-                floatNum
-
-    return when(tweakedValue) {
-        is Int -> NumericLiteralValue.optimalNumeric(value.toInt(), position)
-        is Short -> NumericLiteralValue.optimalNumeric(value.toInt(), position)
-        is Byte -> NumericLiteralValue(DataType.UBYTE, value.toShort(), position)
-        is Double -> NumericLiteralValue(DataType.FLOAT, value.toDouble(), position)
-        is Float -> NumericLiteralValue(DataType.FLOAT, value.toDouble(), position)
-        else -> throw FatalAstException("invalid number type ${value::class}")
-    }
-}

compiler/src/prog8/optimizer/AssignmentTransformer.kt

@@ -1,157 +0,0 @@
-package prog8.optimizer
-
-import prog8.ast.Node
-import prog8.ast.Program
-import prog8.ast.base.ErrorReporter
-import prog8.ast.expressions.BinaryExpression
-import prog8.ast.processing.AstWalker
-import prog8.ast.processing.IAstModification
-import prog8.ast.statements.Assignment
-import prog8.ast.statements.PostIncrDecr
-
-
-
-internal class AssignmentTransformer(val program: Program, val errors: ErrorReporter) : AstWalker() {
-
-    var optimizationsDone: Int = 0
-    private val noModifications = emptyList<IAstModification>()
-
-    override fun before(assignment: Assignment, parent: Node): Iterable<IAstModification> {
-        // modify A = A + 5 back into augmented form A += 5 for easier code generation for optimized in-place assignments
-        // also to put code generation stuff together, single value assignment (A = 5) is converted to a special
-        // augmented form as wel (with the operator "setvalue")
-        if (assignment.aug_op == null) {
-            val binExpr = assignment.value as? BinaryExpression
-            if (binExpr != null) {
-                if (assignment.target.isSameAs(binExpr.left)) {
-                    assignment.value = binExpr.right
-                    assignment.aug_op = binExpr.operator + "="
-                    assignment.value.parent = assignment
-                    optimizationsDone++
-                    return noModifications
-                }
-            }
-            assignment.aug_op = "setvalue"
-            optimizationsDone++
-        } else if(assignment.aug_op == "+=") {
-            val binExpr = assignment.value as? BinaryExpression
-            if (binExpr != null) {
-                val leftnum = binExpr.left.constValue(program)?.number?.toDouble()
-                val rightnum = binExpr.right.constValue(program)?.number?.toDouble()
-                if(binExpr.operator == "+") {
-                    when {
-                        leftnum == 1.0 -> {
-                            optimizationsDone++
-                            return listOf(IAstModification.SwapOperands(binExpr))
-                        }
-                        leftnum == 2.0 -> {
-                            optimizationsDone++
-                            return listOf(IAstModification.SwapOperands(binExpr))
-                        }
-                        rightnum == 1.0 -> {
-                            // x += y + 1  -> x += y ,  x++
-                            return listOf(
-                                    IAstModification.ReplaceNode(assignment.value, binExpr.left, assignment),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "++", assignment.position), parent)
-                            )
-                        }
-                        rightnum == 2.0 -> {
-                            // x += y + 2  -> x += y ,  x++,  x++
-                            return listOf(
-                                    IAstModification.ReplaceNode(assignment.value, binExpr.left, assignment),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "++", assignment.position), parent),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "++", assignment.position), parent)
-                            )
-                        }
-                    }
-                } else if(binExpr.operator == "-") {
-                    when {
-                        leftnum == 1.0 -> {
-                            optimizationsDone++
-                            return listOf(IAstModification.SwapOperands(binExpr))
-                        }
-                        leftnum == 2.0 -> {
-                            optimizationsDone++
-                            return listOf(IAstModification.SwapOperands(binExpr))
-                        }
-                        rightnum == 1.0 -> {
-                            // x += y - 1  -> x += y ,  x--
-                            return listOf(
-                                    IAstModification.ReplaceNode(assignment.value, binExpr.left, assignment),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "--", assignment.position), parent)
-                            )
-                        }
-                        rightnum == 2.0 -> {
-                            // x += y - 2  -> x += y ,  x--,  x--
-                            return listOf(
-                                    IAstModification.ReplaceNode(assignment.value, binExpr.left, assignment),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "--", assignment.position), parent),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "--", assignment.position), parent)
-                            )
-                        }
-                    }
-                }
-            }
-        } else if(assignment.aug_op == "-=") {
-            val binExpr = assignment.value as? BinaryExpression
-            if (binExpr != null) {
-                val leftnum = binExpr.left.constValue(program)?.number?.toDouble()
-                val rightnum = binExpr.right.constValue(program)?.number?.toDouble()
-                if(binExpr.operator == "+") {
-                    when {
-                        leftnum == 1.0 -> {
-                            optimizationsDone++
-                            return listOf(IAstModification.SwapOperands(binExpr))
-                        }
-                        leftnum == 2.0 -> {
-                            optimizationsDone++
-                            return listOf(IAstModification.SwapOperands(binExpr))
-                        }
-                        rightnum == 1.0 -> {
-                            // x -= y + 1  -> x -= y ,  x--
-                            return listOf(
-                                    IAstModification.ReplaceNode(assignment.value, binExpr.left, assignment),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "--", assignment.position), parent)
-                            )
-                        }
-                        rightnum == 2.0 -> {
-                            // x -= y + 2  -> x -= y ,  x--,  x--
-                            return listOf(
-                                    IAstModification.ReplaceNode(assignment.value, binExpr.left, assignment),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "--", assignment.position), parent),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "--", assignment.position), parent)
-                            )
-                        }
-                    }
-                } else if(binExpr.operator == "-") {
-                    when {
-                        leftnum == 1.0 -> {
-                            optimizationsDone++
-                            return listOf(IAstModification.SwapOperands(binExpr))
-                        }
-                        leftnum == 2.0 -> {
-                            optimizationsDone++
-                            return listOf(IAstModification.SwapOperands(binExpr))
-                        }
-                        rightnum == 1.0 -> {
-                            // x -= y - 1  -> x -= y ,  x++
-                            return listOf(
-                                    IAstModification.ReplaceNode(assignment.value, binExpr.left, assignment),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "++", assignment.position), parent)
-                            )
-                        }
-                        rightnum == 2.0 -> {
-                            // x -= y - 2  -> x -= y ,  x++,  x++
-                            return listOf(
-                                    IAstModification.ReplaceNode(assignment.value, binExpr.left, assignment),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "++", assignment.position), parent),
-                                    IAstModification.InsertAfter(assignment, PostIncrDecr(assignment.target, "++", assignment.position), parent)
-                            )
-                        }
-                    }
-                }
-            }
-        }
-        return noModifications
-    }
-}

compiler/src/prog8/optimizer/BinExprSplitter.kt

@@ -0,0 +1,87 @@
+package prog8.optimizer
+
+import prog8.ast.Node
+import prog8.ast.Program
+import prog8.ast.expressions.*
+import prog8.ast.statements.AssignTarget
+import prog8.ast.statements.Assignment
+import prog8.ast.walk.AstWalker
+import prog8.ast.walk.IAstModification
+import prog8.compiler.target.ICompilationTarget
+
+
+internal class BinExprSplitter(private val program: Program, private val compTarget: ICompilationTarget) : AstWalker() {
+    private val noModifications = emptyList<IAstModification>()
+
+//    override fun after(decl: VarDecl, parent: Node): Iterable<IAstModification> {
+// TODO somehow if we do this, the resulting code for some programs (cube3d.p8) gets hundreds of bytes larger...:
+//        if(decl.type==VarDeclType.VAR ) {
+//            val binExpr = decl.value as? BinaryExpression
+//            if (binExpr != null && binExpr.operator in augmentAssignmentOperators) {
+//                // split into a vardecl with just the left expression, and an aug. assignment with the right expression.
+//                val augExpr = BinaryExpression(IdentifierReference(listOf(decl.name), decl.position), binExpr.operator, binExpr.right, binExpr.position)
+//                val target = AssignTarget(IdentifierReference(listOf(decl.name), decl.position), null, null, decl.position)
+//                val assign = Assignment(target, augExpr, binExpr.position)
+//                println("SPLIT VARDECL $decl")
+//                return listOf(
+//                        IAstModification.SetExpression({ decl.value = it }, binExpr.left, decl),
+//                        IAstModification.InsertAfter(decl, assign, parent)
+//                )
+//            }
+//        }
+//        return noModifications
+//    }
+
+    override fun after(assignment: Assignment, parent: Node): Iterable<IAstModification> {
+
+        val binExpr = assignment.value as? BinaryExpression
+        if (binExpr != null) {
+/*
+
+Reduce the complexity of a (binary) expression that has to be evaluated on the eval stack,
+by attempting to splitting it up into individual simple steps.
+We only consider a binary expression *one* level deep (so the operands must not be a combined expression)
+
+
+X =      BinExpr                                    X   =   LeftExpr
+        <operator>                                     followed by
+          /   \             IF 'X' not used         X   =   BinExpr
+         /     \             IN expression ==>             <operator>
+        /       \                                           /   \
+    LeftExpr.  RightExpr.                                  /     \
+                                                          X     RightExpr.
+
+
+ */
+            if(binExpr.operator in augmentAssignmentOperators && isSimpleTarget(assignment.target, program)) {
+                if(assignment.target isSameAs binExpr.left || assignment.target isSameAs binExpr.right)
+                    return noModifications
+
+                if(isSimpleExpression(binExpr.right) && !assignment.isAugmentable) {
+                    val firstAssign = Assignment(assignment.target, binExpr.left, binExpr.left.position)
+                    val targetExpr = assignment.target.toExpression()
+                    val augExpr = BinaryExpression(targetExpr, binExpr.operator, binExpr.right, binExpr.right.position)
+                    return listOf(
+                            IAstModification.InsertBefore(assignment, firstAssign, assignment.definingScope()),
+                            IAstModification.ReplaceNode(assignment.value, augExpr, assignment))
+                }
+            }
+
+            // TODO further unraveling of binary expression trees into flat statements.
+            // however this should probably be done in a more generic way to also service
+            // the expressiontrees that are not used in an assignment statement...
+        }
+
+        return noModifications
+    }
+
+    private fun isSimpleExpression(expr: Expression) =
+            expr is IdentifierReference || expr is NumericLiteralValue || expr is AddressOf || expr is DirectMemoryRead || expr is StringLiteralValue || expr is ArrayLiteralValue || expr is RangeExpr
+
+    private fun isSimpleTarget(target: AssignTarget, program: Program) =
+            if (target.identifier!=null || target.memoryAddress!=null)
+                compTarget.isInRegularRAM(target, program)
+            else
+                false
+
+}

compiler/src/prog8/optimizer/CallGraph.kt

@@ -6,28 +6,29 @@ import prog8.ast.Node
 import prog8.ast.Program
 import prog8.ast.base.DataType
 import prog8.ast.base.ParentSentinel
+import prog8.ast.base.Position
+import prog8.ast.expressions.AddressOf
 import prog8.ast.expressions.FunctionCall
 import prog8.ast.expressions.IdentifierReference
-import prog8.ast.processing.IAstVisitor
 import prog8.ast.statements.*
-import prog8.compiler.loadAsmIncludeFile
+import prog8.ast.walk.IAstVisitor
+import prog8.compiler.IErrorReporter
+import java.nio.file.Path
 
 private val alwaysKeepSubroutines = setOf(
-        Pair("main", "start"),
-        Pair("irq", "irq"),
-        Pair("prog8_lib", "init_system")
+        Pair("main", "start")
 )
 
 private val asmJumpRx = Regex("""[\-+a-zA-Z0-9_ \t]+(jmp|jsr)[ \t]+(\S+).*""", RegexOption.IGNORE_CASE)
 private val asmRefRx = Regex("""[\-+a-zA-Z0-9_ \t]+(...)[ \t]+(\S+).*""", RegexOption.IGNORE_CASE)
 
 
-class CallGraph(private val program: Program) : IAstVisitor {
+class CallGraph(private val program: Program, private val asmFileLoader: (filename: String, source: Path)->String) : IAstVisitor {
 
-    val modulesImporting = mutableMapOf<Module, List<Module>>().withDefault { mutableListOf() }
-    val modulesImportedBy = mutableMapOf<Module, List<Module>>().withDefault { mutableListOf() }
-    val subroutinesCalling = mutableMapOf<INameScope, List<Subroutine>>().withDefault { mutableListOf() }
-    val subroutinesCalledBy = mutableMapOf<Subroutine, List<Node>>().withDefault { mutableListOf() }
+    val imports = mutableMapOf<Module, List<Module>>().withDefault { mutableListOf() }
+    val importedBy = mutableMapOf<Module, List<Module>>().withDefault { mutableListOf() }
+    val calls = mutableMapOf<Subroutine, List<Subroutine>>().withDefault { mutableListOf() }
+    val calledBy = mutableMapOf<Subroutine, List<Node>>().withDefault { mutableListOf() }
 
     // TODO  add dataflow graph: what statements use what variables - can be used to eliminate unused vars
     val usedSymbols = mutableSetOf<Statement>()
@@ -55,17 +56,8 @@ class CallGraph(private val program: Program) : IAstVisitor {
             it.importedBy.clear()
             it.imports.clear()
 
-            it.importedBy.addAll(modulesImportedBy.getValue(it))
-            it.imports.addAll(modulesImporting.getValue(it))
-
-            forAllSubroutines(it) { sub ->
-                sub.calledBy.clear()
-                sub.calls.clear()
-
-                sub.calledBy.addAll(subroutinesCalledBy.getValue(sub))
-                sub.calls.addAll(subroutinesCalling.getValue(sub))
-            }
-
+            it.importedBy.addAll(importedBy.getValue(it))
+            it.imports.addAll(imports.getValue(it))
         }
 
         val rootmodule = program.modules.first()
@@ -85,12 +77,14 @@ class CallGraph(private val program: Program) : IAstVisitor {
         val thisModule = directive.definingModule()
         if (directive.directive == "%import") {
             val importedModule: Module = program.modules.single { it.name == directive.args[0].name }
-            modulesImporting[thisModule] = modulesImporting.getValue(thisModule).plus(importedModule)
-            modulesImportedBy[importedModule] = modulesImportedBy.getValue(importedModule).plus(thisModule)
+            imports[thisModule] = imports.getValue(thisModule).plus(importedModule)
+            importedBy[importedModule] = importedBy.getValue(importedModule).plus(thisModule)
         } else if (directive.directive == "%asminclude") {
-            val asm = loadAsmIncludeFile(directive.args[0].str!!, thisModule.source)
-            val scope = directive.definingScope()
-            scanAssemblyCode(asm, directive, scope)
+            val asm = asmFileLoader(directive.args[0].str!!, thisModule.source)
+            val scope = directive.definingSubroutine()
+            if(scope!=null) {
+                scanAssemblyCode(asm, directive, scope)
+            }
         }
 
         super.visit(directive)
@@ -98,7 +92,7 @@ class CallGraph(private val program: Program) : IAstVisitor {
 
     override fun visit(identifier: IdentifierReference) {
         // track symbol usage
-        val target = identifier.targetStatement(this.program.namespace)
+        val target = identifier.targetStatement(program)
         if (target != null) {
             addNodeAndParentScopes(target)
         }
@@ -126,45 +120,53 @@ class CallGraph(private val program: Program) : IAstVisitor {
     }
 
     override fun visit(decl: VarDecl) {
-        if (decl.autogeneratedDontRemove || decl.definingModule().isLibraryModule) {
-            // make sure autogenerated vardecls are in the used symbols and are never removed as 'unused'
+        if (decl.autogeneratedDontRemove || decl.datatype==DataType.STRUCT)
             addNodeAndParentScopes(decl)
-        }
-
-        if (decl.datatype == DataType.STRUCT)
+        else if(decl.parent is Block && decl.definingModule().isLibraryModule)
             addNodeAndParentScopes(decl)
 
         super.visit(decl)
     }
 
     override fun visit(functionCall: FunctionCall) {
-        val otherSub = functionCall.target.targetSubroutine(program.namespace)
+        val otherSub = functionCall.target.targetSubroutine(program)
         if (otherSub != null) {
             functionCall.definingSubroutine()?.let { thisSub ->
-                subroutinesCalling[thisSub] = subroutinesCalling.getValue(thisSub).plus(otherSub)
-                subroutinesCalledBy[otherSub] = subroutinesCalledBy.getValue(otherSub).plus(functionCall)
+                calls[thisSub] = calls.getValue(thisSub).plus(otherSub)
+                calledBy[otherSub] = calledBy.getValue(otherSub).plus(functionCall)
             }
         }
         super.visit(functionCall)
     }
 
     override fun visit(functionCallStatement: FunctionCallStatement) {
-        val otherSub = functionCallStatement.target.targetSubroutine(program.namespace)
+        val otherSub = functionCallStatement.target.targetSubroutine(program)
         if (otherSub != null) {
             functionCallStatement.definingSubroutine()?.let { thisSub ->
-                subroutinesCalling[thisSub] = subroutinesCalling.getValue(thisSub).plus(otherSub)
-                subroutinesCalledBy[otherSub] = subroutinesCalledBy.getValue(otherSub).plus(functionCallStatement)
+                calls[thisSub] = calls.getValue(thisSub).plus(otherSub)
+                calledBy[otherSub] = calledBy.getValue(otherSub).plus(functionCallStatement)
             }
         }
         super.visit(functionCallStatement)
     }
 
+    override fun visit(addressOf: AddressOf) {
+        val otherSub = addressOf.identifier.targetSubroutine(program)
+        if(otherSub!=null) {
+            addressOf.definingSubroutine()?.let { thisSub ->
+                calls[thisSub] = calls.getValue(thisSub).plus(otherSub)
+                calledBy[otherSub] = calledBy.getValue(otherSub).plus(thisSub)
+            }
+        }
+        super.visit(addressOf)
+    }
+
     override fun visit(jump: Jump) {
-        val otherSub = jump.identifier?.targetSubroutine(program.namespace)
+        val otherSub = jump.identifier?.targetSubroutine(program)
         if (otherSub != null) {
             jump.definingSubroutine()?.let { thisSub ->
-                subroutinesCalling[thisSub] = subroutinesCalling.getValue(thisSub).plus(otherSub)
-                subroutinesCalledBy[otherSub] = subroutinesCalledBy.getValue(otherSub).plus(jump)
+                calls[thisSub] = calls.getValue(thisSub).plus(otherSub)
+                calledBy[otherSub] = calledBy.getValue(otherSub).plus(jump)
             }
         }
         super.visit(jump)
@@ -177,12 +179,12 @@ class CallGraph(private val program: Program) : IAstVisitor {
 
     override fun visit(inlineAssembly: InlineAssembly) {
         // parse inline asm for subroutine calls (jmp, jsr)
-        val scope = inlineAssembly.definingScope()
+        val scope = inlineAssembly.definingSubroutine()
         scanAssemblyCode(inlineAssembly.assembly, inlineAssembly, scope)
         super.visit(inlineAssembly)
     }
 
-    private fun scanAssemblyCode(asm: String, context: Statement, scope: INameScope) {
+    private fun scanAssemblyCode(asm: String, context: Statement, scope: Subroutine?) {
         asm.lines().forEach { line ->
             val matches = asmJumpRx.matchEntire(line)
             if (matches != null) {
@@ -190,14 +192,16 @@ class CallGraph(private val program: Program) : IAstVisitor {
                 if (jumptarget != null && (jumptarget[0].isLetter() || jumptarget[0] == '_')) {
                     val node = program.namespace.lookup(jumptarget.split('.'), context)
                     if (node is Subroutine) {
-                        subroutinesCalling[scope] = subroutinesCalling.getValue(scope).plus(node)
-                        subroutinesCalledBy[node] = subroutinesCalledBy.getValue(node).plus(context)
+                        if(scope!=null)
+                            calls[scope] = calls.getValue(scope).plus(node)
+                        calledBy[node] = calledBy.getValue(node).plus(context)
                     } else if (jumptarget.contains('.')) {
                         // maybe only the first part already refers to a subroutine
                         val node2 = program.namespace.lookup(listOf(jumptarget.substringBefore('.')), context)
                         if (node2 is Subroutine) {
-                            subroutinesCalling[scope] = subroutinesCalling.getValue(scope).plus(node2)
-                            subroutinesCalledBy[node2] = subroutinesCalledBy.getValue(node2).plus(context)
+                            if(scope!=null)
+                                calls[scope] = calls.getValue(scope).plus(node2)
+                            calledBy[node2] = calledBy.getValue(node2).plus(context)
                         }
                     }
                 }
@@ -209,8 +213,9 @@ class CallGraph(private val program: Program) : IAstVisitor {
                         if (target.contains('.')) {
                             val node = program.namespace.lookup(listOf(target.substringBefore('.')), context)
                             if (node is Subroutine) {
-                                subroutinesCalling[scope] = subroutinesCalling.getValue(scope).plus(node)
-                                subroutinesCalledBy[node] = subroutinesCalledBy.getValue(node).plus(context)
+                                if(scope!=null)
+                                    calls[scope] = calls.getValue(scope).plus(node)
+                                calledBy[node] = calledBy.getValue(node).plus(context)
                             }
                         }
                     }
@@ -218,4 +223,55 @@ class CallGraph(private val program: Program) : IAstVisitor {
             }
         }
     }
+
+    fun checkRecursiveCalls(errors: IErrorReporter) {
+        val cycles = recursionCycles()
+        if(cycles.any()) {
+            errors.warn("Program contains recursive subroutine calls. These only works in very specific limited scenarios!", Position.DUMMY)
+            val printed = mutableSetOf<Subroutine>()
+            for(chain in cycles) {
+                if(chain[0] !in printed) {
+                    val chainStr = chain.joinToString(" <-- ") { "${it.name} at ${it.position}" }
+                    errors.warn("Cycle in (a subroutine call in) $chainStr", Position.DUMMY)
+                    printed.add(chain[0])
+                }
+            }
+        }
+    }
+
+    private fun recursionCycles(): List<List<Subroutine>> {
+        val chains = mutableListOf<MutableList<Subroutine>>()
+        for(caller in calls.keys) {
+            val visited = calls.keys.associateWith { false }.toMutableMap()
+            val recStack = calls.keys.associateWith { false }.toMutableMap()
+            val chain = mutableListOf<Subroutine>()
+            if(hasCycle(caller, visited, recStack, chain))
+                chains.add(chain)
+        }
+        return chains
+    }
+
+    private fun hasCycle(sub: Subroutine, visited: MutableMap<Subroutine, Boolean>, recStack: MutableMap<Subroutine, Boolean>, chain: MutableList<Subroutine>): Boolean {
+        // mark current node as visited and add to recursion stack
+        if(recStack[sub]==true)
+            return true
+        if(visited[sub]==true)
+            return false
+
+        // mark visited and add to recursion stack
+        visited[sub] = true
+        recStack[sub] = true
+
+        // recurse for all neighbours
+        for(called in calls.getValue(sub)) {
+            if(hasCycle(called, visited, recStack, chain)) {
+                chain.add(called)
+                return true
+            }
+        }
+
+        // pop from recursion stack
+        recStack[sub] = false
+        return false
+    }
 }

compiler/src/prog8/optimizer/ConstExprEvaluator.kt

@@ -132,11 +132,11 @@ class ConstExprEvaluator {
     private fun bitwiseand(left: NumericLiteralValue, right: NumericLiteralValue): NumericLiteralValue {
         if(left.type== DataType.UBYTE) {
             if(right.type in IntegerDatatypes) {
-                return NumericLiteralValue(DataType.UBYTE, (left.number.toInt() or (right.number.toInt() and 255)).toShort(), left.position)
+                return NumericLiteralValue(DataType.UBYTE, (left.number.toInt() and (right.number.toInt() and 255)).toShort(), left.position)
             }
         } else if(left.type== DataType.UWORD) {
             if(right.type in IntegerDatatypes) {
-                return NumericLiteralValue(DataType.UWORD, left.number.toInt() or right.number.toInt(), left.position)
+                return NumericLiteralValue(DataType.UWORD, left.number.toInt() and right.number.toInt(), left.position)
             }
         }
         throw ExpressionError("cannot calculate $left & $right", left.position)
@@ -163,7 +163,7 @@ class ConstExprEvaluator {
         val error = "cannot add $left and $right"
         return when (left.type) {
             in IntegerDatatypes -> when (right.type) {
-                in IntegerDatatypes -> NumericLiteralValue.optimalNumeric(left.number.toInt() + right.number.toInt(), left.position)
+                in IntegerDatatypes -> NumericLiteralValue.optimalInteger(left.number.toInt() + right.number.toInt(), left.position)
                 DataType.FLOAT -> NumericLiteralValue(DataType.FLOAT, left.number.toInt() + right.number.toDouble(), left.position)
                 else -> throw ExpressionError(error, left.position)
             }
@@ -180,7 +180,7 @@ class ConstExprEvaluator {
         val error = "cannot subtract $left and $right"
         return when (left.type) {
             in IntegerDatatypes -> when (right.type) {
-                in IntegerDatatypes -> NumericLiteralValue.optimalNumeric(left.number.toInt() - right.number.toInt(), left.position)
+                in IntegerDatatypes -> NumericLiteralValue.optimalInteger(left.number.toInt() - right.number.toInt(), left.position)
                 DataType.FLOAT -> NumericLiteralValue(DataType.FLOAT, left.number.toInt() - right.number.toDouble(), left.position)
                 else -> throw ExpressionError(error, left.position)
             }
@@ -197,7 +197,7 @@ class ConstExprEvaluator {
         val error = "cannot multiply ${left.type} and ${right.type}"
         return when (left.type) {
             in IntegerDatatypes -> when (right.type) {
-                in IntegerDatatypes -> NumericLiteralValue.optimalNumeric(left.number.toInt() * right.number.toInt(), left.position)
+                in IntegerDatatypes -> NumericLiteralValue.optimalInteger(left.number.toInt() * right.number.toInt(), left.position)
                 DataType.FLOAT -> NumericLiteralValue(DataType.FLOAT, left.number.toInt() * right.number.toDouble(), left.position)
                 else -> throw ExpressionError(error, left.position)
             }
@@ -220,7 +220,7 @@ class ConstExprEvaluator {
                 in IntegerDatatypes -> {
                     if(right.number.toInt()==0) divideByZeroError(right.position)
                     val result: Int = left.number.toInt() / right.number.toInt()
-                    NumericLiteralValue.optimalNumeric(result, left.position)
+                    NumericLiteralValue.optimalInteger(result, left.position)
                 }
                 DataType.FLOAT -> {
                     if(right.number.toDouble()==0.0) divideByZeroError(right.position)

compiler/src/prog8/optimizer/ConstantFoldingOptimizer.kt

@@ -4,182 +4,16 @@ import prog8.ast.Node
 import prog8.ast.Program
 import prog8.ast.base.*
 import prog8.ast.expressions.*
-import prog8.ast.processing.AstWalker
-import prog8.ast.processing.IAstModification
-import prog8.ast.statements.*
-import prog8.compiler.target.CompilationTarget
+import prog8.ast.statements.Assignment
+import prog8.ast.statements.ForLoop
+import prog8.ast.statements.VarDecl
+import prog8.ast.walk.AstWalker
+import prog8.ast.walk.IAstModification
+import prog8.compiler.target.ICompilationTarget
+import kotlin.math.pow
 
 
-// First thing to do is replace all constant identifiers with their actual value,
-// and the array var initializer values and sizes.
-// This is needed because further constant optimizations depend on those.
-internal class ConstantIdentifierReplacer(private val program: Program, private val errors: ErrorReporter) : AstWalker() {
-    private val noModifications = emptyList<IAstModification>()
-
-    override fun after(identifier: IdentifierReference, parent: Node): Iterable<IAstModification> {
-        // replace identifiers that refer to const value, with the value itself
-        // if it's a simple type and if it's not a left hand side variable
-        if(identifier.parent is AssignTarget)
-            return noModifications
-        var forloop = identifier.parent as? ForLoop
-        if(forloop==null)
-            forloop = identifier.parent.parent as? ForLoop
-        if(forloop!=null && identifier===forloop.loopVar)
-            return noModifications
-
-        val cval = identifier.constValue(program) ?: return noModifications
-        return when (cval.type) {
-            in NumericDatatypes -> listOf(IAstModification.ReplaceNode(identifier, NumericLiteralValue(cval.type, cval.number, identifier.position), identifier.parent))
-            in PassByReferenceDatatypes -> throw FatalAstException("pass-by-reference type should not be considered a constant")
-            else -> noModifications
-        }
-    }
-
-    override fun before(decl: VarDecl, parent: Node): Iterable<IAstModification> {
-        // the initializer value can't refer to the variable itself (recursive definition)
-        // TODO: use call graph for this?
-        if(decl.value?.referencesIdentifiers(decl.name) == true || decl.arraysize?.index?.referencesIdentifiers(decl.name) == true) {
-            errors.err("recursive var declaration", decl.position)
-            return noModifications
-        }
-
-        if(decl.type==VarDeclType.CONST || decl.type==VarDeclType.VAR) {
-            if(decl.isArray){
-                if(decl.arraysize==null) {
-                    // for arrays that have no size specifier (or a non-constant one) attempt to deduce the size
-                    val arrayval = decl.value as? ArrayLiteralValue
-                    if(arrayval!=null) {
-                        return listOf(IAstModification.SetExpression(
-                                { decl.arraysize = ArrayIndex(it, decl.position) },
-                                NumericLiteralValue.optimalInteger(arrayval.value.size, decl.position),
-                                decl
-                        ))
-                    }
-                }
-                else if(decl.arraysize?.size()==null) {
-                    val size = decl.arraysize!!.index.constValue(program)
-                    if(size!=null) {
-                        return listOf(IAstModification.SetExpression(
-                                { decl.arraysize = ArrayIndex(it, decl.position) },
-                                size, decl
-                        ))
-                    }
-                }
-            }
-
-            when(decl.datatype) {
-                DataType.FLOAT -> {
-                    // vardecl: for scalar float vars, promote constant integer initialization values to floats
-                    val litval = decl.value as? NumericLiteralValue
-                    if (litval!=null && litval.type in IntegerDatatypes) {
-                        val newValue = NumericLiteralValue(DataType.FLOAT, litval.number.toDouble(), litval.position)
-                        return listOf(IAstModification.ReplaceNode(decl.value!!, newValue, decl))
-                    }
-                }
-                DataType.ARRAY_UB, DataType.ARRAY_B, DataType.ARRAY_UW, DataType.ARRAY_W -> {
-                    val numericLv = decl.value as? NumericLiteralValue
-                    val rangeExpr = decl.value as? RangeExpr
-                    if(rangeExpr!=null) {
-                        // convert the initializer range expression to an actual array
-                        val declArraySize = decl.arraysize?.size()
-                        if(declArraySize!=null && declArraySize!=rangeExpr.size())
-                            errors.err("range expression size doesn't match declared array size", decl.value?.position!!)
-                        val constRange = rangeExpr.toConstantIntegerRange()
-                        if(constRange!=null) {
-                            val eltType = rangeExpr.inferType(program).typeOrElse(DataType.UBYTE)
-                            val newValue = if(eltType in ByteDatatypes) {
-                                ArrayLiteralValue(InferredTypes.InferredType.known(decl.datatype),
-                                        constRange.map { NumericLiteralValue(eltType, it.toShort(), decl.value!!.position) }.toTypedArray(),
-                                        position = decl.value!!.position)
-                            } else {
-                                ArrayLiteralValue(InferredTypes.InferredType.known(decl.datatype),
-                                        constRange.map { NumericLiteralValue(eltType, it, decl.value!!.position) }.toTypedArray(),
-                                        position = decl.value!!.position)
-                            }
-                            return listOf(IAstModification.ReplaceNode(decl.value!!, newValue, decl))
-                        }
-                    }
-                    if(numericLv!=null && numericLv.type==DataType.FLOAT)
-                        errors.err("arraysize requires only integers here", numericLv.position)
-                    val size = decl.arraysize?.size() ?: return noModifications
-                    if (rangeExpr==null && numericLv!=null) {
-                        // arraysize initializer is empty or a single int, and we know the size; create the arraysize.
-                        val fillvalue = numericLv.number.toInt()
-                        when(decl.datatype){
-                            DataType.ARRAY_UB -> {
-                                if(fillvalue !in 0..255)
-                                    errors.err("ubyte value overflow", numericLv.position)
-                            }
-                            DataType.ARRAY_B -> {
-                                if(fillvalue !in -128..127)
-                                    errors.err("byte value overflow", numericLv.position)
-                            }
-                            DataType.ARRAY_UW -> {
-                                if(fillvalue !in 0..65535)
-                                    errors.err("uword value overflow", numericLv.position)
-                            }
-                            DataType.ARRAY_W -> {
-                                if(fillvalue !in -32768..32767)
-                                    errors.err("word value overflow", numericLv.position)
-                            }
-                            else -> {}
-                        }
-                        // create the array itself, filled with the fillvalue.
-                        val array = Array(size) {fillvalue}.map { NumericLiteralValue(ArrayElementTypes.getValue(decl.datatype), it, numericLv.position) as Expression}.toTypedArray()
-                        val refValue = ArrayLiteralValue(InferredTypes.InferredType.known(decl.datatype), array, position = numericLv.position)
-                        return listOf(IAstModification.ReplaceNode(decl.value!!, refValue, decl))
-                    }
-                }
-                DataType.ARRAY_F  -> {
-                    val size = decl.arraysize?.size() ?: return noModifications
-                    val litval = decl.value as? NumericLiteralValue
-                    val rangeExpr = decl.value as? RangeExpr
-                    if(rangeExpr!=null) {
-                        // convert the initializer range expression to an actual array of floats
-                        val declArraySize = decl.arraysize?.size()
-                        if(declArraySize!=null && declArraySize!=rangeExpr.size())
-                            errors.err("range expression size doesn't match declared array size", decl.value?.position!!)
-                        val constRange = rangeExpr.toConstantIntegerRange()
-                        if(constRange!=null) {
-                            val newValue = ArrayLiteralValue(InferredTypes.InferredType.known(DataType.ARRAY_F),
-                                        constRange.map { NumericLiteralValue(DataType.FLOAT, it.toDouble(), decl.value!!.position) }.toTypedArray(),
-                                        position = decl.value!!.position)
-                            return listOf(IAstModification.ReplaceNode(decl.value!!, newValue, decl))
-                        }
-                    }
-                    if(rangeExpr==null && litval!=null) {
-                        // arraysize initializer is a single int, and we know the size.
-                        val fillvalue = litval.number.toDouble()
-                        if (fillvalue < CompilationTarget.machine.FLOAT_MAX_NEGATIVE || fillvalue > CompilationTarget.machine.FLOAT_MAX_POSITIVE)
-                            errors.err("float value overflow", litval.position)
-                        else {
-                            // create the array itself, filled with the fillvalue.
-                            val array = Array(size) {fillvalue}.map { NumericLiteralValue(DataType.FLOAT, it, litval.position) as Expression}.toTypedArray()
-                            val refValue = ArrayLiteralValue(InferredTypes.InferredType.known(DataType.ARRAY_F), array, position = litval.position)
-                            return listOf(IAstModification.ReplaceNode(decl.value!!, refValue, decl))
-                        }
-                    }
-                }
-                else -> {
-                    // nothing to do for this type
-                    // this includes strings and structs
-                }
-            }
-        }
-
-        val declValue = decl.value
-        if(declValue!=null && decl.type==VarDeclType.VAR
-                && declValue is NumericLiteralValue && !declValue.inferType(program).istype(decl.datatype)) {
-            // cast the numeric literal to the appropriate datatype of the variable
-            return listOf(IAstModification.ReplaceNode(decl.value!!, declValue.cast(decl.datatype), decl))
-        }
-
-        return noModifications
-    }
-}
-
-
-internal class ConstantFoldingOptimizer(private val program: Program, private val errors: ErrorReporter) : AstWalker() {
+internal class ConstantFoldingOptimizer(private val program: Program, private val compTarget: ICompilationTarget) : AstWalker() {
     private val noModifications = emptyList<IAstModification>()
 
     override fun before(memread: DirectMemoryRead, parent: Node): Iterable<IAstModification> {
@@ -203,7 +37,7 @@ internal class ConstantFoldingOptimizer(private val program: Program, private va
                 "-" -> when (subexpr.type) {
                     in IntegerDatatypes -> {
                         listOf(IAstModification.ReplaceNode(expr,
-                                NumericLiteralValue.optimalNumeric(-subexpr.number.toInt(), subexpr.position),
+                                NumericLiteralValue.optimalInteger(-subexpr.number.toInt(), subexpr.position),
                                 parent))
                     }
                     DataType.FLOAT -> {
@@ -214,9 +48,24 @@ internal class ConstantFoldingOptimizer(private val program: Program, private va
                     else -> throw ExpressionError("can only take negative of int or float", subexpr.position)
                 }
                 "~" -> when (subexpr.type) {
-                    in IntegerDatatypes -> {
+                    DataType.BYTE -> {
                         listOf(IAstModification.ReplaceNode(expr,
-                                NumericLiteralValue.optimalNumeric(subexpr.number.toInt().inv(), subexpr.position),
+                                NumericLiteralValue(DataType.BYTE, subexpr.number.toInt().inv(), subexpr.position),
+                                parent))
+                    }
+                    DataType.UBYTE -> {
+                        listOf(IAstModification.ReplaceNode(expr,
+                                NumericLiteralValue(DataType.UBYTE, subexpr.number.toInt().inv() and 255, subexpr.position),
+                                parent))
+                    }
+                    DataType.WORD -> {
+                        listOf(IAstModification.ReplaceNode(expr,
+                                NumericLiteralValue(DataType.WORD, subexpr.number.toInt().inv(), subexpr.position),
+                                parent))
+                    }
+                    DataType.UWORD -> {
+                        listOf(IAstModification.ReplaceNode(expr,
+                                NumericLiteralValue(DataType.UWORD, subexpr.number.toInt().inv() and 65535, subexpr.position),
                                 parent))
                     }
                     else -> throw ExpressionError("can only take bitwise inversion of int", subexpr.position)
@@ -252,35 +101,74 @@ internal class ConstantFoldingOptimizer(private val program: Program, private va
     override fun after(expr: BinaryExpression, parent: Node): Iterable<IAstModification> {
         val leftconst = expr.left.constValue(program)
         val rightconst = expr.right.constValue(program)
+        val modifications = mutableListOf<IAstModification>()
 
-        val subExpr: BinaryExpression? = when {
-            leftconst!=null -> expr.right as? BinaryExpression
-            rightconst!=null -> expr.left as? BinaryExpression
-            else -> null
+        if(expr.operator == "**" && leftconst!=null) {
+            // optimize various simple cases of ** :
+            //  optimize away 1 ** x into just 1 and 0 ** x into just 0
+            //  optimize 2 ** x into (1<<x)  if both operands are integer.
+            val leftDt = leftconst.inferType(program).typeOrElse(DataType.STRUCT)
+            when (leftconst.number.toDouble()) {
+                0.0 -> {
+                    val value = NumericLiteralValue(leftDt, 0, expr.position)
+                    modifications += IAstModification.ReplaceNode(expr, value, parent)
+                }
+                1.0 -> {
+                    val value = NumericLiteralValue(leftDt, 1, expr.position)
+                    modifications += IAstModification.ReplaceNode(expr, value, parent)
+                }
+                2.0 -> {
+                    if(rightconst!=null) {
+                        val value = NumericLiteralValue(leftDt, 2.0.pow(rightconst.number.toDouble()), expr.position)
+                        modifications += IAstModification.ReplaceNode(expr, value, parent)
+                    } else {
+                        val rightDt = expr.right.inferType(program).typeOrElse(DataType.STRUCT)
+                        if(leftDt in IntegerDatatypes && rightDt in IntegerDatatypes) {
+                            val targetDt =
+                                when (parent) {
+                                    is Assignment -> parent.target.inferType(program).typeOrElse(DataType.STRUCT)
+                                    is VarDecl -> parent.datatype
+                                    else -> leftDt
+                                }
+                            val one = NumericLiteralValue(targetDt, 1, expr.position)
+                            val shift = BinaryExpression(one, "<<", expr.right, expr.position)
+                            modifications += IAstModification.ReplaceNode(expr, shift, parent)
+                        }
+                    }
+                }
+            }
         }
-        if(subExpr!=null) {
-            val subleftconst = subExpr.left.constValue(program)
-            val subrightconst = subExpr.right.constValue(program)
-            if ((subleftconst != null && subrightconst == null) || (subleftconst==null && subrightconst!=null)) {
-                // try reordering.
-                val change = groupTwoConstsTogether(expr, subExpr,
+
+        if(expr.inferType(program).istype(DataType.FLOAT)) {
+            val subExpr: BinaryExpression? = when {
+                leftconst != null -> expr.right as? BinaryExpression
+                rightconst != null -> expr.left as? BinaryExpression
+                else -> null
+            }
+            if (subExpr != null) {
+                val subleftconst = subExpr.left.constValue(program)
+                val subrightconst = subExpr.right.constValue(program)
+                if ((subleftconst != null && subrightconst == null) || (subleftconst == null && subrightconst != null)) {
+                    // try reordering.
+                    val change = groupTwoFloatConstsTogether(
+                        expr, subExpr,
                         leftconst != null, rightconst != null,
-                        subleftconst != null, subrightconst != null)
-                return change?.let { listOf(it) } ?: noModifications
+                        subleftconst != null, subrightconst != null
+                    )
+                    if (change != null)
+                        modifications += change
+                }
             }
         }
 
         // const fold when both operands are a const
         if(leftconst != null && rightconst != null) {
             val evaluator = ConstExprEvaluator()
-            return listOf(IAstModification.ReplaceNode(
-                    expr,
-                    evaluator.evaluate(leftconst, expr.operator, rightconst),
-                    parent
-            ))
+            val result = evaluator.evaluate(leftconst, expr.operator, rightconst)
+            modifications += IAstModification.ReplaceNode(expr, result, parent)
         }
 
-        return noModifications
+        return modifications
     }
 
     override fun after(array: ArrayLiteralValue, parent: Node): Iterable<IAstModification> {
@@ -319,21 +207,24 @@ internal class ConstantFoldingOptimizer(private val program: Program, private va
     }
 
     override fun after(forLoop: ForLoop, parent: Node): Iterable<IAstModification> {
-        fun adjustRangeDt(rangeFrom: NumericLiteralValue, targetDt: DataType, rangeTo: NumericLiteralValue, stepLiteral: NumericLiteralValue?, range: RangeExpr): RangeExpr {
-            val newFrom: NumericLiteralValue
-            val newTo: NumericLiteralValue
-            try {
-                newFrom = rangeFrom.cast(targetDt)
-                newTo = rangeTo.cast(targetDt)
-            } catch (x: ExpressionError) {
-                return range
-            }
-            val newStep: Expression = try {
-                stepLiteral?.cast(targetDt)?: range.step
-            } catch(ee: ExpressionError) {
-                range.step
-            }
-            return RangeExpr(newFrom, newTo, newStep, range.position)
+        fun adjustRangeDt(rangeFrom: NumericLiteralValue, targetDt: DataType, rangeTo: NumericLiteralValue, stepLiteral: NumericLiteralValue?, range: RangeExpr): RangeExpr? {
+            val fromCast = rangeFrom.cast(targetDt)
+            val toCast = rangeTo.cast(targetDt)
+            if(!fromCast.isValid || !toCast.isValid)
+                return null
+
+            val newStep =
+                if(stepLiteral!=null) {
+                    val stepCast = stepLiteral.cast(targetDt)
+                    if(stepCast.isValid)
+                        stepCast.valueOrZero()
+                    else
+                        range.step
+                } else {
+                    range.step
+                }
+
+            return RangeExpr(fromCast.valueOrZero(), toCast.valueOrZero(), newStep, compTarget, range.position)
         }
 
         // adjust the datatype of a range expression in for loops to the loop variable.
@@ -342,45 +233,61 @@ internal class ConstantFoldingOptimizer(private val program: Program, private va
         val rangeTo = iterableRange.to as? NumericLiteralValue
         if(rangeFrom==null || rangeTo==null) return noModifications
 
-        val loopvar = forLoop.loopVar?.targetVarDecl(program.namespace)
-        if(loopvar!=null) {
-            val stepLiteral = iterableRange.step as? NumericLiteralValue
-            when(loopvar.datatype) {
-                DataType.UBYTE -> {
-                    if(rangeFrom.type!= DataType.UBYTE) {
-                        // attempt to translate the iterable into ubyte values
-                        val newIter = adjustRangeDt(rangeFrom, loopvar.datatype, rangeTo, stepLiteral, iterableRange)
+        val loopvar = forLoop.loopVar.targetVarDecl(program) ?: throw UndefinedSymbolError(forLoop.loopVar)
+
+        val stepLiteral = iterableRange.step as? NumericLiteralValue
+        when(loopvar.datatype) {
+            DataType.UBYTE -> {
+                if(rangeFrom.type!= DataType.UBYTE) {
+                    // attempt to translate the iterable into ubyte values
+                    val newIter = adjustRangeDt(rangeFrom, loopvar.datatype, rangeTo, stepLiteral, iterableRange)
+                    if(newIter!=null)
                         return listOf(IAstModification.ReplaceNode(forLoop.iterable, newIter, forLoop))
-                    }
                 }
-                DataType.BYTE -> {
-                    if(rangeFrom.type!= DataType.BYTE) {
-                        // attempt to translate the iterable into byte values
-                        val newIter = adjustRangeDt(rangeFrom, loopvar.datatype, rangeTo, stepLiteral, iterableRange)
-                        return listOf(IAstModification.ReplaceNode(forLoop.iterable, newIter, forLoop))
-                    }
-                }
-                DataType.UWORD -> {
-                    if(rangeFrom.type!= DataType.UWORD) {
-                        // attempt to translate the iterable into uword values
-                        val newIter = adjustRangeDt(rangeFrom, loopvar.datatype, rangeTo, stepLiteral, iterableRange)
-                        return listOf(IAstModification.ReplaceNode(forLoop.iterable, newIter, forLoop))
-                    }
-                }
-                DataType.WORD -> {
-                    if(rangeFrom.type!= DataType.WORD) {
-                        // attempt to translate the iterable into word values
-                        val newIter = adjustRangeDt(rangeFrom, loopvar.datatype, rangeTo, stepLiteral, iterableRange)
-                        return listOf(IAstModification.ReplaceNode(forLoop.iterable, newIter, forLoop))
-                    }
-                }
-                else -> throw FatalAstException("invalid loopvar datatype $loopvar")
             }
+            DataType.BYTE -> {
+                if(rangeFrom.type!= DataType.BYTE) {
+                    // attempt to translate the iterable into byte values
+                    val newIter = adjustRangeDt(rangeFrom, loopvar.datatype, rangeTo, stepLiteral, iterableRange)
+                    if(newIter!=null)
+                        return listOf(IAstModification.ReplaceNode(forLoop.iterable, newIter, forLoop))
+                }
+            }
+            DataType.UWORD -> {
+                if(rangeFrom.type!= DataType.UWORD) {
+                    // attempt to translate the iterable into uword values
+                    val newIter = adjustRangeDt(rangeFrom, loopvar.datatype, rangeTo, stepLiteral, iterableRange)
+                    if(newIter!=null)
+                        return listOf(IAstModification.ReplaceNode(forLoop.iterable, newIter, forLoop))
+                }
+            }
+            DataType.WORD -> {
+                if(rangeFrom.type!= DataType.WORD) {
+                    // attempt to translate the iterable into word values
+                    val newIter = adjustRangeDt(rangeFrom, loopvar.datatype, rangeTo, stepLiteral, iterableRange)
+                    if(newIter!=null)
+                        return listOf(IAstModification.ReplaceNode(forLoop.iterable, newIter, forLoop))
+                }
+            }
+            else -> throw FatalAstException("invalid loopvar datatype $loopvar")
         }
 
         return noModifications
     }
 
+    override fun after(decl: VarDecl, parent: Node): Iterable<IAstModification> {
+        val numval = decl.value as? NumericLiteralValue
+        if(decl.type== VarDeclType.CONST && numval!=null) {
+            val valueDt = numval.inferType(program)
+            if(!valueDt.istype(decl.datatype)) {
+                val cast = numval.cast(decl.datatype)
+                if(cast.isValid)
+                    return listOf(IAstModification.ReplaceNode(numval, cast.valueOrZero(), decl))
+            }
+        }
+        return noModifications
+    }
+
     private class ShuffleOperands(val expr: BinaryExpression,
                                   val exprOperator: String?,
                                   val subExpr: BinaryExpression,
@@ -398,18 +305,21 @@ internal class ConstantFoldingOptimizer(private val program: Program, private va
         }
     }
 
-    private fun groupTwoConstsTogether(expr: BinaryExpression,
-                                       subExpr: BinaryExpression,
-                                       leftIsConst: Boolean,
-                                       rightIsConst: Boolean,
-                                       subleftIsConst: Boolean,
-                                       subrightIsConst: Boolean): IAstModification?
+    private fun groupTwoFloatConstsTogether(expr: BinaryExpression,
+       subExpr: BinaryExpression,
+       leftIsConst: Boolean,
+       rightIsConst: Boolean,
+       subleftIsConst: Boolean,
+       subrightIsConst: Boolean): IAstModification?
     {
+        // NOTE: THIS IS ONLY VALID ON FLOATING POINT CONSTANTS
+
         // todo: this implements only a small set of possible reorderings at this time
         if(expr.operator==subExpr.operator) {
             // both operators are the same.
-            // If + or *,  we can simply shuffle the const operands around to optimize.
-            if(expr.operator=="+" || expr.operator=="*") {
+
+            // If associative,  we can simply shuffle the const operands around to optimize.
+            if(expr.operator in associativeOperators) {
                 return if(leftIsConst) {
                     if(subleftIsConst)
                         ShuffleOperands(expr, null, subExpr, subExpr.right, null, null, expr.left)

compiler/src/prog8/optimizer/ConstantIdentifierReplacer.kt

@@ -0,0 +1,214 @@
+package prog8.optimizer
+
+import prog8.ast.Node
+import prog8.ast.Program
+import prog8.ast.base.*
+import prog8.ast.expressions.*
+import prog8.ast.statements.ArrayIndex
+import prog8.ast.statements.AssignTarget
+import prog8.ast.statements.ForLoop
+import prog8.ast.statements.VarDecl
+import prog8.ast.walk.AstWalker
+import prog8.ast.walk.IAstModification
+import prog8.compiler.IErrorReporter
+import prog8.compiler.target.ICompilationTarget
+
+// Fix up the literal value's type to match that of the vardecl
+internal class VarConstantValueTypeAdjuster(private val program: Program, private val errors: IErrorReporter) : AstWalker() {
+    private val noModifications = emptyList<IAstModification>()
+
+    override fun after(decl: VarDecl, parent: Node): Iterable<IAstModification> {
+        try {
+            val declConstValue = decl.value?.constValue(program)
+            if(declConstValue!=null && (decl.type==VarDeclType.VAR || decl.type==VarDeclType.CONST)
+                && !declConstValue.inferType(program).istype(decl.datatype)) {
+                // cast the numeric literal to the appropriate datatype of the variable
+                val cast = declConstValue.cast(decl.datatype)
+                if(cast.isValid)
+                    return listOf(IAstModification.ReplaceNode(decl.value!!, cast.valueOrZero(), decl))
+            }
+        } catch (x: UndefinedSymbolError) {
+            errors.err(x.message, x.position)
+        }
+
+        return noModifications
+    }
+}
+
+
+// Replace all constant identifiers with their actual value,
+// and the array var initializer values and sizes.
+// This is needed because further constant optimizations depend on those.
+internal class ConstantIdentifierReplacer(private val program: Program, private val errors: IErrorReporter, private val compTarget: ICompilationTarget) : AstWalker() {
+    private val noModifications = emptyList<IAstModification>()
+
+    override fun after(identifier: IdentifierReference, parent: Node): Iterable<IAstModification> {
+        // replace identifiers that refer to const value, with the value itself
+        // if it's a simple type and if it's not a left hand side variable
+        if(identifier.parent is AssignTarget)
+            return noModifications
+        var forloop = identifier.parent as? ForLoop
+        if(forloop==null)
+            forloop = identifier.parent.parent as? ForLoop
+        if(forloop!=null && identifier===forloop.loopVar)
+            return noModifications
+
+        try {
+            val cval = identifier.constValue(program) ?: return noModifications
+            return when (cval.type) {
+                in NumericDatatypes -> listOf(
+                    IAstModification.ReplaceNode(
+                        identifier,
+                        NumericLiteralValue(cval.type, cval.number, identifier.position),
+                        identifier.parent
+                    )
+                )
+                in PassByReferenceDatatypes -> throw FatalAstException("pass-by-reference type should not be considered a constant")
+                else -> noModifications
+            }
+        } catch (x: UndefinedSymbolError) {
+            errors.err(x.message, x.position)
+            return noModifications
+        }
+    }
+
+    override fun before(decl: VarDecl, parent: Node): Iterable<IAstModification> {
+        // the initializer value can't refer to the variable itself (recursive definition)
+        // TODO: use call graph for this?
+        if(decl.value?.referencesIdentifier(decl.name) == true || decl.arraysize?.indexVar?.referencesIdentifier(decl.name) == true) {
+            errors.err("recursive var declaration", decl.position)
+            return noModifications
+        }
+
+        if(decl.type== VarDeclType.CONST || decl.type== VarDeclType.VAR) {
+            if(decl.isArray){
+                val arraysize = decl.arraysize
+                if(arraysize==null) {
+                    // for arrays that have no size specifier attempt to deduce the size
+                    val arrayval = decl.value as? ArrayLiteralValue
+                    if(arrayval!=null) {
+                        return listOf(IAstModification.SetExpression(
+                                { decl.arraysize = ArrayIndex(it, decl.position) },
+                                NumericLiteralValue.optimalInteger(arrayval.value.size, decl.position),
+                                decl
+                        ))
+                    }
+                } else if(arraysize.constIndex()==null) {
+                    // see if we can calculate the size from other fields
+                    try {
+                        val cval =  arraysize.indexVar?.constValue(program) ?: arraysize.origExpression?.constValue(program)
+                        if (cval != null) {
+                            arraysize.indexVar = null
+                            arraysize.origExpression = null
+                            arraysize.indexNum = cval
+                        }
+                    } catch (x: UndefinedSymbolError) {
+                        errors.err(x.message, x.position)
+                        return noModifications
+                    }
+                }
+            }
+
+            when(decl.datatype) {
+                DataType.FLOAT -> {
+                    // vardecl: for scalar float vars, promote constant integer initialization values to floats
+                    val litval = decl.value as? NumericLiteralValue
+                    if (litval!=null && litval.type in IntegerDatatypes) {
+                        val newValue = NumericLiteralValue(DataType.FLOAT, litval.number.toDouble(), litval.position)
+                        return listOf(IAstModification.ReplaceNode(decl.value!!, newValue, decl))
+                    }
+                }
+                DataType.ARRAY_UB, DataType.ARRAY_B, DataType.ARRAY_UW, DataType.ARRAY_W -> {
+                    val numericLv = decl.value as? NumericLiteralValue
+                    val rangeExpr = decl.value as? RangeExpr
+                    if(rangeExpr!=null) {
+                        // convert the initializer range expression to an actual array
+                        val declArraySize = decl.arraysize?.constIndex()
+                        if(declArraySize!=null && declArraySize!=rangeExpr.size())
+                            errors.err("range expression size doesn't match declared array size", decl.value?.position!!)
+                        val constRange = rangeExpr.toConstantIntegerRange()
+                        if(constRange!=null) {
+                            val eltType = rangeExpr.inferType(program).typeOrElse(DataType.UBYTE)
+                            val newValue = if(eltType in ByteDatatypes) {
+                                ArrayLiteralValue(InferredTypes.InferredType.known(decl.datatype),
+                                        constRange.map { NumericLiteralValue(eltType, it.toShort(), decl.value!!.position) }.toTypedArray(),
+                                        position = decl.value!!.position)
+                            } else {
+                                ArrayLiteralValue(InferredTypes.InferredType.known(decl.datatype),
+                                        constRange.map { NumericLiteralValue(eltType, it, decl.value!!.position) }.toTypedArray(),
+                                        position = decl.value!!.position)
+                            }
+                            return listOf(IAstModification.ReplaceNode(decl.value!!, newValue, decl))
+                        }
+                    }
+                    if(numericLv!=null && numericLv.type== DataType.FLOAT)
+                        errors.err("arraysize requires only integers here", numericLv.position)
+                    val size = decl.arraysize?.constIndex() ?: return noModifications
+                    if (rangeExpr==null && numericLv!=null) {
+                        // arraysize initializer is empty or a single int, and we know the size; create the arraysize.
+                        val fillvalue = numericLv.number.toInt()
+                        when(decl.datatype){
+                            DataType.ARRAY_UB -> {
+                                if(fillvalue !in 0..255)
+                                    errors.err("ubyte value overflow", numericLv.position)
+                            }
+                            DataType.ARRAY_B -> {
+                                if(fillvalue !in -128..127)
+                                    errors.err("byte value overflow", numericLv.position)
+                            }
+                            DataType.ARRAY_UW -> {
+                                if(fillvalue !in 0..65535)
+                                    errors.err("uword value overflow", numericLv.position)
+                            }
+                            DataType.ARRAY_W -> {
+                                if(fillvalue !in -32768..32767)
+                                    errors.err("word value overflow", numericLv.position)
+                            }
+                            else -> {}
+                        }
+                        // create the array itself, filled with the fillvalue.
+                        val array = Array(size) {fillvalue}.map { NumericLiteralValue(ArrayElementTypes.getValue(decl.datatype), it, numericLv.position) }.toTypedArray<Expression>()
+                        val refValue = ArrayLiteralValue(InferredTypes.InferredType.known(decl.datatype), array, position = numericLv.position)
+                        return listOf(IAstModification.ReplaceNode(decl.value!!, refValue, decl))
+                    }
+                }
+                DataType.ARRAY_F -> {
+                    val size = decl.arraysize?.constIndex() ?: return noModifications
+                    val litval = decl.value as? NumericLiteralValue
+                    val rangeExpr = decl.value as? RangeExpr
+                    if(rangeExpr!=null) {
+                        // convert the initializer range expression to an actual array of floats
+                        val declArraySize = decl.arraysize?.constIndex()
+                        if(declArraySize!=null && declArraySize!=rangeExpr.size())
+                            errors.err("range expression size doesn't match declared array size", decl.value?.position!!)
+                        val constRange = rangeExpr.toConstantIntegerRange()
+                        if(constRange!=null) {
+                            val newValue = ArrayLiteralValue(InferredTypes.InferredType.known(DataType.ARRAY_F),
+                                    constRange.map { NumericLiteralValue(DataType.FLOAT, it.toDouble(), decl.value!!.position) }.toTypedArray(),
+                                    position = decl.value!!.position)
+                            return listOf(IAstModification.ReplaceNode(decl.value!!, newValue, decl))
+                        }
+                    }
+                    if(rangeExpr==null && litval!=null) {
+                        // arraysize initializer is a single int, and we know the size.
+                        val fillvalue = litval.number.toDouble()
+                        if (fillvalue < compTarget.machine.FLOAT_MAX_NEGATIVE || fillvalue > compTarget.machine.FLOAT_MAX_POSITIVE)
+                            errors.err("float value overflow", litval.position)
+                        else {
+                            // create the array itself, filled with the fillvalue.
+                            val array = Array(size) {fillvalue}.map { NumericLiteralValue(DataType.FLOAT, it, litval.position) }.toTypedArray<Expression>()
+                            val refValue = ArrayLiteralValue(InferredTypes.InferredType.known(DataType.ARRAY_F), array, position = litval.position)
+                            return listOf(IAstModification.ReplaceNode(decl.value!!, refValue, decl))
+                        }
+                    }
+                }
+                else -> {
+                    // nothing to do for this type
+                    // this includes strings and structs
+                }
+            }
+        }
+
+        return noModifications
+    }
+}

compiler/src/prog8/optimizer/ExpressionSimplifier.kt

@@ -2,11 +2,14 @@ package prog8.optimizer
 
 import prog8.ast.Node
 import prog8.ast.Program
-import prog8.ast.base.*
+import prog8.ast.base.DataType
+import prog8.ast.base.FatalAstException
+import prog8.ast.base.IntegerDatatypes
+import prog8.ast.base.NumericDatatypes
 import prog8.ast.expressions.*
-import prog8.ast.processing.AstWalker
-import prog8.ast.processing.IAstModification
 import prog8.ast.statements.Assignment
+import prog8.ast.walk.AstWalker
+import prog8.ast.walk.IAstModification
 import kotlin.math.abs
 import kotlin.math.log2
 import kotlin.math.pow
@@ -24,12 +27,6 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
     private val negativePowersOfTwo = powersOfTwo.map { -it }.toSet()
     private val noModifications = emptyList<IAstModification>()
 
-    override fun after(assignment: Assignment, parent: Node): Iterable<IAstModification> {
-        if (assignment.aug_op != null)
-            throw FatalAstException("augmented assignments should have been converted to normal assignments before this optimizer: $assignment")
-        return noModifications
-    }
-
     override fun after(typecast: TypecastExpression, parent: Node): Iterable<IAstModification> {
         val mods = mutableListOf<IAstModification>()
 
@@ -37,19 +34,22 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
         val literal = typecast.expression as? NumericLiteralValue
         if (literal != null) {
             val newLiteral = literal.cast(typecast.type)
-            if (newLiteral !== literal)
-                mods += IAstModification.ReplaceNode(typecast.expression, newLiteral, typecast)
+            if (newLiteral.isValid && newLiteral.valueOrZero() !== literal)
+                mods += IAstModification.ReplaceNode(typecast.expression, newLiteral.valueOrZero(), typecast)
         }
 
-        // remove redundant nested typecasts:
-        // if the typecast casts a value to the same type, remove the cast.
-        // if the typecast contains another typecast, remove the inner typecast.
+        // remove redundant nested typecasts
         val subTypecast = typecast.expression as? TypecastExpression
         if (subTypecast != null) {
-            mods += IAstModification.ReplaceNode(typecast.expression, subTypecast.expression, typecast)
+            // remove the sub-typecast if its datatype is larger than the outer typecast
+            if(subTypecast.type largerThan typecast.type) {
+                mods += IAstModification.ReplaceNode(typecast.expression, subTypecast.expression, typecast)
+            }
         } else {
-            if (typecast.expression.inferType(program).istype(typecast.type))
+            if (typecast.expression.inferType(program).istype(typecast.type)) {
+                // remove duplicate cast
                 mods += IAstModification.ReplaceNode(typecast, typecast.expression, parent)
+            }
         }
 
         return mods
@@ -102,6 +102,12 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
         if (leftVal != null && expr.operator in associativeOperators && rightVal == null)
             return listOf(IAstModification.SwapOperands(expr))
 
+        // NonBinaryExpression  <associativeoperator>  BinaryExpression  -->  BinaryExpression  <associativeoperator>  NonBinaryExpression
+        if (expr.operator in associativeOperators && expr.left !is BinaryExpression && expr.right is BinaryExpression) {
+            if(parent !is Assignment || !(expr.left isSameAs parent.target))
+                return listOf(IAstModification.SwapOperands(expr))
+        }
+
         // X + (-A)  -->  X - A
         if (expr.operator == "+" && (expr.right as? PrefixExpression)?.operator == "-") {
             return listOf(IAstModification.ReplaceNode(
@@ -179,28 +185,6 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
                 // unsigned >= 0 --> true
                 return listOf(IAstModification.ReplaceNode(expr, NumericLiteralValue.fromBoolean(true, expr.position), parent))
             }
-            when(leftDt) {
-                DataType.BYTE -> {
-                    // signed >=0   --> signed ^ $80
-                    return listOf(IAstModification.ReplaceNode(
-                            expr,
-                            BinaryExpression(expr.left, "^", NumericLiteralValue.optimalInteger(0x80, expr.position), expr.position),
-                            parent
-                    ))
-                }
-                DataType.WORD -> {
-                    // signedw >=0   --> msb(signedw) ^ $80
-                    return listOf(IAstModification.ReplaceNode(
-                            expr,
-                            BinaryExpression(FunctionCall(IdentifierReference(listOf("msb"), expr.position),
-                                    mutableListOf(expr.left),
-                                    expr.position
-                            ), "^", NumericLiteralValue.optimalInteger(0x80, expr.position), expr.position),
-                            parent
-                    ))
-                }
-                else -> {}
-            }
         }
 
         if(expr.operator == "<" && rightVal?.number == 0) {
@@ -301,6 +285,49 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
         return noModifications
     }
 
+    override fun after(functionCall: FunctionCall, parent: Node): Iterable<IAstModification> {
+        if(functionCall.target.nameInSource == listOf("lsb")) {
+            val arg = functionCall.args[0]
+            if(arg is TypecastExpression) {
+                val valueDt = arg.expression.inferType(program)
+                if (valueDt.istype(DataType.BYTE) || valueDt.istype(DataType.UBYTE)) {
+                    // useless lsb() of byte value that was casted to word
+                    return listOf(IAstModification.ReplaceNode(functionCall, arg.expression, parent))
+                }
+            } else {
+                val argDt = arg.inferType(program)
+                if (argDt.istype(DataType.BYTE) || argDt.istype(DataType.UBYTE)) {
+                    // useless lsb() of byte value
+                    return listOf(IAstModification.ReplaceNode(functionCall, arg, parent))
+                }
+            }
+        }
+        else if(functionCall.target.nameInSource == listOf("msb")) {
+            val arg = functionCall.args[0]
+            if(arg is TypecastExpression) {
+                val valueDt = arg.expression.inferType(program)
+                if (valueDt.istype(DataType.BYTE) || valueDt.istype(DataType.UBYTE)) {
+                    // useless msb() of byte value that was casted to word, replace with 0
+                    return listOf(IAstModification.ReplaceNode(
+                            functionCall,
+                            NumericLiteralValue(valueDt.typeOrElse(DataType.UBYTE), 0, arg.expression.position),
+                            parent))
+                }
+            } else {
+                val argDt = arg.inferType(program)
+                if (argDt.istype(DataType.BYTE) || argDt.istype(DataType.UBYTE)) {
+                    // useless msb() of byte value, replace with 0
+                    return listOf(IAstModification.ReplaceNode(
+                            functionCall,
+                            NumericLiteralValue(argDt.typeOrElse(DataType.UBYTE), 0, arg.position),
+                            parent))
+                }
+            }
+        }
+
+        return noModifications
+    }
+
     private fun determineY(x: Expression, subBinExpr: BinaryExpression): Expression? {
         return when {
             subBinExpr.left isSameAs x -> subBinExpr.right
@@ -321,7 +348,7 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
         if (leftVal == null && rightVal == null)
             return null
 
-        val (expr2, _, rightVal2) = reorderAssociative(expr, leftVal)
+        val (expr2, _, rightVal2) = reorderAssociativeWithConstant(expr, leftVal)
         if (rightVal2 != null) {
             // right value is a constant, see if we can optimize
             val rightConst: NumericLiteralValue = rightVal2
@@ -334,6 +361,13 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
         }
         // no need to check for left val constant (because of associativity)
 
+        val rnum = rightVal?.number?.toDouble()
+        if(rnum!=null && rnum<0.0) {
+            expr.operator = "-"
+            expr.right = NumericLiteralValue(rightVal.type, -rnum, rightVal.position)
+            return expr
+        }
+
         return null
     }
 
@@ -348,12 +382,16 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
 
         if (rightVal != null) {
             // right value is a constant, see if we can optimize
-            val rightConst: NumericLiteralValue = rightVal
-            when (rightConst.number.toDouble()) {
-                0.0 -> {
-                    // left
-                    return expr.left
-                }
+            val rnum = rightVal.number.toDouble()
+            if (rnum == 0.0) {
+                // left
+                return expr.left
+            }
+
+            if(rnum<0.0) {
+                expr.operator = "+"
+                expr.right = NumericLiteralValue(rightVal.type, -rnum, rightVal.position)
+                return expr
             }
         }
         if (leftVal != null) {
@@ -366,6 +404,7 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
             }
         }
 
+
         return null
     }
 
@@ -448,7 +487,10 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
         when (expr.operator) {
             "%" -> {
                 if (cv == 1.0) {
-                    return NumericLiteralValue(expr.inferType(program).typeOrElse(DataType.STRUCT), 0, expr.position)
+                    val idt = expr.inferType(program)
+                    if(!idt.isKnown)
+                        throw FatalAstException("unknown dt")
+                    return NumericLiteralValue(idt.typeOrElse(DataType.STRUCT), 0, expr.position)
                 } else if (cv == 2.0) {
                     expr.operator = "&"
                     expr.right = NumericLiteralValue.optimalInteger(1, expr.position)
@@ -530,7 +572,7 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
         if (leftVal == null && rightVal == null)
             return null
 
-        val (expr2, _, rightVal2) = reorderAssociative(expr, leftVal)
+        val (expr2, _, rightVal2) = reorderAssociativeWithConstant(expr, leftVal)
         if (rightVal2 != null) {
             // right value is a constant, see if we can optimize
             val leftValue: Expression = expr2.left
@@ -577,8 +619,10 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
         if (amount == 0) {
             return expr.left
         }
-        val targetDt = expr.left.inferType(program).typeOrElse(DataType.STRUCT)
-        when (targetDt) {
+        val targetIDt = expr.left.inferType(program)
+        if(!targetIDt.isKnown)
+            throw FatalAstException("unknown dt")
+        when (val targetDt = targetIDt.typeOrElse(DataType.STRUCT)) {
             DataType.UBYTE, DataType.BYTE -> {
                 if (amount >= 8) {
                     return NumericLiteralValue(targetDt, 0, expr.position)
@@ -588,12 +632,12 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
                 if (amount >= 16) {
                     return NumericLiteralValue(targetDt, 0, expr.position)
                 } else if (amount >= 8) {
-                    val lsb = TypecastExpression(expr.left, DataType.UBYTE, true, expr.position)
+                    val lsb = FunctionCall(IdentifierReference(listOf("lsb"), expr.position), mutableListOf(expr.left), expr.position)
                     if (amount == 8) {
-                        return FunctionCall(IdentifierReference(listOf("mkword"), expr.position), mutableListOf(NumericLiteralValue.optimalInteger(0, expr.position), lsb), expr.position)
+                        return FunctionCall(IdentifierReference(listOf("mkword"), expr.position), mutableListOf(lsb, NumericLiteralValue.optimalInteger(0, expr.position)), expr.position)
                     }
                     val shifted = BinaryExpression(lsb, "<<", NumericLiteralValue.optimalInteger(amount - 8, expr.position), expr.position)
-                    return FunctionCall(IdentifierReference(listOf("mkword"), expr.position), mutableListOf(NumericLiteralValue.optimalInteger(0, expr.position), shifted), expr.position)
+                    return FunctionCall(IdentifierReference(listOf("mkword"), expr.position), mutableListOf(shifted, NumericLiteralValue.optimalInteger(0, expr.position)), expr.position)
                 }
             }
             else -> {
@@ -610,8 +654,10 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
         if (amount == 0) {
             return expr.left
         }
-        val targetDt = expr.left.inferType(program).typeOrElse(DataType.STRUCT)
-        when (targetDt) {
+        val idt = expr.left.inferType(program)
+        if(!idt.isKnown)
+            throw FatalAstException("unknown dt")
+        when (idt.typeOrElse(DataType.STRUCT)) {
             DataType.UBYTE -> {
                 if (amount >= 8) {
                     return NumericLiteralValue.optimalInteger(0, expr.position)
@@ -626,25 +672,21 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
             DataType.UWORD -> {
                 if (amount >= 16) {
                     return NumericLiteralValue.optimalInteger(0, expr.position)
-                } else if (amount >= 8) {
+                }
+                else if (amount >= 8) {
                     val msb = FunctionCall(IdentifierReference(listOf("msb"), expr.position), mutableListOf(expr.left), expr.position)
-                    if (amount == 8)
-                        return msb
-                    return BinaryExpression(msb, ">>", NumericLiteralValue.optimalInteger(amount - 8, expr.position), expr.position)
+                    if (amount == 8) {
+                        // mkword(0, msb(v))
+                        val zero = NumericLiteralValue(DataType.UBYTE, 0, expr.position)
+                        return FunctionCall(IdentifierReference(listOf("mkword"), expr.position), mutableListOf(zero, msb), expr.position)
+                    }
+                    return TypecastExpression(BinaryExpression(msb, ">>", NumericLiteralValue.optimalInteger(amount - 8, expr.position), expr.position), DataType.UWORD, true, expr.position)
                 }
             }
             DataType.WORD -> {
                 if (amount > 16) {
                     expr.right = NumericLiteralValue.optimalInteger(16, expr.right.position)
                     return null
-                } else if (amount >= 8) {
-                    val msbAsByte = TypecastExpression(
-                            FunctionCall(IdentifierReference(listOf("msb"), expr.position), mutableListOf(expr.left), expr.position),
-                            DataType.BYTE,
-                            true, expr.position)
-                    if (amount == 8)
-                        return msbAsByte
-                    return BinaryExpression(msbAsByte, ">>", NumericLiteralValue.optimalInteger(amount - 8, expr.position), expr.position)
                 }
             }
             else -> {
@@ -653,17 +695,17 @@ internal class ExpressionSimplifier(private val program: Program) : AstWalker()
         return null
     }
 
-    private fun reorderAssociative(expr: BinaryExpression, leftVal: NumericLiteralValue?): ReorderedAssociativeBinaryExpr {
+    private fun reorderAssociativeWithConstant(expr: BinaryExpression, leftVal: NumericLiteralValue?): BinExprWithConstants {
         if (expr.operator in associativeOperators && leftVal != null) {
             // swap left and right so that right is always the constant
             val tmp = expr.left
             expr.left = expr.right
             expr.right = tmp
-            return ReorderedAssociativeBinaryExpr(expr, expr.right.constValue(program), leftVal)
+            return BinExprWithConstants(expr, expr.right.constValue(program), leftVal)
         }
-        return ReorderedAssociativeBinaryExpr(expr, leftVal, expr.right.constValue(program))
+        return BinExprWithConstants(expr, leftVal, expr.right.constValue(program))
     }
 
-    private data class ReorderedAssociativeBinaryExpr(val expr: BinaryExpression, val leftVal: NumericLiteralValue?, val rightVal: NumericLiteralValue?)
+    private data class BinExprWithConstants(val expr: BinaryExpression, val leftVal: NumericLiteralValue?, val rightVal: NumericLiteralValue?)
 
 }