prog8/docs/source/todo.rst

TODO
====

For next minor release
^^^^^^^^^^^^^^^^^^^^^^
- try to optimize newexpr a bit more

...


For 9.0 major changes
^^^^^^^^^^^^^^^^^^^^^
- rename builtin function sqrt16 to just sqrt
- copy (not move) the CBM kernal romsubs to a new 'cbm' block so programs on c128 and cx16 can also
  simply refer to cbm.CHROUT rather than c64.CHROUT which looks a bit silly on the non-c64 cbm systems.
  we keep the old definitions intact because of backwards compatibility reasons.
- try to reintroduce builtin functions max/maxw/min/minw that take 2 args and return the largest/smallest of them.
  This is a major change because it will likely break existing code that is now using min and max as variable names.
  Also add optimization that changes the word variant to byte variant if the operands are bytes.
- 6502 codegen: see if we can let for loops skip the loop if startvar>endvar, without adding a lot of code size/duplicating the loop condition.
  It is documented behavior to now loop 'around' $00 but it's too easy to forget about!
  Lot of work because of so many special cases in ForLoopsAsmgen.....
  (vm codegen already behaves like this!)
- ?? get rid of the disknumber parameter everywhere in diskio, make it a configurable variable that defaults to 8.
  the large majority of users will only deal with a single disk drive so why not make it easier for them.
  But see complaint on github https://github.com/irmen/prog8/issues/106
- duplicate diskio for cx16 (get rid of cx16diskio, just copy diskio and tweak everything) + documentation
- get f_seek_w working like in the BASIC program  - this needs the changes to diskio.f_open to use suffixes ,p,m
- add special (u)word array type (or modifier such as @fast? ) that puts the array into memory as 2 separate byte-arrays 1 for LSB 1 for MSB -> allows for word arrays of length 256 and faster indexing
  this is an enormous amout of work, if this type is to be treated equally as existing (u)word , because all expression / lookup / assignment routines need to know about the distinction....
  So maybe only allow the bare essentials? (store, get, bitwise operations?)
- Some more support for (64tass) SEGMENTS ?
    - (What, how, isn't current BSS support enough?)
    - Add a mechanism to allocate variables into golden ram (or segments really) (see GoldenRam class)
    - maybe treat block "golden" in a special way: can only contain vars, every var will be allocated in the Golden ram area?
    - maybe or may not needed: the variables can NOT have initialization values, they will all be set to zero on startup (simple memset)
      just initialize them yourself in start() if you need a non-zero value .
    - OR.... do all this automatically if 'golden' is enabled as a compiler option? So compiler allocates in ZP first, then Golden Ram, then regular ram
    - OR.... make all this more generic and use some %segment option to create real segments for 64tass?
    - (need separate step in codegen and IR to write the "golden" variables)


Need help with
^^^^^^^^^^^^^^
- atari target: more details details about the machine, fixing library routines. I have no clue whatsoever.
- see the :ref:`portingguide` for details on what information is needed.


Future Things and Ideas
^^^^^^^^^^^^^^^^^^^^^^^
Compiler:

- ir: idea: (but LLVM IR simply keeps the variables, so not a good idea then?...): replace all scalar variables by an allocated register. Keep a table of the variable to register mapping (including the datatype)
  global initialization values are simply a list of LOAD instructions.
  Variables replaced include all subroutine parameters!  So the only variables that remain as variables are arrays and strings.
- ir: mechanism to determine for chunks which registers are getting input values from "outside"
- ir: mechanism to determine for chunks which registers are passing values out? (i.e. are used again in another chunk)
- ir: peephole opt: (maybe just integrate this in the variable/register allocator though?) renumber registers in chunks to start with 1 again every time (but keep entry values in mind!)
- ir: peephole opt: (maybe just integrate this in the variable/register allocator though?) reuse registers in chunks (but keep result registers in mind that pass values out! and don't renumber registers above SyscallRegisterBase!)
- ir: add more optimizations in IRPeepholeOptimizer
- ir: for expressions with array indexes that occur multiple times, can we avoid loading them into new virtualregs everytime and just reuse a single virtualreg as indexer? (simple form of common subexpression elimination)
- PtAst/IR: more complex common subexpression eliminations
- generate WASM to eventually run prog8 on a browser canvas? Use binaryen toolkit or my binaryen kotlin library?
- can we get rid of pieces of asmgen.AssignmentAsmGen by just reusing the AugmentableAssignment ? generated code should not suffer
- [problematic due to using 64tass:] better support for building library programs, where unused .proc shouldn't be deleted from the assembly?
  Perhaps replace all uses of .proc/.pend/.endproc by .block/.bend will fix that with a compiler flag?
  But all library code written in asm uses .proc already..... (textual search/replace when writing the actual asm?)
  Once new codegen is written that is based on the IR, this point is mostly moot anyway as that will have its own dead code removal on the IR level.
- Zig-like try-based error handling where the V flag could indicate error condition? and/or BRK to jump into monitor on failure? (has to set BRK vector for that) But the V flag is also set on certain normal instructions
- For c128 target; put floating point variables in bank 1 to make the FP routines work (is this even worth it? very few people will use fp)

Libraries:

- fix the problems in atari target, and flesh out its libraries.
- c128 target: make syslib more complete (missing kernal routines)?
- c64: make the graphics.BITMAP_ADDRESS configurable (VIC banking)
- optimize several inner loops in gfx2 even further?
- add modes 3 and perhaps even 2 to gfx2 (lores 16 color and 4 color)?
- add a flood fill (span fill/scanline fill) routine to gfx2?


Expressions:

- Once the evalstack-free expression codegen is in place, the Eval Stack can be removed from the compiler.
    Machinedefinition, .p8 and .asm library files, all routines operationg on estack, and everything saving/restoring the X register related to this stack.
- Or rewrite expression tree evaluation such that it doesn't use an eval stack but flatten the tree into linear code
  that, for instance, uses a fixed number of predetermined value 'variables'?
  The VM IL solves this already (by using unlimited registers) but that still lacks a translation to 6502.
- this removes the need for the BinExprSplitter? (which is problematic and very limited now)
  and perhaps the assignment splitting in  BeforeAsmAstChanger  too

Optimizations:

- VariableAllocator: can we think of a smarter strategy for allocating variables into zeropage, rather than first-come-first-served?
  for instance, vars used inside loops first, then loopvars, then the rest
- various optimizers skip stuff if compTarget.name==VMTarget.NAME.  Once 6502-codegen is done from IR code,
  those checks should probably be removed, or be made permanent


STRUCTS again?
--------------

What if we were to re-introduce Structs in prog8? Some thoughts:

- can contain only numeric types (byte,word,float) - no nested structs, no reference types (strings, arrays) inside structs
- is just some syntactic sugar for a scoped set of variables -> ast transform to do exactly this before codegen. Codegen doesn't know about struct.
- no arrays of struct -- because too slow on 6502 to access those, rather use struct of arrays instead.
  can we make this a compiler/codegen only issue? i.e. syntax is just as if it was an array of structs?
  or make it explicit in the syntax so that it is clear what the memory layout of it is.
- ability to assign struct variable to another?   this is slow but can be quite handy sometimes.
  however how to handle this in a function that gets the struct passed as reference? Don't allow it there? (there's no pointer dereferencing concept in prog8)
- ability to be passed as argument to a function (by reference)?
  however there is no typed pointer in prog8 at the moment so this can't be implemented in a meaningful way yet,
  because there is no way to reference it as the struct type again. (current ast gets the by-reference parameter
  type replaced by uword)
  So-- maybe don't replace the parameter type in the ast?  Should fix that for str and array types as well then