1
0
mirror of https://github.com/catseye/SixtyPical.git synced 2024-11-04 07:06:23 +00:00
A 6502-oriented low-level programming language supporting advanced static analysis
Go to file
2018-03-27 17:27:09 +01:00
bin Beginnings of modularity. 2018-03-27 16:23:22 +01:00
doc Add symbolic constants. 2018-03-27 12:36:33 +01:00
eg Describe the behaviour just implemented. 2018-03-27 16:49:20 +01:00
images Add screenshot. 2018-03-08 13:36:30 +00:00
src/sixtypical Track the gotos that we have encountered in a routine. 2018-03-27 17:27:09 +01:00
tests Add symbolic constants. 2018-03-27 12:36:33 +01:00
.gitignore Check in test appliance (dcc6502-adapter) that was missed. 2017-12-11 13:46:06 +00:00
HISTORY.md Describe the behaviour just implemented. 2018-03-27 16:49:20 +01:00
LICENSE Update copyright dates. 2018-03-06 15:52:55 +00:00
loadngo.sh Make loadngo.sh able to handle both C64 (x64) and VIC-20 (xvic). 2018-03-07 13:27:57 +00:00
README.md Try to spell out my idea for an algorithm for this. 2018-03-27 17:05:41 +01:00
test.sh The evaluator doesn't add much and keeps falling behind; remove it. 2018-02-05 13:17:23 +00:00

SixtyPical

Version 0.15. Work-in-progress, everything is subject to change.

SixtyPical is a 6502-like programming language with advanced static analysis.

"6502-like" means that it has similar restrictions as programming in 6502 assembly (e.g. the programmer must choose the registers that values will be stored in) and is concomitantly easy for a compiler to translate it to 6502 machine language code.

"Advanced static analysis" includes abstract interpretation, where we go through the program step by step, tracking not just the changes that happen during a specific execution of the program, but sets of changes that could possibly happen in any run of the program. This lets us determine that certain things can never happen, which we can then formulate as safety checks.

In practice, this means it catches things like

  • you forgot to clear carry before adding something to the accumulator
  • a subroutine that you call trashes a register you thought was preserved
  • you tried to read or write a byte beyond the end of a byte array
  • you tried to write the address of something that was not a routine, to a jump vector

and suchlike. It also provides some convenient operations based on machine-language programming idioms, such as

  • copying values from one register to another (via a third register when there are no underlying instructions that directly support it); this includes 16-bit values, which are copied in two steps
  • explicit tail calls
  • indirect subroutine calls

The reference implementation can analyze and compile SixtyPical programs to 6502 machine code.

Quick Start

If you have the VICE emulator installed, from this directory, you can run

./loadngo.sh c64 eg/c64/hearts.60p

and it will compile the hearts.60p source code and automatically start it in the x64 emulator, and you should see:

Screenshot of result of running hearts.60p

You can try the loadngo.sh script on other sources in the eg directory tree, which contains more extensive examples, including an entire game(-like program); see eg/README.md for a listing.

Documentation

TODO

Save registers on stack

This preserves them, so that, semantically, they can be used later even though they are trashed inside the block.

Re-order routines and optimize tail-calls to fallthroughs

Not because it saves 3 bytes, but because it's a neat trick. Doing it optimally is probably NP-complete. But doing it adequately is probably not that hard.

Every routine is falled through to by zero or more routines. Don't consider the main routine. For each routine α that is finally-falled through to by a set of routines R(α), pick a movable routine β from R, move β in front of α, remove the jmp at the end of β and mark β as unmovable. Note this only works if β finally-falls through. If there are multiple tail positions, we can't eliminate all the jmps. Note that if β finally-falls through to α it can't finally-fall through to anything else, so the sets R(α) should be disjoint for every α. (Right?)

And at some point...

  • low and high address operators - to turn word type into byte.
  • Tests, and implementation, ensuring a routine can be assigned to a vector of "wider" type
  • Related: can we simply view a (small) part of a buffer as a byte table? If not, why not?
  • Related: add constant to buffer to get new buffer. (Or to table, but... well, maybe.)
  • Check that the buffer being read or written to through pointer, appears in appropriate inputs or outputs set. (Associate each pointer with the buffer it points into.)
  • static pointers -- currently not possible because pointers must be zero-page, thus @, thus uninitialized.
  • Question the value of the "consistent initialization" principle for if statement analysis.
  • interrupt routines -- to indicate that "the supervisor" has stored values on the stack, so we can trash them.
  • Add absolute addressing in shl/shr, absolute-indexed for add, sub, etc.
  • Automatic tail-call optimization (could be tricky, w/constraints?)
  • Possibly ld x, [ptr] + y, possibly st x, [ptr] + y.
  • Maybe even copy [ptra] + y, [ptrb] + y, which can be compiled to indirect LDA then indirect STA!
  • Optimize ld a, z and st a, z to zero-page operations if address of z < 256.