Nathan Braswell
6602ff3151
The refactor has either caught up, or is close, and is much faster.
...
Need to make partial_eval_helper able to fail though, so that it can fail in the now fallible compiler.
Squashed commit of the following:
commit 283150d1e19cf9f74bc32e9c554f5e041d53582d
Author: Nathan Braswell <miloignis@gmail.com >
Date: Tue Jan 18 00:14:51 2022 -0500
Another bugfix, I think now we're truely running into the partial-eval needs to be able to fail part
commit c62c228a1e89e7922850d2070bc046e5f80af5a5
Author: Nathan Braswell <miloignis@gmail.com >
Date: Mon Jan 17 23:34:19 2022 -0500
Fixed needed check, added note on needing to support failing partial eval
commit d480e486210dd5c7c4842e3a3d3e447dc7a5274e
Author: Nathan Braswell <miloignis@gmail.com >
Date: Mon Jan 17 22:14:58 2022 -0500
Fixed compilation bugs, added debugging prints to running.
commit fc4dc4d3170bf1ceb4cd934cff54dd9ce6c8713a
Author: Nathan Braswell <miloignis@gmail.com >
Date: Mon Jan 17 17:28:29 2022 -0500
The unique id / ctx refactor is starting to work, with all basic test cases compiling & partial evaling, but mis-compile bugs remaining
2022-01-18 00:36:42 -05:00
Nathan Braswell
49c5e9da2f
Added the lazy eval function thing.
...
Keeping it off is 5.3s
Turning it on is 6.8s
Forcing it on always, which I thought would cause compilation problems (and I guess just turns most optimization off, but does enough to remove vaus?): 2.8s
so using it the way I designed makes it worse
and using it to a degree I thought it would be buggy makes is way faster
Indeed, forcing it on all the time does cause it to error for larger cases, but it is interesting
So I'm not entirely sure where to go with that, I think smaller test cases and profiling number of calls or something is warrented
2022-01-10 01:26:56 -05:00
Nathan Braswell
74a7cb8a20
Don't repeatedly evaluate params in different cases - I thought this would be a larger optimization, but it seemed more like 1%...
2022-01-09 00:45:16 -05:00
Nathan Braswell
025b149c28
Implemented each marked node carrying the de Bruijn indicies that it needs to continue evauluating, but now it takes the exact same amount of time that it used to
...
And I think I've realized why - it's being too conservative about what it actually needs to be a value and includes the entire environment chain, which pretty much means anytime it would have been re-evaluated because a parent function was called or re-evaluated it will also be re-evaluated, and none of this changes anything
I think I can change it to more intelligently pull what's necessary based on what's used in the body of the function instead and get the optimization to work as I expected - fingers crossed
2022-01-09 00:17:57 -05:00
Nathan Braswell
6ef60c4cc6
I belive I fixed the bug revealed by the let definition - it was a similar thing where <comb>s were counted as values even if there were fake envs inside, and then thouse fake envs could be moved inside another env_stack and then later resolve to a wrong env, or at lest I think that's what was happening. The let test takes too long to run now - I killed it at 20 minutes and 48GB of RAM taken. Given that, it's now time to move on towards optimization
2022-01-07 22:08:29 -05:00
Nathan Braswell
b559bfdf90
Add note about compiling params meant for Vau & additional partial eval for that case to help in cases where it's legitimate. Eventually it should handle errors gracefully, but non-gracefully can be good enough for now.
2022-01-05 22:58:26 -05:00
Nathan Braswell
1aa9ca972a
Fix & and error checking for compiling environments as code & value, add a todo for things deferred
2022-01-05 01:05:16 -05:00
Nathan Braswell
d1fc4e5d66
Fix another bug where things like (<comb0> <somecomb fake env>) wasn't reevaluating somecomb and fixing up the fake env on calls because comb0 has wrap level 0. In fact, it should still be partially evaluated again, just not stripped. This only comes up in slightly nontrivial examples because there's a good bit of nesting
2022-01-04 23:37:57 -05:00
Nathan Braswell
48cab3cbdf
Found the partial eval bug - a comb with a non-real environment doesn't count as later? which means that array values with them get skipped and not partial evaled further, which means they don't get further refined. I need to think through exactly which definition of later? is correct, or if it should be two concepts, because one of the two uses or the definition must change. For now just commented out the array value return and has it always re-traverse
2022-01-04 01:19:41 -05:00
Nathan Braswell
6816742fd6
Parse comments
2022-01-04 00:14:26 -05:00
Nathan Braswell
d2d4b15afa
Found it - print was dropping it's parameter even though the comment above it *clearly* said that it didn't
2022-01-03 22:22:34 -05:00
Nathan Braswell
321157c311
Added symbol parsing and ' syntax sugar parsing, but found a ref counting problem (unrelated to parsing, though that did expose it). Tracking down the bug now
2022-01-03 22:03:32 -05:00
Nathan Braswell
98bdf32ad8
Not a lot of error checking for mixed or confusing bases, but did add negative, hex, and binary integers
2022-01-03 19:03:36 -05:00
Nathan Braswell
cf72232282
Implement string escapes
2022-01-03 16:15:30 -05:00
Nathan Braswell
6f73be4777
Add ['open parent_fd path <comb(new_fd error_code)>] monad
2022-01-03 14:02:07 -05:00
Nathan Braswell
e633a43f2e
Parsing strings, but naively without escapes yet
2022-01-03 02:17:52 -05:00
Nathan Braswell
26e9a92f74
Can parse arrays!
2022-01-03 01:44:14 -05:00
Nathan Braswell
f8d2e17c90
Started working on parser, can now parse true and false. Also, I realized that most of the array functions have to be adapted to work with strings too (idx, concat, slice - maybe len already works?)
2022-01-02 23:37:45 -05:00
Nathan Braswell
a3ee9291c6
Implement comparison operators = != < <= >= >
2022-01-02 00:16:12 -05:00
Nathan Braswell
6831e76bf3
Add bxor and bnot (rename band bor), remove and and or, add wrap, unwrap, get-text, str-to-symbol
2022-01-01 16:42:57 -05:00
Nathan Braswell
2b08daccd1
Implemented the missing array methods (and some fixes) concat, len, idx, slice
2022-01-01 15:14:47 -05:00
Nathan Braswell
c530405376
Added << and >>
2022-01-01 12:38:49 -05:00
Nathan Braswell
08e3292d92
Added all of the predicate functions
2021-12-31 14:08:29 -05:00
Nathan Braswell
f1d2e0dce2
Add rough version of log, error, str, ptr-equality-rough versions of = and !=, and actual versions for + - * / % & |
2021-12-31 01:28:31 -05:00
Nathan Braswell
ad38628915
Add inlining of cond, which sould be the only non-wrapped vau left in normal programs currently. Started implementation of = to test cond (just ptr equality for now) and it works!
2021-12-30 17:23:15 -05:00
Nathan Braswell
7b7bccb5dd
Implemented unoptimized and naive version of what will become Perceus reference counting (functions hold the env till the end, and the env means everything is borrowed, so all params are dup'd into function calls etc.), along with a very unoptimzied and naive malloc&free (single singlely linked list of blocks that are wasm page sized (4k))
2021-12-30 01:04:07 -05:00
Nathan Braswell
accad76fa9
Add place to handle (calls unreachable for now) for non wrap-1 combiners
2021-12-29 14:38:36 -05:00
Nathan Braswell
6e5a372f07
Add compilation of closures as code
2021-12-28 19:14:02 -05:00
Nathan Braswell
2318b958e8
Implemented partially variadic functions! Remaining: functions in code, refcounting, stdlib
2021-12-28 17:34:17 -05:00
Nathan Braswell
a4a033a72e
Implemented function calls!
2021-12-28 14:24:54 -05:00
Nathan Braswell
78ba54879a
Error checking for monads and number of parameters. Implemented fully variadic, partially variadic deserves a bit more thought to see if it can be implemented efficiently
2021-12-28 00:20:13 -05:00
Nathan Braswell
481cac5070
Add block length to memory header (joining the refcount), add read & write monand impls. Big todo: typechecks in monad impl, length checks for compiled functions, compiling calls, then all the stdlib.
2021-12-27 01:35:15 -05:00
Nathan Braswell
021cae4eea
Implemented Env creation & symbol lookup
2021-12-26 22:25:53 -05:00
Nathan Braswell
4d41c0b535
print monad execution impl with first function call!
2021-12-25 23:39:25 -05:00
Nathan Braswell
8dadf07759
Add missing funcs, add offset support to wasm load/store, start in on monad impl
2021-12-25 21:01:58 -05:00
Nathan Braswell
f376a75f4c
Stubs for compiling function body
2021-12-25 01:38:49 -05:00
Nathan Braswell
5097a11bb6
More work on compiling comb, adding wrap level to comb, placeholders for all builtins, memoization
2021-12-24 17:26:41 -05:00
Nathan Braswell
69b87cbe72
Started sketching functions, swaped env's representation around
2021-12-22 23:42:25 -05:00
Nathan Braswell
eade335e86
off by one, of course
2021-12-18 14:57:52 -05:00
Nathan Braswell
75536f6fb9
Starting compilation / printing of builtin combiners
2021-12-18 14:55:54 -05:00
Nathan Braswell
c4c2d73469
Compilation and pretty-printing of environments. Still need to look at exactly why the additive length constant for the upper-env is 5 when I'm less tired
2021-12-14 00:50:48 -05:00
Nathan Braswell
3538de9498
Add in compiling and printing of array values
2021-12-12 15:32:52 -05:00
Nathan Braswell
f865bccdda
Add support for negative integers
2021-12-08 22:35:44 -05:00
Nathan Braswell
b724493926
Add support for (positive) integer printing, not just a single digit!
2021-12-08 01:22:18 -05:00
Nathan Braswell
7dfe4e0d7c
Add symbol emitting, rearrange encoding to make things nicer, implement more sophisticated to string setup with length etc
2021-12-06 01:58:38 -05:00
Nathan Braswell
b3dfd577cd
String constants working, which involved more work than expected. Lots of fixes for strings, LEB128, etc, and making the top level section func return arrays so that I can programatically generate a concatted array of them and insert it no problem. Technically, you can now compile Hello, World! for Kraken, since it prints the string constant you compile :D
2021-12-04 01:11:55 -05:00
Nathan Braswell
025239947f
Add i32 and i64 condition operations, hex escapes to data strings, and add ability to print out true and false to generated code
2021-12-02 00:24:40 -05:00
Nathan Braswell
25615815b2
bugfixes func type splitter, add more ins, trivial malloc and free, start working on print.
2021-11-29 01:51:54 -05:00
Nathan Braswell
f7f425b32f
Ultra tiny skeleton starting compiler - will read, trivially partially evaluate, then near trivally generate wasm that prints a single digit literal
2021-11-28 01:25:46 -05:00
Nathan Braswell
505dc46998
Add an extra level of lambda and pass around name dict in order to get around not having vau (+ quotation on symbols when referenced)
2021-11-28 00:57:35 -05:00
Nathan Braswell
de8073d1fc
Actually make the demo runnable
2021-11-27 22:13:07 -05:00
Nathan Braswell
4956596f30
Port (slightly hackilly) the rest of wasm.kp
2021-11-27 21:49:41 -05:00
Nathan Braswell
65c9d0b486
Port basic (non vau syntax coolness) WASM emiting code, can output empty file at least (haven't tested more complex yet, but the code is there)
2021-11-26 23:28:06 -05:00
Nathan Braswell
e0244d0489
Hay recursion works!
2021-11-26 12:31:39 -05:00
Nathan Braswell
8ab15fff41
Add hashing (interestingly, arbitrary long hashes with Scheme's infinite precision)
2021-11-25 23:57:23 -05:00
Nathan Braswell
a036936e3b
It all works! I belive all test cases from the prior partial_eval.kp work in partial_eval.csc now :D
2021-11-24 00:45:44 -05:00
Nathan Braswell
f3525def87
Fix, some more tests. Need to figure out the prim comb things
2021-11-23 22:49:54 -05:00
Nathan Braswell
60cd7b1ffa
First Vau call working!
2021-11-23 22:16:24 -05:00
Nathan Braswell
96ea2fad8d
Port over most of the rest, some left commented out. Need to figure out how we actually want to handle prim_combs putting themselves in the output. This comes up a lot with give_up_eval_params. Also, '() = nil counts as true for if in Scheme, need to figure that out :/
2021-11-23 01:54:20 -05:00
Nathan Braswell
7c32c3811a
Fix indexing bug, first partial eval!
2021-11-22 01:46:59 -05:00
Nathan Braswell
d26fcee9b6
Port more, start fixing bugs. Something weird with slice/drop/take
2021-11-22 01:28:05 -05:00
Nathan Braswell
b3261f3db0
Bunch more ported, including all of the actual partial_eval function
2021-11-20 01:13:22 -05:00
Nathan Braswell
551e60cfe1
Destructuring lambda, fix for compiling
2021-11-19 01:08:27 -05:00
Nathan Braswell
d4ff61ade5
Some work from previously, and also now an implementation of destructuring let
2021-11-16 00:15:53 -05:00
Nathan Braswell
627ccb4aa0
Begin port of partial_eval to Chicken Scheme, with a bunch of stuff to define a vaguely kraken-like environment in scheme
2021-11-09 00:35:58 -05:00
Nathan Braswell
437c2c7166
Some bugfixes, a small recursive test (that currently loops forever), and the sierpinski triangle from a while back I forgot to commit
2021-11-07 00:47:44 -04:00
Nathan Braswell
c0a07b54ce
Fixed up!
2021-10-19 02:45:56 -04:00
Nathan Braswell
e0c4366771
With some bugfixes, sucessfully got let4.7 working, which is like let5 but written in macro style! Stiil need to finish up strip for the cases like let5 that we can't finish for now.
2021-10-19 02:26:19 -04:00
Nathan Braswell
92ac879c6d
Finally fixed let_test4. Let_test5 will require more thinking to see if it's even possible, and what macro like means etc
2021-10-18 00:46:39 -04:00
Nathan Braswell
923c4565fb
prep for useing de bruijn
2021-10-17 17:39:38 -04:00
Nathan Braswell
e322fc7cd7
Finally correct again, but with the latest safety fixes no longer even evaluates the easier lets. The issue is that we're not executing combiners that use de if we don't have a real de (normally the enclosing function's se) because that's subject to the alpha-renaming problem. de is essentially a parameter, which is subtle and easy to miss when looking at the safety conditions, and indeed I did miss it earlier
2021-10-15 23:10:58 -04:00
Nathan Braswell
64c793d9f4
Flesh out the closes over test. Still doesnt' partial eval like we want, need to investigate why - have idea for specalizing function but do have to be careful with actual_function
2021-09-14 02:01:31 -04:00
Nathan Braswell
ec9083a958
Fixed bug with y, but still not partially evaluating the lets with later :/
2021-09-12 01:37:07 -04:00
Nathan Braswell
f89b21420a
Ok, figured out why let5 is failing and wrote down a plan of attack
2021-09-07 00:06:19 -04:00
Nathan Braswell
12271ff27c
Bug fixed and back to state before refactor/rewrite. Need to get that darn let5 working now! That's why we did this
2021-09-06 23:00:04 -04:00
Nathan Braswell
873e7c4244
Working through the bugs, got the first few working
2021-09-06 12:29:05 -04:00
Nathan Braswell
6a47375d28
Ok, refactored the whole thing! It's now always operating on marked values with a mark step at the beginning, which should solve our stripping problems. It parses, but crashes, but it's 3am, so I'm saving and going to bed
2021-09-06 03:00:33 -04:00
Nathan Braswell
49d5a196aa
2 steps forward, 1 step back - I have arrays (and I think varargs) working, but let5_test is now not partial evaling as far as it should because of my fix. Detailed explanation in the comments, but it looks like we are going to have to insert that eval_strip hack, or somesuch
2021-09-05 00:25:33 -04:00
Nathan Braswell
b44ff104fb
Implement variadic paremters, found additional bug for array, still need to handle recursion too
2021-08-26 01:03:36 -04:00
Nathan Braswell
26e9c5a41f
Fix a bug in eval strip not actually stripping in the proper eval way - f there's a later anywhere in there, it has to desist, which is now implemented. Added do tests - we have to add support for & params, which I had forgotten, and I belive it now loops infinately because it tries to partially evaluate a recursive function!
2021-08-25 01:32:24 -04:00
Nathan Braswell
649e0107d4
And lambda is working nicely too, once I added support for val to the function call evaluaator (just looking it up in comb_to_mark_map)
2021-08-24 00:33:29 -04:00
Nathan Braswell
634812cea4
Got it working!
2021-08-23 23:43:47 -04:00
Nathan Braswell
2cd85a552d
Some more tests and cleanup, looks like we're going to have to do either a try-eval thing or CPS-style partial evaluator, which I'm reading up on
2021-08-22 20:27:48 -04:00
Nathan Braswell
7700f0b709
FIXED THE BUG!
2021-08-22 13:03:33 -04:00
Nathan Braswell
c9e2e4659b
Completely untested, but added partial eval for all array methods and and/or - this is pretty much everything! We don't bother with meta or error/recover right now though
2021-08-17 23:09:42 -04:00
Nathan Braswell
04f1c76bd7
Factored out parameter eval for some combinators and then added combiner? and env?
2021-08-17 20:31:07 -04:00
Nathan Braswell
47d5149400
Implement smart cond partial-eval
2021-08-17 18:17:42 -04:00
Nathan Braswell
772b984de0
Ok, more bugfixes & comments, and got eval working!
2021-08-17 01:19:38 -04:00
Nathan Braswell
e8747ef899
Bugfix and improvment, eval one works even though eval isn't implemented yet!
2021-08-16 11:54:05 -04:00
Nathan Braswell
62a238e041
Added wrap/unwrap and a vau test that uses wrap!
2021-08-16 01:20:10 -04:00
Nathan Braswell
550565a235
More vau usage working
2021-08-16 00:37:56 -04:00
Nathan Braswell
05092cba76
First vau-involved partial evals working!
2021-08-15 01:27:53 -04:00
Nathan Braswell
b3c02c82b3
Create and pass through comb_to_mark_map to allow us to re-mark stripped expressions so we can evaluate them mutliple times, etc, etc
2021-08-11 23:30:49 -04:00
Nathan Braswell
f09962ddc4
Add 2 parameter vau that discards dynamic env
2021-08-10 23:26:22 -04:00
Nathan Braswell
95b37da44e
Ok, I think this is a correct? skeleton, implemented as much as I did yesterday
2021-08-10 22:56:12 -04:00
Nathan Braswell
e1fa65deaf
Starting work on a partial evaluator - just sketching for now, needs a bunch of work and likely some foundational changes
2021-08-10 00:23:14 -04:00
Nathan Braswell
d2215c2831
Fix some of the massive speed drop from destructuring lambda by checking to see if it actually needs to destrucure first. Still slower than it was, but not 10x anymore. Maybe 1.5-2x?
2021-08-05 01:11:14 -04:00
Nathan Braswell
ae82af6636
Bugfix & added an extra lambda wrapping around term/nterm so that their value can be made recursive with let-rec, and with that the parser works! On the other hand, it takes 38 seconds to parse 'a,a,a' with the grammer A='a'|'a' ',' A .... so that could be a lot better.
2021-08-05 00:39:56 -04:00
Nathan Braswell
c8c876e1bc
Gonna checkpoint here - got basic fungll working, did a mod to allow strings as terminals, working on getting recursion working.
2021-08-05 00:07:44 -04:00
Nathan Braswell
5a61d5f90c
Added destructuring lambda/rec-lambda, changed let to use the same [] array destructuring syntax, added basic multiset & set-foldl. Fixed a bunch of bugs in fungll, hopefully close to working, but just realized that < is only defined for ints, while it's how the RB-Tree set/map sort their values/keys, so I'll need to extend it like = for all types. Tomorrow!
2021-08-04 00:56:04 -04:00
Nathan Braswell