 Nathan Braswell
|
5ebb6d5482
|
Start work on structs. Can declare structs and access members. Copying structs and passing them as parameters or returning them doesn't work yet
|
2018-03-15 00:29:21 -04:00 |
|
|
Nathan Braswell
|
8de92e1c3b
|
After reflecting on the difficulties of lowering the increment operators, decided just to implement them in bytecode
|
2018-03-14 23:43:52 -04:00 |
|
|
Nathan Braswell
|
c6a43728a5
|
Fix function calls by giving own reg for return and saving registers first. Use reset_reg more often to try to keep the number of registers that need to be saved down.
|
2018-03-14 01:16:00 -04:00 |
|
|
Nathan Braswell
|
364f4d8a3e
|
Sketch out places for struct involvement, write while and for loops
|
2018-03-13 00:39:16 -04:00 |
|
|
Nathan Braswell
|
fc6c87f56a
|
Finished the operators, besides struct access
|
2018-03-12 23:34:35 -04:00 |
|
|
Nathan Braswell
|
378a81602a
|
Add in signed and unsigned mul div, and add a mod
|
2018-03-12 20:17:26 -04:00 |
|
|
Nathan Braswell
|
ef516891ca
|
add xor
|
2018-03-12 00:43:51 -04:00 |
|
|
Nathan Braswell
|
5464555e0d
|
addressof, dereference, and [] implemented
|
2018-03-10 15:32:51 -05:00 |
|
|
Nathan Braswell
|
3c4b1864c7
|
Add in addi to reduce use of imm, add and or and not to put in negate/subtract (might need it's own opcode, really) as well as and or and not themselves
|
2018-03-10 00:27:16 -05:00 |
|
|
Nathan Braswell
|
55a448a3b5
|
Small work towards operators. + workes now
|
2018-03-08 23:25:48 -05:00 |
|
|
Nathan Braswell
|
19e97113f1
|
Accidentally messed up bootstrapping by returning a long in a function with return type int, which was an error in earlier versions of kraken, but the type check got removed during one of the simplifications, so would error when bootstrapping. Really, this should still be an error - I will try to fix this tonight. To make the boostrap process work, though, I've added an extra bootstrapping stage that builds the version without the return type check along the way.
|
2018-03-08 09:48:48 -05:00 |
|
|
Nathan Braswell
|
fc67aa7ce8
|
Added support for parameters
|
2018-03-07 22:57:46 -05:00 |
|
|
Nathan Braswell
|
8da84b56c2
|
working bytecode with call and a real stack and psudo abi. Found a bug where -= doesn't behave correctly when the lhs is a function returning a reference because it's lowered to a = a - b where a is the same ast_node, I think
|
2018-03-07 01:58:19 -05:00 |
|
|
Nathan Braswell
|
5b46089694
|
Stop declaring variable before assignment - i don't think necessary for recursive closure, def not in current code, and messes up globals with values
|
2018-03-06 23:30:00 -05:00 |
|
|
Nathan Braswell
|
8a22263b1d
|
make JNZ JZ (if was backwards, and this is more convient right now
|
2018-02-27 21:41:57 -05:00 |
|
|
Nathan Braswell
|
dad97a25a5
|
If statements work with the addition of jmp and jnz
|
2018-02-27 21:27:29 -05:00 |
|
|
Nathan Braswell
|
adcebb3284
|
Move towards more legit stack setup api and registers as registers
|
2018-02-27 23:53:08 +00:00 |
|
|
Nathan Braswell
|
00bb4e928c
|
Can eval basic block with variables and return
|
2018-02-03 22:47:21 -05:00 |
|
|
Nathan Braswell
|
aad363f631
|
Bytecode work
|
2018-02-03 18:53:13 -05:00 |
|
|
Nathan Braswell
|
7850f11297
|
very basic groundwork for bytecode
|
2018-02-02 00:26:31 -05:00 |
|
|
Nathan Braswell
|
c0209118e5
|
Small cleanup
|
2018-01-02 23:22:46 -05:00 |
|
|
Nathan Braswell
|
241764f1f4
|
Clean up
|
2017-12-27 16:49:08 -05:00 |
|
|
Nathan Braswell
|
d414459313
|
Fixed it
|
2017-12-27 16:10:36 -05:00 |
|
|
Nathan Braswell
|
7ddbecaf1d
|
debugging
|
2017-11-13 22:39:14 -05:00 |
|
|
Nathan Braswell
|
cb720e5cd6
|
working on moving rest of obj stuff into obj_lower
|
2017-11-03 00:39:58 -04:00 |
|
|
Nathan Braswell
|
5b8ef4ce2e
|
some cleanup
|
2017-10-29 17:53:30 -04:00 |
|
|
Nathan Braswell
|
0132ade7ed
|
10% or so speed up
|
2017-10-28 15:28:34 -04:00 |
|
|
Nathan Braswell
|
48b21c54ae
|
Missed a change last time - also made all_types in function_value_lower a hash_set for a ~20 sec speedup
|
2017-10-23 09:50:18 -04:00 |
|
|
Nathan Braswell
|
13f1e9df89
|
More clean up; need to add return type checking pass so that can clean up c_generator more.
|
2017-10-23 01:08:25 -04:00 |
|
|
Nathan Braswell
|
6964df5ac8
|
Remove code_triple entirely!
|
2017-10-23 00:06:25 -04:00 |
|
|
Nathan Braswell
|
e70d31a51c
|
Add in address_of_ensure_variable_lower.krak that makes sure & can always take a reference, even of a constant, etc, and start the cleanup of c_generator that this allows
|
2017-10-22 21:41:58 -04:00 |
|
|
Nathan Braswell
|
73faf50053
|
tiny cleanup
|
2017-09-02 09:12:25 -04:00 |
|
|
Nathan Braswell
|
00f29fe6a5
|
small clean up
|
2017-08-27 14:15:09 -04:00 |
|
|
Nathan Braswell
|
e4326233dc
|
a bit more cleanup
|
2017-08-23 10:12:00 -04:00 |
|
|
Nathan Braswell
|
36d7581cb0
|
further simplify the function call function
|
2017-08-22 20:52:27 -04:00 |
|
|
Nathan Braswell
|
4921371afe
|
Cleaned up generate_function_call. It's quite understandable now!
|
2017-08-18 10:05:12 -04:00 |
|
|
Nathan Braswell
|
61feff50a8
|
Update captain.sh to be able to build the new function_value_lower version
|
2017-08-16 01:23:17 -04:00 |
|
|
Nathan Braswell
|
9621ce85a6
|
Always error, remove some chaff
|
2017-08-15 19:53:17 -04:00 |
|
|
Nathan Braswell
|
0cf9beb057
|
Function value lower can self-host!
|
2017-08-15 01:55:44 -04:00 |
|
|
Nathan Braswell
|
bdb9e313b5
|
push forward. Current problem is double closed things
|
2017-06-21 02:12:00 -04:00 |
|
|
Nathan Braswell
|
18c53eb9be
|
MOve it forwards a bit
|
2017-06-13 01:29:56 -04:00 |
|
|
Nathan Braswell
|
6f659ece49
|
Fix closing over raw function pointers
|
2017-06-13 00:40:29 -04:00 |
|
|
Nathan Braswell
|
bd83156c12
|
work in progress
|
2017-06-12 23:52:12 -04:00 |
|
|
Nathan Braswell
|
8166db37ac
|
Closures working! (in the basic case, anyway)
|
2017-06-06 01:33:18 -04:00 |
|
|
Nathan Braswell
|
7a2a76e6da
|
wip
|
2017-05-16 11:12:05 -04:00 |
|
|
Nathan Braswell
|
bfe65fd854
|
Now support parameters for lambdas
|
2017-05-15 21:25:40 -04:00 |
|
|
Nathan Braswell
|
265a994858
|
Can call non-closure no parameter lambdas
|
2017-04-17 21:15:14 -04:00 |
|
|
Nathan Braswell
|
524247d0d1
|
Work on creating the call function
|
2017-04-13 01:05:36 -04:00 |
|
|
Nathan Braswell
|
a0c040da36
|
more lowering
|
2017-04-08 16:10:57 -04:00 |
|
|
Nathan Braswell
|
9662197c21
|
comment out the as value part
|
2017-03-16 19:05:45 -04:00 |
|
|
Nathan Braswell
|
8a676a1b5b
|
More work. Doesn't compile at this point
|
2017-02-23 01:24:22 -05:00 |
|
|
Nathan Braswell
|
cb8124afc0
|
baby steps
|
2017-02-17 01:22:27 -05:00 |
|
|
Nathan Braswell
|
425d75675e
|
babier steps
|
2017-02-08 00:01:29 +00:00 |
|
|
Nathan Braswell
|
3fc0ea9bec
|
baby steps
|
2017-02-04 01:29:22 -05:00 |
|
|
Nathan Braswell
|
1b0dce9ed1
|
Work on function value lower
|
2017-02-02 00:46:36 -05:00 |
|
|
Nathan Braswell
|
754ff41226
|
C_generator can call raw function pointers now
|
2017-01-28 00:09:13 -05:00 |
|
|
Nathan Braswell
|
dad0f780bb
|
Groundwork for closure lowering, run raw function pointer type
|
2017-01-26 23:58:48 -05:00 |
|
|
Nathan Braswell
|
caba8b310f
|
Small speed improvement int trim by using stack (and DFS) instead of queue (and BFS) for remove and collapse, finally fix the unknown escape '\*' error
|
2017-01-25 01:47:05 -05:00 |
|
|
Nathan Braswell
|
987e6eb9a5
|
Cleaning up to bring into main. Speed problem seems to have dissapeard, but we have fixes and the ^ operator in here
|
2017-01-24 22:11:33 -05:00 |
|
|
Nathan Braswell
|
f0a657e80f
|
Fix invalid malloc on slice
|
2017-01-23 23:00:26 -05:00 |
|
|
Nathan Braswell
|
3a7f73b711
|
sigh
|
2017-01-23 01:09:31 -05:00 |
|
|
Nathan Braswell
|
beb50b8e25
|
3 sec laptop speed improvement in hash_map
|
2017-01-22 16:36:04 -05:00 |
|
|
Nathan Braswell
|
ebb34d5ba3
|
Replace visited list for pass_common with hash_set (which isn't complete, but has the basics) for a massive speedup (6xish)
|
2017-01-22 10:13:06 -05:00 |
|
|
Nathan Braswell
|
896e8a936c
|
Merge branch 'ref_lower'
|
2017-01-21 22:50:59 -05:00 |
|
|
Nathan Braswell
|
d9814a278a
|
It does work! Commented out ctce_pass test because it chokes on the new iteration styrategery
|
2017-01-21 22:50:29 -05:00 |
|
|
Nathan Braswell
|
5a6f498043
|
Working!
|
2017-01-21 22:01:47 -05:00 |
|
|
Nathan Braswell
|
21f957195a
|
Fix dreferencing function_call refs. Now need to make sure new pass method works with defer_lower, which currently does very crazy defer double stack chains
|
2017-01-20 01:31:28 -05:00 |
|
|
Nathan Braswell
|
e2639989c9
|
ref_lower now generates C, though a ton of syntax errors
|
2017-01-20 01:11:06 -05:00 |
|
|
Nathan Braswell
|
2a53d5539d
|
Modify captain.sh to use /usr/bin/env bash
|
2016-11-23 23:57:01 +00:00 |
|
|
Nathan Braswell
|
b0d2a6918d
|
Regex rewrite for big speed gain, some extras (--no-c-compile), -O2 is default now
|
2016-08-02 01:33:16 -07:00 |
|
|
Nathan Braswell
|
523526f40e
|
Improve arg parsing a bit, including parsing arbitrary flags for later use. Got (after a bunch of linker pain solved by just linking in LLVM-3.8 plain and not messing with llvm-config at all) a LLVM example working
|
2016-07-09 23:52:32 -07:00 |
|
|
Nathan Braswell
|
c633cb5b04
|
CTCE Pass first implementation, but it's very slow.
|
2016-07-09 15:33:15 -07:00 |
|
|
Nathan Braswell
|
fb63eee9e8
|
Commit pre enabling CTCE pass (which is quite slow, but does work). Had to add walking through cast nodes for finding variables to close over. Also had to remove the static in front of functions to prevent gcc compiling it so that it segfaults (doesn't segfault if compiled with clang, but wanted to make sure.)
|
2016-07-09 15:08:57 -07:00 |
|
|
Nathan Braswell
|
ddd250e7f3
|
Fixed baking the right integer types into values from #ctce, added the rest of the integer type literals (10ul, etc), fixed/added struct padding/alignment in interpreter
|
2016-07-09 00:45:40 -07:00 |
|
|
Nathan Braswell
|
dc5fe39083
|
Added static to non extern non main functions in c_generator giving a 12 second compilation speedup - added a nodes touched pass, commented out in main, and fixed the test_extern test to use the right printf arguments.
|
2016-07-06 22:46:57 -07:00 |
|
|
Nathan Braswell
|
a5deef8c30
|
Transitioned to using CTCE for Kraken compile information
|
2016-07-06 00:22:33 -07:00 |
|
|
Nathan Braswell
|
f71b5f3576
|
Init globals before lowering CTCE, allows CTCE to use imported functions, etc.
|
2016-07-06 00:16:39 -07:00 |
|
|
Nathan Braswell
|
cd21e7a1cc
|
Add in simple CTCE test
|
2016-07-03 23:07:34 -07:00 |
|
|
Nathan Braswell
|
067c72f77c
|
Finish transition to the new #link syntax
|
2016-07-03 22:53:05 -07:00 |
|
|
Nathan Braswell
|
0f2ac1421a
|
Basic CTCE working! In between commit because #link(a) syntax changed to #link(a)
|
2016-07-03 22:50:42 -07:00 |
|
|
Nathan Braswell
|
6fee942756
|
Refactored interpreter into just functions, added a REPL to the main kraken.
|
2016-07-03 15:32:45 -07:00 |
|
|
Nathan Braswell
|
87c2b1d2c1
|
Interpreter up to 71/73 tests passing, the only ones that don't yet are the future ones. Struct size is still wrong though
|
2016-07-03 01:55:32 -07:00 |
|
|
Nathan Braswell
|
2e80682f01
|
small double fix for interpreter
|
2016-07-01 09:37:14 -07:00 |
|
|
Nathan Braswell
|
9291003af7
|
Fixed interpreter not returning ref correctly, 65/73 tests passing
|
2016-06-30 22:41:32 -07:00 |
|
|
Nathan Braswell
|
b4d0d20883
|
Fix returning functions
|
2016-07-01 00:08:41 -04:00 |
|
|
Nathan Braswell
|
4bf8572d21
|
Move functions to use a pointer to a map for closures instead of a raw map, as we cannot currently ensure that the object is made with the right size (type_size doesn't do padding, etc)
|
2016-06-30 17:06:34 -04:00 |
|
|
Nathan Braswell
|
76e78a2ac0
|
transfer work on interpreted closures
|
2016-06-30 09:57:41 -07:00 |
|
|
Nathan Braswell
|
4195da6492
|
Added union and ref = ptr workaround to interpreter (fixing adt matching), allowing test_union and test_adt to pass.
|
2016-06-28 01:35:54 -07:00 |
|
|
Nathan Braswell
|
284d8ac9c7
|
Fixed up interpreter, ADT is close to working but something's weird about what match statements are translated into
|
2016-06-27 01:21:24 -07:00 |
|
|
Nathan Braswell
|
15fe2aeba4
|
Moved copy_constructing into functions and destructing the results into obj_lower. Removed all defer_stack from c_generator. The last thing should be to move ref handling into obj_lower or maybe a pass after, then the rest of c_generator can be cleaned up and fixing interpreter (except for closures) should be easy.
|
2016-06-26 04:44:54 -07:00 |
|
|
Nathan Braswell
|
4cc0d26c4c
|
Now && and || short circuiting is done in obj_lower
|
2016-06-25 23:56:07 -07:00 |
|
|
Nathan Braswell
|
dd8fbc0489
|
moved munging while and for into obj_lower
|
2016-06-25 16:02:53 -07:00 |
|
|
Nathan Braswell
|
9de3e07715
|
forgot to add a dependency in captain.sh
|
2016-06-24 01:45:10 -07:00 |
|
|
Nathan Braswell
|
e3607beabe
|
Refactor parameter destruction into obj_lower
|
2016-06-24 01:42:48 -07:00 |
|
|
Nathan Braswell
|
29eff2a23e
|
Fix defer_lower to handle return statements, starting on obj_lower which has now taken over calling destruct for declared varaibles. Some of the code from the c_generator has been commented out or bypassed to use the new system - it should be removed when it's complete.
|
2016-06-22 01:41:57 -07:00 |
|
|
Nathan Braswell
|
26e7ee249a
|
Now using adt_lower, which also makes the backing deconstruct a reference, which makes sense
|
2016-06-20 01:52:28 -07:00 |
|
|
Nathan Braswell
|
4a33a94b15
|
Fixed, working unions. Had to move the is_union member to the end. I'm not sure if this is a C bug or an effect of me doing something weird somewhere
|
2016-06-16 23:06:38 -07:00 |
|
|
Nathan Braswell
|
59f5c6f667
|
found segfault member
|
2016-06-17 01:01:59 -04:00 |
|
|
Nathan Braswell
|
3f89cb1e37
|
partial rollback, might still not work
|
2016-06-16 09:53:15 -07:00 |
|
|
Nathan Braswell
|
4de7dd1210
|
Added support for unions as a step towards lowering ADTs in the next pass to be written.
|
2016-06-15 22:26:03 -07:00 |
|