jyapayne/llvmpy
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
52 commits 8 branches 17 tags 4.2 MiB
Commit graph

52 commits

This branch
This branch
All branches
Author SHA1 Message Date
Travis E. Oliphant
ab0ce0d49a Update numbapro to handle latest numba changes. 2012-09-27 01:11:35 -05:00
Mark Florisson
20a896b206 Start on array expressions 2012-09-24 12:21:12 +01:00
Siu Kwan Lam
b9584e1071 Fix assertion error in python source due to invalid lineno increment inside llvm_cbuilder.translator. 2012-09-21 19:37:10 -07:00
Mark Florisson
bc03c9a8b7 Fix GC head for debug builds for gufunc fake arrays 2012-09-21 22:46:48 +02:00
Mark Florisson
10b3f188cc Some more progress towards CUDA gufuncs 2012-09-21 22:46:48 +02:00
Mark Florisson
b0137289b9 Use meta to decompile python functions into ASTs 2012-09-21 22:39:48 +02:00
Siu Kwan Lam
8a159b35f0 refactor llvm_cbuilder 2012-09-07 17:53:22 -07:00
Siu Kwan Lam
af44641bb9 add noalias attribute to parameters 2012-09-07 17:53:22 -07:00
Siu Kwan Lam
1e4633c44d add invariant variable 2012-09-07 17:53:22 -07:00
Siu Kwan Lam
b941ae5f1f Use non-temporal hint at memory store for vectorizer. 
Note: LLVM optimizer is smart enough to not use NT write if it is slower.
 2012-09-07 17:53:22 -07:00
Siu Kwan Lam
8505256510 Inline ufunc kernel in vectorize. (Better io throughput) 2012-09-04 10:08:07 -07:00
Mark Wiebe
e7b1af043f Use the function's calling convention when doing the call in cbuilder 2012-08-30 18:13:03 -07:00
Siu Kwan Lam
a046685b2c Add calling_conv. 
Set win32 api to use CC_X86_STDCALL. (need testing)
 2012-08-30 12:23:53 -07:00
Siu Kwan Lam
d390960657 Add llvm_cbuilder translate features: 
* handle while loop
 * handle for loop
 * handle immediate values

Todo: handle "break" and "continue"
 2012-08-23 11:43:24 -05:00
Siu Kwan Lam
eeda441dac Added stream vectorize. 2012-08-22 13:23:52 -05:00
Siu Kwan Lam
4d773ed5ed Add alignment control to vector_load/store. 
E.g: align=1 for universal alignment. default to aligned.
 2012-08-21 18:32:44 -05:00
Siu Kwan Lam
2218f1d49e Improve vector load store. 
Changes make llvm emit parallel ld/st instead of scalar ld/st.
 2012-08-21 17:26:12 -05:00
Siu Kwan Lam
5513a34e80 Add some doc and cleanup 2012-08-21 14:17:14 -05:00
Siu Kwan Lam
5319fb8c98 Add initial vector support 2012-08-21 12:33:21 -05:00
Siu Kwan Lam
562e3c1e46 Fix: if-else branch should terminate automatically. Insert unreachable to if.end block if necessary. 2012-08-21 12:30:35 -05:00
Siu Kwan Lam
d55bb95d4d Add translator to convert python code to llvm cbuilder construct 2012-08-21 12:30:35 -05:00
Siu Kwan Lam
f9d7195413 reorganize 2012-08-13 15:53:09 -07:00
Siu Kwan Lam
1009daf3fa Add notebook: parallel_vectorize 2012-08-11 14:55:35 -07:00
Siu Kwan Lam
c1d3044c77 Fix parallel_vectorize_from_func: 
* use signed integer;
* use multiprocessing.cpu_count() to determine # of threads.
 2012-08-11 14:06:13 -07:00
Siu Kwan Lam
7f84ecbe3b Add notebook: llvm_cbuilder quick tour 2012-08-10 17:58:54 -07:00
Siu Kwan Lam
9ccc6782c7 Add doc for parallel_vectorize_from_func. 
Add FunctionAlreadyExists exception.
 2012-08-08 17:46:20 -07:00
Siu Kwan Lam
c6d5538ead Add usecase tests 2012-08-08 17:33:34 -07:00
Siu Kwan Lam
753f74d485 Fix tests 2012-08-08 17:33:24 -07:00
Siu Kwan Lam
b23d7909fb add parallel_vectorize_from_func 2012-08-08 17:33:00 -07:00
Siu Kwan Lam
e35b8ca636 Add some float casting 2012-08-08 17:32:39 -07:00
Siu Kwan Lam
3b0aeea6f2 change CDeclare to CFuncRef 2012-08-08 16:27:57 -07:00
Siu Kwan Lam
3009d8468f more cleanup and doc 2012-08-08 15:29:16 -07:00
Siu Kwan Lam
68e2359ad2 cleanup and doc 2012-08-08 15:17:31 -07:00
Siu Kwan Lam
8b45094e20 Add test that uses function pointer. 2012-08-07 16:56:28 -07:00
Siu Kwan Lam
2acc070984 Refactor. 
Modify the way CDefinition is specialized.
 2012-08-07 16:55:33 -07:00
Siu Kwan Lam
01054c726f Add specialize() for cleaner specialization of CDefinition. 2012-08-06 20:25:36 -07:00
Siu Kwan Lam
b5893d4ebb Add test to use numpy.fromfunc. 
Depends on numpy branch "ufunc-from-function-pointer" in git://github.com/jayvius/numpy.git
 2012-08-06 17:02:04 -07:00
Siu Kwan Lam
150f497ed0 Add specializable parallel ufunc. 
Update test.
Some fixes.
 2012-08-06 15:33:36 -07:00
Siu Kwan Lam
fef9c2efbf Add CBuilder.depends to automatically insert depending function. 2012-08-06 14:46:27 -07:00
Siu Kwan Lam
c43b27356b Fix 2012-08-06 14:23:43 -07:00
Siu Kwan Lam
1594e5456a Got test_parallel_vectorize working. 2012-08-06 11:35:03 -07:00
Siu Kwan Lam
891daa1164 Added test implementation of parallel-vectorize (incomplete). 
Various fix.
 2012-08-05 21:35:02 -07:00
Siu Kwan Lam
0654e00ab1 Add printf() and friends. 
Disable logging in old tests.
 2012-08-04 19:13:35 -07:00
Siu Kwan Lam
673ada6a4e Fix and add tests. 2012-08-04 15:18:41 -07:00
Siu Kwan Lam
b45c768cc5 Fix 2012-08-04 15:14:09 -07:00
Siu Kwan Lam
184e4ea1a1 Add API for structure 2012-08-03 22:30:54 -07:00
Siu Kwan Lam
f1e10e849f Fix __div__ for py3. 
Add test for nested-loops.
 2012-08-03 17:02:01 -07:00
Siu Kwan Lam
aebb9857eb Drop volatile option in atomic ops. 
Add volatile option to store.
 2012-08-03 16:31:23 -07:00
Siu Kwan Lam
8381b25f66 Add atomic ops 2012-08-03 16:26:46 -07:00
Siu Kwan Lam
7238fecc7f add build & test instruction to README 2012-08-02 14:20:02 -07:00