gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
582 Commits 14 Branches 9 Tags
20100a7254e62efd0fced864b52a877c520fc38a
Commit Graph

582 Commits

This Branch
This Branch
All Branches
Author SHA1 Message Date
Jason Ansel
e647402fd3 Fix warning in generated C code (#667) 2022-09-18 12:57:32 -07:00
Philippe Tillet
4a77dfb042 [FRONTEND] Complete rewrite of the runtime (#644) 
This PR completely rewrites the runtime of Triton to be more lean and
clearly separate the compilation step from the just-in-time caching logic.
This should substantially reduce launch overhead.
 2022-09-18 08:51:48 -07:00
Ian Bearman
889d9e34a1 [REPO] update gitignore (#666) 
Update `.gitignore` to include `.vs` and `.vscode`
 2022-09-17 14:25:28 -07:00
Shintaro Iwasaki
c668d6596e [DOCS] Fix spelling (#664) 
This PR applies minor spelling fix in comments and string literals to
`master`. It shouldn't hurt anything.
 2022-09-16 12:26:40 -07:00
Sophia Wisdom
4580a04710 [FRONTEND] Improve error message for CPU tensors (#654) 
Redo of #651 against master. Fixes #525 by catching CUDA error when we
check pytorch tensor size and rethrowing a more informative error that
says why we failed.
 2022-09-14 14:26:42 -07:00
Philippe Tillet
cfbbc7b43a [CI] Added V100 tag to disambiguate self-hosted runners (#653) 2022-09-14 13:47:50 -07:00
Yunxing Dai
59a8e25f43 [DOCS] Fix typo (#650) 2022-09-14 12:17:05 -07:00
Da Yan
437ced38c2 fp8 <> bf16 conversion (#637) 
Co-authored-by: Philippe Tillet <phil@openai.com>
 2022-08-30 14:20:12 -07:00
Da Yan
210a296699 [BACKEND] bf16 flash-attention (#636) 2022-08-26 20:40:55 -07:00
Daniil Fukalov
fe0c29b9ec Fix inconsistent struct declaration instead of class. (#632) 
Looks like typo.
 2022-08-26 16:20:21 -07:00
Phil Wang
7394d732ad [DOCS] support for variable head dimensions in flash attention triton tutorial (#623) 2022-08-15 19:16:49 -07:00
Da Yan
3e2953f357 Allow multiple_of and max_contiguous to accept n-d values (#617) 2022-08-10 09:59:32 -07:00
Daniil Fukalov
cc79376222 Fix deprectaion warning on CreateGEP(Value *, ArrayRef<Value *>, const Twine &) (#608) 
This variant of CreateGEP() is already removed in LLVM 14.
 2022-08-07 17:10:18 -07:00
Daniil Fukalov
7b91c7befd Fix "warning: control reaches end of non-void function". (#607) 2022-08-02 16:12:48 -07:00
Sharad Vikram
968f59027e Expose module.print in pybind (#604) 2022-07-29 21:36:08 -07:00
Anton Kostin
923d468187 Update LICENSE (#602) 2022-07-25 09:30:03 -07:00
Jason Ansel
027321cdcf [FRONTEND] Make tl.rand() 1-exclusive (#601) 2022-07-24 17:47:23 -07:00
Jason Ansel
e02e56dc63 [FRONTEND] Add missing rfloordiv (#598) 
* [FRONTEND] Add missing rfloordiv

* fix tests
 2022-07-23 21:54:12 -07:00
Philippe Tillet
ab56d310dd [BACKEND][IR] Fixed up internal dtype size for booleans (1bit -> 8bit) (#600) 2022-07-23 20:08:03 -07:00
Da Yan
f28caddbf8 [FRONTEND] Allow tl.where to select pointers (#595) 2022-07-21 09:54:27 -07:00
Keren Zhou
af85f5fa46 [FRONTEND] Refresh cache when the source code of outlined functions are changed (#590) 2022-07-20 17:34:07 -07:00
daadaada
9b2bc88d11 [BACKEND] Better bf16 support (#588) 2022-07-19 21:22:37 -07:00
Philippe Tillet
86cab58d89 [CI] Changed dev wheel date to UTC time to match CRON schedule (#587) 2022-07-18 14:54:13 -07:00
Phil Tillet
5b04331dd2 [TUTORIALS] Added more credits in fused attention tutorial 2022-07-13 23:48:58 -07:00
Jason Ansel
0a3f3d5f25 [PACKAGING] Include triton/language/libdevice.10.bc in package data (#582) 2022-07-13 23:45:27 -07:00
Keren Zhou
4912916c11 [FRONTEND] Added support for element-wise function defined in external LLVM bitcode (e.g., libdevice) (#562) 2022-07-13 15:52:21 -07:00
Phil Tillet
971f5782b4 [tutorials] Added flash attention credits in tutorial 2022-07-11 18:56:48 -07:00
Philippe Tillet
d5eb9bc230 [tutorial] Added bwd in fused attention example (#579) 
Doesn't work on V100
 2022-07-11 15:43:46 -07:00
Jason Ansel
c9a2b9c7d4 [FRONTEND] Add missing args to get_simd_tflops() (#578) 2022-07-11 14:37:59 -07:00
Philippe Tillet
4a399a7e40 [BACKEND] Fix some bugs (atomics, a segfault...) (#577) 
This should fix #558 , #573 and #574
 2022-07-06 20:03:04 -07:00
vesuppi
22105bc33b [FRONTEND] Added type check in semantic arange (#572) 2022-07-03 15:25:37 -07:00
Keren Zhou
4bf509889b [BUILD] Change the default build type to Release (#571) 2022-07-01 12:17:22 -07:00
Keren Zhou
a74cce375f [FRONTEND] Raise broadcast error (#555) 2022-06-30 17:32:07 -07:00
Philippe Tillet
f733327ba4 [BACKEND][CODEGEN] Disabling L2 residency control by default (#570) 2022-06-29 17:05:13 -07:00
Natalia Gimelshein
1bbb2430d9 [TUTORIALS] adjust heuristics for dwdb kernel (#565) 2022-06-29 17:00:22 -07:00
Kashif Rasul
1895ceaa2d [TUTORIAL] Fix f-string for older python (#569) 
fixes issue #568
 2022-06-29 09:39:10 -07:00
Philippe Tillet
feb7a2a0dc [FRONTEND] Hotfix for store argument order (#567) 2022-06-28 00:24:02 -07:00
Philippe Tillet
5b4c8f221e [BACKEND] Compiler improvements (#557) 
This PR adds several optimization capabilities in the compiler backend:
- Now using inline PTX for `tl.store`, making it possible to use things like evict_last
- For A100, mma layout can be directly converted to shared memory
- For A100, an additional "transpose" argument in `dot` allows tensors to be loaded once and used both row- and col- major.
- Fixed liveness analysis; this was broken.
- Now can load/store directly mma layout without converting. Useful for when tl.dot accumulator is initialized with DRAM data inside of an inner loop.
- `tl.dot` can now take LHS inputs in registers when it comes from a previous `tl.dot` instruction. Useful for e.g. fused attention.
 2022-06-27 11:49:19 -07:00
Keren Zhou
87413bc925 [BACKEND] Fix layout convert for non-contiguous input (#564) 2022-06-25 23:12:03 -07:00
Keren Zhou
d345ddf837 [DOCS] Separate atomic cas from other atomic operations since operands are very different (#559) 2022-06-22 17:51:17 -07:00
Keren Zhou
b02bac41ba [CI] Change cache dir (#561) 2022-06-22 11:44:35 -07:00
Keren Zhou
a428cf0bb2 [FRONTEND] Fix pytorch warning. (#560) 
UserWarning: __floordiv__ is deprecated, and its behavior will change in a future version of pytorch. It currently rounds toward 0 (like the 'trunc' function NOT 'floor'). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mode='trunc').
 2022-06-20 20:12:09 -07:00
Keren Zhou
b5e728cb14 Add argmin argmax (#552) 2022-06-15 13:55:20 -07:00
Jason Ansel
6b9756532f [BACKEND] Remove print in coalesce.cc (#551) 2022-06-15 13:13:20 -07:00
Madeleine Thompson
8ce2c12e33 [PYTHON] move ephemeral files to homedir (#549) 
This prevents potential conflicts with other users on shared machines.
 2022-06-13 19:37:52 -07:00
Keren Zhou
93209c07e0 [BACKEND][CODEGEN] Fix reduce uint (#547) 2022-06-13 16:43:57 -07:00
Philippe Tillet
58c8889235 [FRONTEND] Fix scanline layout (#548) 2022-06-13 16:21:10 -07:00
Natalia Gimelshein
7094657aa9 [FRONTEND] fix bool conversion of floating types (#545) 2022-06-13 15:52:37 -07:00
Keren Zhou
38573d1261 [FRONTEND] Return allocated registers and spilled registers for users (#541) 2022-06-07 18:37:12 -07:00
Mengchi Zhang
2cdc6d35c4 [FRONTEND] Give col_per_thread an initial value to make the compiler happy (#535) 
Signed-off-by: Mengchi Zhang <mengchi@fb.com>
 2022-06-06 12:48:23 -07:00