gaspersic/triton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
138 Commits 14 Branches 9 Tags
269ebc12e5f96355e5ebee510cf882ceed73502f
Commit Graph

32 Commits

Author SHA1 Message Date
Philippe Tillet
269ebc12e5 [PYTHON][TESTS][DOC] Various improvement of the API and code quality: 
* Simplified `triton.kernel` API to achieve lower latency:
  > .data_ptr() must now be passed as kernel argument. No more implicit
conversion from torch.tensor
  > compilation options are now constant attributes, i.e., opt.d('VAR')
becomes opt.VAR
  > torch.device must now be passed explicitly to triton.kernel (no
longer inferred from torch.tensor arguments)
* C++ tests moved to `python/tests/`
* C++ tutorial created in `tutorials/`
* Python tutorial created in python/tutorials/
* Version changed to 1.0alpha
* No longer copying C++ headers into the Python package
* added python/triton/ops/ package for pre-written Triton ops
 2021-07-27 12:38:48 -07:00
Philippe Tillet
a5a477c36b [CODEGEN] Fixed bug in recoalesce_inst LLVM codegen 2021-07-27 12:38:48 -07:00
Philippe Tillet
376c876eb8 [RUNTIME] Disable error on spills 2021-07-27 12:38:48 -07:00
Philippe Tillet
083bbd1e8d [GENERAL] Merged v1.0alpha into master. Added features are: 
- A100 support via mma.16816
- Thread swizzling for conflict-free shared memory accesses without
padding
- Complete overhaul of the LLVM code generation in
codegen/selection/generator.cc to remove overengineering
- Added debugging capabilities in the Python binding
- Compilation error for kernels that spill
 2021-07-27 12:38:48 -07:00
Philippe Tillet
c4fceeea49 [LANG] Added hacky min/max 2021-07-27 12:38:48 -07:00
Philippe Tillet
fd5c72d6a0 [LANG] Added some more atomic_add support 2021-07-27 12:38:48 -07:00
Philippe Tillet
44ca2c0cb8 [DRIVER] Removed deprecated files and functions 2021-07-27 12:38:48 -07:00
Philippe Tillet
4f08d87fed [DRIVER] Simplified Driver API by substantially removing reliance on driver::context 2021-07-27 12:38:48 -07:00
Philippe Tillet
baa858aa74 [CODEGEN] Fixed bug in atomic_add 2021-07-27 12:38:48 -07:00
Philippe Tillet
7d095ec686 [LANG] Added sqrtf support 2021-07-27 12:38:48 -07:00
Philippe Tillet
a77c925dfd [DRIVER] Improved performance of Host driver code 2021-07-27 12:38:48 -07:00
Philippe Tillet
8f8d36c7a4 [GENERAL] Various bugfixes 2021-07-27 12:38:48 -07:00
Philippe Tillet
50587bbf4b [General] LLVM-9 -> LLVM-10 2021-07-27 12:38:48 -07:00
Philippe Tillet
8f3ee53f24 [PYTHON] Added option to show PTX source code in Python 2021-07-27 12:38:48 -07:00
Philippe Tillet
f152150e7d [LANG] Added log intrinsic 2021-07-27 12:38:48 -07:00
Philippe Tillet
02a6e81b88 [PYTHON] Cleaning C++ bindings 2021-07-27 12:38:48 -07:00
Philippe Tillet
049ab989b5 [GENERAL] Various improvements: 
* Sparse einsum in triton.ops.einsum
* Hacky support for fixed-tile-size atomic-add
* Various bugfixes in parser
 2021-07-27 12:38:48 -07:00
Philippe Tillet
664d3cae89 [DRIVER] Removed OpenCL support 
There is no plan to support OpenCL anytime soon (Vulkan would be preferred). Removing the adequate portion of the driver code
 2021-07-27 12:38:48 -07:00
Philippe Tillet
840308ab5d [CODEGEN] More work on the CPU backend 2021-07-27 12:38:48 -07:00
Philippe Tillet
acff1b5e05 [RUNTIME] Lower-level interface for executing functions 2021-07-27 12:38:48 -07:00
Philippe Tillet
5995cbff8e [CORE] Auto-tuning now copies scalar buffers. Still needs to copy all buffers that are both read from and written to. 2021-07-27 12:38:48 -07:00
Philippe Tillet
e7461a862b [CODEGEN] Bugfix in Disassociate pass; Added fp32 atomic_add support 2021-07-27 12:38:48 -07:00
Philippe Tillet
0516ea96d0 [CODEGEN] Fixed bug that caused missing recoalescing for some transpose 
operations
 2021-07-27 12:38:48 -07:00
Philippe Tillet
677ccfb44e [CORE][RUNTIME] Better error message on internal compilation error 2021-07-27 12:38:48 -07:00
Philippe Tillet
5943baa53f [GENERAL] Error messages now no longer make terminal color green 2021-07-27 12:38:48 -07:00
Philippe Tillet
3304629de9 [CORE] Fixed several issues that arose in the development of the 
torch-blocksparse package:

* Now using warp shuffle in reductions when possible
* Various bugfixes in layout inference
* Added INFINITY, exponential and select
* Better error messages for unimplemented constructs
 2021-07-27 12:38:48 -07:00
Philippe Tillet
268894a5ce [PYTHON] Merged blocksparse branch: 
* Example for blocksparse matrix multiplication
* Simplified Triton kernel API
* Revived auto-tuning in einsum
 2021-07-27 12:38:48 -07:00
Philippe Tillet
dfb844bf41 [GENERAL] Improved caching mechanism: 
* Now computing hash in libtriton
* Now only compiling a single pytorch hook per function signature
 2021-07-27 12:38:48 -07:00
Philippe Tillet
4181f9f2af [CODEGEN][TRANSFORM][PEEPHOLE] Fixed bug in *1 multiplication 2021-07-27 12:38:48 -07:00
Philippe Tillet
3816f2f259 [PYTHON][EINSUM] Now handling reduction sizes that are not a multiple of 
TK
 2021-07-27 12:38:48 -07:00
Philippe Tillet
d22cf4f717 [TRITON][LANG] Added support for bitcast 2021-07-27 12:38:48 -07:00
Philippe Tillet
6d7cf35123 History prior to this date belonged to the now deprecated ISAAC project, and was deleted to save space 2021-07-27 12:38:38 -07:00