jsh-20 fc5d7e6e7c [FRONTEND] Improve grid calculation for persistent kernels to hoist pe… (#2283) ...

…rf on problems that need few blocks.

constrain the number of launched blocks to what it exactely needs for
persistent warp specialized kernel. It's useful when problems need very
few blocks.
e.g. MxNxK=800x800x60000, f16_f16_f32, block size=128x128x64,
non-split-k. Experiments show it can achieve ~16% speedup.

2023-09-12 09:14:47 +00:00

..

backend

[FRONTEND] Enable triton to support register thirdparty backend at runtime (#1643)

2023-06-09 09:09:59 -07:00

kernel_comparison

[CI] Update kernels names (#2093)

2023-08-14 19:41:41 -07:00

regression

[TESTS] replace deprecated torch.testing.assert_allclose (#2250)

2023-09-11 15:31:17 -04:00

tools

[CI] upload only test/unit/operators cache to artifacts and rely on kernel names in cache to compare artifacts (#2111)

2023-08-16 20:34:40 -07:00

unit

[FRONTEND] Improve grid calculation for persistent kernels to hoist pe… (#2283)

2023-09-12 09:14:47 +00:00