Weixing Zhang 76858bd917 [OPTIMIZER] Tweak warpsPerCTA based on the shape of MMA output (#2485) ...

In current implementation, warpsPerCTA is always set to [numWarps, 1]
for 2 tt.dot fusion scenario. But, it is not optimal for cases such that
tt.dot doesn't have enough parallelism on row dimension but on column
dimension.

2023-10-12 22:25:42 -07:00

..

test_functional_regressions.py

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

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

test_performance.py

[OPTIMIZER] Tweak warpsPerCTA based on the shape of MMA output (#2485)

2023-10-12 22:25:42 -07:00