Atomics Reduce Counting Optimization for SplitK Skinny GEMMs. (#29843)

Signed-off-by: Hashem Hashemi <hashem.hashemi@amd.com>
This commit is contained in:
Hashem Hashemi
2026-01-16 11:45:04 -06:00
committed by GitHub
parent 9fd918e510
commit 7a1030431a
6 changed files with 635 additions and 10 deletions
+6
View File
@@ -2072,6 +2072,12 @@ def wvSplitK(
return torch.ops._rocm_C.wvSplitK(a, b, bias, cu_count)
def wvSplitKrc(
a: torch.Tensor, b: torch.Tensor, cu_count: int, bias: torch.Tensor = None
) -> torch.Tensor:
return torch.ops._rocm_C.wvSplitKrc(a, b, bias, cu_count)
def wvSplitKQ(
a: torch.Tensor,
b: torch.Tensor,