Matthias Gehre a9ec46d4b7 [ROCm][Perf] Support N=5 in wvSplitK skinny GEMM kernels for speculative decoding (#40687) ...

Signed-off-by: Matthias Gehre <matthias.gehre@amd.com>

2026-05-28 16:28:21 +00:00

..

attention.cu

[ROCm] Add gfx1102/gfx1103 support (#40037)

2026-04-23 01:32:04 -07:00

ops.h

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

2026-01-16 11:45:04 -06:00

skinny_gemms.cu

[ROCm][Perf] Support N=5 in wvSplitK skinny GEMM kernels for speculative decoding (#40687)

2026-05-28 16:28:21 +00:00

torch_bindings.cpp

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

2026-01-16 11:45:04 -06:00