biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Build][Kernel] Update CUTLASS to v3.6.0 (#11607)

Browse Source 
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
This commit is contained in:
Tyler Michael Smith
2024-12-30 04:22:13 -05:00
committed by GitHub
parent 628ec6c17b
commit 970d6d0776
6 changed files with 25 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
csrc/quantization/machete/machete_mainloop.cuh
View File

@@ -66,13 +66,11 @@ struct MacheteCollectiveMma {
using Schedule = KernelScheduleType;
static_assert(
cute::is_same_v<Schedule, KernelTmaWarpSpecialized> ||
cute::is_same_v<Schedule, KernelTmaWarpSpecializedMixedInput> ||
cute::is_same_v<Schedule, KernelTmaWarpSpecialized> ||
cute::is_same_v<Schedule, KernelTmaWarpSpecializedPingpong> ||
cute::is_same_v<Schedule, KernelTmaWarpSpecializedPingpong> ||
cute::is_same_v<Schedule,
KernelTmaWarpSpecializedPingpongMixedInput> ||
cute::is_same_v<Schedule, KernelTmaWarpSpecializedCooperative> ||
cute::is_same_v<Schedule,
KernelTmaWarpSpecializedCooperativeMixedInput>,
cute::is_same_v<Schedule, KernelTmaWarpSpecializedCooperative>,
"KernelSchedule must be one of the warp specialized policies");
public:
@@ -113,8 +111,7 @@ struct MacheteCollectiveMma {
// For coop schedules we have two warp groups cooperatively issuing wgmma
// instructions so we use 2 atoms along the M dim (one for each warpgroup)
using AtomLayoutMNK = cute::conditional_t<
cute::is_same_v<KernelScheduleType,
KernelTmaWarpSpecializedCooperativeMixedInput>,
cute::is_same_v<KernelScheduleType, KernelTmaWarpSpecializedCooperative>,
Layout<Shape<_2, _1, _1>>, Layout<Shape<_1, _1, _1>>>;
using TiledMma = decltype(cute::make_tiled_mma(