leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 24 Packages Projects Releases Wiki Activity

v4.4 tag release update. (#3032)

Browse Source
This commit is contained in:
Junkai-Wu
2026-02-13 23:27:58 -05:00
committed by GitHub
parent 01687cfba1
commit d4bbf728ca
140 changed files with 41622 additions and 3689 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/cutlass/conv/kernel/sm100_implicit_gemm_tma_warpspecialized.hpp
+5 -1
View File
@@ -276,7 +276,11 @@ public:
static constexpr int MaxClusterSize = 16;
implementable &= size(args.hw_info.cluster_shape) <= MaxClusterSize;
implementable &= size(args.hw_info.cluster_shape_fallback) <= MaxClusterSize;
implementable &= cutlass::detail::preferred_cluster_can_implement<AtomThrShapeMNK>(args.hw_info.cluster_shape, args.hw_info.cluster_shape_fallback);
// Early return if cluster shape validation failed to avoid division by zero below
if (not cutlass::detail::preferred_cluster_can_implement<AtomThrShapeMNK>(args.hw_info.cluster_shape, args.hw_info.cluster_shape_fallback)) {
CUTLASS_TRACE_HOST(" CAN IMPLEMENT: Invalid dynamic cluster shape\n");
return false;
}
}
auto cluster_shape = cutlass::detail::select_cluster_shape(ClusterShape{}, args.hw_info.cluster_shape);