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

CUTLASS 3.2.1 (#1113)

Browse Source 
* Updates for 3.2.1 release.

* Minor fix in gemm op profiler for raster order.

* Add scheduler mapping for raster order in the kernels.
This commit is contained in:
ANIKET SHIVAM
2023-09-26 14:24:26 -07:00
committed by GitHub
parent e0aaa3c3b3
commit 90d3b0fb18
428 changed files with 22253 additions and 21762 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
include/cutlass/transform/collective/sm90_wgmma_transpose.hpp
View File

@@ -383,7 +383,7 @@ public:
make_fragment_like(tmp_tCsB)
};
int step = current_step * NumMathWarpGroup;
[[maybe_unused]] int step = current_step * NumMathWarpGroup;
if constexpr (NumMathWarpGroup == 2) {
// For 2 math warpgroup, warp idx4~7 is 1st warp group and 8~9 is 2nd, so decide if 2nd warpgroup need warp idx divide 8.
step += warp_idx / (NumWarpsPerWarpGroup * 2);

12
include/cutlass/transform/threadblock/predicated_tile_access_iterator_params.h
View File

@@ -32,9 +32,21 @@
\brief
*/
/*
Note: CUTLASS 3x increases the host compiler requirements to C++17. However, certain
existing integrations of CUTLASS require C++11 host compilers.
Until this requirement can be lifted, certain headers with this annotation are required
to be remain consistent with C++11 syntax.
C++11 compatibility is enforced by this unit test: `cutlass_test_unit_core_cpp11`.
*/
#pragma once
#include "cutlass/cutlass.h"
#include "cutlass/array.h"
#include "cutlass/layout/matrix.h"
#include "cutlass/layout/pitch_linear.h"
/////////////////////////////////////////////////////////////////////////////////////////////////