NVIDIA/cutlass
1
0
Fork 0
mirror of https://github.com/NVIDIA/cutlass.git synced 2026-04-19 22:38:56 +00:00
Code Issues Packages Projects Releases Wiki Activity

v4.2 release. (#2587)

Browse Source 
* Fix default cluster callback values to 1 to avoid profiler failure when these values are not set in command line.

* v4.2 release.
This commit is contained in:
Junkai-Wu
2025-08-23 06:11:24 +08:00
committed by GitHub
parent 11cad1f67b
commit a49a78ffef
351 changed files with 28182 additions and 2032 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
include/cute/algorithm/copy.hpp
View File

@@ -254,11 +254,11 @@ copy(AutoVectorizingCopyWithAssumedAlignment<MaxVecBits> const&,
if constexpr (common_elem > 1)
{
constexpr int align_bits = CUTE_STATIC_V(gcd(max_alignment(src), max_alignment(dst), Int<MaxVecBits>{}));
constexpr int vec_bits = gcd(common_elem * sizeof_bits_v<typename SrcEngine::value_type>, align_bits);
constexpr int vec_bits = gcd(common_elem * sizeof_bits_v<typename DstEngine::value_type>, align_bits);
if constexpr ((vec_bits % 8) == 0)
if constexpr ((vec_bits % 8) == 0 && sizeof_bits_v<typename DstEngine::value_type> < Int<vec_bits>{})
{
// If more than one element vectorizes to 8bits or more, then recast and copy
// If more than one element vectorizes to a multiple of 8bits that is larger than the value_type, then recast and copy
using VecType = uint_bit_t<vec_bits>;
// Recast