leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 27 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
2025-08-22 18:11:24 -04:00
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
tools/library/include/cutlass/library/arch_mappings.h
+15 -2
View File
@@ -127,12 +127,25 @@ template <typename OperatorClass> struct ArchMap<arch::Sm100, OperatorClass> {
template <> struct ArchMap<arch::Sm100, arch::OpClassTensorOp> {
static int const kMin = 100;
static int const kMax = 101;
#if (__CUDACC_VER_MAJOR__ >= 13)
static int const kMax = 110;
#else
static int const kMax = 103;
#endif // __CUDACC_VER_MAJOR__ >= 13
};
template <typename OperatorClass> struct ArchMap<arch::Sm103, OperatorClass> {
static int const kMin = 103;
static int const kMax = 1024;
};
template <> struct ArchMap<arch::Sm103, arch::OpClassTensorOp> {
static int const kMin = 103;
static int const kMax = 103;
};
template <typename OperatorClass> struct ArchMap<arch::Sm120, OperatorClass> {
static int const kMin = 120;
static int const kMax = 120;
static int const kMax = 121;
};
/////////////////////////////////////////////////////////////////////////////////////////////////