leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 28 Packages Projects Releases Wiki Activity

CUTLASS 2.6 (#298)

Browse Source 
CUTLASS 2.6
This commit is contained in:
Manish Gupta
2021-07-23 00:40:53 -04:00
committed by GitHub
parent 6c29fe20ba
commit e5d51840e8
308 changed files with 32408 additions and 4722 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/cutlass/cutlass.h
-20
View File
@@ -141,26 +141,6 @@ static const int NUM_THREADS_PER_HALF_WARP = NUM_THREADS_PER_WARP / 2;
static const int NUM_THREADS_PER_QUAD = 4;
static const int NUM_THREADS_PER_QUAD_PAIR = NUM_THREADS_PER_QUAD * 2;
#if defined(__NVCC__) || (defined(__clang__) && defined(__CUDA__))
/// Computes laneId within a warp
CUTLASS_DEVICE
int LaneId() {
int ret;
asm ("mov.u32 %0, %%laneid;" : "=r"(ret) : );
return ret;
}
/// Computes SM number the thread is running on
CUTLASS_DEVICE
int SmId() {
int ret;
asm ("mov.u32 %0, %%smid;" : "=r"(ret) : );
return ret;
}
#endif
////////////////////////////////////////////////////////////////////////////////////////////////////
} // namespace cutlass