leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 26 Packages Projects Releases Wiki Activity

CUTLASS 2.3 initial commit (#134)

Browse Source 
CUTLASS 2.3 adds GEMMs targeting Sparse Tensor Cores on the NVIDIA Ampere Architecture, fast SGEMM, and small matrix classes, bug fixes, and performance enhancements.
This commit is contained in:
Andrew Kerr
2020-09-23 14:00:58 -07:00
committed by GitHub
parent 4dac7490e6
commit c53f3339bb
209 changed files with 46922 additions and 1677 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/cutlass/gemm/kernel/gemm_splitk_parallel.h
+4 -2
View File
@@ -128,7 +128,8 @@ struct GemmSplitKParallel {
// Compute threadblock location
ThreadblockSwizzle threadblock_swizzle;
cutlass::gemm::GemmCoord threadblock_tile_offset = threadblock_swizzle.get_tile_offset();
cutlass::gemm::GemmCoord threadblock_tile_offset =
threadblock_swizzle.get_tile_offset(params.grid_tiled_shape);
// Early exit if CTA is out of range
if (params.grid_tiled_shape.m() <= threadblock_tile_offset.m() ||
@@ -205,7 +206,8 @@ struct GemmSplitKParallel {
// Masked tile iterators constructed from members
//
threadblock_tile_offset = threadblock_swizzle.get_tile_offset();
threadblock_tile_offset =
threadblock_swizzle.get_tile_offset(params.grid_tiled_shape);
//assume identity swizzle
MatrixCoord threadblock_offset(