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
test/unit/gemm/device/simt_sm50.py
+9 -8
View File
@@ -45,14 +45,15 @@ warpShapeMin = 8*8
threadblockEdgeMax = 256
# char, type bits/elem, max tile, L0 threadblock tiles
# char, type bits/elem, max tile, L0 threadblock tiles
precisions = [
["c", "cutlass::complex<float>", 64, 64*128, [ [ 64, 128], [ 64, 32] ] ],
["d", "double", 64, 64*64, [ [ 64, 64], [ 32, 32] ] ],
["h", "cutlass::half_t", 16, 128*256, [ [256, 128], [ 64, 128], [ 64, 32] ] ],
["i", "int", 32, 128*128, [ [128, 64], [ 16, 32] ] ],
["s", "float", 32, 128*128, [ [128, 256], [128, 128], [ 64, 64] ] ],
["z", "cutlass::complex<double>", 128, 64*64, [ [ 32, 64], [ 16, 32] ] ],
["c", "cutlass::complex<float>", 64, 64*128, [ [ 64, 128], [ 64, 32] ] ],
["q", "cutlass::Quaternion<float>", 64, 64*128, [ [ 64, 128], [ 64, 32] ] ],
["d", "double", 64, 64*64, [ [ 64, 64], [ 32, 32] ] ],
["h", "cutlass::half_t", 16, 128*256, [ [256, 128], [ 64, 128], [ 64, 32] ] ],
["i", "int", 32, 128*128, [ [128, 64], [ 16, 32] ] ],
["s", "float", 32, 128*128, [ [128, 256], [128, 128], [ 64, 64] ] ],
["z", "cutlass::complex<double>", 128, 64*64, [ [ 32, 64], [ 16, 32] ] ],
]
# L1 will have a single kernel for every unique shape
# L2 will have everything else
@@ -313,7 +314,7 @@ for precision in precisions:
" cutlass::arch::Sm50,\n"
" ThreadblockShape, WarpShape, InstructionShape,\n"
" EpilogueOutputOp,\n"
" cutlass::gemm::threadblock::GemmIdentityThreadblockSwizzle,\n"
" cutlass::gemm::threadblock::GemmIdentityThreadblockSwizzle<>,\n"
" 2 // Stages\n"
" >;\n" % (
"Column" if columnMajorA else "Row",