NVIDIA/cutlass
1
0
Fork 0
mirror of https://github.com/NVIDIA/cutlass.git synced 2026-05-11 17:00:05 +00:00
Code Issues Packages Projects Releases Wiki Activity

Update generator.py (#192)

Browse Source
This commit is contained in:
Haicheng Wu
2021-03-02 15:21:48 -05:00
committed by GitHub
parent 8f09b82b12
commit 34a42e5620
1 changed files with 8 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
tools/library/scripts/generator.py
View File

@@ -1712,7 +1712,14 @@ def GenerateSM80_TensorOp_16864_TN(manifest, args):
operations += CreateGemmOperator(manifest, layouts, tile_descriptions, \
data_type_mixed, alignment_constraints, None, EpilogueFunctor.LinearCombinationClamp)
conv_layout = (LayoutType.TensorNHWC, LayoutType.TensorNHWC, LayoutType.TensorNHWC)
CreateConv2dOperator(manifest, conv_layout, tile_descriptions,
data_type, 32, [ConvKind.Fprop], EpilogueFunctor.LinearCombinationClamp)
operations += CreateConv2dOperator(manifest, conv_layout, tile_descriptions,
data_type_mixed, 32, [ConvKind.Fprop], EpilogueFunctor.LinearCombinationClamp)
for op in operations:
if op.tile_description.threadblock_shape[1] >= 128:
op.C.alignment = 8