CUTLASS 3.1 (#915)

Co-authored-by: Aniket Shivam <ashivam@nvidia.com>
This commit is contained in:
ANIKET SHIVAM
2023-04-14 23:19:34 -04:00
committed by GitHub
co-authored by Aniket Shivam
parent 9b8166e3f0
commit d572cc1aab
482 changed files with 37175 additions and 16410 deletions
@@ -117,7 +117,7 @@ struct DefaultGemmWithReduction {
SharedMemoryClearOption::kClearLastStage
>::GemmKernel;
// Replace epilogue
// Define epilogue
using Epilogue = typename cutlass::epilogue::threadblock::DefaultEpilogueWithReductionTensorOp<
typename GemmBase::Epilogue::Shape,
typename GemmBase::Epilogue::WarpMmaOperator,
@@ -218,7 +218,7 @@ struct DefaultGemmWithReduction<
Operator
>::GemmKernel;
// Replace epilogue
// Define epilogue
using Epilogue = typename cutlass::epilogue::threadblock::DefaultEpilogueWithReductionVoltaTensorOp<
typename GemmBase::Epilogue::Shape,
typename GemmBase::Epilogue::WarpMmaOperator,