![]() |
CUTLASS
CUDA Templates for Linear Algebra Subroutines and Solvers
|
#include <default_gemm_splitk_parallel.h>
Public Types | |
| using | Default = DefaultGemm< ElementA_, LayoutA_, kAlignmentA, ElementB_, LayoutB_, kAlignmentB, ElementAccumulator, LayoutC_, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ThreadblockSwizzle, Stages, false, Operator > |
| using | Mma = typename Default::Mma |
| Define the matrix multiply operator. More... | |
| using | Epilogue = typename Default::Epilogue |
| Define the epilogue. More... | |
| using | GemmKernel = kernel::GemmSplitKParallel< Mma, Epilogue, ThreadblockSwizzle > |
| Define the kernel-level GEMM operator. More... | |
| using cutlass::gemm::kernel::DefaultGemmSplitKParallel< ElementA_, LayoutA_, kAlignmentA, ElementB_, LayoutB_, kAlignmentB, ElementC_, LayoutC_, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ThreadblockSwizzle, Stages, Operator >::Default = DefaultGemm< ElementA_, LayoutA_, kAlignmentA, ElementB_, LayoutB_, kAlignmentB, ElementAccumulator, LayoutC_, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ThreadblockSwizzle, Stages, false, Operator > |
Define the threadblock-scoped matrix multiply-accumulate using the basic GEMM's mainloop.
| using cutlass::gemm::kernel::DefaultGemmSplitKParallel< ElementA_, LayoutA_, kAlignmentA, ElementB_, LayoutB_, kAlignmentB, ElementC_, LayoutC_, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ThreadblockSwizzle, Stages, Operator >::Epilogue = typename Default::Epilogue |
| using cutlass::gemm::kernel::DefaultGemmSplitKParallel< ElementA_, LayoutA_, kAlignmentA, ElementB_, LayoutB_, kAlignmentB, ElementC_, LayoutC_, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ThreadblockSwizzle, Stages, Operator >::GemmKernel = kernel::GemmSplitKParallel<Mma, Epilogue, ThreadblockSwizzle> |
| using cutlass::gemm::kernel::DefaultGemmSplitKParallel< ElementA_, LayoutA_, kAlignmentA, ElementB_, LayoutB_, kAlignmentB, ElementC_, LayoutC_, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ThreadblockSwizzle, Stages, Operator >::Mma = typename Default::Mma |
1.8.11