CUDA Templates for Linear Algebra Subroutines and Solvers
Classes | Namespaces
default_gemm_splitk_parallel.h File Reference

Default kernel-level GEMM definitions combine threadblock-scoped matrix multiply-add with the appropriate threadblock-scoped epilogue. More...

#include "cutlass/cutlass.h"
#include "cutlass/gemm/kernel/default_gemm.h"
#include "cutlass/gemm/kernel/gemm_splitk_parallel.h"
Include dependency graph for default_gemm_splitk_parallel.h:
This graph shows which files directly or indirectly include this file:

Go to the source code of this file.


struct  cutlass::gemm::kernel::DefaultGemmSplitKParallel< ElementA_, LayoutA_, kAlignmentA, ElementB_, LayoutB_, kAlignmentB, ElementC_, LayoutC_, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ThreadblockSwizzle, Stages, Operator >



Detailed Description

Note, CUTLASS epilogues universally target row-major outputs. Column-major outputs are accommodated by exchanging A and B operands and assuming transposed layouts. Partial specializations here choose 'device::GemmTransposed' to implement this functionality.