Cutlass: cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ > Struct Template Reference

Cutlass

CUDA Templates for Linear Algebra Subroutines and Solvers

Template performing matrix multiply-add operation within a thread.

#include <thread_multiply_add.h>

Public Types
typedef Shape< 1, 1, 1, 1 >	InstructionShape
	The shape of the instruction. More...

typedef AccumulatorsPerThread_	AccumulatorsPerThread
	The number of accumulators per thread. More...

typedef ThreadsPerWarp_	ThreadsPerWarp
	The number of threads per warp. More...

typedef ShapeMul< AccumulatorsPerThread, ThreadsPerWarp >::Shape	AccumulatorsPerWarp
	The number of accumulators per warp. More...

typedef ScalarA_	ScalarA
	The type for A. More...

typedef Fragment< ScalarA, AccumulatorsPerThread::kW >	FragmentA
	The fragment for A. More...

typedef ScalarB_	ScalarB
	The type for B. More...

typedef Fragment< ScalarB, AccumulatorsPerThread::kH >	FragmentB
	The fragment for B. More...

typedef ScalarC_	ScalarC
	The type for C and D. More...

typedef Fragment< ScalarC, AccumulatorsPerThread::kH *AccumulatorsPerThread::kW, 16 >	Accumulators
	The accumulators. More...

Public Member Functions
CUTLASS_DEVICE	ThreadMultiplyAdd ()
	Ctor. More...

CUTLASS_DEVICE void	multiply_add (FragmentA const &a, FragmentB const &b, Accumulators const &c, Accumulators &d)
	Multiply : d = a*b + c. More...

Member Typedef Documentation

◆ Accumulators

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

typedef Fragment<ScalarC, AccumulatorsPerThread::kH * AccumulatorsPerThread::kW, 16> cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::Accumulators

◆ AccumulatorsPerThread

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

typedef AccumulatorsPerThread_ cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::AccumulatorsPerThread

◆ AccumulatorsPerWarp

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

typedef ShapeMul<AccumulatorsPerThread, ThreadsPerWarp>::Shape cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::AccumulatorsPerWarp

◆ FragmentA

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

typedef Fragment<ScalarA, AccumulatorsPerThread::kW> cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::FragmentA

◆ FragmentB

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

typedef Fragment<ScalarB, AccumulatorsPerThread::kH> cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::FragmentB

◆ InstructionShape

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

typedef Shape<1, 1, 1, 1> cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::InstructionShape

◆ ScalarA

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

typedef ScalarA_ cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::ScalarA

◆ ScalarB

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

typedef ScalarB_ cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::ScalarB

◆ ScalarC

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

typedef ScalarC_ cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::ScalarC

◆ ThreadsPerWarp

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

typedef ThreadsPerWarp_ cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::ThreadsPerWarp

Constructor & Destructor Documentation

◆ ThreadMultiplyAdd()

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

CUTLASS_DEVICE cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::ThreadMultiplyAdd ( )

inline

Member Function Documentation

◆ multiply_add()

template<typename AccumulatorsPerThread_ , typename ThreadsPerWarp_ , typename ScalarA_ , typename ScalarB_ , typename ScalarC_ >

CUTLASS_DEVICE void cutlass::gemm::ThreadMultiplyAdd< AccumulatorsPerThread_, ThreadsPerWarp_, ScalarA_, ScalarB_, ScalarC_ >::multiply_add	(	FragmentA const &	a,
		FragmentB const &	b,
		Accumulators const &	c,
		Accumulators &	d
	)

inline

The documentation for this struct was generated from the following file:

thread_multiply_add.h