Public Types
using	ThreadblockShape = GemmShape< 128, 256, 32 >

using	WarpShape = GemmShape< 64, 64, 32 >

using	InstructionShape = GemmShape< 16, 8, 8 >

using	EpilogueOutputOp = epilogue::thread::LinearCombination< ElementC, 128/sizeof_bits< ElementC >::value, ElementAccumulator, ElementAccumulator >

using	Operator = typename platform::conditional< (platform::is_same< ElementA, int8_t >::value\|\|platform::is_same< ElementA, int4b_t >::value\|\|platform::is_same< ElementA, uint8_t >::value\|\|platform::is_same< ElementA, uint4b_t >::value), arch::OpMultiplyAddSaturate, arch::OpMultiplyAdd >::type

Static Public Attributes
static int const	kAlignmentA = 128 / sizeof_bits<ElementA>::value

static int const	kAlignmentB = 128 / sizeof_bits<ElementA>::value

static int const	kStages = 2

Member Typedef Documentation

template<typename ElementA , typename ElementB , typename ElementC , typename ElementAccumulator >

using cutlass::gemm::device::DefaultGemmConfiguration< arch::OpClassTensorOp, arch::Sm75, ElementA, ElementB, ElementC, ElementAccumulator >::EpilogueOutputOp = epilogue::thread::LinearCombination< ElementC, 128 / sizeof_bits<ElementC>::value, ElementAccumulator, ElementAccumulator >

template<typename ElementA , typename ElementB , typename ElementC , typename ElementAccumulator >

using cutlass::gemm::device::DefaultGemmConfiguration< arch::OpClassTensorOp, arch::Sm75, ElementA, ElementB, ElementC, ElementAccumulator >::InstructionShape = GemmShape<16, 8, 8>

template<typename ElementA , typename ElementB , typename ElementC , typename ElementAccumulator >

using cutlass::gemm::device::DefaultGemmConfiguration< arch::OpClassTensorOp, arch::Sm75, ElementA, ElementB, ElementC, ElementAccumulator >::Operator = typename platform::conditional< (platform::is_same<ElementA, int8_t>::value || platform::is_same<ElementA, int4b_t>::value || platform::is_same<ElementA, uint8_t>::value || platform::is_same<ElementA, uint4b_t>::value), arch::OpMultiplyAddSaturate, arch::OpMultiplyAdd>::type

template<typename ElementA , typename ElementB , typename ElementC , typename ElementAccumulator >

using cutlass::gemm::device::DefaultGemmConfiguration< arch::OpClassTensorOp, arch::Sm75, ElementA, ElementB, ElementC, ElementAccumulator >::ThreadblockShape = GemmShape<128, 256, 32>

template<typename ElementA , typename ElementB , typename ElementC , typename ElementAccumulator >

using cutlass::gemm::device::DefaultGemmConfiguration< arch::OpClassTensorOp, arch::Sm75, ElementA, ElementB, ElementC, ElementAccumulator >::WarpShape = GemmShape<64, 64, 32>

Member Data Documentation

template<typename ElementA , typename ElementB , typename ElementC , typename ElementAccumulator >

int const cutlass::gemm::device::DefaultGemmConfiguration< arch::OpClassTensorOp, arch::Sm75, ElementA, ElementB, ElementC, ElementAccumulator >::kAlignmentA = 128 / sizeof_bits<ElementA>::value

static

template<typename ElementA , typename ElementB , typename ElementC , typename ElementAccumulator >

int const cutlass::gemm::device::DefaultGemmConfiguration< arch::OpClassTensorOp, arch::Sm75, ElementA, ElementB, ElementC, ElementAccumulator >::kAlignmentB = 128 / sizeof_bits<ElementA>::value

static

template<typename ElementA , typename ElementB , typename ElementC , typename ElementAccumulator >

int const cutlass::gemm::device::DefaultGemmConfiguration< arch::OpClassTensorOp, arch::Sm75, ElementA, ElementB, ElementC, ElementAccumulator >::kStages = 2

static

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