Skip to content

GitLab

Commit c44818e7 authored by rocking's avatar rocking
Browse files

Rename DxsReduceAccElementwiseOperation to DxsReduceAccElementwiseOperation

parent 46eca0a1
Hide whitespace changes
Inline Side-by-side
Showing with 103 additions and 99 deletions
include/ck/tensor_operation/gpu/device/device_batched_gemm_reduce_xdl_cshuffle.hpp
View file @ c44818e7
......@@ -22,7 +22,7 @@ template <typename GridwiseGemm,
typename BElementwiseOperation,
typename CElementwiseOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation,
typename DxsReduceAccElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
......@@ -44,7 +44,7 @@ __global__ void
const BElementwiseOperation b_element_op,
const CElementwiseOperation c_element_op,
const DxsInElementwiseOperation dxs_in_element_op,
const DxsAccElementwiseOperation dxs_out_element_op,
const DxsReduceAccElementwiseOperation dxs_out_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
......@@ -126,7 +126,7 @@ template <typename ALayout,
typename CElementwiseOperation,
typename DxsReduceOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation,
typename DxsReduceAccElementwiseOperation,
typename DGlobalMemoryDataOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
......@@ -162,12 +162,13 @@ template <typename ALayout,
index_t CReduceThreadLds2VGprCopySrcDstScalarPerVector_NPerBlock,
index_t CReduceThreadVgpr2GlobalCopySrcDstScalarPerVector_MPerBlock,
LoopScheduler LoopSched = make_default_loop_scheduler()>
struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGlobal,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation>
struct DeviceBatchedGemmReduce_Xdl_CShuffle
: public DeviceGemmReduce<DPtrsGlobal,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
DxsInElementwiseOperation,
DxsReduceAccElementwiseOperation>
{
using DeviceOp = DeviceBatchedGemmReduce_Xdl_CShuffle;
......@@ -527,7 +528,7 @@ struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGloba
CElementwiseOperation,
DxsReduceOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation,
DxsReduceAccElementwiseOperation,
InMemoryDataOperationEnum::Set,
DGlobalMemoryDataOperation,
AGridDesc_AK0_M_AK1,
......@@ -587,7 +588,7 @@ struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGloba
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op,
DxsReduceAccElementwiseOperation dxs_out_element_op,
index_t BatchCount)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
......@@ -645,7 +646,7 @@ struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGloba
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
DxsInElementwiseOperation dxs_in_element_op_;
DxsAccElementwiseOperation dxs_out_element_op_;
DxsReduceAccElementwiseOperation dxs_out_element_op_;
};
// Invoker
......@@ -703,7 +704,7 @@ struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGloba
BElementwiseOperation,
CElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation,
DxsReduceAccElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
......@@ -746,7 +747,7 @@ struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGloba
BElementwiseOperation,
CElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation,
DxsReduceAccElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
......@@ -832,7 +833,7 @@ struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGloba
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op,
DxsReduceAccElementwiseOperation dxs_out_element_op,
index_t BatchCount)
{
return Argument{p_a,
......@@ -856,22 +857,23 @@ struct DeviceBatchedGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGloba
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
DPtrsGlobal p_dxs,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op,
index_t BatchCount) override
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
DPtrsGlobal p_dxs,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsReduceAccElementwiseOperation dxs_out_element_op,
index_t BatchCount) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
......
include/ck/tensor_operation/gpu/device/device_gemm_bias_add_reduce_xdl_cshuffle.hpp
View file @ c44818e7
......@@ -35,7 +35,7 @@ template <typename ALayout,
typename C1ElementwiseOperation,
typename DxsReduceOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation,
typename DxsReduceAccElementwiseOperation,
typename DGlobalMemoryDataOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
......@@ -78,7 +78,7 @@ struct DeviceGemmBiasAddReduce_Xdl_CShuffle
CElementwiseOperation,
C1ElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation>
DxsReduceAccElementwiseOperation>
{
using DeviceOp = DeviceGemmBiasAddReduce_Xdl_CShuffle;
......@@ -399,7 +399,7 @@ struct DeviceGemmBiasAddReduce_Xdl_CShuffle
C1ElementwiseOperation,
DxsReduceOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation,
DxsReduceAccElementwiseOperation,
InMemoryDataOperationEnum::Set,
DGlobalMemoryDataOperation,
AGridDesc_AK0_M_AK1,
......@@ -465,7 +465,7 @@ struct DeviceGemmBiasAddReduce_Xdl_CShuffle
CElementwiseOperation c_element_op,
C1ElementwiseOperation c1_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op)
DxsReduceAccElementwiseOperation dxs_out_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
......@@ -538,7 +538,7 @@ struct DeviceGemmBiasAddReduce_Xdl_CShuffle
CElementwiseOperation c_element_op_;
C1ElementwiseOperation c1_element_op_;
DxsInElementwiseOperation dxs_in_element_op_;
DxsAccElementwiseOperation dxs_out_element_op_;
DxsReduceAccElementwiseOperation dxs_out_element_op_;
};
// Invoker
......@@ -577,7 +577,7 @@ struct DeviceGemmBiasAddReduce_Xdl_CShuffle
CElementwiseOperation,
C1ElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation,
DxsReduceAccElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
......@@ -627,7 +627,7 @@ struct DeviceGemmBiasAddReduce_Xdl_CShuffle
CElementwiseOperation,
C1ElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation,
DxsReduceAccElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
......@@ -713,7 +713,7 @@ struct DeviceGemmBiasAddReduce_Xdl_CShuffle
CElementwiseOperation c_element_op,
C1ElementwiseOperation c1_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op)
DxsReduceAccElementwiseOperation dxs_out_element_op)
{
return Argument{p_a,
p_b,
......@@ -739,26 +739,27 @@ struct DeviceGemmBiasAddReduce_Xdl_CShuffle
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
const void* p_c0,
const void* p_c1,
DPtrsGlobal p_dxs,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
index_t StrideC1,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
C1ElementwiseOperation c1_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op,
index_t /* KBatch */ = 1) override
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
const void* p_c0,
const void* p_c1,
DPtrsGlobal p_dxs,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
index_t StrideC1,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
C1ElementwiseOperation c1_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsReduceAccElementwiseOperation dxs_out_element_op,
index_t /* KBatch */ = 1) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
......
include/ck/tensor_operation/gpu/device/device_gemm_reduce.hpp
View file @ c44818e7
......@@ -11,7 +11,7 @@ template <typename DPtrsGlobal,
typename BElementwiseOperation,
typename CElementwiseOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation>
typename DxsReduceAccElementwiseOperation>
struct DeviceGemmReduce : public BaseOperator
{
virtual std::unique_ptr<BaseArgument>
......@@ -29,7 +29,7 @@ struct DeviceGemmReduce : public BaseOperator
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op,
DxsReduceAccElementwiseOperation dxs_out_element_op,
ck::index_t BatchCount = 1) = 0;
virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
......@@ -40,13 +40,13 @@ template <typename DPtrsGlobal,
typename BElementwiseOperation,
typename CElementwiseOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation>
typename DxsReduceAccElementwiseOperation>
using DeviceGemmReducePtr = std::unique_ptr<DeviceGemmReduce<DPtrsGlobal,
AElementwiseOperation,
BElementwiseOperation,
CElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation>>;
DxsReduceAccElementwiseOperation>>;
template <typename DPtrsGlobal,
typename AElementwiseOperation,
......@@ -54,7 +54,7 @@ template <typename DPtrsGlobal,
typename CElementwiseOperation,
typename C1ElementwiseOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation>
typename DxsReduceAccElementwiseOperation>
struct DeviceGemmBiasAddReduce : public BaseOperator
{
virtual std::unique_ptr<BaseArgument>
......@@ -76,7 +76,7 @@ struct DeviceGemmBiasAddReduce : public BaseOperator
CElementwiseOperation c_element_op,
C1ElementwiseOperation c1_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op,
DxsReduceAccElementwiseOperation dxs_out_element_op,
ck::index_t BatchCount = 1) = 0;
virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
......@@ -88,7 +88,7 @@ template <typename DPtrsGlobal,
typename CElementwiseOperation,
typename C1ElementwiseOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation>
typename DxsReduceAccElementwiseOperation>
using DeviceGemmBiasAddReducePtr =
std::unique_ptr<DeviceGemmBiasAddReduce<DPtrsGlobal,
AElementwiseOperation,
......@@ -96,7 +96,7 @@ using DeviceGemmBiasAddReducePtr =
CElementwiseOperation,
C1ElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation>>;
DxsReduceAccElementwiseOperation>>;
} // namespace device
} // namespace tensor_operation
......
include/ck/tensor_operation/gpu/device/device_gemm_reduce_xdl_cshuffle.hpp
View file @ c44818e7
......@@ -32,7 +32,7 @@ template <typename ALayout,
typename CElementwiseOperation,
typename DxsReduceOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation,
typename DxsReduceAccElementwiseOperation,
typename DGlobalMemoryDataOperation,
GemmSpecialization GemmSpec,
index_t NumGemmKPrefetchStage,
......@@ -73,7 +73,7 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGlobal,
BElementwiseOperation,
CElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation>
DxsReduceAccElementwiseOperation>
{
using DeviceOp = DeviceGemmReduce_Xdl_CShuffle;
......@@ -389,7 +389,7 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGlobal,
CElementwiseOperation,
DxsReduceOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation,
DxsReduceAccElementwiseOperation,
InMemoryDataOperationEnum::Set,
DGlobalMemoryDataOperation,
AGridDesc_AK0_M_AK1,
......@@ -449,7 +449,7 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGlobal,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op)
DxsReduceAccElementwiseOperation dxs_out_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_c_grid_{p_c_grid},
......@@ -498,7 +498,7 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGlobal,
BElementwiseOperation b_element_op_;
CElementwiseOperation c_element_op_;
DxsInElementwiseOperation dxs_in_element_op_;
DxsAccElementwiseOperation dxs_out_element_op_;
DxsReduceAccElementwiseOperation dxs_out_element_op_;
};
// Invoker
......@@ -554,7 +554,7 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGlobal,
BElementwiseOperation,
CElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation,
DxsReduceAccElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
......@@ -594,7 +594,7 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGlobal,
BElementwiseOperation,
CElementwiseOperation,
DxsInElementwiseOperation,
DxsAccElementwiseOperation,
DxsReduceAccElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
typename GridwiseGemm::CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
......@@ -669,7 +669,7 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGlobal,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op)
DxsReduceAccElementwiseOperation dxs_out_element_op)
{
return Argument{p_a,
p_b,
......@@ -691,22 +691,23 @@ struct DeviceGemmReduce_Xdl_CShuffle : public DeviceGemmReduce<DPtrsGlobal,
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
DPtrsGlobal p_dxs,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsAccElementwiseOperation dxs_out_element_op,
index_t /* KBatch */ = 1) override
std::unique_ptr<BaseArgument>
MakeArgumentPointer(const void* p_a,
const void* p_b,
void* p_c,
DPtrsGlobal p_dxs,
index_t MRaw,
index_t NRaw,
index_t KRaw,
index_t StrideA,
index_t StrideB,
index_t StrideC,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
CElementwiseOperation c_element_op,
DxsInElementwiseOperation dxs_in_element_op,
DxsReduceAccElementwiseOperation dxs_out_element_op,
index_t /* KBatch */ = 1) override
{
return std::make_unique<Argument>(static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
......
include/ck/tensor_operation/gpu/grid/gridwise_gemm_bias_add_reduce_xdl_cshuffle_v1.hpp
View file @ c44818e7
......@@ -24,7 +24,7 @@ template <typename GridwiseGemm,
typename CElementwiseOperation,
typename C1ElementwiseOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation,
typename DxsReduceAccElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
......@@ -49,7 +49,7 @@ __global__ void
const CElementwiseOperation c_element_op,
const C1ElementwiseOperation c1_element_op,
const DxsInElementwiseOperation dxs_in_element_op,
const DxsAccElementwiseOperation dxs_out_element_op,
const DxsReduceAccElementwiseOperation dxs_out_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
......@@ -121,7 +121,7 @@ template <typename FloatAB,
typename C1ElementwiseOperation,
typename DxsReduceOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation,
typename DxsReduceAccElementwiseOperation,
InMemoryDataOperationEnum CGlobalMemoryDataOperation,
typename DGlobalMemoryDataOperation,
typename AGridDesc_AK0_M_AK1,
......@@ -366,7 +366,7 @@ struct GridwiseGemmBiasAddReduce_k0mk1_k0nk1_mn_xdl_cshuffle_v1
const CElementwiseOperation& c_element_op,
const C1ElementwiseOperation& c1_element_op,
const DxsInElementwiseOperation& dxs_in_element_op,
const DxsAccElementwiseOperation& dxs_out_element_op,
const DxsReduceAccElementwiseOperation& dxs_out_element_op,
const AGridDesc_AK0_M_AK1& a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1& b_grid_desc_bk0_n_bk1,
const CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock&
......
include/ck/tensor_operation/gpu/grid/gridwise_gemm_reduce_xdl_cshuffle_v1.hpp
View file @ c44818e7
......@@ -21,7 +21,7 @@ template <typename GridwiseGemm,
typename BElementwiseOperation,
typename CElementwiseOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation,
typename DxsReduceAccElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
......@@ -41,7 +41,7 @@ __global__ void
const BElementwiseOperation b_element_op,
const CElementwiseOperation c_element_op,
const DxsInElementwiseOperation dxs_in_element_op,
const DxsAccElementwiseOperation dxs_out_element_op,
const DxsReduceAccElementwiseOperation dxs_out_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
......@@ -96,7 +96,7 @@ template <typename FloatAB,
typename CElementwiseOperation,
typename DxsReduceOperation,
typename DxsInElementwiseOperation,
typename DxsAccElementwiseOperation,
typename DxsReduceAccElementwiseOperation,
InMemoryDataOperationEnum CGlobalMemoryDataOperation,
typename DGlobalMemoryDataOperation,
typename AGridDesc_AK0_M_AK1,
......@@ -329,7 +329,7 @@ struct GridwiseGemmReduce_k0mk1_k0nk1_mn_xdl_cshuffle_v1
const BElementwiseOperation& b_element_op,
const CElementwiseOperation& c_element_op,
const DxsInElementwiseOperation& dxs_in_element_op,
const DxsAccElementwiseOperation& dxs_out_element_op,
const DxsReduceAccElementwiseOperation& dxs_out_element_op,
const AGridDesc_AK0_M_AK1& a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1& b_grid_desc_bk0_n_bk1,
const CGridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock&
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment