Skip to content

GitLab

Unverified Commit 5903efe7 authored by arai713's avatar arai713 Committed by GitHub
Browse files

Merge branch 'develop' into transpose_5d

parents 2100ea4b e1fa0091
Hide whitespace changes
Inline Side-by-side
Showing with 643 additions and 531 deletions
library/include/ck/library/tensor_operation_instance/gpu/quantization/grouped_convolution_bias_forward_perchannel_quantization.hpp
View file @ 5903efe7
...@@ -7,7 +7,7 @@ ...@@ -7,7 +7,7 @@
#include "ck/ck.hpp" #include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp" #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_d.hpp" #include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_abd.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp" #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp" #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
...@@ -20,96 +20,96 @@ namespace instance { ...@@ -20,96 +20,96 @@ namespace instance {
// grouped conv2d forward, NHWGC/GKYXC/NHWGK // grouped conv2d forward, NHWGC/GKYXC/NHWGK
void add_device_conv2d_dl_bias_perchannel_quantization_int8_instances( void add_device_conv2d_dl_bias_perchannel_quantization_int8_instances(
std::vector< std::vector<
std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_GK_Tuple, GK_GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
I32_F32_Tuple, I32_F32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Add_Activation_Mul2_Clamp<PassThrough>>>>& Add_Activation_Mul2_Clamp<PassThrough>>>>&
instances); instances);
void add_device_conv2d_dl_bias_relu_perchannel_quantization_int8_instances( void add_device_conv2d_dl_bias_relu_perchannel_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_GK_Tuple, GK_GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
I32_F32_Tuple, I32_F32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Add_Activation_Mul2_Clamp<Relu>>>>& Add_Activation_Mul2_Clamp<Relu>>>>&
instances); instances);
void add_device_conv2d_dl_bias_tanh_perchannel_quantization_int8_instances( void add_device_conv2d_dl_bias_tanh_perchannel_quantization_int8_instances(
std::vector< std::vector<
std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_GK_Tuple, GK_GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
I32_F32_Tuple, I32_F32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Add_Mul2_Activation_Mul_Clamp<TanH>>>>& Add_Mul2_Activation_Mul_Clamp<TanH>>>>&
instances); instances);
#endif #endif
void add_device_conv2d_xdl_bias_perchannel_quantization_int8_instances( void add_device_conv2d_xdl_bias_perchannel_quantization_int8_instances(
std::vector< std::vector<
std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_GK_Tuple, GK_GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
I32_F32_Tuple, I32_F32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Add_Activation_Mul2_Clamp<PassThrough>>>>& Add_Activation_Mul2_Clamp<PassThrough>>>>&
instances); instances);
void add_device_conv2d_xdl_bias_relu_perchannel_quantization_int8_instances( void add_device_conv2d_xdl_bias_relu_perchannel_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_GK_Tuple, GK_GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
I32_F32_Tuple, I32_F32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Add_Activation_Mul2_Clamp<Relu>>>>& Add_Activation_Mul2_Clamp<Relu>>>>&
instances); instances);
void add_device_conv2d_xdl_bias_tanh_perchannel_quantization_int8_instances( void add_device_conv2d_xdl_bias_tanh_perchannel_quantization_int8_instances(
std::vector< std::vector<
std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_GK_Tuple, GK_GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
I32_F32_Tuple, I32_F32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Add_Mul2_Activation_Mul_Clamp<TanH>>>>& Add_Mul2_Activation_Mul_Clamp<TanH>>>>&
instances); instances);
// piecewise activation function // piecewise activation function
...@@ -123,7 +123,7 @@ template <ck::index_t NumDimSpatial, ...@@ -123,7 +123,7 @@ template <ck::index_t NumDimSpatial,
typename DsDataType, typename DsDataType,
typename OutDataType, typename OutDataType,
typename Activation> typename Activation>
struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD< struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
NumDimSpatial, NumDimSpatial,
InLayout, InLayout,
WeiLayout, WeiLayout,
...@@ -137,18 +137,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe ...@@ -137,18 +137,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Add_Activation_Mul2_Clamp<Activation>>> Add_Activation_Mul2_Clamp<Activation>>>
{ {
using DeviceOp = DeviceGroupedConvFwdMultipleD<NumDimSpatial, using DeviceOp =
InLayout, DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
WeiLayout, InLayout,
DsLayout, WeiLayout,
OutLayout, DsLayout,
InDataType, OutLayout,
WeiDataType, InDataType,
DsDataType, WeiDataType,
OutDataType, DsDataType,
ck::tensor_operation::element_wise::PassThrough, OutDataType,
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Add_Activation_Mul2_Clamp<Activation>>; ck::tensor_operation::element_wise::PassThrough,
Add_Activation_Mul2_Clamp<Activation>>;
static auto GetInstances() static auto GetInstances()
{ {
...@@ -193,7 +194,7 @@ template <ck::index_t NumDimSpatial, ...@@ -193,7 +194,7 @@ template <ck::index_t NumDimSpatial,
typename DsDataType, typename DsDataType,
typename OutDataType, typename OutDataType,
typename Activation> typename Activation>
struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD< struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
NumDimSpatial, NumDimSpatial,
InLayout, InLayout,
WeiLayout, WeiLayout,
...@@ -207,18 +208,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe ...@@ -207,18 +208,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Add_Mul2_Activation_Mul_Clamp<Activation>>> Add_Mul2_Activation_Mul_Clamp<Activation>>>
{ {
using DeviceOp = DeviceGroupedConvFwdMultipleD<NumDimSpatial, using DeviceOp =
InLayout, DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
WeiLayout, InLayout,
DsLayout, WeiLayout,
OutLayout, DsLayout,
InDataType, OutLayout,
WeiDataType, InDataType,
DsDataType, WeiDataType,
OutDataType, DsDataType,
ck::tensor_operation::element_wise::PassThrough, OutDataType,
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Add_Mul2_Activation_Mul_Clamp<Activation>>; ck::tensor_operation::element_wise::PassThrough,
Add_Mul2_Activation_Mul_Clamp<Activation>>;
static auto GetInstances() static auto GetInstances()
{ {
......
library/include/ck/library/tensor_operation_instance/gpu/quantization/grouped_convolution_bias_forward_perlayer_quantization.hpp
View file @ 5903efe7
...@@ -7,7 +7,7 @@ ...@@ -7,7 +7,7 @@
#include "ck/ck.hpp" #include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp" #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_d.hpp" #include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_abd.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp" #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp" #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
...@@ -20,94 +20,96 @@ namespace instance { ...@@ -20,94 +20,96 @@ namespace instance {
// grouped conv2d forward, NHWGC/GKYXC/NHWGK // grouped conv2d forward, NHWGC/GKYXC/NHWGK
void add_device_conv2d_dl_bias_perlayer_quantization_int8_instances( void add_device_conv2d_dl_bias_perlayer_quantization_int8_instances(
std::vector< std::vector<
std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_Tuple, GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
I32_Tuple, I32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Add_Activation_Mul_Clamp<PassThrough>>>>& Add_Activation_Mul_Clamp<PassThrough>>>>&
instances); instances);
void add_device_conv2d_dl_bias_relu_perlayer_quantization_int8_instances( void add_device_conv2d_dl_bias_relu_perlayer_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_Tuple, GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
I32_Tuple, I32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Add_Activation_Mul_Clamp<Relu>>>>& Add_Activation_Mul_Clamp<Relu>>>>&
instances); instances);
void add_device_conv2d_dl_bias_tanh_perlayer_quantization_int8_instances( void add_device_conv2d_dl_bias_tanh_perlayer_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<
NHWGC, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GKYXC, NHWGC,
GK_Tuple, GKYXC,
NHWGK, GK_Tuple,
int8_t, NHWGK,
int8_t, int8_t,
I32_Tuple, int8_t,
int8_t, I32_Tuple,
PassThrough, int8_t,
PassThrough, PassThrough,
Add_Mul_Activation_Mul_Clamp<TanH>>>>& PassThrough,
Add_Mul_Activation_Mul_Clamp<TanH>>>>&
instances); instances);
#endif #endif
void add_device_conv2d_xdl_bias_perlayer_quantization_int8_instances( void add_device_conv2d_xdl_bias_perlayer_quantization_int8_instances(
std::vector< std::vector<
std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_Tuple, GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
I32_Tuple, I32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Add_Activation_Mul_Clamp<PassThrough>>>>& Add_Activation_Mul_Clamp<PassThrough>>>>&
instances); instances);
void add_device_conv2d_xdl_bias_relu_perlayer_quantization_int8_instances( void add_device_conv2d_xdl_bias_relu_perlayer_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_Tuple, GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
I32_Tuple, I32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Add_Activation_Mul_Clamp<Relu>>>>& Add_Activation_Mul_Clamp<Relu>>>>&
instances); instances);
void add_device_conv2d_xdl_bias_tanh_perlayer_quantization_int8_instances( void add_device_conv2d_xdl_bias_tanh_perlayer_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<
NHWGC, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GKYXC, NHWGC,
GK_Tuple, GKYXC,
NHWGK, GK_Tuple,
int8_t, NHWGK,
int8_t, int8_t,
I32_Tuple, int8_t,
int8_t, I32_Tuple,
PassThrough, int8_t,
PassThrough, PassThrough,
Add_Mul_Activation_Mul_Clamp<TanH>>>>& PassThrough,
Add_Mul_Activation_Mul_Clamp<TanH>>>>&
instances); instances);
// piecewise activation function // piecewise activation function
...@@ -121,7 +123,7 @@ template <ck::index_t NumDimSpatial, ...@@ -121,7 +123,7 @@ template <ck::index_t NumDimSpatial,
typename DsDataType, typename DsDataType,
typename OutDataType, typename OutDataType,
typename Activation> typename Activation>
struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD< struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
NumDimSpatial, NumDimSpatial,
InLayout, InLayout,
WeiLayout, WeiLayout,
...@@ -135,18 +137,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe ...@@ -135,18 +137,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Add_Activation_Mul_Clamp<Activation>>> Add_Activation_Mul_Clamp<Activation>>>
{ {
using DeviceOp = DeviceGroupedConvFwdMultipleD<NumDimSpatial, using DeviceOp =
InLayout, DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
WeiLayout, InLayout,
DsLayout, WeiLayout,
OutLayout, DsLayout,
InDataType, OutLayout,
WeiDataType, InDataType,
DsDataType, WeiDataType,
OutDataType, DsDataType,
ck::tensor_operation::element_wise::PassThrough, OutDataType,
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Add_Activation_Mul_Clamp<Activation>>; ck::tensor_operation::element_wise::PassThrough,
Add_Activation_Mul_Clamp<Activation>>;
static auto GetInstances() static auto GetInstances()
{ {
...@@ -191,7 +194,7 @@ template <ck::index_t NumDimSpatial, ...@@ -191,7 +194,7 @@ template <ck::index_t NumDimSpatial,
typename DsDataType, typename DsDataType,
typename OutDataType, typename OutDataType,
typename Activation> typename Activation>
struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD< struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
NumDimSpatial, NumDimSpatial,
InLayout, InLayout,
WeiLayout, WeiLayout,
...@@ -205,18 +208,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe ...@@ -205,18 +208,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Add_Mul_Activation_Mul_Clamp<Activation>>> Add_Mul_Activation_Mul_Clamp<Activation>>>
{ {
using DeviceOp = DeviceGroupedConvFwdMultipleD<NumDimSpatial, using DeviceOp =
InLayout, DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
WeiLayout, InLayout,
DsLayout, WeiLayout,
OutLayout, DsLayout,
InDataType, OutLayout,
WeiDataType, InDataType,
DsDataType, WeiDataType,
OutDataType, DsDataType,
ck::tensor_operation::element_wise::PassThrough, OutDataType,
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Add_Mul_Activation_Mul_Clamp<Activation>>; ck::tensor_operation::element_wise::PassThrough,
Add_Mul_Activation_Mul_Clamp<Activation>>;
static auto GetInstances() static auto GetInstances()
{ {
......
library/include/ck/library/tensor_operation_instance/gpu/quantization/grouped_convolution_forward_perchannel_quantization.hpp
View file @ 5903efe7
...@@ -7,7 +7,7 @@ ...@@ -7,7 +7,7 @@
#include "ck/ck.hpp" #include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp" #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_d.hpp" #include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_abd.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp" #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp" #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
...@@ -19,63 +19,65 @@ namespace instance { ...@@ -19,63 +19,65 @@ namespace instance {
#ifdef DL_KERNELS #ifdef DL_KERNELS
// grouped conv2d forward, NHWGC/GKYXC/NHWGK // grouped conv2d forward, NHWGC/GKYXC/NHWGK
void add_device_conv2d_dl_perchannel_quantization_int8_instances( void add_device_conv2d_dl_perchannel_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<
NHWGC, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GKYXC, NHWGC,
GK_Tuple, GKYXC,
NHWGK, GK_Tuple,
int8_t, NHWGK,
int8_t, int8_t,
F32_Tuple, int8_t,
int8_t, F32_Tuple,
PassThrough, int8_t,
PassThrough, PassThrough,
Activation_Mul2_Clamp<PassThrough>>>>& PassThrough,
Activation_Mul2_Clamp<PassThrough>>>>&
instances); instances);
void add_device_conv2d_dl_relu_perchannel_quantization_int8_instances( void add_device_conv2d_dl_relu_perchannel_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_Tuple, GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
F32_Tuple, F32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Activation_Mul2_Clamp<Relu>>>>& Activation_Mul2_Clamp<Relu>>>>&
instances); instances);
#endif #endif
void add_device_conv2d_xdl_perchannel_quantization_int8_instances( void add_device_conv2d_xdl_perchannel_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<
NHWGC, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GKYXC, NHWGC,
GK_Tuple, GKYXC,
NHWGK, GK_Tuple,
int8_t, NHWGK,
int8_t, int8_t,
F32_Tuple, int8_t,
int8_t, F32_Tuple,
PassThrough, int8_t,
PassThrough, PassThrough,
Activation_Mul2_Clamp<PassThrough>>>>& PassThrough,
Activation_Mul2_Clamp<PassThrough>>>>&
instances); instances);
void add_device_conv2d_xdl_relu_perchannel_quantization_int8_instances( void add_device_conv2d_xdl_relu_perchannel_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
GK_Tuple, GK_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
F32_Tuple, F32_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Activation_Mul2_Clamp<Relu>>>>& Activation_Mul2_Clamp<Relu>>>>&
instances); instances);
template <ck::index_t NumDimSpatial, template <ck::index_t NumDimSpatial,
...@@ -88,7 +90,7 @@ template <ck::index_t NumDimSpatial, ...@@ -88,7 +90,7 @@ template <ck::index_t NumDimSpatial,
typename DsDataType, typename DsDataType,
typename OutDataType, typename OutDataType,
typename Activation> typename Activation>
struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD< struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
NumDimSpatial, NumDimSpatial,
InLayout, InLayout,
WeiLayout, WeiLayout,
...@@ -102,18 +104,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe ...@@ -102,18 +104,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Activation_Mul2_Clamp<Activation>>> Activation_Mul2_Clamp<Activation>>>
{ {
using DeviceOp = DeviceGroupedConvFwdMultipleD<NumDimSpatial, using DeviceOp =
InLayout, DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
WeiLayout, InLayout,
GK_Tuple, WeiLayout,
OutLayout, GK_Tuple,
InDataType, OutLayout,
WeiDataType, InDataType,
F32_Tuple, WeiDataType,
OutDataType, F32_Tuple,
ck::tensor_operation::element_wise::PassThrough, OutDataType,
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Activation_Mul2_Clamp<Activation>>; ck::tensor_operation::element_wise::PassThrough,
Activation_Mul2_Clamp<Activation>>;
static auto GetInstances() static auto GetInstances()
{ {
......
library/include/ck/library/tensor_operation_instance/gpu/quantization/grouped_convolution_forward_perlayer_quantization.hpp
View file @ 5903efe7
...@@ -7,7 +7,7 @@ ...@@ -7,7 +7,7 @@
#include "ck/ck.hpp" #include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp" #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_d.hpp" #include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_abd.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp" #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp" #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
...@@ -19,63 +19,65 @@ namespace instance { ...@@ -19,63 +19,65 @@ namespace instance {
#ifdef DL_KERNELS #ifdef DL_KERNELS
// grouped conv2d forward, NHWGC/GKYXC/NHWGK // grouped conv2d forward, NHWGC/GKYXC/NHWGK
void add_device_conv2d_dl_perlayer_quantization_int8_instances( void add_device_conv2d_dl_perlayer_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<
NHWGC, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GKYXC, NHWGC,
Empty_Tuple, GKYXC,
NHWGK, Empty_Tuple,
int8_t, NHWGK,
int8_t, int8_t,
Empty_Tuple, int8_t,
int8_t, Empty_Tuple,
PassThrough, int8_t,
PassThrough, PassThrough,
Activation_Mul_Clamp<PassThrough>>>>& PassThrough,
Activation_Mul_Clamp<PassThrough>>>>&
instances); instances);
void add_device_conv2d_dl_relu_perlayer_quantization_int8_instances( void add_device_conv2d_dl_relu_perlayer_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
Empty_Tuple, Empty_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Activation_Mul_Clamp<Relu>>>>& Activation_Mul_Clamp<Relu>>>>&
instances); instances);
#endif #endif
void add_device_conv2d_xdl_perlayer_quantization_int8_instances( void add_device_conv2d_xdl_perlayer_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<
NHWGC, std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GKYXC, NHWGC,
Empty_Tuple, GKYXC,
NHWGK, Empty_Tuple,
int8_t, NHWGK,
int8_t, int8_t,
Empty_Tuple, int8_t,
int8_t, Empty_Tuple,
PassThrough, int8_t,
PassThrough, PassThrough,
Activation_Mul_Clamp<PassThrough>>>>& PassThrough,
Activation_Mul_Clamp<PassThrough>>>>&
instances); instances);
void add_device_conv2d_xdl_relu_perlayer_quantization_int8_instances( void add_device_conv2d_xdl_relu_perlayer_quantization_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
NHWGK, NHWGK,
int8_t, int8_t,
int8_t, int8_t,
Empty_Tuple, Empty_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
Activation_Mul_Clamp<Relu>>>>& Activation_Mul_Clamp<Relu>>>>&
instances); instances);
template <ck::index_t NumDimSpatial, template <ck::index_t NumDimSpatial,
...@@ -86,7 +88,7 @@ template <ck::index_t NumDimSpatial, ...@@ -86,7 +88,7 @@ template <ck::index_t NumDimSpatial,
typename WeiDataType, typename WeiDataType,
typename OutDataType, typename OutDataType,
typename Activation> typename Activation>
struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD< struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
NumDimSpatial, NumDimSpatial,
InLayout, InLayout,
WeiLayout, WeiLayout,
...@@ -100,18 +102,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe ...@@ -100,18 +102,19 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Activation_Mul_Clamp<Activation>>> Activation_Mul_Clamp<Activation>>>
{ {
using DeviceOp = DeviceGroupedConvFwdMultipleD<NumDimSpatial, using DeviceOp =
InLayout, DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
WeiLayout, InLayout,
Empty_Tuple, WeiLayout,
OutLayout, Empty_Tuple,
InDataType, OutLayout,
WeiDataType, InDataType,
Empty_Tuple, WeiDataType,
OutDataType, Empty_Tuple,
ck::tensor_operation::element_wise::PassThrough, OutDataType,
ck::tensor_operation::element_wise::PassThrough, ck::tensor_operation::element_wise::PassThrough,
Activation_Mul_Clamp<Activation>>; ck::tensor_operation::element_wise::PassThrough,
Activation_Mul_Clamp<Activation>>;
static auto GetInstances() static auto GetInstances()
{ {
......
library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instance.cpp
View file @ 5903efe7
...@@ -7,6 +7,8 @@ ...@@ -7,6 +7,8 @@
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp" #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp" #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp" #include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp" #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck { namespace ck {
...@@ -27,29 +29,75 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough; ...@@ -27,29 +29,75 @@ using PassThrough = ck::tensor_operation::element_wise::PassThrough;
static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default; static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
static constexpr auto MNPadding = ck::tensor_operation::device::GemmSpecialization::MNPadding;
static constexpr auto MNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
// Compilation parameters for a[m, k] * b[k, n] = c[m, n] // Compilation parameters for a[m, k] * b[k, n] = c[m, n]
template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances = std::tuple< using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances = std::tuple<
// clang-format off // clang-format off
//#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle| A| B| C| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle| A| B| C| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| LoopScheduler| Pipeline|
//#####################| | | | Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| //#####################| | | | Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| | |
//#####################| | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| //#####################| | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| | |
//#####################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | //#####################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 256, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8>, // pipeline v1, 1 wave
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 128, 256, 32, 8, 2, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 2, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 128, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 128, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 128, 64, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 4>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 4>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 64, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 128, 64, 32, 8, 2, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<16,16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 2, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<16,16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 64, 128, 32, 8, 2, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8> DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 2, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>
#if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
// pipeline v1, 2 waves
,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 2, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 2, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<16,16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 2, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>
#endif
#if CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES
// pipeline v2, 1 wave
,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 2, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 2, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 2, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 2, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<16,16, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 2, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<8, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 2, 0, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Row, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>
#endif
// clang-format on // clang-format on
>; >;
...@@ -58,8 +106,14 @@ void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances( ...@@ -58,8 +106,14 @@ void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances(
DeviceGemm<Row, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>& DeviceGemm<Row, Row, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
instances) instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(
device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances{}); instances, device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances<GemmDefault>{});
add_device_operation_instances(
instances, device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances<MNPadding>{});
add_device_operation_instances(
instances, device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_kn_mn_instances<MNKPadding>{});
} }
} // namespace instance } // namespace instance
......
library/src/tensor_operation_instance/gpu/gemm/device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instance.cpp
View file @ 5903efe7
...@@ -7,6 +7,8 @@ ...@@ -7,6 +7,8 @@
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp" #include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp" #include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp" #include "ck/tensor_operation/gpu/device/impl/device_gemm_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp" #include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
namespace ck { namespace ck {
...@@ -26,27 +28,65 @@ using S = ck::Sequence<Is...>; ...@@ -26,27 +28,65 @@ using S = ck::Sequence<Is...>;
using PassThrough = ck::tensor_operation::element_wise::PassThrough; using PassThrough = ck::tensor_operation::element_wise::PassThrough;
static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default; static constexpr auto GemmDefault = ck::tensor_operation::device::GemmSpecialization::Default;
static constexpr auto MNPadding = ck::tensor_operation::device::GemmSpecialization::MNPadding;
static constexpr auto MNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
template <ck::tensor_operation::device::GemmSpecialization GemmSpec>
// Compilation parameters for a[m, k] * b[n, k] = c[m, n] // Compilation parameters for a[m, k] * b[n, k] = c[m, n]
using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances = std::tuple< using device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances = std::tuple<
// clang-format off // clang-format off
//#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle| A| B| C| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| //#####################| ALayout| BLayout| CLayout| AData| BData| CData| AccData| CShuffle| A| B| C| GEMM| NumGemmK| Block| MPer| NPer| KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds| BBlockTransfer| BBlockTransfer| BBlockTransfer| BlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds| CShuffle| CShuffle| CBlockTransferClusterLengths| CBlockTransfer| LoopScheduler| Pipeline|
//#####################| | | | Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| //#####################| | | | Type| Type| Type| Type| DataType| Elementwise| Elementwise| Elementwise| Specialization| Prefetch| Size| Block| Block| Block| | | XDL| XDL| Per| Per| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraM| ThreadCluster| ThreadCluster| SrcAccessOrder| SrcVectorDim| SrcScalar| DstScalar| AddExtraN| MXdlPerWave| NXdlPerWave| _MBlock_MWaveMPerXdl| ScalarPerVector| | |
//#####################| | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| //#####################| | | | | | | | | Operation| Operation| Operation| | Stage| | | | | | | | | Wave| Wave| Lengths_K0_M_K1| ArrangeOrder| | | PerVector| PerVector_K1| | Lengths_K0_N_K1| ArrangeOrder| | | PerVector| PerVector_K1| | PerShuffle| PerShuffle| _NBlock_NWaveNPerXdl| _NWaveNPerXdl| | |
//#####################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | //#####################| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>, // pipeline v1, 1 wave
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 64, 64, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 64, 64, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 128, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 128, 32, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 64, 64, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8>, DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 32, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmDefault, 1, 64, 32, 64, 32, 8, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8> DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 64, 64, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 64, 32, 64, 32, 8, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v1>
#if CK_EXPERIMENTAL_INTER_WAVE_INSTANCES
// pipeline v1, 2 waves
,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 64, 64, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 32, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 64, 64, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 64, 32, 64, 32, 8, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8, LoopScheduler::Interwave, PipelineVersion::v1>
#endif
#if CK_EXPERIMENTAL_PIPELINE_V2_INSTANCES
// pipeline v2, 1 wave
,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 256, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 256, 32, 8, 8, 32, 32, 2, 4, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 128, 32, 8, 8, 32, 32, 4, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 64, 128, 32, 8, 8, 32, 32, 2, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 64, 64, 64, 32, 8, 8, 32, 32, 2, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 128, 64, 32, 8, 8, 32, 32, 2, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 256, 64, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 64, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 128, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 32, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 128, 32, 128, 32, 8, 8, 32, 32, 1, 2, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 32, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 8>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 64, 64, 32, 32, 8, 8, 32, 32, 2, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v2>,
DeviceGemm_Xdl_CShuffle< Row, Col, Row, BF16, BF16, BF16, F32, BF16, PassThrough, PassThrough, PassThrough, GemmSpec, 1, 64, 32, 64, 32, 8, 8, 32, 32, 1, 2, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, S<4, 16, 1>, S<1, 0, 2>, S<1, 0, 2>, 2, 8, 8, 1, 1, 1, S<1, 16, 1, 4>, 8, LoopScheduler::Default, PipelineVersion::v2>
#endif
// clang-format on // clang-format on
>; >;
...@@ -55,8 +95,14 @@ void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances( ...@@ -55,8 +95,14 @@ void add_device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances(
DeviceGemm<Row, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>& DeviceGemm<Row, Col, Row, BF16, BF16, BF16, PassThrough, PassThrough, PassThrough>>>&
instances) instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(
device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances{}); instances, device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances<GemmDefault>{});
add_device_operation_instances(
instances, device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances<MNPadding>{});
add_device_operation_instances(
instances, device_gemm_xdl_c_shuffle_bf16_bf16_bf16_mk_nk_mn_instances<MNKPadding>{});
} }
} // namespace instance } // namespace instance
......
library/src/tensor_operation_instance/gpu/grouped_conv1d_fwd/xdl/device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_bf16_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_bf16_instances( void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_bf16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<1, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
GNWC, GNWC,
GKXC, GKXC,
Empty_Tuple, Empty_Tuple,
GNWK, GNWK,
BF16, BF16,
BF16, BF16,
Empty_Tuple, Empty_Tuple,
BF16, BF16,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_bf16_instances<1, device_grouped_conv_fwd_xdl_bf16_instances<1,
......
library/src/tensor_operation_instance/gpu/grouped_conv1d_fwd/xdl/device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f16_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f16_instances( void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<1, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
GNWC, GNWC,
GKXC, GKXC,
Empty_Tuple, Empty_Tuple,
GNWK, GNWK,
F16, F16,
F16, F16,
Empty_Tuple, Empty_Tuple,
F16, F16,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_f16_instances<1, device_grouped_conv_fwd_xdl_f16_instances<1,
......
library/src/tensor_operation_instance/gpu/grouped_conv1d_fwd/xdl/device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f32_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f32_instances( void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_f32_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<1, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
GNWC, GNWC,
GKXC, GKXC,
Empty_Tuple, Empty_Tuple,
GNWK, GNWK,
F32, F32,
F32, F32,
Empty_Tuple, Empty_Tuple,
F32, F32,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_f32_instances<1, device_grouped_conv_fwd_xdl_f32_instances<1,
......
library/src/tensor_operation_instance/gpu/grouped_conv1d_fwd/xdl/device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_int8_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_int8_instances( void add_device_grouped_conv1d_fwd_xdl_gnwc_gkxc_gnwk_int8_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<1, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<1,
GNWC, GNWC,
GKXC, GKXC,
Empty_Tuple, Empty_Tuple,
GNWK, GNWK,
int8_t, int8_t,
int8_t, int8_t,
Empty_Tuple, Empty_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv_fwd_xdl_int8_instances<1, device_grouped_conv_fwd_xdl_int8_instances<1,
......
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/dl/device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f16_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
void add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f16_instances( void add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GNHWC, GNHWC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
GNHWK, GNHWK,
F16, F16,
F16, F16,
Empty_Tuple, Empty_Tuple,
F16, F16,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv2d_fwd_dl_f16_instances<GNHWC, device_grouped_conv2d_fwd_dl_f16_instances<GNHWC,
......
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/dl/device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f32_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
void add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f32_instances( void add_device_grouped_conv2d_fwd_dl_gnhwc_gkyxc_gnhwk_f32_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GNHWC, GNHWC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
GNHWK, GNHWK,
F32, F32,
F32, F32,
Empty_Tuple, Empty_Tuple,
F32, F32,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv2d_fwd_dl_f32_instances<GNHWC, device_grouped_conv2d_fwd_dl_f32_instances<GNHWC,
......
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/dl/device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f16_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
void add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f16_instances( void add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
NHWGK, NHWGK,
F16, F16,
F16, F16,
Empty_Tuple, Empty_Tuple,
F16, F16,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv2d_fwd_dl_f16_instances<NHWGC, device_grouped_conv2d_fwd_dl_f16_instances<NHWGC,
......
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/dl/device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f32_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
void add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f32_instances( void add_device_grouped_conv2d_fwd_dl_nhwgc_gkyxc_nhwgk_f32_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
NHWGC, NHWGC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
NHWGK, NHWGK,
F32, F32,
F32, F32,
Empty_Tuple, Empty_Tuple,
F32, F32,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv2d_fwd_dl_f32_instances<NHWGC, device_grouped_conv2d_fwd_dl_f32_instances<NHWGC,
......
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1p0_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k] // Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1p0_instances( void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GNHWC, GNHWC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
GNHWK, GNHWK,
F16, F16,
F16, F16,
Empty_Tuple, Empty_Tuple,
F16, F16,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv_fwd_wmma_f16_instances<2, device_grouped_conv_fwd_wmma_f16_instances<2,
......
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1s1p0_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k] // Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1s1p0_instances( void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_1x1s1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GNHWC, GNHWC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
GNHWK, GNHWK,
F16, F16,
F16, F16,
Empty_Tuple, Empty_Tuple,
F16, F16,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv_fwd_wmma_f16_instances<2, device_grouped_conv_fwd_wmma_f16_instances<2,
......
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k] // Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_instances( void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GNHWC, GNHWC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
GNHWK, GNHWK,
F16, F16,
F16, F16,
Empty_Tuple, Empty_Tuple,
F16, F16,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv_fwd_wmma_f16_instances<2, device_grouped_conv_fwd_wmma_f16_instances<2,
......
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_oddc_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k] // Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_oddc_instances( void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_f16_oddc_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GNHWC, GNHWC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
GNHWK, GNHWK,
F16, F16,
F16, F16,
Empty_Tuple, Empty_Tuple,
F16, F16,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv_fwd_wmma_f16_instances<2, device_grouped_conv_fwd_wmma_f16_instances<2,
......
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1p0_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k] // Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1p0_instances( void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GNHWC, GNHWC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
GNHWK, GNHWK,
int8_t, int8_t,
int8_t, int8_t,
Empty_Tuple, Empty_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv_fwd_wmma_i8_instances<2, device_grouped_conv_fwd_wmma_i8_instances<2,
......
library/src/tensor_operation_instance/gpu/grouped_conv2d_fwd/wmma/device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1s1p0_instance.cpp
View file @ 5903efe7
...@@ -10,18 +10,18 @@ namespace device { ...@@ -10,18 +10,18 @@ namespace device {
namespace instance { namespace instance {
// Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k] // Compilation parameters for in[g, n, hi, wi, c] * wei[g, k, y, x, c] = out[g, n, ho, wo, k]
void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1s1p0_instances( void add_device_grouped_conv2d_fwd_wmma_gnhwc_gkyxc_gnhwk_i8_1x1s1p0_instances(
std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleD<2, std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<2,
GNHWC, GNHWC,
GKYXC, GKYXC,
Empty_Tuple, Empty_Tuple,
GNHWK, GNHWK,
int8_t, int8_t,
int8_t, int8_t,
Empty_Tuple, Empty_Tuple,
int8_t, int8_t,
PassThrough, PassThrough,
PassThrough, PassThrough,
PassThrough>>>& instances) PassThrough>>>& instances)
{ {
add_device_operation_instances(instances, add_device_operation_instances(instances,
device_grouped_conv_fwd_wmma_i8_instances<2, device_grouped_conv_fwd_wmma_i8_instances<2,
......
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