Skip to content

GitLab

Unverified Commit e521823c authored by arai713's avatar arai713 Committed by GitHub
Browse files

Merge branch 'develop' into hip_tensor_permute

parents adce1006 c7d5c772
Hide whitespace changes
Inline Side-by-side
Showing with 108 additions and 108 deletions
Config.cmake.in
View file @ e521823c
@PACKAGE_INIT@
set(_composable_kernel_supported_components device_operations utility)
set(_composable_kernel_supported_components device_other_operations device_gemm_operations device_conv_operations device_mha_operations device_contraction_operations device_reduction_operations utility)
foreach(_comp ${composable_kernel_FIND_COMPONENTS})
if(NOT _comp IN_LIST _composable_kernel_supported_components)
......
Jenkinsfile
View file @ e521823c
def rocmnode(name) {
return 'rocmtest && miopen && ' + name
return '(rocmtest || miopen) && ' + name
}
def show_node_info() {
......
client_example/01_gemm/CMakeLists.txt
View file @ e521823c
add_executable(client_gemm gemm.cpp)
target_link_libraries(client_gemm PRIVATE composable_kernel::device_operations)
target_link_libraries(client_gemm PRIVATE composable_kernel::device_other_operations composable_kernel::device_gemm_operations)
client_example/02_gemm_add_add_fastgelu/CMakeLists.txt
View file @ e521823c
add_custom_target(client_gemm_fastgelu_examples)
add_executable(client_gemm_add_add_fastgelu gemm_add_add_fastgelu.cpp)
target_link_libraries(client_gemm_add_add_fastgelu PRIVATE composable_kernel::device_operations)
target_link_libraries(client_gemm_add_add_fastgelu PRIVATE composable_kernel::device_gemm_operations)
add_executable(client_gemm_add_fastgelu gemm_add_fastgelu.cpp)
target_link_libraries(client_gemm_add_fastgelu PRIVATE composable_kernel::device_operations)
target_link_libraries(client_gemm_add_fastgelu PRIVATE composable_kernel::device_gemm_operations)
add_executable(client_gemm_fastgelu gemm_fastgelu.cpp)
target_link_libraries(client_gemm_fastgelu PRIVATE composable_kernel::device_operations)
target_link_libraries(client_gemm_fastgelu PRIVATE composable_kernel::device_gemm_operations)
add_dependencies(client_gemm_fastgelu_examples client_gemm_add_add_fastgelu client_gemm_add_fastgelu
client_gemm_fastgelu)
......@@ -15,13 +15,13 @@ add_dependencies(client_gemm_fastgelu_examples client_gemm_add_add_fastgelu clie
add_custom_target(client_gemm_fastgelu_generic_examples)
add_executable(client_gemm_add_add_fastgelu_generic gemm_add_add_fastgelu_generic.cpp)
target_link_libraries(client_gemm_add_add_fastgelu_generic PRIVATE composable_kernel::device_operations)
target_link_libraries(client_gemm_add_add_fastgelu_generic composable_kernel::device_gemm_operations)
add_executable(client_gemm_add_fastgelu_generic gemm_add_fastgelu_generic.cpp)
target_link_libraries(client_gemm_add_fastgelu_generic PRIVATE composable_kernel::device_operations)
target_link_libraries(client_gemm_add_fastgelu_generic PRIVATE composable_kernel::device_gemm_operations)
add_executable(client_gemm_fastgelu_generic gemm_fastgelu_generic.cpp)
target_link_libraries(client_gemm_fastgelu_generic PRIVATE composable_kernel::device_operations)
target_link_libraries(client_gemm_fastgelu_generic PRIVATE composable_kernel::device_gemm_operations)
add_dependencies(client_gemm_fastgelu_generic_examples client_gemm_add_add_fastgelu_generic
client_gemm_add_fastgelu_generic client_gemm_fastgelu_generic)
client_example/03_gemm_layernorm/CMakeLists.txt
View file @ e521823c
add_executable(client_gemm_add_add_layernorm_naive gemm_add_add_layernorm_naive.cpp)
target_link_libraries(client_gemm_add_add_layernorm_naive PRIVATE composable_kernel::device_operations)
target_link_libraries(client_gemm_add_add_layernorm_naive PRIVATE composable_kernel::device_gemm_operations composable_kernel::device_other_operations)
add_executable(client_gemm_add_relu_add_layernorm_welford gemm_add_relu_add_layernorm_welford.cpp)
target_link_libraries(client_gemm_add_relu_add_layernorm_welford PRIVATE composable_kernel::device_operations)
target_link_libraries(client_gemm_add_relu_add_layernorm_welford PRIVATE composable_kernel::device_gemm_operations composable_kernel::device_other_operations)
client_example/04_contraction/CMakeLists.txt
View file @ e521823c
add_executable(client_contraction_scale_fp32 contraction_scale_fp32.cpp)
target_link_libraries(client_contraction_scale_fp32 PRIVATE composable_kernel::device_operations)
target_link_libraries(client_contraction_scale_fp32 PRIVATE composable_kernel::device_other_operations composable_kernel::device_contraction_operations composable_kernel::device_gemm_operations)
add_executable(client_contraction_bilinear_fp32 contraction_bilinear_fp32.cpp)
target_link_libraries(client_contraction_bilinear_fp32 PRIVATE composable_kernel::device_operations)
target_link_libraries(client_contraction_bilinear_fp32 PRIVATE composable_kernel::device_other_operations composable_kernel::device_contraction_operations composable_kernel::device_gemm_operations)
add_executable(client_contraction_scale_fp64 contraction_scale_fp64.cpp)
target_link_libraries(client_contraction_scale_fp64 PRIVATE composable_kernel::device_operations)
target_link_libraries(client_contraction_scale_fp64 PRIVATE composable_kernel::device_other_operations composable_kernel::device_contraction_operations composable_kernel::device_gemm_operations)
add_executable(client_contraction_bilinear_fp64 contraction_bilinear_fp64.cpp)
target_link_libraries(client_contraction_bilinear_fp64 PRIVATE composable_kernel::device_operations)
target_link_libraries(client_contraction_bilinear_fp64 PRIVATE composable_kernel::device_other_operations composable_kernel::device_contraction_operations composable_kernel::device_gemm_operations)
add_executable(contraction_g1m2n3k1_add_xdl_fp16 contraction_g1m2n3k1_add_xdl_fp16.cpp)
target_link_libraries(contraction_g1m2n3k1_add_xdl_fp16 PRIVATE composable_kernel::device_operations)
target_link_libraries(contraction_g1m2n3k1_add_xdl_fp16 PRIVATE composable_kernel::device_other_operations composable_kernel::device_contraction_operations composable_kernel::device_gemm_operations)
client_example/05_layernorm/CMakeLists.txt
View file @ e521823c
add_executable(client_layernorm2d_fwd layernorm2d_fwd.cpp)
target_link_libraries(client_layernorm2d_fwd PRIVATE composable_kernel::device_operations)
target_link_libraries(client_layernorm2d_fwd PRIVATE composable_kernel::device_other_operations)
add_executable(client_layernorm4d_fwd layernorm4d_fwd.cpp)
target_link_libraries(client_layernorm4d_fwd PRIVATE composable_kernel::device_operations)
target_link_libraries(client_layernorm4d_fwd PRIVATE composable_kernel::device_other_operations)
client_example/06_softmax/CMakeLists.txt
View file @ e521823c
add_executable(client_softmax4d softmax4d.cpp)
target_link_libraries(client_softmax4d PRIVATE composable_kernel::device_operations)
target_link_libraries(client_softmax4d PRIVATE composable_kernel::device_other_operations composable_kernel::device_reduction_operations)
client_example/07_grouped_convnd_fwd/CMakeLists.txt
View file @ e521823c
add_executable(client_grouped_conv2d_fwd grouped_conv2d_fwd.cpp)
target_link_libraries(client_grouped_conv2d_fwd PRIVATE composable_kernel::device_operations)
target_link_libraries(client_grouped_conv2d_fwd PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_conv1d_fwd grouped_conv1d_fwd.cpp)
target_link_libraries(client_grouped_conv1d_fwd PRIVATE composable_kernel::device_operations)
target_link_libraries(client_grouped_conv1d_fwd PRIVATE composable_kernel::device_conv_operations)
client_example/07_grouped_convnd_fwd/grouped_conv1d_fwd.cpp
View file @ e521823c
......@@ -100,18 +100,18 @@ int main()
SimpleDeviceMem wei(sizeof(WeiDataType) * G * K * X * C);
SimpleDeviceMem out(sizeof(OutDataType) * G * N * Wo * K);
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<>,
OutDataType,
PassThrough,
PassThrough,
PassThrough>;
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<>,
OutDataType,
PassThrough,
PassThrough,
PassThrough>;
// get device op instances
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
......
client_example/07_grouped_convnd_fwd/grouped_conv2d_fwd.cpp
View file @ e521823c
......@@ -71,18 +71,18 @@ int main()
SimpleDeviceMem wei(sizeof(WeiDataType) * G * K * Y * X * C);
SimpleDeviceMem out(sizeof(OutDataType) * N * Ho * Wo * G * K);
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<>,
OutDataType,
PassThrough,
PassThrough,
PassThrough>;
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<>,
OutDataType,
PassThrough,
PassThrough,
PassThrough>;
// get device op instances
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
......
client_example/08_fused_attention/CMakeLists.txt
View file @ e521823c
add_executable(client_fused_attention fused_attention.cpp)
target_link_libraries(client_fused_attention PRIVATE composable_kernel::device_operations)
target_link_libraries(client_fused_attention PRIVATE composable_kernel::device_other_operations composable_kernel::device_gemm_operations)
add_executable(client_fused_attention_bias fused_attention_bias.cpp)
target_link_libraries(client_fused_attention_bias PRIVATE composable_kernel::device_operations)
target_link_libraries(client_fused_attention_bias PRIVATE composable_kernel::device_other_operations composable_kernel::device_gemm_operations)
client_example/09_quantization/CMakeLists.txt
View file @ e521823c
if(DTYPES MATCHES "int8" OR NOT DEFINED DTYPES)
add_executable(client_conv2d_fwd_bias_tanh_perchannel_quantization conv2d_fwd_bias_tanh_perchannel_quantization.cpp)
target_link_libraries(client_conv2d_fwd_bias_tanh_perchannel_quantization PRIVATE composable_kernel::device_operations)
target_link_libraries(client_conv2d_fwd_bias_tanh_perchannel_quantization PRIVATE composable_kernel::device_conv_operations composable_kernel::device_other_operations composable_kernel::device_gemm_operations)
add_executable(client_conv2d_fwd_bias_relu_perchannel_quantization conv2d_fwd_bias_relu_perchannel_quantization.cpp)
target_link_libraries(client_conv2d_fwd_bias_relu_perchannel_quantization PRIVATE composable_kernel::device_operations)
target_link_libraries(client_conv2d_fwd_bias_relu_perchannel_quantization PRIVATE composable_kernel::device_conv_operations composable_kernel::device_other_operations composable_kernel::device_gemm_operations)
add_executable(client_conv2d_fwd_bias_tanh_perlayer_quantization conv2d_fwd_bias_tanh_perlayer_quantization.cpp)
target_link_libraries(client_conv2d_fwd_bias_tanh_perlayer_quantization PRIVATE composable_kernel::device_operations)
target_link_libraries(client_conv2d_fwd_bias_tanh_perlayer_quantization PRIVATE composable_kernel::device_conv_operations composable_kernel::device_other_operations composable_kernel::device_gemm_operations)
add_executable(client_conv2d_fwd_bias_relu_perlayer_quantization conv2d_fwd_bias_relu_perlayer_quantization.cpp)
target_link_libraries(client_conv2d_fwd_bias_relu_perlayer_quantization PRIVATE composable_kernel::device_operations)
target_link_libraries(client_conv2d_fwd_bias_relu_perlayer_quantization PRIVATE composable_kernel::device_conv_operations composable_kernel::device_other_operations composable_kernel::device_gemm_operations)
add_executable(client_conv2d_fwd_perchannel_quantization conv2d_fwd_perchannel_quantization.cpp)
target_link_libraries(client_conv2d_fwd_perchannel_quantization PRIVATE composable_kernel::device_operations)
target_link_libraries(client_conv2d_fwd_perchannel_quantization PRIVATE composable_kernel::device_conv_operations composable_kernel::device_other_operations composable_kernel::device_gemm_operations)
add_executable(client_conv2d_fwd_perlayer_quantization conv2d_fwd_perlayer_quantization.cpp)
target_link_libraries(client_conv2d_fwd_perlayer_quantization PRIVATE composable_kernel::device_operations)
target_link_libraries(client_conv2d_fwd_perlayer_quantization PRIVATE composable_kernel::device_conv_operations composable_kernel::device_other_operations composable_kernel::device_gemm_operations)
add_executable(client_gemm_quantization gemm_quantization.cpp)
target_link_libraries(client_gemm_quantization PRIVATE composable_kernel::device_operations)
target_link_libraries(client_gemm_quantization PRIVATE composable_kernel::device_conv_operations composable_kernel::device_other_operations composable_kernel::device_gemm_operations)
endif()
client_example/09_quantization/conv2d_fwd_bias_relu_perchannel_quantization.cpp
View file @ e521823c
......@@ -80,7 +80,7 @@ int main(int argc, char* argv[])
SimpleDeviceMem requant_scale(sizeof(RequantScaleDataType) * G * K);
SimpleDeviceMem out(sizeof(OutDataType) * N * Ho * Wo * G * K);
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD<
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
NumDimSpatial,
InLayout,
WeiLayout,
......
client_example/09_quantization/conv2d_fwd_bias_relu_perlayer_quantization.cpp
View file @ e521823c
......@@ -78,18 +78,18 @@ int main(int argc, char* argv[])
SimpleDeviceMem out(sizeof(OutDataType) * N * Ho * Wo * G * K);
using DeviceOp =
ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<BiasLayout>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<BiasDataType>,
OutDataType,
PassThrough,
PassThrough,
OutElementOp>;
ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<BiasLayout>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<BiasDataType>,
OutDataType,
PassThrough,
PassThrough,
OutElementOp>;
// get device op instances
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
DeviceOp>::GetInstances();
......
client_example/09_quantization/conv2d_fwd_bias_tanh_perchannel_quantization.cpp
View file @ e521823c
......@@ -83,7 +83,7 @@ int main(int argc, char* argv[])
SimpleDeviceMem requant_scale(sizeof(RequantScaleDataType) * G * K);
SimpleDeviceMem out(sizeof(OutDataType) * N * Ho * Wo * G * K);
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD<
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
NumDimSpatial,
InLayout,
WeiLayout,
......
client_example/09_quantization/conv2d_fwd_bias_tanh_perlayer_quantization.cpp
View file @ e521823c
......@@ -79,18 +79,18 @@ int main(int argc, char* argv[])
SimpleDeviceMem out(sizeof(OutDataType) * N * Ho * Wo * G * K);
using DeviceOp =
ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<BiasLayout>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<BiasDataType>,
OutDataType,
PassThrough,
PassThrough,
OutElementOp>;
ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<BiasLayout>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<BiasDataType>,
OutDataType,
PassThrough,
PassThrough,
OutElementOp>;
// get device op instances
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
DeviceOp>::GetInstances();
......
client_example/09_quantization/conv2d_fwd_perchannel_quantization.cpp
View file @ e521823c
......@@ -76,19 +76,19 @@ int main(int argc, char* argv[])
SimpleDeviceMem requant_scale(sizeof(RequantScaleDataType) * G * K);
SimpleDeviceMem out(sizeof(OutDataType) * N * Ho * Wo * G * K);
using DeviceOp =
ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<RequantScaleLayout>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<RequantScaleDataType>,
OutDataType,
PassThrough,
PassThrough,
OutElementOp>;
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<
NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<RequantScaleLayout>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<RequantScaleDataType>,
OutDataType,
PassThrough,
PassThrough,
OutElementOp>;
// get device op instances
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
DeviceOp>::GetInstances();
......
client_example/09_quantization/conv2d_fwd_perlayer_quantization.cpp
View file @ e521823c
......@@ -72,18 +72,18 @@ int main(int argc, char* argv[])
SimpleDeviceMem wei(sizeof(WeiDataType) * G * K * Y * X * C);
SimpleDeviceMem out(sizeof(OutDataType) * N * Ho * Wo * G * K);
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<>,
OutDataType,
PassThrough,
PassThrough,
OutElementOp>;
using DeviceOp = ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
InLayout,
WeiLayout,
ck::Tuple<>,
OutLayout,
InDataType,
WeiDataType,
ck::Tuple<>,
OutDataType,
PassThrough,
PassThrough,
OutElementOp>;
// get device op instances
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
DeviceOp>::GetInstances();
......
client_example/10_grouped_convnd_bwd_data/CMakeLists.txt
View file @ e521823c
add_executable(client_grouped_conv2d_bwd_data grouped_conv2d_bwd_data.cpp)
target_link_libraries(client_grouped_conv2d_bwd_data PRIVATE composable_kernel::device_operations)
target_link_libraries(client_grouped_conv2d_bwd_data PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_conv3d_bwd_data grouped_conv3d_bwd_data.cpp)
target_link_libraries(client_grouped_conv3d_bwd_data PRIVATE composable_kernel::device_operations)
target_link_libraries(client_grouped_conv3d_bwd_data PRIVATE composable_kernel::device_conv_operations)
add_executable(client_grouped_conv3d_bwd_data_input_fp16_comp_bf8f8 grouped_conv3d_bwd_data_input_fp16_comp_bf8f8.cpp)
target_link_libraries(client_grouped_conv3d_bwd_data_input_fp16_comp_bf8f8 PRIVATE composable_kernel::device_operations)
target_link_libraries(client_grouped_conv3d_bwd_data_input_fp16_comp_bf8f8 PRIVATE composable_kernel::device_conv_operations)
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