Merge branch 'lib_gemm_softmax_gemm_type' of...

Merge branch 'lib_gemm_softmax_gemm_type' of https://github.com/ROCmSoftwarePlatform/composable_kernel into lib_gemm_softmax_gemm_type

Merge branch 'lib_gemm_softmax_gemm_type' of...
Merge branch 'lib_gemm_softmax_gemm_type' of https://github.com/ROCmSoftwarePlatform/composable_kernel into lib_gemm_softmax_gemm_type
a483948c · ltqin · 54e599ca · e288457d · a483948c · a483948c
Commit a483948c authored Mar 31, 2023 by ltqin
14 changed files
--- a/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_bilinear/device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance.cpp
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2022, Advanced Micro Devices, Inc. All rights reserved.
+// This (ifndef) is a hack to use customized behavior for buffer load rather than using default
+// setting Don't use this hack unless absolutely necessary!
+// FIXME: make the behavior of buffer load a configurable (template) parameter of each device op
+#define CK_EXPERIMENTAL_USE_BUFFER_LOAD_OOB_CHECK_OFFSET_TRICK 1
+#include <cstdlib>
+#include "ck/ck.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/impl/device_contraction_multiple_d_xdl_cshuffle.hpp"
+#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
+#include "ck/library/tensor_operation_instance/add_device_operation_instance.hpp"
+namespace ck {
+namespace tensor_operation {
+namespace device {
+namespace instance {
+using F64       = double;
+using F64_Tuple = ck::Tuple<F64>;
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+using PassThrough = ck::tensor_operation::element_wise::PassThrough;
+using Bilinear    = ck::tensor_operation::element_wise::Bilinear;
+static constexpr auto GemmMNKPadding = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+// A[m0, m1, k0, k1] * B[n0, n1, k0, k1] + D[m0, m1, n0, n1] = E[m0, m1, n0, n1]
+// m/n/n/n are the fast changing dimension for A/B/D/E
+using device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance = std::tuple<
+    // clang-format off
+        //#####################################| NumDimM| NumDimN| NumDimK| AData| BData| AccData| CShuffle|    DsData| EData|            A|           B|         CDE|           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|
+        //#####################################|        |        |        |  Type|  Type|    Type| DataType|      Type|  Type|  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|
+        //#####################################|        |        |        |      |      |        |         |          |      |             |            |            |               |         |      |      |      |      |    |    |     |     |     |     |                |               |               |               |               |               |          |                |               |               |              |               |               |          |            |            |                             |                |
+        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,   F64,   F64,     F64,      F64, F64_Tuple,   F64,  PassThrough, PassThrough,    Bilinear, GemmMNKPadding,        1,   256,   128,   128,    16,   1,   1,   16,   16,    4,    4,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1>,
+        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,   F64,   F64,     F64,      F64, F64_Tuple,   F64,  PassThrough, PassThrough,    Bilinear, GemmMNKPadding,        1,   256,   128,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1>,
+        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,   F64,   F64,     F64,      F64, F64_Tuple,   F64,  PassThrough, PassThrough,    Bilinear, GemmMNKPadding,        1,   128,   128,    64,    16,   1,   1,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1,  8>,               1>,
+        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,   F64,   F64,     F64,      F64, F64_Tuple,   F64,  PassThrough, PassThrough,    Bilinear, GemmMNKPadding,        1,   128,   128,    64,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1,  8>,               1>,
+        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,   F64,   F64,     F64,      F64, F64_Tuple,   F64,  PassThrough, PassThrough,    Bilinear, GemmMNKPadding,        1,   128,    64,   128,    16,   1,   1,   16,   16,    4,    4,     S<8, 16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1,  8, 1, 16>,               1>,
+        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,   F64,   F64,     F64,      F64, F64_Tuple,   F64,  PassThrough, PassThrough,    Bilinear, GemmMNKPadding,        1,   128,    64,   128,    16,   2,   2,   16,   16,    4,    4,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1,  8, 1, 16>,               1>,
+        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,   F64,   F64,     F64,      F64, F64_Tuple,   F64,  PassThrough, PassThrough,    Bilinear, GemmMNKPadding,        1,   256,   128,    64,    16,   1,   1,   16,   16,    4,    2,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1>,
+        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,   F64,   F64,     F64,      F64, F64_Tuple,   F64,  PassThrough, PassThrough,    Bilinear, GemmMNKPadding,        1,   256,   128,    64,    16,   2,   2,   16,   16,    4,    2,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1>,
+        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,   F64,   F64,     F64,      F64, F64_Tuple,   F64,  PassThrough, PassThrough,    Bilinear, GemmMNKPadding,        1,   256,    64,   128,    16,   1,   1,   16,   16,    2,    4,     S<16,16, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         0,     S<8, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         0,           1,           1,              S<1, 16, 1, 16>,               1>,
+        DeviceContractionMultipleD_Xdl_CShuffle<       2,       2,       2,   F64,   F64,     F64,      F64, F64_Tuple,   F64,  PassThrough, PassThrough,    Bilinear, GemmMNKPadding,        1,   256,    64,   128,    16,   2,   2,   16,   16,    2,    4,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              1,         1,     S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,             1,              1,              1,         1,           1,           1,              S<1, 16, 1, 16>,               1>
+    // clang-format on
+    >;
+void add_device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance(
+    std::vector<std::unique_ptr<DeviceContractionMultipleD<2,
+                                                           2,
+                                                           2,
+                                                           F64,
+                                                           F64,
+                                                           F64_Tuple,
+                                                           F64,
+                                                           PassThrough,
+                                                           PassThrough,
+                                                           Bilinear>>>& instances)
+{
+    add_device_operation_instances(
+        instances,
+        device_contraction_bilinear_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_f64_mnnn_instance{});
+}
+} // namespace instance
+} // namespace device
+} // namespace tensor_operation
+} // namespace ck
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/CMakeLists.txt
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/CMakeLists.txt
 add_instance_library(device_contraction_scale_instance
+    #float
    device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_kkn_instance.cpp
    device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_knn_instance.cpp
    device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mkn_instance.cpp
    device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f32_f32_f32_mnn_instance.cpp
+    #double
+    device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
+    device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
+    device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
+    device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp
 )
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_kkn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_knn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mkn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/contraction_scale/device_contraction_scale_m2_n2_k2_xdl_c_shuffle_f64_f64_f64_mnn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_splitk/device_gemm_xdl_splitk_f16_f16_f16_mk_kn_mn_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/quantization/conv2d_fwd/conv2d_quantization_common.hpp
+++ b/library/src/tensor_operation_instance/gpu/quantization/conv2d_fwd/conv2d_quantization_common.hpp
--- a/library/src/tensor_operation_instance/gpu/quantization/conv2d_fwd/device_conv2d_dl_bias_perchannel_quantization_int8_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/quantization/conv2d_fwd/device_conv2d_dl_bias_perchannel_quantization_int8_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/quantization/conv2d_fwd/device_conv2d_dl_bias_perlayer_quantization_int8_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/quantization/conv2d_fwd/device_conv2d_dl_bias_perlayer_quantization_int8_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/quantization/conv2d_fwd/device_conv2d_xdl_bias_perchannel_quantization_int8_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/quantization/conv2d_fwd/device_conv2d_xdl_bias_perchannel_quantization_int8_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/quantization/conv2d_fwd/device_conv2d_xdl_bias_perlayer_quantization_int8_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/quantization/conv2d_fwd/device_conv2d_xdl_bias_perlayer_quantization_int8_instance.cpp
--- a/script/cmake-ck-dev.sh
+++ b/script/cmake-ck-dev.sh
@@ -8,8 +8,8 @@ MY_PROJECT_SOURCE=$1
 cmake                                                                                             \
 -D CMAKE_PREFIX_PATH=/opt/rocm                                                                    \
 -D CMAKE_CXX_COMPILER=/opt/rocm/bin/hipcc                                                         \
-D CMAKE_CXX_FLAGS="-std=c++17 -O3 -ftemplate-backtrace-limit=0 -fPIE -mcumode                    \
+-D CMAKE_CXX_FLAGS="-std=c++17 -O3 -ftemplate-backtrace-limit=0  -fPIE  -Wno-gnu-line-marker      \
-mno-wavefrontsize64 -Wno-gnu-line-marker -save-temps=$PWD"                                       \
+-save-temps=$PWD"                                                                                 \
 -D CMAKE_BUILD_TYPE=Release                                                                       \
 -D BUILD_DEV=ON                                                                                   \
 -D GPU_TARGETS="gfx908;gfx90a"                                                                    \

--- a/test/grouped_convnd_bwd_weight/grouped_convnd_bwd_weight.cpp
+++ b/test/grouped_convnd_bwd_weight/grouped_convnd_bwd_weight.cpp