Merge pull request #138 from ROCm/merge_from_public

Merge from public

CMakeLists.txt

@@ -553,12 +553,7 @@ if(NOT DEFINED INSTANCES_ONLY)
         PACKAGE_NAME examples
    )
    add_subdirectory(example)
-   if(GPU_TARGETS MATCHES "gfx9" AND NOT INSTANCES_ONLY)
-      add_subdirectory(codegen)
-   endif()
-   if(BUILD_TESTING)
    add_subdirectory(test)
-   endif()
 
    rocm_package_setup_component(profiler
         LIBRARY_NAME composablekernel
@@ -575,6 +570,10 @@ if(NOT DEFINED INSTANCES_ONLY)
   endif()
 endif()
 
+if(NOT DEFINED PROFILER_ONLY AND (GPU_TARGETS MATCHES "gfx9" OR DEFINED INSTANCES_ONLY))
+  add_subdirectory(codegen)
+endif()
+
 #Create an interface target for the include only files and call it "composablekernels"
 include(CMakePackageConfigHelpers)
 

client_example/24_grouped_conv_activation/CMakeLists.txt

@@ -55,6 +55,14 @@ target_link_libraries(client_conv3d_fwd_convscale_relu_amax_fp8
                               composable_kernel::device_other_operations
                               composable_kernel::device_reduction_operations
                               utility)
+# Fwd convscale + AMAX
+add_executable(client_conv3d_fwd_convscale_amax_fp8
+               grouped_convnd_fwd_convscale_reduce/conv3d_fwd_convscale_amax_fp8.cpp)
+target_link_libraries(client_conv3d_fwd_convscale_amax_fp8
+                      PRIVATE composable_kernel::device_conv_operations
+                              composable_kernel::device_other_operations
+                              composable_kernel::device_reduction_operations
+                              utility)
 # Fwd convscale
 add_executable(client_conv3d_fwd_convscale_fp8
                grouped_convnd_fwd_convscale/conv3d_fwd_convscale_fp8.cpp)

client_example/24_grouped_conv_activation/grouped_convnd_fwd_convscale_reduce/common.hpp

@@ -15,21 +15,18 @@
 #include "ck/tensor_operation/gpu/device/device_elementwise.hpp"
 #include "ck/tensor_operation/gpu/device/device_reduce.hpp"
 #include "ck/tensor_operation/gpu/device/reduction_operator_mapping.hpp"
-#include "ck/tensor_operation/gpu/element/combined_element_wise_operation.hpp"
-#include "ck/tensor_operation/gpu/element/unary_element_wise_operation.hpp"
 #include "ck/utility/tuple.hpp"
 #include "ck/utility/type.hpp"
 #include "ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_convscale_relu.hpp"
+#include "ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_convscale.hpp"
 #include "ck/utility/reduction_enums.hpp"
 #include "ck/library/tensor_operation_instance/gpu/permute_scale.hpp"
 #include "ck/library/tensor_operation_instance/gpu/reduce/reduce.hpp"
 #include "ck/library/utility/host_tensor.hpp"
 
-namespace ew = ck::tensor_operation::element_wise;
-
-using PassThrough   = ew::PassThrough;
-using ConvScaleRelu = ew::UnaryCombinedOp<ew::Scale, ew::Scale, ew::Relu>;
-using ConvScale     = ew::UnaryCombinedOp<ew::Scale, ew::Scale, PassThrough>;
+using PassThrough   = ck::tensor_operation::element_wise::PassThrough;
+using ConvScaleRelu = ck::tensor_operation::element_wise::ScaleScaleRelu;
+using ConvScale     = ck::tensor_operation::element_wise::ScaleScalePass;
 
 struct SimpleDeviceMem
 {
@@ -221,7 +218,9 @@ bool run_grouped_conv_fwd_convscale_reduce(
      * FP8 Convolution with Scaling
      */
     std::cout << "\n\nConvolution with scale Benchmarking:" << std::endl;
-    auto elementwise_op = ConvElementOp{ew::Scale{scale_in}, ew::Scale{scale_wei}, {}};
+    auto elementwise_op = ConvElementOp{ck::tensor_operation::element_wise::Scale{scale_in},
+                                        ck::tensor_operation::element_wise::Scale{scale_wei},
+                                        {}};
     auto conv_ok        = ConvolutionScale<InDataType,
                                     WeiDataType,
                                     ConvOutDataType,
@@ -465,7 +464,7 @@ bool TensorScaleConvert(SimpleDeviceMem& in,
     using DeviceScaleConvert =
         ck::tensor_operation::device::DeviceElementwise<ck::Tuple<InDataType>,
                                                         ck::Tuple<OutDataType>,
-                                                        ew::Scale,
+                                                        ck::tensor_operation::element_wise::Scale,
                                                         NumDimSpatial + NumNonSpatialDim>;
 
     // get device op instances
@@ -483,7 +482,7 @@ bool TensorScaleConvert(SimpleDeviceMem& in,
     // profile device operation instances
     std::cout << "Run all DeviceScaleConvert instances and do timing" << std::endl;
 
-    auto scale_convert = ew::Scale{scale_out};
+    auto scale_convert = ck::tensor_operation::element_wise::Scale{scale_out};
 
     for(int i = 0; i < op_ptrs.size(); ++i)
     {

client_example/24_grouped_conv_activation/grouped_convnd_fwd_convscale_reduce/conv3d_fwd_convscale_amax_fp8.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "common.hpp"
+
+#include "ck/ck.hpp"
+#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
+
+using InDataType       = ck::f8_t;
+using WeiDataType      = ck::f8_t;
+using CShuffleDataType = float;
+using ConvOutDataType  = float;    // data type of convolution result
+using OutDataType      = ck::f8_t; // data type of final result
+using AComputeDataType = ck::f8_t;
+using BComputeDataType = ck::f8_t;
+
+using ConvElementOp = ConvScale;
+
+using InLayout  = ck::tensor_layout::convolution::NDHWGC;
+using WeiLayout = ck::tensor_layout::convolution::GKZYXC;
+using OutLayout = ck::tensor_layout::convolution::NDHWGK;
+
+constexpr auto ReduceOpId = ck::ReduceTensorOp::AMAX;
+
+static constexpr ck::index_t NumDimSpatial = 3;
+static constexpr ck::index_t G             = 1;
+static constexpr ck::index_t N             = 64;
+static constexpr ck::index_t K             = 128;
+static constexpr ck::index_t C             = 64;
+static constexpr ck::index_t Z             = 3;
+static constexpr ck::index_t Y             = 3;
+static constexpr ck::index_t X             = 3;
+static constexpr ck::index_t Di            = 28;
+static constexpr ck::index_t Hi            = 28;
+static constexpr ck::index_t Wi            = 3;
+static constexpr ck::index_t Do            = 28;
+static constexpr ck::index_t Ho            = 28;
+static constexpr ck::index_t Wo            = 3;
+
+int main()
+{
+    return run_grouped_conv_fwd_convscale_reduce<NumDimSpatial,
+                                                 InDataType,
+                                                 WeiDataType,
+                                                 ConvOutDataType,
+                                                 OutDataType,
+                                                 ConvElementOp,
+                                                 ReduceOpId,
+                                                 InLayout,
+                                                 WeiLayout,
+                                                 OutLayout,
+                                                 3,
+                                                 AComputeDataType,
+                                                 BComputeDataType>(
+               {N, Di, Hi, Wi, G, C}, {G, K, Z, Y, X, C}, {N, Do, Ho, Wo, G, K})
+               ? EXIT_SUCCESS
+               : EXIT_FAILURE;
+}

codegen/CMakeLists.txt

@@ -27,6 +27,8 @@ file(GLOB_RECURSE KERNEL_FILES CONFIGURE_DEPENDS
 add_embed_library(ck_headers ${KERNEL_FILES} RELATIVE ${CK_ROOT}/include)
 
 file(GLOB SOURCES CONFIGURE_DEPENDS src/*.cpp)
+
+##message(STATUS "SOURCE_FILES: ${SOURCES}")
 # TODO: Use object library
 add_library(ck_host STATIC ${SOURCES})
 target_link_libraries(ck_host PRIVATE ck_headers)
@@ -48,6 +50,4 @@ rocm_install(
 )
 rocm_install(DIRECTORY include/ck DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
 
-if(BUILD_TESTING)
 add_subdirectory(test)
-endif()

codegen/test/CMakeLists.txt

 list(APPEND CMAKE_PREFIX_PATH /opt/rocm)
 add_subdirectory(rtc)
 file(GLOB TEST_SRCS CONFIGURE_DEPENDS *.cpp)
-foreach(TEST_SRC ${TEST_SRCS})
+if(NOT INSTANCES_ONLY)
+  foreach(TEST_SRC ${TEST_SRCS})
     set_source_files_properties(${TEST_SRC} PROPERTIES LANGUAGE HIP)
     get_filename_component(BASE_NAME ${TEST_SRC} NAME_WE)
-  add_executable(test_host_${BASE_NAME} ${TEST_SRC})
-  add_dependencies(codegen test_host_${BASE_NAME})
-  add_test(NAME codegen_test_${BASE_NAME} COMMAND test_host_${BASE_NAME})
-  target_link_libraries(test_host_${BASE_NAME} ck_rtc ck_host)
-  # target_link_libraries(test_host_${BASE_NAME} ${CK_ROOT}/build/lib/libutility.a)
-  target_include_directories(test_host_${BASE_NAME} PUBLIC include())
-  target_include_directories(test_host_${BASE_NAME} PUBLIC ${CK_ROOT}/include)
-  target_include_directories(test_host_${BASE_NAME} PUBLIC ${CK_ROOT}/library/include)
-endforeach()
+    add_executable(codegen_test_${BASE_NAME} ${TEST_SRC})
+    add_dependencies(codegen codegen_test_${BASE_NAME})
+    add_dependencies(tests codegen_test_${BASE_NAME})
+    add_dependencies(check codegen_test_${BASE_NAME})
+    add_test(NAME codegen_test_${BASE_NAME} COMMAND codegen_test_${BASE_NAME})
+    message("adding test codegen_test_${BASE_NAME}")
+    target_link_libraries(codegen_test_${BASE_NAME} ck_rtc ck_host)
+    target_include_directories(codegen_test_${BASE_NAME} PUBLIC ${CK_ROOT}/codegen/test/include)
+    target_include_directories(codegen_test_${BASE_NAME} PUBLIC ${CK_ROOT}/include)
+    target_include_directories(codegen_test_${BASE_NAME} PUBLIC ${CK_ROOT}/library/include)
+  endforeach()
+endif()

codegen/test/rtc/CMakeLists.txt

-
-find_package(hip)
 file(GLOB RTC_SOURCES CONFIGURE_DEPENDS src/*.cpp)
 add_library(ck_rtc ${RTC_SOURCES})
 target_include_directories(ck_rtc PUBLIC include)

docs/sphinx/requirements.in

-rocm-docs-core==1.7.1
+rocm-docs-core==1.7.2
 sphinxcontrib-bibtex==2.6.2

docs/sphinx/requirements.txt

@@ -103,7 +103,7 @@ requests==2.32.3
     # via
     #   pygithub
     #   sphinx
-rocm-docs-core==1.7.1
+rocm-docs-core==1.7.2
     # via -r requirements.in
 six==1.16.0
     # via pybtex

example/01_gemm/gemm_xdl_fp8.cpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #include "common.hpp"
 
@@ -7,7 +7,7 @@
 
 using ADataType        = ck::f8_t;
 using BDataType        = ck::f8_t;
-using CDataType        = ck::half_t;
+using CDataType        = ck::f8_t;
 using AccDataType      = float;
 using CShuffleDataType = float;
 

example/01_gemm/run_gemm_example.inc

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -34,11 +34,11 @@ inline __host__ __device__ constexpr double get_rtol()
     }
     else if constexpr(std::is_same_v<DataType, ck::f8_t>)
     {
-        return 1e-1; // 240 and 224 are acceptable
+        return 2e-1;
     }
     else if constexpr(std::is_same_v<DataType, ck::bf8_t>)
     {
-        return 1.5e-1; // 57344 and 49152 are acceptable
+        return 2e-1;
     }
     else
     {
@@ -75,11 +75,11 @@ inline __host__ __device__ constexpr double get_atol()
     }
     else if constexpr(std::is_same_v<DataType, ck::f8_t>)
     {
-        return 16.1; // 240 and 224 are acceptable
+        return 2e-1;
     }
     else if constexpr(std::is_same_v<DataType, ck::bf8_t>)
     {
-        return 8192.1; // 57344 and 49152 are acceptable
+        return 2e-1;
     }
     else
     {

example/62_convnd_activ/convscale_reduce/CMakeLists.txt

@@ -4,7 +4,10 @@ foreach(gpu IN LISTS GPU_TARGETS)
  if(gpu IN_LIST gpu_list AND target EQUAL 0)
      add_custom_target(example_convnd_activ_xdl_convscale_reduce)
      add_example_executable(example_convnd_fwd_xdl_convscale_relu_amax_fp8 convnd_fwd_xdl_convscale_relu_amax_fp8.cpp)
-     add_example_dependencies(example_convnd_activ_xdl_convscale_reduce example_convnd_fwd_xdl_convscale_relu_amax_fp8 )
+     add_example_dependencies(example_convnd_activ_xdl_convscale_reduce example_convnd_fwd_xdl_convscale_relu_amax_fp8)
+
+     add_example_executable(example_convnd_fwd_xdl_convscale_amax_fp8 convnd_fwd_xdl_convscale_amax_fp8.cpp)
+     add_example_dependencies(example_convnd_activ_xdl_convscale_reduce example_convnd_fwd_xdl_convscale_amax_fp8)
 
      set(target 1)
  endif()

example/62_convnd_activ/convscale_reduce/convnd_fwd_xdl_convscale_amax_fp8.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "convnd_fwd_convscale_reduce_common.hpp"
+
+#include "ck/tensor_operation/gpu/device/impl/device_grouped_conv_fwd_multiple_abd_xdl_cshuffle.hpp"
+
+using InDataType       = ck::f8_t;
+using WeiDataType      = ck::f8_t;
+using AccDataType      = float;
+using CShuffleDataType = float;
+using ConvOutDataType  = float;    // data type of convolution result
+using OutDataType      = ck::f8_t; // data type of final result
+using AComputeDataType = ck::f8_t;
+using BComputeDataType = ck::f8_t;
+
+template <ck::index_t... Is>
+using S = ck::Sequence<Is...>;
+
+using InElementOp  = PassThrough;
+using WeiElementOp = PassThrough;
+using OutElementOp = ConvScale;
+
+static constexpr auto ConvSpec =
+    ck::tensor_operation::device::ConvolutionForwardSpecialization::Default;
+
+static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKPadding;
+
+template <ck::index_t NDimSpatial, typename InLayout, typename WeiLayout, typename OutLayout>
+using DeviceGroupedConvNDFwdInstance =
+    ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD_Xdl_CShuffle<
+        NDimSpatial,
+        InLayout,
+        WeiLayout,
+        ck::Tuple<>,
+        OutLayout,
+        InDataType,
+        WeiDataType,
+        AccDataType,
+        CShuffleDataType,
+        ck::Tuple<>,
+        ConvOutDataType,
+        InElementOp,
+        WeiElementOp,
+        OutElementOp,
+        ConvSpec,    // ConvForwardSpecialization
+        GemmSpec,    // GemmSpecialization
+        1,           //
+        256,         // BlockSize
+        128,         // MPerBlock
+        256,         // NPerBlock
+        32,          // KPerBlock
+        8,           // AK1
+        8,           // BK1
+        32,          // MPerXdl
+        32,          // NPerXdl
+        2,           // MXdlPerWave
+        4,           // NXdlPerWave
+        S<4, 64, 1>, // ABlockTransferThreadClusterLengths_AK0_M_AK1
+        S<1, 0, 2>,  // ABlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,  // ABlockTransferSrcAccessOrder
+        2,           // ABlockTransferSrcVectorDim
+        8,           // ABlockTransferSrcScalarPerVector
+        8,           // ABlockTransferDstScalarPerVector_AK1
+        1,           // ABlockLdsExtraM
+        S<4, 64, 1>, // BBlockTransferThreadClusterLengths_BK0_N_BK1
+        S<1, 0, 2>,  // BBlockTransferThreadClusterArrangeOrder
+        S<1, 0, 2>,  // BBlockTransferSrcAccessOrder
+        2,           // BBlockTransferSrcVectorDim
+        8,           // BBlockTransferSrcScalarPerVector
+        8,           // BBlockTransferDstScalarPerVector_BK1
+        1,           // BBlockLdsExtraN
+        1,
+        1,
+        S<1, 32, 1, 8>,
+        8,
+        AComputeDataType,
+        BComputeDataType>;
+
+#include "run_convnd_fwd_example.inc"
+
+int main(int argc, char* argv[]) { return run_convnd_fwd_example(argc, argv) ? 0 : 1; }

include/ck/ck.hpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -153,8 +153,8 @@ CK_DECLARE_ENV_VAR_BOOL(CK_LOGGING)
 // LDS direct loads using inline assembly
 #define CK_USE_AMD_LDS_DIRECT_LOAD_INLINE_ASM 0
 
-// set stochastic rounding as default for f8 conversions
-#define CK_USE_SR_F8_CONVERSION 1
+// set rounding to nearest even as default for f8 conversions
+#define CK_USE_SR_F8_CONVERSION 0
 
 // block synchronization only s_wait lgkmcnt(0), not vmcnt(0)
 #define CK_EXPERIMENTAL_BLOCK_SYNC_LDS_WITHOUT_SYNC_VMEM 1

include/ck/tensor_operation/gpu/element/combined_element_wise_operation.hpp

@@ -3,7 +3,6 @@
 
 #pragma once
 
-#include "ck/utility/data_type.hpp"
 #include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
 
 namespace ck {
@@ -107,6 +106,9 @@ struct TrinaryWithUnaryCombinedOp
     UnaryOp2 unary_op2_{};
 };
 
+using ScaleScalePass = UnaryCombinedOp<Scale, Scale, PassThrough>;
+using ScaleScaleRelu = UnaryCombinedOp<Scale, Scale, Relu>;
+
 } // namespace element_wise
 } // namespace tensor_operation
 } // namespace ck

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_convscale.hpp

@@ -8,9 +8,7 @@
 
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_abd.hpp"
-#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
-#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
-
+#include "ck/tensor_operation/gpu/element/combined_element_wise_operation.hpp"
 #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
 
 namespace ck {
@@ -177,6 +175,88 @@ struct DeviceOperationInstanceFactory<ck::tensor_operation::device::DeviceGroupe
     }
 };
 
+using CombConvScale = ck::tensor_operation::element_wise::ScaleScalePass;
+
+#ifdef CK_ENABLE_FP8
+void add_device_grouped_conv3d_fwd_xdl_combconvscale_ndhwgc_gkzyxc_ndhwgk_f8_f8_f32_instances(
+    std::vector<std::unique_ptr<DeviceGroupedConvFwdMultipleABD<3,
+                                                                NDHWGC,
+                                                                GKZYXC,
+                                                                ck::Tuple<>,
+                                                                NDHWGK,
+                                                                F8,
+                                                                F8,
+                                                                ck::Tuple<>,
+                                                                F32,
+                                                                PassThrough,
+                                                                PassThrough,
+                                                                CombConvScale,
+                                                                F8,
+                                                                F8>>>& instances);
+#endif
+
+template <ck::index_t NumDimSpatial,
+          typename InLayout,
+          typename WeiLayout,
+          typename DLayouts,
+          typename OutLayout,
+          typename InDataType,
+          typename WeiDataType,
+          typename DDataTypes,
+          typename OutDataType,
+          typename AComputeType,
+          typename BComputeType>
+struct DeviceOperationInstanceFactory<
+    ck::tensor_operation::device::DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
+                                                                  InLayout,
+                                                                  WeiLayout,
+                                                                  DLayouts,
+                                                                  OutLayout,
+                                                                  InDataType,
+                                                                  WeiDataType,
+                                                                  DDataTypes,
+                                                                  OutDataType,
+                                                                  PassThrough,
+                                                                  PassThrough,
+                                                                  CombConvScale,
+                                                                  AComputeType,
+                                                                  BComputeType>>
+{
+    using DeviceOp = DeviceGroupedConvFwdMultipleABD<NumDimSpatial,
+                                                     InLayout,
+                                                     WeiLayout,
+                                                     DLayouts,
+                                                     OutLayout,
+                                                     InDataType,
+                                                     WeiDataType,
+                                                     DDataTypes,
+                                                     OutDataType,
+                                                     PassThrough,
+                                                     PassThrough,
+                                                     CombConvScale,
+                                                     AComputeType,
+                                                     BComputeType>;
+
+    static auto GetInstances()
+    {
+        std::vector<std::unique_ptr<DeviceOp>> op_ptrs;
+        if constexpr(NumDimSpatial == 3 && is_same_v<InLayout, NDHWGC> &&
+                     is_same_v<WeiLayout, GKZYXC> && is_same_v<OutLayout, NDHWGK>)
+        {
+#ifdef CK_ENABLE_FP8
+            if constexpr(is_same_v<InDataType, f8_t> && is_same_v<WeiDataType, f8_t> &&
+                         is_same_v<OutDataType, F32> && is_same_v<AComputeType, f8_t> &&
+                         is_same_v<BComputeType, f8_t>)
+            {
+                add_device_grouped_conv3d_fwd_xdl_combconvscale_ndhwgc_gkzyxc_ndhwgk_f8_f8_f32_instances(
+                    op_ptrs);
+            }
+#endif
+        }
+        return op_ptrs;
+    }
+};
+
 } // namespace instance
 } // namespace device
 } // namespace tensor_operation

library/include/ck/library/tensor_operation_instance/gpu/grouped_convolution_forward_convscale_relu.hpp

@@ -9,7 +9,6 @@
 #include "ck/ck.hpp"
 #include "ck/tensor_operation/gpu/device/device_grouped_conv_fwd_multiple_abd.hpp"
 #include "ck/tensor_operation/gpu/element/combined_element_wise_operation.hpp"
-#include "ck/tensor_operation/gpu/element/unary_element_wise_operation.hpp"
 #include "ck/library/tensor_operation_instance/device_operation_instance_factory.hpp"
 
 namespace ck {
@@ -100,8 +99,7 @@ struct DeviceOperationInstanceFactory<
     }
 };
 
-namespace ew            = ck::tensor_operation::element_wise;
-using CombConvScaleRelu = ew::UnaryCombinedOp<ew::Scale, ew::Scale, ew::Relu>;
+using CombConvScaleRelu = ck::tensor_operation::element_wise::ScaleScaleRelu;
 
 #ifdef CK_ENABLE_FP8
 void add_device_grouped_conv3d_fwd_xdl_combconvscale_relu_ndhwgc_gkzyxc_ndhwgk_f8_f8_f32_instances(

library/include/ck/library/utility/check_err.hpp

 // SPDX-License-Identifier: MIT
-// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
 
 #pragma once
 
@@ -272,7 +272,8 @@ check_err(const Range& out,
     }
     if(!res)
     {
-        std::cerr << std::setw(12) << std::setprecision(7) << "max err: " << max_err << std::endl;
+        std::cerr << std::setw(12) << std::setprecision(7) << "max err: " << max_err
+                  << " number of errors: " << err_count << std::endl;
     }
     return res;
 }

library/src/tensor_operation_instance/gpu/grouped_conv3d_fwd_convscale/CMakeLists.txt

@@ -3,6 +3,7 @@ set(GROUPED_CONV3D_FWD_CONVSCALE
    xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_f8_instance.cpp
    xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_bf8_instance.cpp
    xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_f8_bf8_instance.cpp
-   xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_bf8_f8_instance.cpp)
+   xdl/device_grouped_conv3d_fwd_xdl_convscale_ndhwgc_gkzyxc_ndhwgk_bf8_f8_instance.cpp
+   xdl/device_grouped_conv3d_fwd_xdl_combconvscale_ndhwgc_gkzyxc_ndhwgk_f8_f8_f32_instance.cpp)
 
 add_instance_library(device_grouped_conv3d_fwd_convscale_instance ${GROUPED_CONV3D_FWD_CONVSCALE})