Merge branch 'develop' into gemm_elementwise_gemm

client_example/25_wrapper/README.md

+[Back to the main page](../../README.md)
 # Composable Kernel wrapper GEMM tutorial
 
-This tutorial demonstrates how to implement matrix multiplication using Composable Kernel (CK)
-wrapper. We present the base version of GEMM without most of the available optimizations; however,
-it's worth noting that CK has kernels with different optimizations.
+This tutorial demonstrates how to implement matrix multiplication using Composable Kernel (CK) wrapper. We present the base version of GEMM without most of the available optimizations; however, it's worth noting that CK has kernels with different optimizations.
 
-To implement these optimizations, you can use the CK wrapper or directly use available instances in
-CK. You can also refer to the
-[optimized GEMM example](https://github.com/ROCm/composable_kernel/blob/develop/client_example/25_wrapper/wrapper_optimized_gemm.cpp),
-that uses CK wrapper based on the
-[`gridwise_gemm_xdlops_v2r3`](https://github.com/ROCm/composable_kernel/blob/develop/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r3.hpp) implementation.
+To implement these optimizations, you can use the CK wrapper or directly use available instances in CK. You can also refer to the [optimized GEMM example](https://github.com/ROCm/composable_kernel/blob/develop/client_example/25_wrapper/wrapper_optimized_gemm.cpp), that uses CK wrapper based on the [`gridwise_gemm_xdlops_v2r3`](https://github.com/ROCm/composable_kernel/blob/develop/include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v2r3.hpp) implementation.
 
 The kernel definition should look similar to:
 

client_example/31_grouped_gemm_bf16Aint8B/grouped_gemm_multiply_bias_fastgelu_xdl_bf16_i8.cpp

@@ -121,7 +121,7 @@ bool run_grouped_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
     constexpr ck::index_t NumDTensor = 2;
 
     using GroupedGemmKernelArgument =
-        ck::tensor_operation::device::GroupedGemmTileLoopKernelArguments<NumDTensor>;
+        ck::tensor_operation::device::GroupedGemmKernelArgument<NumDTensor>;
 
     std::vector<GroupedGemmKernelArgument> grouped_gemm_kernel_args_;
     grouped_gemm_kernel_args_.reserve(group_count);

client_example/31_grouped_gemm_bf16Aint8B/grouped_gemm_multiply_xdl_bf16_i8.cpp

@@ -120,7 +120,7 @@ bool run_grouped_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
     constexpr ck::index_t NumDTensor = 1;
 
     using GroupedGemmKernelArgument =
-        ck::tensor_operation::device::GroupedGemmTileLoopKernelArguments<NumDTensor>;
+        ck::tensor_operation::device::GroupedGemmKernelArgument<NumDTensor>;
 
     std::vector<GroupedGemmKernelArgument> grouped_gemm_kernel_args_;
     grouped_gemm_kernel_args_.reserve(group_count);

client_example/CMakeLists.txt

@@ -56,13 +56,21 @@ if (GPU_TARGETS)
         add_definitions(-DCK_USE_WMMA)
         set(CK_USE_WMMA "ON")
     endif()
+    if (GPU_TARGETS MATCHES "gfx12" OR GPU_TARGETS MATCHES "gfx950")
+        add_definitions(-DCK_USE_OCP_FP8)
+        set(CK_USE_OCP_FP8 "ON")
+    endif()
+    if (GPU_TARGETS MATCHES "gfx90a" OR GPU_TARGETS MATCHES "gfx94")
+        add_definitions(-DCK_USE_FNUZ_FP8)
+        set(CK_USE_FNUZ_FP8 "ON")
+    endif()
 else()
     add_definitions(-DCK_USE_WMMA -DCK_USE_XDL)
     set(CK_USE_XDL "ON")
     set(CK_USE_WMMA "ON")
 endif()
 
-find_package(composable_kernel COMPONENTS device_other_operations device_gemm_operations device_conv_operations  device_reduction_operations)
+find_package(composable_kernel COMPONENTS device_other_operations device_gemm_operations device_conv_operations  device_reduction_operations utility)
 if(GPU_TARGETS MATCHES "gfx9")
     find_package(composable_kernel COMPONENTS device_contraction_operations)
 endif()

client_example/README.md

+[Back to the main page](../README.md)
+# Composable Kernel client examples
 ##
 Client application links to CK library, and therefore CK library needs to be installed before building client applications.
 

cmake/EnableCompilerWarnings.cmake

@@ -66,7 +66,7 @@ else()
             -Wunreachable-code
             -Wunused
             -Wno-reserved-identifier
-	    -Werror
+            -Werror
             -Wno-option-ignored
             -Wsign-compare
             -Wno-extra-semi-stmt

codegen/CMakeLists.txt

@@ -7,6 +7,7 @@ set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/lib)
 set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/lib)
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)
 set(CK_ROOT ${CMAKE_CURRENT_SOURCE_DIR}/..)
+configure_file(${CK_ROOT}/include/ck/config.h.in ${CK_ROOT}/include/ck/config.h)
 
 find_package(ROCM)
 include(ROCMInstallTargets)

codegen/README.md

+[Back to the main page](../README.md)
+# Composable Kernel codegen
\ No newline at end of file

codegen/driver/main.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
 
 #include <functional>
 #include <iostream>

codegen/src/headers.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #include "ck/host/headers.hpp"
 #include "ck_headers.hpp"
 

codegen/src/types.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #include "ck/host/types.hpp"
 #include "ck/host/stringutils.hpp"
 #include <algorithm>

codegen/test/gemm_multiple_d.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #include "ck/host/device_gemm_multiple_d/problem.hpp"
 #include "ck/host/device_gemm_multiple_d/operation.hpp"
 #include "ck/host/headers.hpp"

codegen/test/grouped_conv_fwd_multiple_d_v1.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #include "ck/host/device_grouped_conv_fwd_multiple_d/conv_fwd_op.hpp"
 #include "ck/host/device_grouped_conv_fwd_multiple_d/conv_fwd_problem.hpp"
 #include "ck/host/headers.hpp"

codegen/test/grouped_conv_fwd_multiple_d_v2.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #include "ck/host/device_grouped_conv_fwd_multiple_d/conv_fwd_op.hpp"
 #include "ck/host/device_grouped_conv_fwd_multiple_d/conv_fwd_problem.hpp"
 #include "ck/host/headers.hpp"

codegen/test/grouped_conv_fwd_multiple_d_v3.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #include "ck/host/device_grouped_conv_fwd_multiple_d/conv_fwd_op.hpp"
 #include "ck/host/device_grouped_conv_fwd_multiple_d/conv_fwd_problem.hpp"
 #include "ck/host/headers.hpp"

codegen/test/grouped_conv_fwd_multiple_d_v4.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #include "ck/host/device_grouped_conv_fwd_multiple_d/conv_fwd_op.hpp"
 #include "ck/host/device_grouped_conv_fwd_multiple_d/conv_fwd_problem.hpp"
 #include "ck/host/headers.hpp"

codegen/test/include/common.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #pragma once
 #include <algorithm>
 #include <cmath>

codegen/test/rtc/include/rtc/compile_kernel.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #ifndef GUARD_HOST_TEST_RTC_INCLUDE_RTC_COMPILE_KERNEL
 #define GUARD_HOST_TEST_RTC_INCLUDE_RTC_COMPILE_KERNEL
 

codegen/test/rtc/include/rtc/hip.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #ifndef GUARD_HOST_TEST_RTC_INCLUDE_RTC_HIP
 #define GUARD_HOST_TEST_RTC_INCLUDE_RTC_HIP
 
 #include <hip/hip_runtime_api.h>
 #include <memory>
+#include <stdexcept>
 #include <string>
 #include <stdexcept>
 

codegen/test/rtc/include/rtc/kernel.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2024-2025, Advanced Micro Devices, Inc. All rights reserved.
+
 #ifndef GUARD_HOST_TEST_RTC_INCLUDE_RTC_KERNEL
 #define GUARD_HOST_TEST_RTC_INCLUDE_RTC_KERNEL