Merge branch 'develop' into navi3_rel

.github/CODEOWNERS

-* @zjing14 @asroy @junliume @illsilin @carlushuang @aosewski
+* @zjing14 @junliume @illsilin @carlushuang @aosewski
 # Documentation files
-docs/* @saadrahim @LisaDelaney
-*.md  @saadrahim @LisaDelaney
-*.rst  @saadrahim @LisaDelaney
-# Header directory
-library/include/*  @saadrahim @LisaDelaney
+docs/* @ROCm/rocm-documentation
+*.md @ROCm/rocm-documentation
+*.rst @ROCm/rocm-documentation
+# Header directory for Doxygen documentation
+library/include/* @ROCm/rocm-documentation

Jenkinsfile

 def rocmnode(name) {
-    return '(rocmtest || miopen) && ' + name
+    return '(rocmtest || miopen) && (' + name + ')'
 }
 
 def show_node_info() {
@@ -7,6 +7,7 @@ def show_node_info() {
         echo "NODE_NAME = \$NODE_NAME"
         lsb_release -sd
         uname -r
+        cat /sys/module/amdgpu/version
         ls /opt/ -la
     """
 }
@@ -33,6 +34,10 @@ def runShell(String command){
 
 def getDockerImageName(){
     def img
+    if (params.USE_CUSTOM_DOCKER != ""){
+        img = "${params.USE_CUSTOM_DOCKER}"
+    }
+    else{
     if (params.ROCMVERSION != "6.0.1"){
        if (params.COMPILER_VERSION == "") {
            img = "${env.CK_DOCKERHUB}:ck_ub20.04_rocm${params.ROCMVERSION}"
@@ -61,6 +66,7 @@ def getDockerImageName(){
           }
        }
     }
+    }
     return img
 }
 
@@ -258,18 +264,24 @@ def cmake_build(Map conf=[:]){
             """)
         sh cmd3
     }
-
-    def setup_cmd = conf.get("setup_cmd", "${cmake_envs} cmake ${setup_args}   .. ")
     // reduce parallelism when compiling, clang uses too much memory
     def nt = nthreads()
-    def build_cmd = conf.get("build_cmd", "${build_envs} dumb-init make  -j${nt} ${config_targets}")
+    def cmd
     def execute_cmd = conf.get("execute_cmd", "")
-
-    def cmd = conf.get("cmd", """
+    if(!setup_args.contains("NO_CK_BUILD")){
+        def setup_cmd = conf.get("setup_cmd", "${cmake_envs} cmake ${setup_args}   .. ")
+        def build_cmd = conf.get("build_cmd", "${build_envs} dumb-init make  -j${nt} ${config_targets}")
+        cmd = conf.get("cmd", """
             ${setup_cmd}
             ${build_cmd}
             ${execute_cmd}
         """)
+    }
+    else{
+        cmd = conf.get("cmd", """
+            ${execute_cmd}
+        """)
+    }
 
     echo cmd
 
@@ -365,8 +377,8 @@ def runCKProfiler(Map conf=[:]){
                 (retimage, image) = getDockerImage(conf)
                 withDockerContainer(image: image, args: dockerOpts) {
                     timeout(time: 5, unit: 'MINUTES'){
-                        sh 'PATH="/opt/rocm/opencl/bin:/opt/rocm/opencl/bin/x86_64:$PATH" clinfo | tee clinfo.log'
-                        if ( runShell('grep -n "Number of devices:.*. 0" clinfo.log') ){
+                        sh 'rocminfo | tee rocminfo.log'
+                        if ( !runShell('grep -n "gfx" rocminfo.log') ){
                             throw new Exception ("GPU not found")
                         }
                         else{
@@ -379,20 +391,6 @@ def runCKProfiler(Map conf=[:]){
                 echo "The job was cancelled or aborted"
                 throw e
             }
-            catch(Exception ex) {
-                retimage = docker.build("${image}", dockerArgs + " --no-cache .")
-                withDockerContainer(image: image, args: dockerOpts) {
-                    timeout(time: 5, unit: 'MINUTES'){
-                        sh 'PATH="/opt/rocm/opencl/bin:/opt/rocm/opencl/bin/x86_64:$PATH" clinfo | tee clinfo.log'
-                        if ( runShell('grep -n "Number of devices:.*. 0" clinfo.log') ){
-                            throw new Exception ("GPU not found")
-                        }
-                        else{
-                            echo "GPU is OK"
-                        }
-                    }
-                }
-            }
 
             withDockerContainer(image: image, args: dockerOpts + ' -v=/var/jenkins/:/var/jenkins') {
                 timeout(time: 24, unit: 'HOURS')
@@ -473,6 +471,7 @@ def Build_CK(Map conf=[:]){
         show_node_info()
 
         env.HSA_ENABLE_SDMA=0
+        env.DOCKER_BUILDKIT=1
         checkout scm
 
         def image = getDockerImageName() 
@@ -487,25 +486,35 @@ def Build_CK(Map conf=[:]){
         if (params.COMPILER_VERSION == "amd-staging" || params.COMPILER_VERSION == "amd-mainline-open" || params.COMPILER_COMMIT != ""){
             dockerOpts = dockerOpts + " --env HIP_CLANG_PATH='/llvm-project/build/bin' "
         }
+        def video_id = sh(returnStdout: true, script: 'getent group video | cut -d: -f3')
+        def render_id = sh(returnStdout: true, script: 'getent group render | cut -d: -f3')
+        dockerOpts = dockerOpts + " --group-add=${video_id} --group-add=${render_id} "
+        echo "Docker flags: ${dockerOpts}"
 
         def variant = env.STAGE_NAME
         def retimage
         def navi_node = 0
+        def mi300_node = 0
 
-        gitStatusWrapper(credentialsId: "${status_wrapper_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
+        gitStatusWrapper(credentialsId: "${env.status_wrapper_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
             try {
                 (retimage, image) = getDockerImage(conf)
                 withDockerContainer(image: image, args: dockerOpts) {
                     timeout(time: 5, unit: 'MINUTES'){
-                        sh 'PATH="/opt/rocm/opencl/bin:/opt/rocm/opencl/bin/x86_64:$PATH" clinfo | tee clinfo.log'
-                        if ( runShell('grep -n "Number of devices:.*. 0" clinfo.log') ){
+                        sh 'rocminfo | tee rocminfo.log'
+                        if ( !runShell('grep -n "gfx" rocminfo.log') ){
                             throw new Exception ("GPU not found")
                         }
                         else{
                             echo "GPU is OK"
                         }
-                        if ( runShell('grep -n "gfx1030" clinfo.log') || runShell('grep -n "gfx1101" clinfo.log') ){
+                        if ( runShell('grep -n "gfx1030" rocminfo.log') || runShell('grep -n "gfx1101" rocminfo.log') ){
                             navi_node = 1
+                            echo "This is a Navi node"
+                        }
+                        if ( runShell('grep -n "gfx942" rocminfo.log') ){
+                            mi300_node = 1
+                            echo "This is MI300 node"
                         }
                     }
                 }
@@ -514,23 +523,6 @@ def Build_CK(Map conf=[:]){
                 echo "The job was cancelled or aborted"
                 throw e
             }
-            catch(Exception ex) {
-                retimage = docker.build("${image}", dockerArgs + " --no-cache .")
-                withDockerContainer(image: image, args: dockerOpts) {
-                    timeout(time: 5, unit: 'MINUTES'){
-                        sh 'PATH="/opt/rocm/opencl/bin:/opt/rocm/opencl/bin/x86_64:$PATH" clinfo |tee clinfo.log'
-                        if ( runShell('grep -n "Number of devices:.*. 0" clinfo.log') ){
-                            throw new Exception ("GPU not found")
-                        }
-                        else{
-                            echo "GPU is OK"
-                        }
-                        if ( runShell('grep -n "gfx1030" clinfo.log') || runShell('grep -n "gfx1101" clinfo.log') ){
-                            navi_node = 1
-                        }
-                    }
-                }
-            }
             withDockerContainer(image: image, args: dockerOpts + ' -v=/var/jenkins/:/var/jenkins') {
                 timeout(time: 24, unit: 'HOURS')
                 {
@@ -544,8 +536,8 @@ def Build_CK(Map conf=[:]){
                            sh 'tar -zcvf ckProfiler.tar.gz bin/ckProfiler'
                            stash "ckProfiler.tar.gz"
                         }
-                        if (params.RUN_FULL_QA){
-                           // build deb packages
+                        if (params.RUN_FULL_QA && mi300_node == 0 ){
+                           // build deb packages for all MI100/200/300 targets and prepare to export
                            sh 'make -j package'
                            archiveArtifacts artifacts: 'composablekernel-ckprofiler_*.deb'
                            archiveArtifacts artifacts: 'composablekernel-tests_*.deb'
@@ -610,7 +602,7 @@ def process_results(Map conf=[:]){
     def variant = env.STAGE_NAME
     def retimage
 
-    gitStatusWrapper(credentialsId: "${status_wrapper_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
+    gitStatusWrapper(credentialsId: "${env.status_wrapper_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
         try {
             (retimage, image) = getDockerImage(conf)
         }
@@ -678,6 +670,10 @@ pipeline {
             name: "BUILD_DOCKER",
             defaultValue: false,
             description: "Force building docker image (default: false), set to true if docker image needs to be updated.")
+        string(
+            name: 'USE_CUSTOM_DOCKER',
+            defaultValue: '',
+            description: 'If you want to use a custom docker image, please specify it here (default: leave blank).')
         string(
             name: 'ROCMVERSION', 
             defaultValue: '6.0', 
@@ -722,6 +718,10 @@ pipeline {
             name: "RUN_PERFORMANCE_TESTS",
             defaultValue: false,
             description: "Run the performance tests (default: OFF)")
+        booleanParam(
+            name: "RUN_CODEGEN_TESTS",
+            defaultValue: true,
+            description: "Run the codegen tests (default: ON)")
     }
     environment{
         dbuser = "${dbuser}"
@@ -800,7 +800,34 @@ pipeline {
                 }
             }
         }
-    
+        stage("Run Codegen Tests")
+        {
+            parallel
+            {
+                stage("Run Codegen Tests on MI100/MI200")
+                {
+                    when {
+                        beforeAgent true
+                        expression { params.RUN_CODEGEN_TESTS.toBoolean() }
+                    }
+                    options { retry(2) }
+                    agent{ label rocmnode("gfx908 || gfx90a")}
+                    environment{
+                        setup_args = "NO_CK_BUILD"
+                        execute_args = """ cd ../codegen && rm -rf build && mkdir build && cd build && \
+                                           cmake -D CMAKE_PREFIX_PATH=/opt/rocm \
+                                           -D CMAKE_CXX_COMPILER=/opt/rocm/llvm/bin/clang++ \
+                                           -D CMAKE_BUILD_TYPE=Release \
+                                           -D GPU_TARGETS="gfx908;gfx90a" \
+                                           -DCMAKE_CXX_FLAGS=" -O3 " .. && make -j check"""
+                   }
+                    steps{
+                        buildHipClangJobAndReboot(setup_args:setup_args, no_reboot:true, build_type: 'Release', execute_cmd: execute_args)
+                        cleanWs()
+                    }
+                }
+            }
+        }
 		stage("Build CK and run Tests")
         {
             parallel
@@ -828,6 +855,26 @@ pipeline {
                         cleanWs()
                     }
                 }
+                stage("Build CK and run Tests on MI300")
+                {
+                    when {
+                        beforeAgent true
+                        expression { params.RUN_FULL_QA.toBoolean() }
+                    }
+                    agent{ label rocmnode("gfx942") }
+                    environment{
+                        setup_args = """ -DCMAKE_INSTALL_PREFIX=../install -DGPU_TARGETS="gfx942" -DCMAKE_CXX_FLAGS=" -O3 " """
+                        execute_args = """ cd ../client_example && rm -rf build && mkdir build && cd build && \
+                                           cmake -DCMAKE_PREFIX_PATH="${env.WORKSPACE}/install;/opt/rocm" \
+                                           -DGPU_TARGETS="gfx942" \
+                                           -DCMAKE_CXX_COMPILER="${build_compiler()}" \
+                                           -DCMAKE_CXX_FLAGS=" -O3 " .. && make -j """
+                    }
+                    steps{
+                        Build_CK_and_Reboot(setup_args: setup_args, config_targets: "install", no_reboot:true, build_type: 'Release', execute_cmd: execute_args, prefixpath: '/usr/local')
+                        cleanWs()
+                    }
+                }
                 stage("Build CK and run Tests on MI100/MI200")
                 {
                     when {

cmake/Embed.cmake

+#####################################################################################
+# The MIT License (MIT)
+#
+# Copyright (c) 2015-2024 Advanced Micro Devices, Inc. All rights reserved.
+#
+# Permission is hereby granted, free of charge, to any person obtaining a copy
+# of this software and associated documentation files (the "Software"), to deal
+# in the Software without restriction, including without limitation the rights
+# to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+# copies of the Software, and to permit persons to whom the Software is
+# furnished to do so, subject to the following conditions:
+#
+# The above copyright notice and this permission notice shall be included in
+# all copies or substantial portions of the Software.
+#
+# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL THE
+# AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+# OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+# THE SOFTWARE.
+#####################################################################################
+
+if(WIN32)
+    set(EMBED_USE RC CACHE STRING "Use RC or CArrays to embed data files")
+    set_property(CACHE EMBED_USE PROPERTY STRINGS "RC;CArrays")
+else()
+    if(BUILD_SHARED_LIBS)
+        set(EMBED_USE LD CACHE STRING "Use LD or CArrays to embed data files")
+    else()
+        set(EMBED_USE CArrays CACHE STRING "Use LD or CArrays to embed data files")
+    endif()
+    set_property(CACHE EMBED_USE PROPERTY STRINGS "LD;CArrays")
+endif()
+
+if(EMBED_USE STREQUAL "LD")
+    find_program(EMBED_LD ld REQUIRED)
+    find_program(EMBED_OBJCOPY objcopy REQUIRED)
+endif()
+
+function(embed_wrap_string)
+    set(options)
+    set(oneValueArgs VARIABLE AT_COLUMN)
+    set(multiValueArgs)
+    cmake_parse_arguments(PARSE "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+
+    string(LENGTH ${${PARSE_VARIABLE}} string_length)
+    math(EXPR offset "0")
+
+    while(string_length GREATER 0)
+
+        if(string_length GREATER ${PARSE_AT_COLUMN})
+            math(EXPR length "${PARSE_AT_COLUMN}")
+        else()
+            math(EXPR length "${string_length}")
+        endif()
+
+        string(SUBSTRING ${${PARSE_VARIABLE}} ${offset} ${length} line)
+        set(lines "${lines}\n${line}")
+
+        math(EXPR string_length "${string_length} - ${length}")
+        math(EXPR offset "${offset} + ${length}")
+    endwhile()
+
+    set(${PARSE_VARIABLE} "${lines}" PARENT_SCOPE)
+endfunction()
+
+function(generate_embed_source EMBED_NAME EMBED_DIR BASE_DIRECTORY)
+    set(options)
+    set(oneValueArgs)
+    set(multiValueArgs SYMBOLS FILES)
+    cmake_parse_arguments(PARSE "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+
+    set(RESOURCE_ID 100)
+
+    list(LENGTH PARSE_SYMBOLS SYMBOLS_LEN)
+    list(LENGTH PARSE_FILES FILES_LEN)
+    if(NOT ${SYMBOLS_LEN} EQUAL ${FILES_LEN})
+        message(FATAL_ERROR "Symbols and objects dont match: ${SYMBOLS_LEN} != ${FILES_LEN}")
+    endif()
+    math(EXPR LEN "${SYMBOLS_LEN} - 1")
+
+    foreach(idx RANGE ${LEN})
+        list(GET PARSE_SYMBOLS ${idx} SYMBOL)
+        list(GET PARSE_FILES ${idx} FILE)
+        file(RELATIVE_PATH BASE_NAME "${BASE_DIRECTORY}" ${FILE})
+        if(EMBED_USE STREQUAL "RC")
+            string(TOUPPER "${SYMBOL}" SYMBOL)
+            string(APPEND FILE_IDS "#define IDR_${SYMBOL} ${RESOURCE_ID}\n")
+            file(TO_NATIVE_PATH "${FILE}" NATIVE_FILE)
+            string(REPLACE "\\" "\\\\" NATIVE_FILE "${NATIVE_FILE}")
+            string(APPEND RC_FILE_MAPPING "IDR_${SYMBOL} TEXTFILE \"${NATIVE_FILE}\"\n")
+            string(APPEND INIT_KERNELS "\n        {\"${BASE_NAME}\", resource::read(IDR_${SYMBOL})},")
+            math(EXPR RESOURCE_ID "${RESOURCE_ID} + 1" OUTPUT_FORMAT DECIMAL)
+        else()
+            set(START_SYMBOL "_binary_${SYMBOL}_start")
+            set(LENGTH_SYMBOL "_binary_${SYMBOL}_length")
+            if(EMBED_USE STREQUAL "LD")
+                string(APPEND EXTERNS "
+extern const char ${START_SYMBOL}[];
+extern const size_t _binary_${SYMBOL}_size;
+const auto ${LENGTH_SYMBOL} = reinterpret_cast<size_t>(&_binary_${SYMBOL}_size);
+")
+            else()
+                string(APPEND EXTERNS "
+extern const char ${START_SYMBOL}[];
+extern const size_t ${LENGTH_SYMBOL};
+")
+            endif()
+            string(APPEND INIT_KERNELS "
+        { \"${BASE_NAME}\", { ${START_SYMBOL}, ${LENGTH_SYMBOL}} },")
+        endif()
+    endforeach()
+    if(EMBED_USE STREQUAL "RC")
+       file(WRITE "${EMBED_DIR}/include/resource.h" "
+#define TEXTFILE 256
+
+${FILE_IDS}
+")
+        file(WRITE "${EMBED_DIR}/resource.rc" "
+#include \"resource.h\"
+
+${RC_FILE_MAPPING}
+")
+        set(EXTERNS "
+#include <Windows.h>
+#include \"resource.h\"
+
+namespace resource {
+std::string_view read(int id)
+{
+    HMODULE handle = GetModuleHandle(nullptr);
+    HRSRC rc = FindResource(handle, MAKEINTRESOURCE(id), MAKEINTRESOURCE(TEXTFILE));
+    HGLOBAL data = LoadResource(handle, rc);
+    return {static_cast<const char*>(LockResource(data)), SizeofResource(handle, rc)};
+}
+}
+")
+        set(EMBED_FILES ${EMBED_DIR}/include/resource.h ${EMBED_DIR}/resource.rc)
+    endif()
+    file(WRITE "${EMBED_DIR}/include/${EMBED_NAME}.hpp" "
+#include <string_view>
+#include <unordered_map>
+#include <utility>
+std::unordered_map<std::string_view, std::string_view> ${EMBED_NAME}();
+")
+
+    file(WRITE "${EMBED_DIR}/${EMBED_NAME}.cpp" "
+#include <${EMBED_NAME}.hpp>
+${EXTERNS}
+std::unordered_map<std::string_view, std::string_view> ${EMBED_NAME}()
+{
+    static std::unordered_map<std::string_view, std::string_view> result = {${INIT_KERNELS}
+    };
+    return result;
+}
+")
+    list(APPEND EMBED_FILES ${EMBED_DIR}/${EMBED_NAME}.cpp ${EMBED_DIR}/include/${EMBED_NAME}.hpp)
+    set(EMBED_FILES ${EMBED_FILES} PARENT_SCOPE)
+endfunction()
+
+function(embed_file FILE BASE_DIRECTORY)
+    message(STATUS "    ${FILE}")
+    file(RELATIVE_PATH REL_FILE "${BASE_DIRECTORY}" ${FILE})
+    string(MAKE_C_IDENTIFIER "${REL_FILE}" OUTPUT_SYMBOL)
+    get_filename_component(OUTPUT_FILE_DIR "${REL_FILE}" DIRECTORY)
+    file(MAKE_DIRECTORY "${CMAKE_CURRENT_BINARY_DIR}/${OUTPUT_FILE_DIR}")
+    if(EMBED_USE STREQUAL "LD")
+        set(OUTPUT_FILE "${CMAKE_CURRENT_BINARY_DIR}/${REL_FILE}.o")
+        add_custom_command(
+            OUTPUT "${OUTPUT_FILE}"
+            COMMAND ${EMBED_LD} -r -o "${OUTPUT_FILE}" -z noexecstack --format=binary "${REL_FILE}"
+            COMMAND ${EMBED_OBJCOPY} --rename-section .data=.rodata,alloc,load,readonly,data,contents "${OUTPUT_FILE}"
+            WORKING_DIRECTORY "${BASE_DIRECTORY}"
+            DEPENDS "${FILE}"
+            VERBATIM)
+        set(OUTPUT_FILE ${OUTPUT_FILE} PARENT_SCOPE)
+    elseif(EMBED_USE STREQUAL "CArrays")
+        set_property(DIRECTORY APPEND PROPERTY CMAKE_CONFIGURE_DEPENDS ${FILE})
+        set(OUTPUT_FILE "${CMAKE_CURRENT_BINARY_DIR}/${REL_FILE}.cpp")
+        # reads source file contents as hex string
+        file(READ ${FILE} HEX_STRING HEX)
+        # wraps the hex string into multiple lines
+        embed_wrap_string(VARIABLE HEX_STRING AT_COLUMN 80)
+        # adds '0x' prefix and comma suffix before and after every byte respectively
+        string(REGEX REPLACE "([0-9a-f][0-9a-f])" "0x\\1, " ARRAY_VALUES ${HEX_STRING})
+        # removes trailing comma
+        string(REGEX REPLACE ", $" "" ARRAY_VALUES ${ARRAY_VALUES})
+        file(WRITE "${OUTPUT_FILE}" "
+#include <cstddef>
+extern const char _binary_${OUTPUT_SYMBOL}_start[] = { ${ARRAY_VALUES} };
+extern const size_t _binary_${OUTPUT_SYMBOL}_length = sizeof(_binary_${OUTPUT_SYMBOL}_start);
+")
+        set(OUTPUT_FILE ${OUTPUT_FILE} PARENT_SCOPE)
+    endif()
+    set(OUTPUT_SYMBOL ${OUTPUT_SYMBOL} PARENT_SCOPE)
+endfunction()
+
+function(add_embed_library EMBED_NAME)
+    set(options)
+    set(oneValueArgs RELATIVE)
+    set(multiValueArgs)
+    cmake_parse_arguments(PARSE "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
+
+    set(EMBED_DIR ${CMAKE_CURRENT_BINARY_DIR}/embed/${EMBED_NAME})
+    file(MAKE_DIRECTORY ${EMBED_DIR})
+    message(STATUS "Embedding kernel files:")
+    foreach(FILE ${PARSE_UNPARSED_ARGUMENTS})
+        embed_file(${FILE} ${PARSE_RELATIVE})
+        list(APPEND OUTPUT_FILES ${OUTPUT_FILE})
+        list(APPEND SYMBOLS ${OUTPUT_SYMBOL})
+    endforeach()
+    message(STATUS "Generating embedding library '${EMBED_NAME}'")
+    generate_embed_source(${EMBED_NAME} ${EMBED_DIR} "${PARSE_RELATIVE}" SYMBOLS ${SYMBOLS} FILES ${PARSE_UNPARSED_ARGUMENTS})
+    set(INTERNAL_EMBED_LIB embed_lib_${EMBED_NAME})
+    if(EMBED_USE STREQUAL "LD")
+        add_library(${INTERNAL_EMBED_LIB} STATIC ${EMBED_FILES} ${OUTPUT_FILES})
+    else()
+        add_library(${INTERNAL_EMBED_LIB} OBJECT ${EMBED_FILES})
+    endif()
+    if(EMBED_USE STREQUAL "CArrays")
+        target_sources(${INTERNAL_EMBED_LIB} PRIVATE ${OUTPUT_FILES})
+    endif()
+    target_include_directories(${INTERNAL_EMBED_LIB} PRIVATE "${EMBED_DIR}/include")
+    target_compile_options(${INTERNAL_EMBED_LIB} PRIVATE -Wno-reserved-identifier -Wno-extern-initializer -Wno-missing-variable-declarations)
+    set_target_properties(${INTERNAL_EMBED_LIB} PROPERTIES POSITION_INDEPENDENT_CODE On)
+    add_library(${EMBED_NAME} INTERFACE)
+    if(EMBED_USE STREQUAL "RC")
+        target_link_libraries(${EMBED_NAME} INTERFACE $<TARGET_OBJECTS:${INTERNAL_EMBED_LIB}>)
+    elseif(EMBED_USE STREQUAL "LD")
+        target_link_libraries(${EMBED_NAME} INTERFACE ${INTERNAL_EMBED_LIB})
+    else()
+        target_sources(${EMBED_NAME} INTERFACE $<TARGET_OBJECTS:${INTERNAL_EMBED_LIB}>)
+    endif()
+    target_include_directories(${EMBED_NAME} INTERFACE "${EMBED_DIR}/include")
+endfunction()
+

codegen/CMakeLists.txt

+cmake_minimum_required(VERSION 3.16)
+project(composable_kernel_host)
+
+set(CMAKE_EXPORT_COMPILE_COMMANDS ON)
+
+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}/..)
+
+find_package(ROCM)
+include(ROCMInstallTargets)
+include(ROCMTest)
+
+list(APPEND CMAKE_MODULE_PATH ${CK_ROOT}/cmake)
+include(Embed)
+file(GLOB_RECURSE KERNEL_FILES CONFIGURE_DEPENDS
+    ${CK_ROOT}/include/ck/*.hpp)
+message(STATUS "KERNEL_FILES: ${KERNEL_FILES}")
+message(STATUS "RELATIVE: ${CK_ROOT}/include")
+add_embed_library(ck_headers ${KERNEL_FILES} RELATIVE ${CK_ROOT}/include)
+
+add_definitions(-std=c++17)
+
+file(GLOB SOURCES CONFIGURE_DEPENDS src/*.cpp)
+# TODO: Use object library
+add_library(ck_host STATIC ${SOURCES})
+target_link_libraries(ck_host PRIVATE ck_headers)
+
+set_target_properties(ck_host PROPERTIES 
+    LINKER_LANGUAGE CXX
+    POSITION_INDEPENDENT_CODE ON)
+
+target_include_directories(ck_host PUBLIC
+    $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/include>
+)
+
+add_executable(ck-template-driver driver/main.cpp)
+target_link_libraries(ck-template-driver ck_host)
+
+rocm_install(
+    TARGETS ck_host ck_headers
+    EXPORT ck_hostTargets
+)
+rocm_install(DIRECTORY include/ck DESTINATION ${CMAKE_INSTALL_INCLUDEDIR})
+
+if(BUILD_TESTING)
+add_subdirectory(test)
+endif()

codegen/driver/main.cpp

+
+#include <functional>
+#include <iostream>
+#include <string>
+#include <unordered_map>
+#include <vector>
+#include "ck/host/device_gemm_multiple_d/operation.hpp"
+#include "ck/host/stringutils.hpp"
+
+using ck::host::Transform;
+
+struct Emitters
+{
+    std::unordered_map<std::string, std::function<std::vector<std::string>()>> m;
+
+    template <class T>
+    void Register(const std::string& name)
+    {
+        m[name] = [] {
+            auto configs = T::CreateOperations();
+
+            return Transform(configs, [](const auto& ops) { return ToTuple(ops); });
+        };
+    }
+
+    template <class T>
+    static std::string ToTuple(const T& ops)
+    {
+        auto templates = Transform(
+            ops, [](const auto& op) { return "    " + op.ToSolution().ToTemplateString(); });
+        return "std::tuple<\n" + ck::host::JoinStrings(templates, ",\n") + ">";
+    }
+
+    std::string Emit(const std::string& name) { return ck::host::JoinStrings(m.at(name)(), "\n"); }
+
+    std::vector<std::string> List() const
+    {
+        return Transform(m, [](auto&& p) { return p.first; });
+    }
+};
+
+int main(int argc, const char* argv[])
+{
+    std::string prog = argv[0];
+    std::vector<std::string> args(argv + 1, argv + argc);
+    Emitters e;
+    e.Register<ck::host::device_gemm_multiple_d::Operation_Xdl_CShuffle>(
+        "DeviceGemmMultipleD_Xdl_CShuffle");
+
+    if(args.empty() or std::any_of(args.begin(), args.end(), [](auto arg) {
+           return arg == "-h" or arg == "--help";
+       }))
+    {
+        std::cout << "USAGE:" << std::endl;
+        std::cout << "    " << prog << " [TEMPLATE]" << std::endl;
+        std::cout << std::endl;
+        std::cout << "FLAGS:" << std::endl;
+        std::cout << "    -h, --help                     Show help" << std::endl;
+        std::cout << std::endl;
+        std::cout << "TEMPLATES:" << std::endl;
+        for(auto x : e.List())
+            std::cout << "    " << x << std::endl;
+        std::cout << std::endl;
+        return 0;
+    }
+
+    for(auto name : args)
+        std::cout << e.Emit(name) << std::endl;
+
+    return 0;
+}

codegen/include/ck/host/device_gemm_multiple_d.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <cstdlib>
+#include <vector>
+#include <memory>
+#include <sstream>
+#include <iterator>
+#include <numeric>
+#include "ck/host/types.hpp"
+
+namespace ck {
+namespace host {
+namespace device_gemm_multiple_d {
+
+struct Problem
+{
+    std::size_t M                    = 0;
+    std::size_t N                    = 0;
+    std::size_t K                    = 0;
+    bool TransA                      = false;
+    bool TransB                      = false;
+    bool TransE                      = false;
+    std::vector<bool> DsTrans        = {};
+    DataType ADataType               = DataType::Half;
+    DataType BDataType               = DataType::Half;
+    DataType EDataType               = DataType::Half;
+    std::vector<DataType> DsDataType = {};
+    std::string AElementOp           = "ck::tensor_operation::element_wise::PassThrough";
+    std::string BElementOp           = "ck::tensor_operation::element_wise::PassThrough";
+    std::string CDEElementOp         = "ck::Tuple<>";
+
+    std::string GetIncludeHeader() const;
+
+    std::vector<Solution> GetSolutions(const std::string& arch) const;
+};
+
+} // namespace device_gemm_multiple_d
+} // namespace host
+} // namespace ck

codegen/include/ck/host/device_gemm_multiple_d/operation.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <cstdlib>
+#include <vector>
+#include <string>
+#include "ck/host/types.hpp"
+#include "ck/host/operation/gemm.hpp"
+#include "ck/host/device_gemm_multiple_d/problem.hpp"
+
+namespace ck {
+namespace host {
+namespace device_gemm_multiple_d {
+
+struct Operation_Xdl_CShuffle
+{
+    static std::vector<std::vector<Operation_Xdl_CShuffle>> CreateOperations();
+    static std::vector<Operation_Xdl_CShuffle> CreateOperations(const Problem& prob);
+    TensorDesc A{};
+    TensorDesc B{};
+    DataType acc               = DataType::Float;
+    DataType cs_type           = DataType::Half;
+    std::vector<TensorDesc> Ds = {};
+    TensorDesc E{};
+    std::string a_elem_op           = PassThrough;
+    std::string b_elem_op           = PassThrough;
+    std::string cde_elem_op         = Bilinear;
+    std::string gemm_specialization = "ck::tensor_operation::device::GemmSpecialization::Default";
+    operation::TileDesc tile_desc{};
+    operation::BlockTransferDesc a_block_transfer{};
+    operation::BlockTransferDesc b_block_transfer{};
+    operation::CShuffleDesc cshuffle{};
+    operation::CBlockTransferDesc c_block_transfer{};
+
+    Solution ToSolution() const;
+};
+
+} // namespace device_gemm_multiple_d
+} // namespace host
+} // namespace ck

codegen/include/ck/host/device_gemm_multiple_d/problem.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <cstdlib>
+#include <vector>
+#include <string>
+#include "ck/host/types.hpp"
+
+namespace ck {
+namespace host {
+namespace device_gemm_multiple_d {
+
+struct Problem
+{
+    std::size_t M                    = 0;
+    std::size_t N                    = 0;
+    std::size_t K                    = 0;
+    bool TransA                      = false;
+    bool TransB                      = false;
+    bool TransE                      = false;
+    std::vector<bool> DsTrans        = {};
+    DataType ADataType               = DataType::Half;
+    DataType BDataType               = DataType::Half;
+    DataType EDataType               = DataType::Half;
+    std::vector<DataType> DsDataType = {};
+    std::string AElementOp           = PassThrough;
+    std::string BElementOp           = PassThrough;
+    std::string CDEElementOp         = PassThrough;
+
+    std::string GetIncludeHeader() const;
+
+    std::vector<Solution> GetSolutions(const std::string& arch) const;
+};
+
+} // namespace device_gemm_multiple_d
+} // namespace host
+} // namespace ck

codegen/include/ck/host/headers.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <string>
+#include <string_view>
+#include <utility>
+#include <unordered_map>
+#include <vector>
+
+namespace ck {
+namespace host {
+
+std::unordered_map<std::string_view, std::string_view> GetHeaders();
+
+} // namespace host
+} // namespace ck

codegen/include/ck/host/operation/gemm.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <string>
+
+namespace ck {
+namespace host {
+namespace operation {
+
+struct TileDesc
+{
+    int block_size               = 0;
+    int m_per_block              = 0;
+    int n_per_block              = 0;
+    int k_per_block              = 0;
+    int ak1                      = 0;
+    int bk1                      = 0;
+    int m_per_XDL                = 0;
+    int n_per_XDL                = 0;
+    int m_Xdl_per_wave           = 0;
+    int n_Xdl_per_wave           = 0;
+    int num_gemmk_prefetch_stage = 0;
+};
+struct BlockTransferDesc
+{
+    std::string thread_cluster_length        = "";
+    std::string thread_cluster_arrange_order = "";
+    std::string src_access_order             = "";
+    int src_vec_dim                          = 0;
+    int src_scalar_per_vector                = 0;
+    int dst_scalar_per_vector_k1             = 0;
+    int lds_add_extra_dim                    = 0;
+};
+struct CShuffleDesc
+{
+    int m_Xdl_per_wave_per_shuffle = 0;
+    int n_Xdl_per_wave_per_shuffle = 0;
+};
+struct CBlockTransferDesc
+{
+    std::string cluster_lengths_m_block_m_wave_m_per_Xdl_n_block_n_wave_n_per_Xdl = "";
+    int scalar_per_vector_n_wave_n_per_Xdl                                        = 0;
+};
+
+} // namespace operation
+} // namespace host
+} // namespace ck

codegen/include/ck/host/stringutils.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <algorithm>
+#include <cassert>
+#include <numeric>
+#include <string>
+#include <utility>
+#include <unordered_map>
+#include <vector>
+
+namespace ck {
+namespace host {
+
+template <class F>
+std::string trim(const std::string& s, F f)
+{
+    auto start = std::find_if_not(s.begin(), s.end(), f);
+    auto last  = std::find_if_not(s.rbegin(), std::string::const_reverse_iterator(start), f).base();
+    return {start, last};
+}
+
+inline std::string trim(const std::string& s)
+{
+    return trim(s, [](unsigned char c) { return std::isspace(c); });
+}
+
+template <class Strings>
+inline std::string JoinStrings(Strings strings, const std::string& delim)
+{
+    auto it = strings.begin();
+    if(it == strings.end())
+        return "";
+
+    auto nit = std::next(it);
+    return std::accumulate(nit, strings.end(), *it, [&](std::string x, std::string y) {
+        return std::move(x) + delim + std::move(y);
+    });
+}
+
+template <class F>
+inline std::string
+InterpolateString(const std::string& input, F f, std::string start = "${", std::string end = "}")
+{
+    std::string result = "";
+    result.reserve(input.size());
+    auto it = input.begin();
+    while(it != input.end())
+    {
+        auto next_start = std::search(it, input.end(), start.begin(), start.end());
+        auto next_end   = std::search(next_start, input.end(), end.begin(), end.end());
+        result.append(it, next_start);
+        if(next_start == input.end())
+            break;
+        if(next_end == input.end())
+        {
+            throw std::runtime_error("Unbalanced brackets");
+        }
+        auto r = f(next_start + start.size(), next_end);
+        result.append(r.begin(), r.end());
+        it = next_end + end.size();
+    }
+    return result;
+}
+inline std::string InterpolateString(const std::string& input,
+                                     const std::unordered_map<std::string, std::string>& vars,
+                                     std::string start = "${",
+                                     std::string end   = "}")
+{
+    return InterpolateString(
+        input,
+        [&](auto start_it, auto last_it) {
+            auto key = trim({start_it, last_it});
+            auto it  = vars.find(key);
+            if(it == vars.end())
+                throw std::runtime_error("Unknown key: " + key);
+            return it->second;
+        },
+        std::move(start),
+        std::move(end));
+}
+
+template <class Range, class F>
+inline auto Transform(const Range& r, F f) -> std::vector<decltype(f(*r.begin()))>
+{
+    std::vector<decltype(f(*r.begin()))> result;
+    std::transform(r.begin(), r.end(), std::back_inserter(result), f);
+    return result;
+}
+
+template <class Range1, class Range2, class F>
+inline auto Transform(const Range1& r1, const Range2& r2, F f)
+    -> std::vector<decltype(f(*r1.begin(), *r2.begin()))>
+{
+    std::vector<decltype(f(*r1.begin(), *r2.begin()))> result;
+    assert(std::distance(r1.begin(), r1.end()) == std::distance(r2.begin(), r2.end()));
+    std::transform(r1.begin(), r1.end(), r2.begin(), std::back_inserter(result), f);
+    return result;
+}
+
+} // namespace host
+} // namespace ck

codegen/include/ck/host/types.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <string>
+#include <sstream>
+#include <utility>
+#include <unordered_map>
+#include <vector>
+
+namespace ck {
+namespace host {
+
+struct Solution
+{
+
+    Solution() = default;
+    Solution(std::string str, std::unordered_map<std::string, std::string> values);
+    std::string ToTemplateString() const;
+    std::string GetTemplateParameter(const std::string& name) const;
+    template <class T>
+    T GetTemplateParameter(const std::string& name) const
+    {
+        T result;
+        std::stringstream ss(GetTemplateParameter(name));
+        ss >> result;
+        return result;
+    }
+
+    private:
+    std::string template_str;
+    std::unordered_map<std::string, std::string> template_values;
+};
+
+enum class DataType
+{
+    Half,
+    Float,
+    Int8,
+    Int32
+};
+
+std::string ToString(DataType dt);
+
+enum class Layout
+{
+    Row,
+    Column
+};
+
+std::string ToString(Layout dl);
+
+enum class GemmType
+{
+    Default
+};
+
+std::string ToString(GemmType gt);
+
+struct TensorDesc
+{
+    DataType element;
+    Layout layout;
+};
+
+std::string SequenceStr(const std::vector<int>& v);
+
+std::string MakeTuple(const std::vector<std::string>& v);
+
+template <int... xs>
+const std::string S = SequenceStr({xs...});
+
+constexpr const char* PassThrough = "ck::tensor_operation::element_wise::PassThrough";
+constexpr const char* Bilinear    = "ck::tensor_operation::element_wise::Bilinear";
+
+} // namespace host
+} // namespace ck

codegen/include/ck/host/utils.hpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#pragma once
+
+#include <cstdint>
+#include <unordered_set>
+
+namespace ck {
+namespace host {
+
+std::size_t integer_divide_ceil(std::size_t x, std::size_t y);
+
+const std::unordered_set<std::string>& get_xdlop_archs();
+
+} // namespace host
+} // namespace ck

codegen/src/device_gemm_multiple_d.cpp

+
+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, 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/utils.hpp"
+#include <algorithm>
+
+namespace ck {
+namespace host {
+namespace device_gemm_multiple_d {
+
+std::string Problem::GetIncludeHeader() const
+{
+    return "ck/tensor_operation/gpu/device/impl/device_gemm_multiple_d_xdl_cshuffle.hpp";
+}
+
+std::vector<Solution> Problem::GetSolutions(const std::string& arch) const
+{
+    if(get_xdlop_archs().count(arch) == 0)
+        return {};
+    auto ops = ck::host::device_gemm_multiple_d::Operation_Xdl_CShuffle::CreateOperations(*this);
+    std::vector<Solution> result;
+    std::transform(ops.begin(), ops.end(), std::back_inserter(result), [&](const auto& op) {
+        return op.ToSolution();
+    });
+    return result;
+}
+
+} // namespace device_gemm_multiple_d
+} // namespace host
+} // namespace ck
\ No newline at end of file

codegen/src/device_gemm_multiple_d_operation_xdl_cshuffle.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2024, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/host/device_gemm_multiple_d/operation.hpp"
+#include "ck/host/stringutils.hpp"
+#include "ck/host/utils.hpp"
+#include <cassert>
+
+namespace ck {
+namespace host {
+namespace device_gemm_multiple_d {
+
+static std::string GetGemmSpec(const std::size_t m,
+                               const std::size_t n,
+                               const std::size_t k,
+                               const std::size_t m_per_block,
+                               const std::size_t n_per_block,
+                               const std::size_t k_per_block)
+{
+    std::string spec = "";
+    if(integer_divide_ceil(m, m_per_block) * m_per_block - m != 0)
+        spec += "M";
+    if(integer_divide_ceil(n, n_per_block) * n_per_block - n != 0)
+        spec += "N";
+    if(integer_divide_ceil(k, k_per_block) * k_per_block - k != 0)
+        spec += "K";
+    if(spec == "")
+        return "ck::tensor_operation::device::GemmSpecialization::Default";
+
+    return "ck::tensor_operation::device::GemmSpecialization::" + spec + "Padding";
+}
+
+static Layout ToLayout(bool Trans) { return Trans ? Layout::Column : Layout::Row; }
+
+std::vector<Operation_Xdl_CShuffle> Operation_Xdl_CShuffle::CreateOperations(const Problem& prob)
+{
+    std::vector<Operation_Xdl_CShuffle> result;
+
+    std::vector<operation::TileDesc> tile_descriptions = {
+        // clang-format off
+//  Block|  MPer|  NPer|  KPer| AK1| BK1| MPer| NPer| MXdl| NXdl| NumGemmK|
+//   Size| Block| Block| Block|    |    |  XDL|  XDL|  Per|  Per| Prefetch|
+//       |      |      |      |    |    |     |     | Wave| Wave|    Stage|
+//       |      |      |      |    |    |     |     |     |     |         |
+  {   256,   256,   128,    32,   8,   8,   32,   32,    4,    2,        1},
+  {   256,   128,   256,    32,   8,   8,   32,   32,    2,    4,        1},
+  {   128,   128,   128,    32,   8,   8,   32,   32,    4,    2,        1},
+  {   256,   128,   128,    32,   8,   8,   32,   32,    2,    2,        1},
+  {   128,   128,    64,    32,   8,   8,   32,   32,    2,    2,        1},
+  {   128,    64,   128,    32,   8,   8,   32,   32,    2,    2,        1},
+  {   256,   128,    64,    32,   8,   8,   32,   32,    2,    1,        1},
+  {   256,    64,   128,    32,   8,   8,   32,   32,    1,    2,        1},
+        // clang-format on
+    };
+
+    std::vector<operation::BlockTransferDesc> a_block_descriptions_rowmajor = {
+        // clang-format off
+//  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|
+//   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|
+// Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          |
+//                |               |               |               |               |               |          |
+  {    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},
+  {    S<4, 32, 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},
+  {    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},
+  {    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},
+        // clang-format on
+    };
+
+    std::vector<operation::BlockTransferDesc> a_block_descriptions_colmajor = {
+        // clang-format off
+//  ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockTransfer| ABlockLds|
+//   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraM|
+// Lengths_K0_M_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          |
+//                |               |               |               |               |               |          |
+        // clang-format on
+        {S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1},
+        {S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1},
+        {S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1},
+        {S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1},
+        {S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 4, 8, 1},
+        {S<4, 32, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1},
+        {S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 2, 8, 1},
+        {S<4, 64, 1>, S<0, 2, 1>, S<0, 2, 1>, 1, 1, 8, 1},
+    };
+
+    std::vector<operation::BlockTransferDesc> b_block_descriptions_rowmajor = {
+        // clang-format off
+//  BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|
+//   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN|
+// Lengths_K0_N_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          |
+//                |               |               |               |               |               |          |
+  {    S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1},
+  {    S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1},
+  {    S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1},
+  {    S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1},
+  {    S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1},
+  {    S<4, 32, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              4,              8,         1},
+  {    S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              1,              8,         1},
+  {    S<4, 64, 1>,     S<0, 2, 1>,     S<0, 2, 1>,              1,              2,              8,         1},
+        // clang-format on
+    };
+
+    std::vector<operation::BlockTransferDesc> b_block_descriptions_colmajor = {
+        // clang-format off
+//  BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockTransfer| BBlockLds|
+//   ThreadCluster|  ThreadCluster| SrcAccessOrder|   SrcVectorDim|      SrcScalar|      DstScalar| AddExtraN|
+// Lengths_K0_N_K1|   ArrangeOrder|               |               |      PerVector|   PerVector_K1|          |
+//                |               |               |               |               |               |          |
+  {    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},
+  {    S<4, 32, 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},
+  {    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},
+  {    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},
+        // clang-format on
+    };
+
+    std::vector<operation::CShuffleDesc> cshuffle_descriptions = {
+        // clang-format off
+//    CShuffle|    CShuffle|
+// MXdlPerWave| NXdlPerWave|
+//  PerShuffle|  PerShuffle|
+//            |            |
+  {          1,           1},
+  {          1,           1},
+  {          1,           1},
+  {          1,           1},
+  {          1,           1},
+  {          1,           1},
+  {          1,           1},
+  {          1,           1},
+        // clang-format on
+    };
+
+    std::vector<operation::CBlockTransferDesc> c_block_descriptions = {
+        // clang-format off
+// CBlockTransferClusterLengths|  CBlockTransfer
+//         _MBlock_MWaveMPerXdl| ScalarPerVector
+//         _NBlock_NWaveNPerXdl|   _NWaveNPerXdl
+//                             |                
+  {              S<1, 32, 1, 8>,               8},
+  {              S<1, 32, 1, 8>,               8},
+  {              S<1, 16, 1, 8>,               8},
+  {              S<1, 32, 1, 8>,               8},
+  {              S<1, 32, 1, 4>,               8},
+  {              S<1, 16, 1, 8>,               8},
+  {              S<1, 32, 1, 8>,               8},
+  {              S<1, 32, 1, 8>,               8},
+        // clang-format on
+    };
+
+    const auto a_block_descriptions =
+        prob.TransA ? a_block_descriptions_colmajor : a_block_descriptions_rowmajor;
+    const auto b_block_descriptions =
+        prob.TransB ? b_block_descriptions_colmajor : b_block_descriptions_rowmajor;
+
+    assert(tile_descriptions.size() == a_block_descriptions.size());
+    assert(tile_descriptions.size() == b_block_descriptions.size());
+    assert(tile_descriptions.size() == cshuffle_descriptions.size());
+    assert(tile_descriptions.size() == c_block_descriptions.size());
+
+    for(std::size_t i = 0; i < tile_descriptions.size(); i++)
+    {
+        Operation_Xdl_CShuffle x;
+        x.tile_desc           = tile_descriptions[i];
+        x.a_block_transfer    = a_block_descriptions[i];
+        x.b_block_transfer    = b_block_descriptions[i];
+        x.cshuffle            = cshuffle_descriptions[i];
+        x.c_block_transfer    = c_block_descriptions[i];
+        x.A                   = TensorDesc{prob.ADataType, ToLayout(prob.TransA)};
+        x.B                   = TensorDesc{prob.BDataType, ToLayout(prob.TransB)};
+        x.E                   = TensorDesc{prob.EDataType, ToLayout(prob.TransE)};
+        x.Ds                  = Transform(prob.DsTrans, prob.DsDataType, [](auto trans, auto dt) {
+            return TensorDesc{dt, ToLayout(trans)};
+        });
+        x.a_elem_op           = prob.AElementOp;
+        x.b_elem_op           = prob.BElementOp;
+        x.cde_elem_op         = prob.CDEElementOp;
+        x.gemm_specialization = GetGemmSpec(prob.M,
+                                            prob.N,
+                                            prob.K,
+                                            x.tile_desc.m_per_block,
+                                            x.tile_desc.n_per_block,
+                                            x.tile_desc.k_per_block);
+        result.push_back(x);
+    }
+    return result;
+}
+
+std::vector<std::vector<Operation_Xdl_CShuffle>> Operation_Xdl_CShuffle::CreateOperations()
+{
+    std::vector<Problem> problems;
+    for(bool TransA : {true, false})
+        for(bool TransB : {true, false})
+        {
+            Problem prob;
+            prob.TransA = TransA;
+            prob.TransB = TransB;
+            problems.push_back(prob);
+        }
+    return Transform(problems, [](const Problem& p) { return CreateOperations(p); });
+}
+
+static const char* const DeviceGemmMultipleD_Xdl_CShuffleTemplate =
+    "ck::tensor_operation::device::DeviceGemmMultipleD_Xdl_CShuffle<${LayoutA}, ${LayoutB}, "
+    "${LayoutDs}, ${LayoutE}, ${ADataType}, ${BDataType}, ${AccDataType}, ${CShuffleDataType}, "
+    "${DsDataType}, ${EDataType}, ${AElementwiseOperation}, ${BElementwiseOperation}, "
+    "${CDEElementwiseOperation}, ${GemmSpecialization}, ${NumGemmkPrefetchStage}, ${BlockSize}, "
+    "${MPerBlock}, ${NPerBlock}, ${KPerBlock}, ${AK1}, ${BK1}, ${MPerXDL}, ${NPerXDL}, "
+    "${MXdlPerWave}, ${NXdlPerWave}, ${ABlockTransferThreadClusterLengths_AK0_M_AK1}, "
+    "${ABlockTransferThreadClusterArrangeOrder}, ${ABlockTransferSrcAccessOrder}, "
+    "${ABlockTransferSrcVectorDim}, ${ABlockTransferSrcScalarPerVector}, "
+    "${ABlockTransferDstScalarPerVector_AK1}, ${ABlockLdsExtraM}, "
+    "${BBlockTransferThreadClusterLengths_BK0_N_BK1}, ${BBlockTransferThreadClusterArrangeOrder}, "
+    "${BBlockTransferSrcAccessOrder}, ${BBlockTransferSrcVectorDim}, "
+    "${BBlockTransferSrcScalarPerVector}, ${BBlockTransferDstScalarPerVector_BK1}, "
+    "${BBlockLdsExtraN}, ${CShuffleMXdlPerWavePerShuffle}, ${CShuffleNXdlPerWavePerShuffle}, "
+    "${CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock}, "
+    "${CDEBlockTransferScalarPerVector_NPerBlock}>";
+
+Solution Operation_Xdl_CShuffle::ToSolution() const
+{
+    std::unordered_map<std::string, std::string> values = {
+        {"LayoutA", ToString(this->A.layout)},
+        {"LayoutB", ToString(this->B.layout)},
+        {"LayoutDs",
+         MakeTuple(Transform(this->Ds, [](auto tensor) { return ToString(tensor.layout); }))},
+        {"LayoutE", ToString(this->E.layout)},
+        {"ADataType", ToString(this->A.element)},
+        {"BDataType", ToString(this->B.element)},
+        {"AccDataType", ToString(this->acc)},
+        {"CShuffleDataType", ToString(this->cs_type)},
+        {"DsDataType",
+         MakeTuple(Transform(this->Ds, [](auto tensor) { return ToString(tensor.element); }))},
+        {"EDataType", ToString(this->E.element)},
+        {"AElementwiseOperation", this->a_elem_op},
+        {"BElementwiseOperation", this->b_elem_op},
+        {"CDEElementwiseOperation", this->cde_elem_op},
+        {"GemmSpecialization", this->gemm_specialization},
+        {"NumGemmkPrefetchStage", std::to_string(this->tile_desc.num_gemmk_prefetch_stage)},
+        {"BlockSize", std::to_string(this->tile_desc.block_size)},
+        {"MPerBlock", std::to_string(this->tile_desc.m_per_block)},
+        {"NPerBlock", std::to_string(this->tile_desc.n_per_block)},
+        {"KPerBlock", std::to_string(this->tile_desc.k_per_block)},
+        {"AK1", std::to_string(this->tile_desc.ak1)},
+        {"BK1", std::to_string(this->tile_desc.bk1)},
+        {"MPerXDL", std::to_string(this->tile_desc.m_per_XDL)},
+        {"NPerXDL", std::to_string(this->tile_desc.n_per_XDL)},
+        {"MXdlPerWave", std::to_string(this->tile_desc.m_Xdl_per_wave)},
+        {"NXdlPerWave", std::to_string(this->tile_desc.n_Xdl_per_wave)},
+        {"ABlockTransferThreadClusterLengths_AK0_M_AK1",
+         this->a_block_transfer.thread_cluster_length},
+        {"ABlockTransferThreadClusterArrangeOrder",
+         this->a_block_transfer.thread_cluster_arrange_order},
+        {"ABlockTransferSrcAccessOrder", this->a_block_transfer.src_access_order},
+        {"ABlockTransferSrcVectorDim", std::to_string(this->a_block_transfer.src_vec_dim)},
+        {"ABlockTransferSrcScalarPerVector",
+         std::to_string(this->a_block_transfer.src_scalar_per_vector)},
+        {"ABlockTransferDstScalarPerVector_AK1",
+         std::to_string(this->a_block_transfer.dst_scalar_per_vector_k1)},
+        {"ABlockLdsExtraM", std::to_string(this->a_block_transfer.lds_add_extra_dim)},
+        {"BBlockTransferThreadClusterLengths_BK0_N_BK1",
+         this->b_block_transfer.thread_cluster_length},
+        {"BBlockTransferThreadClusterArrangeOrder",
+         this->b_block_transfer.thread_cluster_arrange_order},
+        {"BBlockTransferSrcAccessOrder", this->b_block_transfer.src_access_order},
+        {"BBlockTransferSrcVectorDim", std::to_string(this->b_block_transfer.src_vec_dim)},
+        {"BBlockTransferSrcScalarPerVector",
+         std::to_string(this->b_block_transfer.src_scalar_per_vector)},
+        {"BBlockTransferDstScalarPerVector_BK1",
+         std::to_string(this->b_block_transfer.dst_scalar_per_vector_k1)},
+        {"BBlockLdsExtraN", std::to_string(this->b_block_transfer.lds_add_extra_dim)},
+        {"CShuffleMXdlPerWavePerShuffle",
+         std::to_string(this->cshuffle.m_Xdl_per_wave_per_shuffle)},
+        {"CShuffleNXdlPerWavePerShuffle",
+         std::to_string(this->cshuffle.n_Xdl_per_wave_per_shuffle)},
+        {"CDEBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock",
+         this->c_block_transfer.cluster_lengths_m_block_m_wave_m_per_Xdl_n_block_n_wave_n_per_Xdl},
+        {"CDEBlockTransferScalarPerVector_NPerBlock",
+         std::to_string(this->c_block_transfer.scalar_per_vector_n_wave_n_per_Xdl)},
+    };
+
+    return Solution{InterpolateString(DeviceGemmMultipleD_Xdl_CShuffleTemplate, values),
+                    std::move(values)};
+}
+
+} // namespace device_gemm_multiple_d
+} // namespace host
+} // namespace ck

codegen/src/headers.cpp

+#include "ck/host/headers.hpp"
+#include "ck_headers.hpp"
+
+namespace ck {
+namespace host {
+
+const std::string config_header = "";
+
+std::unordered_map<std::string_view, std::string_view> GetHeaders()
+{
+    auto headers = ck_headers();
+    headers.insert(std::make_pair("ck/config.h", config_header));
+    return headers;
+}
+
+} // namespace host
+} // namespace ck
\ No newline at end of file

codegen/src/types.cpp

+#include "ck/host/types.hpp"
+#include "ck/host/stringutils.hpp"
+#include <algorithm>
+#include <stdexcept>
+
+namespace ck {
+namespace host {
+
+Solution::Solution(std::string str, std::unordered_map<std::string, std::string> values)
+    : template_str(std::move(str)), template_values(std::move(values))
+{
+}
+
+std::string Solution::ToTemplateString() const { return this->template_str; }
+std::string Solution::GetTemplateParameter(const std::string& name) const
+{
+    return this->template_values.at(name);
+}
+
+std::string ToString(DataType dt)
+{
+    switch(dt)
+    {
+    case DataType::Float: return "float";
+    case DataType::Half: return "ck::half_t";
+    case DataType::Int8: return "int8_t";
+    case DataType::Int32: return "int32_t";
+    }
+    throw std::runtime_error("Incorrect data type");
+}
+
+std::string ToString(Layout dl)
+{
+    switch(dl)
+    {
+    case Layout::Row: return "ck::tensor_layout::gemm::RowMajor";
+    case Layout::Column: return "ck::tensor_layout::gemm::ColumnMajor";
+    }
+    throw std::runtime_error("Incorrect layout");
+}
+
+std::string ToString(GemmType gt)
+{
+    switch(gt)
+    {
+    case GemmType::Default: return "ck::tensor_operation::device::GemmSpecialization::Default";
+    }
+    throw std::runtime_error("Incorrect gemm type");
+}
+
+std::string SequenceStr(const std::vector<int>& v)
+{
+    return "ck::Sequence<" +
+           JoinStrings(Transform(v, [](int x) { return std::to_string(x); }), ", ") + ">";
+}
+
+std::string MakeTuple(const std::vector<std::string>& v)
+{
+    return "ck::Tuple<" + JoinStrings(v, ", ") + ">";
+}
+
+} // namespace host
+} // namespace ck

codegen/src/utils.cpp

+// SPDX-License-Identifier: MIT
+// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
+
+#include "ck/host/utils.hpp"
+
+namespace ck {
+namespace host {
+
+std::size_t integer_divide_ceil(std::size_t x, std::size_t y)
+{
+    return (x + y - std::size_t{1}) / y;
+}
+
+const std::unordered_set<std::string>& get_xdlop_archs()
+{
+    static std::unordered_set<std::string> supported_archs{"gfx90a", "gfx908", "gfx940", "gfx942"};
+    return supported_archs;
+}
+
+} // namespace host
+} // namespace ck

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})
+get_filename_component(BASE_NAME ${TEST_SRC} NAME_WE)
+rocm_add_test_executable(test_host_${BASE_NAME} ${TEST_SRC})
+target_link_libraries(test_host_${BASE_NAME} ck_rtc ck_host)
+target_include_directories(test_host_${BASE_NAME} PUBLIC include())
+endforeach()

codegen/test/gemm_multiple_d.cpp

+#include "ck/host/device_gemm_multiple_d/problem.hpp"
+#include "ck/host/device_gemm_multiple_d/operation.hpp"
+#include "ck/host/headers.hpp"
+#include "ck/host/stringutils.hpp"
+#include "ck/host/utils.hpp"
+#include <algorithm>
+#include <cmath>
+#include <iterator>
+#include <random>
+#include <test.hpp>
+#include <rtc/compile_kernel.hpp>
+#include <rtc/hip.hpp>
+
+using half = _Float16;
+// using half = __fp16;
+
+std::vector<rtc::src_file> get_headers_for_test()
+{
+    std::vector<rtc::src_file> result;
+    auto hs = ck::host::GetHeaders();
+    std::transform(
+        hs.begin(), hs.end(), std::back_inserter(result), [&](const auto& p) -> rtc::src_file {
+            return {p.first, p.second};
+        });
+    return result;
+}
+
+template <class T>
+rtc::buffer<T> generate_buffer(std::size_t n, std::size_t seed = 0)
+{
+    rtc::buffer<T> result(n);
+    std::mt19937 gen(seed);
+    std::uniform_real_distribution<double> dis(-1.0);
+    std::generate(result.begin(), result.end(), [&] { return dis(gen); });
+    return result;
+}
+
+template <class T, class U>
+bool allclose(const T& a, const U& b, double atol = 0.01, double rtol = 0.01)
+{
+    return std::equal(a.begin(), a.end(), b.begin(), b.end(), [&](double x, double y) {
+        return fabs(x - y) < atol + rtol * fabs(y);
+    });
+}
+
+std::string classify(double x)
+{
+    switch(std::fpclassify(x))
+    {
+    case FP_INFINITE: return "inf";
+    case FP_NAN: return "nan";
+    case FP_NORMAL: return "normal";
+    case FP_SUBNORMAL: return "subnormal";
+    case FP_ZERO: return "zero";
+    default: return "unknown";
+    }
+}
+
+template <class Buffer>
+void print_classification(const Buffer& x)
+{
+    std::unordered_set<std::string> result;
+    for(const auto& i : x)
+        result.insert(classify(i));
+    for(const auto& c : result)
+        std::cout << c << ", ";
+    std::cout << std::endl;
+}
+
+template <class Buffer>
+void print_statistics(const Buffer& x)
+{
+    std::cout << "Min value: " << *std::min_element(x.begin(), x.end()) << ", ";
+    std::cout << "Max value: " << *std::max_element(x.begin(), x.end()) << ", ";
+    double num_elements = x.size();
+    auto mean =
+        std::accumulate(x.begin(), x.end(), double{0.0}, std::plus<double>{}) / num_elements;
+    auto stddev = std::sqrt(
+        std::accumulate(x.begin(),
+                        x.end(),
+                        double{0.0},
+                        [&](double r, double v) { return r + std::pow((v - mean), 2.0); }) /
+        num_elements);
+    std::cout << "Mean: " << mean << ", ";
+    std::cout << "StdDev: " << stddev << "\n";
+}
+
+template <class Buffer>
+void print_preview(const Buffer& x)
+{
+    if(x.size() <= 10)
+    {
+        std::for_each(x.begin(), x.end(), [&](double i) { std::cout << i << ", "; });
+    }
+    else
+    {
+        std::for_each(x.begin(), x.begin() + 5, [&](double i) { std::cout << i << ", "; });
+        std::cout << "..., ";
+        std::for_each(x.end() - 5, x.end(), [&](double i) { std::cout << i << ", "; });
+    }
+    std::cout << std::endl;
+}
+
+template <class T>
+struct check_all
+{
+    rtc::buffer<T> data{};
+    bool operator()(const rtc::buffer<T>& x)
+    {
+        if(data.empty())
+        {
+            data = x;
+            return true;
+        }
+        if(std::any_of(x.begin(), x.end(), [](double y) { return std::isnan(y); }))
+            return false;
+        return allclose(data, x);
+    }
+};
+
+template <class Solution>
+auto report(const Solution& solution, bool pass)
+{
+    return test::make_predicate(solution.ToTemplateString(), [=] { return pass; });
+}
+
+const std::string gemm_compile_check = R"__ck__(
+#include <${include}>
+
+extern "C" __global__ void f(const ck::half_t* a, const ck::half_t* b, ck::half_t* c) {
+    using G = ${template};
+    constexpr auto desc = ${template}::make_descriptor(ck::make_naive_tensor_descriptor_packed(ck::make_tuple(${m}, ${k})),
+                                             ck::make_naive_tensor_descriptor(ck::make_tuple(${n}, ${k}), ck::make_tuple(1, ${n})),
+                                             ck::make_tuple(),
+                                             ck::make_naive_tensor_descriptor_packed(ck::make_tuple(${m}, ${n})));
+
+    static_assert(desc.IsValid(), "Invalid ck gemm.");
+
+    if constexpr(desc.IsValid())
+    {
+        ${template}::Run(desc,
+               a,
+               b,
+               ck::make_tuple(),
+               c);
+    }
+}
+
+)__ck__";
+
+TEST_CASE(test_problem_kernel)
+{
+    ck::host::device_gemm_multiple_d::Problem prob;
+    prob.M = 1024;
+    prob.N = 1024;
+    prob.K = 1024;
+    check_all<half> check;
+    auto a = to_gpu(generate_buffer<half>(1024 * 1024, 0));
+    auto b = to_gpu(generate_buffer<half>(1024 * 1024, 1));
+    auto c = to_gpu(generate_buffer<half>(1024 * 1024, 2));
+
+    for(auto solution : prob.GetSolutions("gfx90a"))
+    {
+        auto src  = ck::host::InterpolateString(gemm_compile_check,
+                                               {{"include", prob.GetIncludeHeader()},
+                                                {"template", solution.ToTemplateString()},
+                                                {"m", std::to_string(prob.M)},
+                                                {"n", std::to_string(prob.N)},
+                                                {"k", std::to_string(prob.K)}});
+        auto srcs = get_headers_for_test();
+        srcs.push_back({"main.cpp", src});
+        rtc::compile_options options;
+        options.kernel_name = "f";
+        auto k              = rtc::compile_kernel(srcs, options);
+        auto block_size     = solution.GetTemplateParameter<std::size_t>("BlockSize");
+        auto m_per_block    = solution.GetTemplateParameter<std::size_t>("MPerBlock");
+        auto n_per_block    = solution.GetTemplateParameter<std::size_t>("NPerBlock");
+        auto grid_size      = ck::host::integer_divide_ceil(prob.M, m_per_block) *
+                         ck::host::integer_divide_ceil(prob.N, n_per_block);
+        k.launch(nullptr, grid_size * block_size, block_size)(a.data(), b.data(), c.data());
+        CHECK(report(solution, check(rtc::from_gpu(c))));
+    }
+}
+
+int main(int argc, const char* argv[]) { test::run(argc, argv); }