Merge remote-tracking branch 'origin/develop' into vpietila/ggemm-profiling

5dee1b64 · Ville Pietilä · 870c3a76 · 355893cd · 5dee1b64 · 5dee1b64
Commit 5dee1b64 authored Dec 09, 2024 by Ville Pietilä
20 changed files
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -185,13 +185,22 @@ if (SUPPORTED_GPU_TARGETS MATCHES "gfx9")
    add_definitions(-DCK_USE_XDL)
 endif()
 if (SUPPORTED_GPU_TARGETS MATCHES "gfx94")
-    message("Enabling FP8 gemms in ckProfiler")
+    message("Enabling FP8 gemms on native architectures")
    add_definitions(-DCK_USE_GFX94)
 endif()
 if (SUPPORTED_GPU_TARGETS MATCHES "gfx11" OR SUPPORTED_GPU_TARGETS MATCHES "gfx12")
    message("Enabling WMMA instances")
    add_definitions(-DCK_USE_WMMA)
 endif()
+if (SUPPORTED_GPU_TARGETS MATCHES "gfx12")
+    add_definitions(-DCK_USE_OCP_FP8)
+    set(CK_USE_OCP_FP8 "ON")
+endif()
+if (SUPPORTED_GPU_TARGETS MATCHES "gfx90a" OR SUPPORTED_GPU_TARGETS MATCHES "gfx94")
+    add_definitions(-DCK_USE_FNUZ_FP8)
+    set(CK_USE_FNUZ_FP8 "ON")
+endif()
+
 option(CK_USE_FP8_ON_UNSUPPORTED_ARCH "Enable FP8 GEMM instances on older architectures" OFF)
 if(CK_USE_FP8_ON_UNSUPPORTED_ARCH AND (SUPPORTED_GPU_TARGETS MATCHES "gfx90a" OR SUPPORTED_GPU_TARGETS MATCHES "gfx908"))
    add_definitions(-DCK_USE_FP8_ON_UNSUPPORTED_ARCH)

--- a/CONTRIBUTORS.md
+++ b/CONTRIBUTORS.md
+[Back to the main page](./README.md)
 # Composable Kernel Developers and Contributors

 This is the list of developers and contributors to Composable Kernel library

--- a/Dockerfile
+++ b/Dockerfile
 FROM ubuntu:20.04
 ARG DEBIAN_FRONTEND=noninteractive
-ARG ROCMVERSION=6.2
+ARG ROCMVERSION=6.3
 ARG compiler_version=""
 ARG compiler_commit=""
 ARG CK_SCCACHE=""
@@ -13,17 +13,12 @@ RUN set -xe && \
    apt-get update && apt-get install -y --allow-unauthenticated apt-utils wget gnupg2 curl && \
    curl -fsSL https://repo.radeon.com/rocm/rocm.gpg.key | gpg --dearmor -o /etc/apt/trusted.gpg.d/rocm-keyring.gpg

-RUN if [ "$ROCMVERSION" != "6.3" ]; then \
-        sh -c "wget https://repo.radeon.com/amdgpu-install/$ROCMVERSION/ubuntu/focal/amdgpu-install_6.2.60200-1_all.deb  --no-check-certificate" && \
-        apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated ./amdgpu-install_6.2.60200-1_all.deb && \
+RUN if [ "$ROCMVERSION" != "6.4" ]; then \
+        sh -c "wget https://repo.radeon.com/amdgpu-install/$ROCMVERSION/ubuntu/focal/amdgpu-install_6.3.60300-1_all.deb  --no-check-certificate" && \
+        apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --allow-unauthenticated ./amdgpu-install_6.3.60300-1_all.deb && \
        wget -qO - http://repo.radeon.com/rocm/rocm.gpg.key | apt-key add - && \
        sh -c "echo deb [arch=amd64 signed-by=/etc/apt/trusted.gpg.d/rocm-keyring.gpg] $DEB_ROCM_REPO focal main > /etc/apt/sources.list.d/rocm.list" && \
        sh -c 'echo deb [arch=amd64 signed-by=/etc/apt/trusted.gpg.d/rocm-keyring.gpg] https://repo.radeon.com/amdgpu/$ROCMVERSION/ubuntu focal main > /etc/apt/sources.list.d/amdgpu.list'; \
-    elif [ "$ROCMVERSION" = "6.3" ] && [ "$compiler_version" = "rc1" ]; then \
-        sh -c "wget http://artifactory-cdn.amd.com/artifactory/list/amdgpu-deb/amdgpu-install-internal_6.3-20.04-1_all.deb --no-check-certificate" && \
-        apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install dialog libpopt0 rsync && DEBIAN_FRONTEND=noninteractive apt-get install ./amdgpu-install-internal_6.3-20.04-1_all.deb && \
-        sh -c 'echo deb [arch=amd64 trusted=yes] http://compute-artifactory.amd.com/artifactory/list/rocm-release-archive-20.04-deb/ 6.3 rel-20 > /etc/apt/sources.list.d/rocm-build.list' && \
-        amdgpu-repo --amdgpu-build=2074281; \
    fi

 RUN sh -c "echo deb http://mirrors.kernel.org/ubuntu focal main universe | tee -a /etc/apt/sources.list" && \

--- a/Dockerfile.compiler
+++ b/Dockerfile.compiler
-ARG BASE_DOCKER="rocm/composable_kernel:ck_ub20.04_rocm6.2"
+ARG BASE_DOCKER="rocm/composable_kernel:ck_ub20.04_rocm6.3"
 FROM $BASE_DOCKER
 ARG compiler_version=""
 ARG compiler_commit=""

--- a/Jenkinsfile
+++ b/Jenkinsfile
@@ -38,7 +38,8 @@ def getBaseDockerImageName(){
        img = "${params.USE_CUSTOM_DOCKER}"
    }
    else{
-    if (params.ROCMVERSION != "6.3"){
+        def ROCM_numeric = "${params.ROCMVERSION}" as float
+        if ( ROCM_numeric < 6.4 ){
            img = "${env.CK_DOCKERHUB}:ck_ub20.04_rocm${params.ROCMVERSION}"
            }
        else{
@@ -329,10 +330,8 @@ def cmake_build(Map conf=[:]){
        try{
            archiveArtifacts "perf_fmha_fwd_*.log"
            archiveArtifacts "perf_fmha_bwd_*.log"
-            stash name: "perf_fmha_fwd_gfx942.log"
-            stash name: "perf_fmha_bwd_gfx942.log"
-            stash name: "perf_fmha_fwd_gfx90a.log"
-            stash name: "perf_fmha_bwd_gfx90a.log"
+            stash includes: "perf_fmha_**_gfx942.log", name: "perf_fmha_log_gfx942"
+            stash includes: "perf_fmha_**_gfx90a.log", name: "perf_fmha_log_gfx90a"
        }
        catch(Exception err){
            echo "could not locate the requested artifacts: ${err.getMessage()}. will skip the stashing."
@@ -407,128 +406,6 @@ def buildHipClangJobAndReboot(Map conf=[:]){
    }
 }

-def runCKProfiler(Map conf=[:]){
-        show_node_info()
-
-        env.HSA_ENABLE_SDMA=0
-        checkout scm
-
-        def image = getDockerImageName()
-        def prefixpath = conf.get("prefixpath", "/opt/rocm")
-
-        // Jenkins is complaining about the render group 
-        def dockerOpts="--device=/dev/kfd --device=/dev/dri --group-add video --group-add render --cap-add=SYS_PTRACE --security-opt seccomp=unconfined"
-        if (conf.get("enforce_xnack_on", false)) {
-            dockerOpts = dockerOpts + " --env HSA_XNACK=1 "
-        }
-        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: ${dockerOpts}"
-
-        def dockerArgs = "--build-arg PREFIX=${prefixpath} --build-arg compiler_version='${params.COMPILER_VERSION}' --build-arg compiler_commit='${params.COMPILER_COMMIT}' --build-arg ROCMVERSION='${params.ROCMVERSION}' "
-
-        def variant = env.STAGE_NAME
-        def retimage
-
-        gitStatusWrapper(credentialsId: "${env.ck_git_creds}", gitHubContext: "Jenkins - ${variant}", account: 'ROCm', repo: 'composable_kernel') {
-            try {
-                (retimage, image) = getDockerImage(conf)
-                withDockerContainer(image: image, args: dockerOpts) {
-                    timeout(time: 5, unit: 'MINUTES'){
-                        sh 'rocminfo | tee rocminfo.log'
-                        if ( !runShell('grep -n "gfx" rocminfo.log') ){
-                            throw new Exception ("GPU not found")
-                        }
-                        else{
-                            echo "GPU is OK"
-                        }
-                    }
-                }
-            }
-            catch (org.jenkinsci.plugins.workflow.steps.FlowInterruptedException e){
-                echo "The job was cancelled or aborted"
-                throw e
-            }
-
-            withDockerContainer(image: image, args: dockerOpts + ' -v=/var/jenkins/:/var/jenkins') {
-                timeout(time: 24, unit: 'HOURS')
-                {
-                    sh """
-                        rm -rf build
-                        mkdir build
-                    """
-                    dir("build"){
-                        unstash 'ckProfiler.tar.gz'
-                        sh 'tar -xvf ckProfiler.tar.gz'
-                    }
-
-					dir("script"){
-                        if (params.RUN_FULL_QA){
-                            sh "./run_full_performance_tests.sh 0 QA_${params.COMPILER_VERSION} ${env.BRANCH_NAME} ${NODE_NAME}"
-                            archiveArtifacts "perf_gemm.log"
-                            archiveArtifacts "perf_resnet50_N256.log"
-                            archiveArtifacts "perf_resnet50_N4.log"
-                            archiveArtifacts "perf_batched_gemm.log"
-                            archiveArtifacts "perf_grouped_gemm.log"
-                            archiveArtifacts "perf_grouped_conv_fwd.log"
-                            archiveArtifacts "perf_grouped_conv_bwd_data.log"
-                            archiveArtifacts "perf_grouped_conv_bwd_weight.log"
-                            archiveArtifacts "perf_gemm_bilinear.log"
-                            archiveArtifacts "perf_reduction.log"
-                            archiveArtifacts "perf_splitK_gemm.log"
-                            archiveArtifacts "perf_onnx_gemm.log"
-                            archiveArtifacts "perf_mixed_gemm.log"
-                           // stash perf files to master
-                            stash name: "perf_gemm.log"
-                            stash name: "perf_resnet50_N256.log"
-                            stash name: "perf_resnet50_N4.log"
-                            stash name: "perf_batched_gemm.log"
-                            stash name: "perf_grouped_gemm.log"
-                            stash name: "perf_grouped_conv_fwd.log"
-                            stash name: "perf_grouped_conv_bwd_data.log"
-                            stash name: "perf_grouped_conv_bwd_weight.log"
-                            stash name: "perf_gemm_bilinear.log"
-                            stash name: "perf_reduction.log"
-                            stash name: "perf_splitK_gemm.log"
-                            stash name: "perf_onnx_gemm.log"
-                            stash name: "perf_mixed_gemm.log"
-                            //we will process results on the master node
-                        }
-                        else{
-                            sh "./run_performance_tests.sh 0 CI_${params.COMPILER_VERSION} ${env.BRANCH_NAME} ${NODE_NAME}"
-                            archiveArtifacts "perf_gemm.log"
-                            archiveArtifacts "perf_resnet50_N256.log"
-                            archiveArtifacts "perf_resnet50_N4.log"
-                            // stash perf files to master
-                            stash name: "perf_gemm.log"
-                            stash name: "perf_resnet50_N256.log"
-                            stash name: "perf_resnet50_N4.log"
-                            //we will process the results on the master node
-                        }
-					}
-                }
-            }
-        }
-        return retimage
-}
-
-def runPerfTest(Map conf=[:]){
-    try{
-        runCKProfiler(conf)
-    }
-    catch(e){
-        echo "throwing error exception in performance tests"
-        echo 'Exception occurred: ' + e.toString()
-        throw e
-    }
-    finally{
-        if (!conf.get("no_reboot", false)) {
-            reboot()
-        }
-    }
-}
-
 def Build_CK(Map conf=[:]){
        show_node_info()

@@ -588,36 +465,95 @@ def Build_CK(Map conf=[:]){
                throw e
            }
            withDockerContainer(image: image, args: dockerOpts + ' -v=/var/jenkins/:/var/jenkins') {
-                timeout(time: 24, unit: 'HOURS')
+                timeout(time: 12, unit: 'HOURS')
                {
                    //check whether to run performance tests on this node
-                    def do_perf_tests = 0
+                    def arch_type = 0
                    sh 'rocminfo | tee rocminfo.log'
-                    if ( runShell('grep -n "gfx1030" rocminfo.log') || runShell('grep -n "gfx1101" rocminfo.log') || runShell('grep -n "gfx1201" rocminfo.log') || runShell('grep -n "gfx942" rocminfo.log') ){
-                        do_perf_tests = 1
-                        echo "Stash profiler and run performance tests"
+                    if ( runShell('grep -n "gfx90a" rocminfo.log') ){
+                        arch_type = 1
+                    }
+                    else if ( runShell('grep -n "gfx942" rocminfo.log') ) {
+                        arch_type = 2
+                    }
+                    else if ( runShell('grep -n "gfx1030" rocminfo.log') ) {
+                        arch_type = 3
+                    }
+                    else if ( runShell('grep -n "gfx1101" rocminfo.log') ) {
+                        arch_type = 4
+                    }
+                    else if ( runShell('grep -n "gfx1201" rocminfo.log') ) {
+                        arch_type = 5
                    }
                    cmake_build(conf)
                    dir("build"){
-                        //run tests and examples
-                        //sh 'make -j check'
-                        if (params.RUN_PERFORMANCE_TESTS && do_perf_tests == 0 ){
-                            //we only need the ckProfiler to run the performance tests, so we pack and stash it
-                            //do not stash profiler on nodes where we don't need to run performance tests
-                            sh 'tar -zcvf ckProfiler.tar.gz bin/ckProfiler'
-                            stash name: "ckProfiler.tar.gz"
-                        }
-                        if (params.RUN_FULL_QA && do_perf_tests == 0 ){
-                            // build deb packages for all gfx9 targets and prepare to export
+                        if (params.RUN_FULL_QA && arch_type == 1 ){
+                            // build deb packages for all gfx9 targets on gfx90a system and prepare to export
+                            echo "Build ckProfiler package"
                            sh 'make -j package'
                            archiveArtifacts artifacts: 'composablekernel-ckprofiler_*.deb'
-                            archiveArtifacts artifacts: 'composablekernel-tests_*.deb'
                            sh 'mv composablekernel-ckprofiler_*.deb ckprofiler_0.2.0_amd64.deb'
-                            stash name: "ckprofiler_0.2.0_amd64.deb"
+                            stash includes: "ckprofiler_0.2.0_amd64.deb", name: "ckprofiler_0.2.0_amd64.deb"
                        }
                    }
-                    if (params.hipTensor_test && do_perf_tests == 0 ){
-                        //build and test hipTensor
+                    // run performance tests, stash the logs, results will be processed on the master node
+					dir("script"){
+                        if (params.RUN_PERFORMANCE_TESTS){
+                        if (params.RUN_FULL_QA && arch_type == 1){
+                            // run full tests on gfx90a
+                            echo "Run full performance tests"
+                            sh "./run_full_performance_tests.sh 0 QA_${params.COMPILER_VERSION} ${env.BRANCH_NAME} ${NODE_NAME}"
+                            archiveArtifacts "perf_gemm.log"
+                            archiveArtifacts "perf_resnet50_N256.log"
+                            archiveArtifacts "perf_resnet50_N4.log"
+                            archiveArtifacts "perf_batched_gemm.log"
+                            archiveArtifacts "perf_grouped_gemm.log"
+                            archiveArtifacts "perf_grouped_conv_fwd.log"
+                            archiveArtifacts "perf_grouped_conv_bwd_data.log"
+                            archiveArtifacts "perf_grouped_conv_bwd_weight.log"
+                            archiveArtifacts "perf_gemm_bilinear.log"
+                            archiveArtifacts "perf_reduction.log"
+                            archiveArtifacts "perf_splitK_gemm.log"
+                            archiveArtifacts "perf_onnx_gemm.log"
+                            archiveArtifacts "perf_mixed_gemm.log"
+                            stash includes: "perf_**.log", name: "perf_log"
+                        }
+                        else if ( arch_type == 1 ){
+                            // run standard tests on gfx90a
+                            echo "Run performance tests"
+                            sh "./run_performance_tests.sh 0 CI_${params.COMPILER_VERSION} ${env.BRANCH_NAME} ${NODE_NAME}"
+                            archiveArtifacts "perf_gemm.log"
+                            archiveArtifacts "perf_onnx_gemm.log"
+                            archiveArtifacts "perf_resnet50_N256.log"
+                            archiveArtifacts "perf_resnet50_N4.log"
+                            stash includes: "perf_**.log", name: "perf_log"
+                        }
+                        // disable performance tests on gfx1030 for now.
+                        //else if ( arch_type == 3){
+                            // run basic tests on gfx1030
+                        //    echo "Run gemm performance tests"
+                        //    sh "./run_gemm_performance_tests.sh 0 CI_${params.COMPILER_VERSION} ${env.BRANCH_NAME} ${NODE_NAME} gfx10"
+                        //    archiveArtifacts "perf_onnx_gemm_gfx10.log"
+                        //    stash includes: "perf_onnx_gemm_gfx10.log", name: "perf_log_gfx10"
+                        //}
+                        else if ( arch_type == 4){
+                            // run basic tests on gfx11
+                            echo "Run gemm performance tests"
+                            sh "./run_gemm_performance_tests.sh 0 CI_${params.COMPILER_VERSION} ${env.BRANCH_NAME} ${NODE_NAME} gfx11"
+                            archiveArtifacts "perf_onnx_gemm_gfx11.log"
+                            stash includes: "perf_onnx_gemm_gfx11.log", name: "perf_log_gfx11"
+                        }
+                        else if ( arch_type == 5 ){
+                            // run basic tests on gfx12
+                            echo "Run gemm performance tests"
+                            sh "./run_gemm_performance_tests.sh 0 CI_${params.COMPILER_VERSION} ${env.BRANCH_NAME} ${NODE_NAME} gfx12"
+                            archiveArtifacts "perf_onnx_gemm_gfx12.log"
+                            stash includes: "perf_onnx_gemm_gfx12.log", name: "perf_log_gfx12"
+                        }                        
+                        }
+                    }
+                    if (params.hipTensor_test && arch_type == 1 ){
+                        // build and test hipTensor on gfx90a node
                        sh """#!/bin/bash
                            rm -rf "${params.hipTensor_branch}".zip
                            rm -rf hipTensor-"${params.hipTensor_branch}"
@@ -689,10 +625,8 @@ def process_results(Map conf=[:]){
                dir("script"){
                    if (params.RUN_CK_TILE_FMHA_TESTS){
                        try{
-                            unstash "perf_fmha_fwd_gfx942.log"
-                            unstash "perf_fmha_bwd_gfx942.log"
-                            unstash "perf_fmha_fwd_gfx90a.log"
-                            unstash "perf_fmha_bwd_gfx90a.log"
+                            unstash "perf_fmha_log_gfx942"
+                            unstash "perf_fmha_log_gfx90a"
                        }
                        catch(Exception err){
                            echo "could not locate the FMHA performance logs: ${err.getMessage()}."
@@ -702,26 +636,26 @@ def process_results(Map conf=[:]){
                        // unstash perf files to master
                        unstash "ckprofiler_0.2.0_amd64.deb"
                        sh "sshpass -p ${env.ck_deb_pw} scp -o StrictHostKeyChecking=no ckprofiler_0.2.0_amd64.deb ${env.ck_deb_user}@${env.ck_deb_ip}:/var/www/html/composable_kernel/"
-                        unstash "perf_gemm.log"
-                        unstash "perf_resnet50_N256.log"
-                        unstash "perf_resnet50_N4.log"
-                        unstash "perf_batched_gemm.log"
-                        unstash "perf_grouped_gemm.log"
-                        unstash "perf_grouped_conv_fwd.log"
-                        unstash "perf_grouped_conv_bwd_data.log"
-                        unstash "perf_grouped_conv_bwd_weight.log"
-                        unstash "perf_gemm_bilinear.log"
-                        unstash "perf_reduction.log"
-                        unstash "perf_splitK_gemm.log"
-                        unstash "perf_onnx_gemm.log"
-                        unstash "perf_mixed_gemm.log"
+                        unstash "perf_log"
+                        try{
+                            unstash "perf_log_gfx11"
+                            unstash "perf_log_gfx12"
+                        }
+                        catch(Exception err){
+                            echo "could not locate the GEMM gfx11/gfx12 performance logs: ${err.getMessage()}."
+                        }
                        sh "./process_qa_data.sh"
                    }
                    else{
                        // unstash perf files to master
-                        unstash "perf_gemm.log"
-                        unstash "perf_resnet50_N256.log"
-                        unstash "perf_resnet50_N4.log"
+                        unstash "perf_log"
+                        try{
+                            unstash "perf_log_gfx11"
+                            unstash "perf_log_gfx12"
+                        }
+                        catch(Exception err){
+                            echo "could not locate the GEMM gfx11/gfx12 performance logs: ${err.getMessage()}."
+                        }
                        sh "./process_perf_data.sh"
                    }
                }
@@ -739,8 +673,8 @@ def process_results(Map conf=[:]){
 }

 //launch develop branch daily at 23:00 UT in FULL_QA mode and at 19:00 UT with latest staging compiler version
-CRON_SETTINGS = BRANCH_NAME == "develop" ? '''0 23 * * * % RUN_FULL_QA=true;ROCMVERSION=6.2;RUN_CK_TILE_FMHA_TESTS=true;RUN_CK_TILE_GEMM_TESTS=true
-                                              0 21 * * * % ROCMVERSION=6.2;hipTensor_test=true;RUN_CODEGEN_TESTS=true
+CRON_SETTINGS = BRANCH_NAME == "develop" ? '''0 23 * * * % RUN_FULL_QA=true;ROCMVERSION=6.3;RUN_CK_TILE_FMHA_TESTS=true;RUN_CK_TILE_GEMM_TESTS=true
+                                              0 21 * * * % ROCMVERSION=6.3;hipTensor_test=true;RUN_CODEGEN_TESTS=true
                                              0 19 * * * % BUILD_DOCKER=true;DL_KERNELS=true;COMPILER_VERSION=amd-staging;BUILD_COMPILER=/llvm-project/build/bin/clang++;BUILD_GFX12=true;USE_SCCACHE=false;NINJA_BUILD_TRACE=true
                                              0 17 * * * % BUILD_DOCKER=true;DL_KERNELS=true;COMPILER_VERSION=amd-mainline;BUILD_COMPILER=/llvm-project/build/bin/clang++;BUILD_GFX12=true;USE_SCCACHE=false;NINJA_BUILD_TRACE=true
                                              0 15 * * * % BUILD_INSTANCES_ONLY=true;RUN_PERFORMANCE_TESTS=false;USE_SCCACHE=false
@@ -765,8 +699,8 @@ pipeline {
            description: 'If you want to use a custom docker image, please specify it here (default: leave blank).')
        string(
            name: 'ROCMVERSION', 
-            defaultValue: '6.2', 
-            description: 'Specify which ROCM version to use: 6.2 (default).')
+            defaultValue: '6.3',
+            description: 'Specify which ROCM version to use: 6.3 (default).')
        string(
            name: 'COMPILER_VERSION', 
            defaultValue: '', 
@@ -1240,29 +1174,6 @@ pipeline {
                }
            }
        }
-
-        stage("Performance Tests")
-        {
-            parallel
-            {
-                stage("Run ckProfiler: gfx90a")
-                {
-                    when {
-                        beforeAgent true
-                        expression { params.RUN_PERFORMANCE_TESTS.toBoolean() && !params.BUILD_LEGACY_OS.toBoolean() }
-                    }
-                    options { retry(1) }
-                    agent{ label rocmnode("gfx90a")}
-                    environment{
-                        setup_args = "NO_CK_BUILD"
-                    }
-                    steps{
-                        runPerfTest(setup_args:setup_args, config_targets: "ckProfiler", no_reboot:true, build_type: 'Release')
-                        cleanWs()
-                    }
-                }
-            }
-        }
        stage("Process Performance Test Results")
        {
            parallel

--- a/README.md
+++ b/README.md
@@ -26,23 +26,15 @@ The current CK library is structured into four layers:

 ## General information

-To build our documentation locally, use the following code:
-
-``` bash
-cd docs
-pip3 install -r sphinx/requirements.txt
-python3 -m sphinx -T -E -b html -d _build/doctrees -D language=en . _build/html
-```
-
-You can find a list of our developers and contributors on our [Contributors](/CONTRIBUTORS.md) page.
-
-```note
-If you use CK, cite us as follows:
-
-* [Realizing Tensor Operators Using Coordinate Transformations and Tile Based Programming](???):
-  This paper will be available on arXiv soon.
-* [CITATION.cff](/CITATION.cff)
-```
+* [CK supported operations](include/ck/README.md)
+* [CK Tile supported operations](include/ck_tile/README.md)
+* [CK wrapper](client_example/25_wrapper/README.md)
+* [CK codegen](codegen/README.md)
+* [CK profiler](profiler/README.md)
+* [Examples (Custom use of CK supported operations)](example/README.md)
+* [Client examples (Use of CK supported operations with instance factory)](client_example/README.md)
+* [Terminology](/TERMINOLOGY.md)
+* [Contributors](/CONTRIBUTORS.md)

 CK is released under the **[MIT license](/LICENSE)**.

@@ -137,6 +129,14 @@ Docker images are available on [DockerHub](https://hub.docker.com/r/rocm/composa

    You can find instructions for running ckProfiler in [profiler](/profiler).

+* Build our documentation locally:
+
+    ``` bash
+    cd docs
+    pip3 install -r sphinx/requirements.txt
+    python3 -m sphinx -T -E -b html -d _build/doctrees -D language=en . _build/html
+    ```
+
 Note the `-j` option for building with multiple threads in parallel, which speeds up the build significantly.
 However, `-j` launches unlimited number of threads, which can cause the build to run out of memory and
 crash. On average, you should expect each thread to use ~2Gb of RAM.

--- a/TERMINOLOGY.md
+++ b/TERMINOLOGY.md
+[Back to the main page](./README.md)
+# Composable Kernel terminology
\ No newline at end of file
--- a/client_example/25_wrapper/README.md
+++ b/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:


--- a/client_example/CMakeLists.txt
+++ b/client_example/CMakeLists.txt
@@ -56,6 +56,14 @@ if (GPU_TARGETS)
        add_definitions(-DCK_USE_WMMA)
        set(CK_USE_WMMA "ON")
    endif()
+    if (GPU_TARGETS MATCHES "gfx12")
+        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")

--- a/client_example/README.md
+++ b/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.


--- a/codegen/README.md
+++ b/codegen/README.md
+[Back to the main page](../README.md)
+# Composable Kernel codegen
\ No newline at end of file
--- a/docs/sphinx/requirements.in
+++ b/docs/sphinx/requirements.in
-rocm-docs-core==1.9.2
+rocm-docs-core==1.11.0
 sphinxcontrib-bibtex==2.6.3
--- a/docs/sphinx/requirements.txt
+++ b/docs/sphinx/requirements.txt
@@ -103,7 +103,7 @@ requests==2.32.3
    # via
    #   pygithub
    #   sphinx
-rocm-docs-core==1.9.2
+rocm-docs-core==1.11.0
    # via -r requirements.in
 six==1.16.0
    # via pybtex

--- a/example/01_gemm/common.hpp
+++ b/example/01_gemm/common.hpp
@@ -76,7 +76,7 @@ struct ProblemSizeSplitK final
 struct ExecutionConfig final
 {
    // 0 - no verification, 1 - CPU, 2 - GPU, 3 - CPU + GPU
-    int do_verification = 3;
+    int do_verification = 1;
    int init_method     = 2;
    bool time_kernel    = false;
 };

--- a/example/01_gemm/run_gemm_example.inc
+++ b/example/01_gemm/run_gemm_example.inc
@@ -143,8 +143,8 @@ bool run_gemm(const ProblemType& problem_size, const ExecutionConfig& config)
    switch(config.init_method)
    {
    case 0:
-        ck::utils::FillConstant<ADataType>{static_cast<ADataType>(1.f)}(a_m_k);
-        ck::utils::FillConstant<BDataType>{static_cast<BDataType>(1.f)}(b_k_n);
+        ck::utils::FillConstant<ADataType>{ck::type_convert<ADataType>(1.f)}(a_m_k);
+        ck::utils::FillConstant<BDataType>{ck::type_convert<BDataType>(1.f)}(b_k_n);
        break;
    case 1:
        ck::utils::FillUniformDistributionIntegerValue<ADataType>{-5.f, 5.f}(a_m_k);

--- a/example/15_grouped_gemm/grouped_gemm_multiple_d_splitk_xdl_fp16.cpp
+++ b/example/15_grouped_gemm/grouped_gemm_multiple_d_splitk_xdl_fp16.cpp
@@ -186,15 +186,15 @@ bool run_grouped_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
            b_tensors[i].GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
            for(int j = 0; j < NumDMatrices; ++j)
            {
-                d_tensors[i][j].GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
+                d_tensors[i][j].GenerateTensorValue(GeneratorTensor_3<DDataType>{0.0, 1.0});
            }
            break;
        default:
-            a_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<0>{});
-            b_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<1>{});
+            a_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<ADataType, 0>{});
+            b_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<BDataType, 1>{});
            for(int j = 0; j < NumDMatrices; ++j)
            {
-                d_tensors[i][j].GenerateTensorValue(GeneratorTensor_Sequential<0>{});
+                d_tensors[i][j].GenerateTensorValue(GeneratorTensor_Sequential<DDataType, 0>{});
            }
        }
    }

--- a/example/15_grouped_gemm/grouped_gemm_multiple_d_xdl_fp16.cpp
+++ b/example/15_grouped_gemm/grouped_gemm_multiple_d_xdl_fp16.cpp
@@ -190,15 +190,15 @@ bool run_grouped_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
            b_tensors[i].GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
            for(int j = 0; j < NumDs; ++j)
            {
-                d_tensors[i][j].GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
+                d_tensors[i][j].GenerateTensorValue(GeneratorTensor_3<DDataType>{0.0, 1.0});
            }
            break;
        default:
-            a_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<0>{});
-            b_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<1>{});
+            a_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<ADataType, 0>{});
+            b_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<BDataType, 1>{});
            for(int j = 0; j < NumDs; ++j)
            {
-                d_tensors[i][j].GenerateTensorValue(GeneratorTensor_Sequential<0>{});
+                d_tensors[i][j].GenerateTensorValue(GeneratorTensor_Sequential<DDataType, 0>{});
            }
        }
    }

--- a/example/15_grouped_gemm/grouped_gemm_xdl_fixed_nk_bias_fp16.cpp
+++ b/example/15_grouped_gemm/grouped_gemm_xdl_fixed_nk_bias_fp16.cpp
@@ -167,11 +167,11 @@ bool run_grouped_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
            b_tensors[i].GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
            break;
        default:
-            a_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<0>{});
-            b_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<1>{});
+            a_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<ADataType, 0>{});
+            b_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<BDataType, 1>{});
        }

-        d0_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<1>{});
+        d0_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<D0DataType, 1>{});
    }

    using GroupedGemmKernelArgument = ck::tensor_operation::device::GroupedGemmKernelArgument<1>;

--- a/example/15_grouped_gemm/grouped_gemm_xdl_fixed_nk_fp16.cpp
+++ b/example/15_grouped_gemm/grouped_gemm_xdl_fixed_nk_fp16.cpp
@@ -157,8 +157,8 @@ bool run_grouped_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
            b_tensors[i].GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
            break;
        default:
-            a_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<0>{});
-            b_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<1>{});
+            a_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<ADataType, 0>{});
+            b_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<BDataType, 1>{});
        }
    }


--- a/example/15_grouped_gemm/grouped_gemm_xdl_fixed_nk_fp16_fp8.cpp
+++ b/example/15_grouped_gemm/grouped_gemm_xdl_fixed_nk_fp16_fp8.cpp
@@ -158,8 +158,8 @@ bool run_grouped_gemm(const ProblemSize& problem_size, const ExecutionConfig& co
            b_tensors[i].GenerateTensorValue(GeneratorTensor_3<BDataType>{-0.5, 0.5});
            break;
        default:
-            a_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<0>{});
-            b_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<1>{});
+            a_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<ADataType, 0>{});
+            b_tensors[i].GenerateTensorValue(GeneratorTensor_Sequential<BDataType, 1>{});
        }
    }