Merge branch 'main' into cl_khr_unified_svm

.github/workflows/presubmit.yml

@@ -5,34 +5,26 @@ jobs:
   build:
     name: Build ${{ matrix.os }} ${{ matrix.arch }}${{ matrix.extra }}
     runs-on: ${{ matrix.os }}
-    env:
-      JOB_ARCHITECTURE: ${{ matrix.arch }}
-      JOB_ENABLE_GL: ${{ matrix.gl }}
-      JOB_ENABLE_DEBUG: ${{ matrix.debug }}
     strategy:
       fail-fast: false
       matrix:
-        mainmatrix: [true]
+        build-type: [Release]
+        gl: [0]
         os: [ubuntu-22.04, macos-latest, windows-latest]
         include:
           - os: ubuntu-22.04
-            mainmatrix: true
             gl: 1
             extra: " gl"
           - os: ubuntu-22.04
-            mainmatrix: false
             arch: arm
           - os: ubuntu-22.04
-            mainmatrix: false
             arch: aarch64
-            debug: 1
+            build-type: Debug
             extra: " debug"
           - os: ubuntu-22.04
-            mainmatrix: false
             arch: android-arm
             android_arch_abi: armeabi-v7a
           - os: ubuntu-22.04
-            mainmatrix: false
             arch: android-aarch64
             android_arch_abi: arm64-v8a
     steps:
@@ -84,9 +76,101 @@ jobs:
           echo "ANDROID_NDK=$ANDROID_NDK" >> $GITHUB_ENV
           export ANDROID_ARCH_ABI=${{ matrix.android_arch_abi }}
           echo "ANDROID_ARCH_ABI=$ANDROID_ARCH_ABI" >> $GITHUB_ENV
-      - name: Build
+      - name: Prepare CMake Toolchain file
         shell: bash
-        run: ./presubmit.sh
+        run: |
+          if [[ '${{ matrix.arch }}' == android-* ]]; then
+              TOOLCHAIN_FILE=${ANDROID_NDK}/build/cmake/android.toolchain.cmake
+          else
+              TOOLCHAIN_FILE='${{ github.workspace }}'/toolchain.cmake
+              touch ${TOOLCHAIN_FILE}
+          fi
+
+          TOOLCHAIN_PREFIX_arm=arm-linux-gnueabihf
+          TOOLCHAIN_PREFIX_aarch64=aarch64-linux-gnu
+
+          if [[ '${{ matrix.arch }}' != android-* ]]; then
+            if [[ '${{ matrix.arch }}' != "" && ${RUNNER_OS} != "Windows" ]]; then
+              TOOLCHAIN_PREFIX_VAR=TOOLCHAIN_PREFIX_${{ matrix.arch }}
+              TOOLCHAIN_PREFIX=${!TOOLCHAIN_PREFIX_VAR}
+
+              echo "SET(CMAKE_SYSTEM_NAME Linux)" >> ${TOOLCHAIN_FILE}
+              echo "SET(CMAKE_SYSTEM_PROCESSOR ${{ matrix.arch }})" >> ${TOOLCHAIN_FILE}
+              echo "SET(CMAKE_C_COMPILER   ${TOOLCHAIN_PREFIX}-gcc)" >> ${TOOLCHAIN_FILE}
+              echo "SET(CMAKE_CXX_COMPILER ${TOOLCHAIN_PREFIX}-g++)" >> ${TOOLCHAIN_FILE}
+              echo "SET(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)" >> ${TOOLCHAIN_FILE}
+              echo "SET(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)" >> ${TOOLCHAIN_FILE}
+              echo "SET(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)" >> ${TOOLCHAIN_FILE}
+            fi
+          fi
+          echo "TOOLCHAIN_FILE=${TOOLCHAIN_FILE}" >> $GITHUB_ENV
+      - name: Prepare Android CMake arguments
+        if: ${{ matrix.arch == 'android-arm' || matrix.arch == 'android-aarch64' }}
+        shell: bash
+        run: |
+          echo "CMAKE_CONFIG_ARGS_ANDROID=-DCMAKE_ANDROID_ARCH_ABI=${ANDROID_ARCH_ABI}" >> $GITHUB_ENV
+      - name: Fetch and build OpenCL ICD Loader
+        shell: bash
+        run: |
+          git clone https://github.com/KhronosGroup/OpenCL-ICD-Loader.git
+          cd OpenCL-ICD-Loader
+          mkdir build
+          cd build
+          cmake .. -G Ninja \
+                -DCMAKE_BUILD_TYPE=Release \
+                -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_FILE} \
+                -DOPENCL_ICD_LOADER_HEADERS_DIR='${{ github.workspace }}'/OpenCL-Headers/ \
+                "${CMAKE_CONFIG_ARGS_ANDROID}"
+          cmake --build . --parallel
+      - name: Fetch Vulkan Headers
+        shell: bash
+        run: |
+          git clone https://github.com/KhronosGroup/Vulkan-Headers.git
+      - name: Fetch and build Vulkan Loader
+        if: ${{ matrix.arch != 'android-arm' && matrix.arch != 'android-aarch64' }}
+        shell: bash
+        run: |
+          git clone https://github.com/KhronosGroup/Vulkan-Loader.git
+          cd Vulkan-Loader
+          mkdir build
+          cd build
+          python3 ../scripts/update_deps.py
+          cmake .. -G Ninja \
+                -DCMAKE_BUILD_TYPE=Release \
+                -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_FILE} \
+                -DBUILD_WSI_XLIB_SUPPORT=OFF \
+                -DBUILD_WSI_XCB_SUPPORT=OFF \
+                -DBUILD_WSI_WAYLAND_SUPPORT=OFF \
+                -C helper.cmake ..
+          cmake --build . --parallel
+      - name: Configure and build
+        shell: bash
+        run: |
+          mkdir build
+          cd build
+          if [[ ${RUNNER_OS} == "Windows" ]]; then
+              CMAKE_OPENCL_LIBRARIES_OPTION="OpenCL"
+          else
+              CMAKE_OPENCL_LIBRARIES_OPTION="-lOpenCL"
+              if [[ '${{ matrix.arch }}' != android-* ]]; then
+                  CMAKE_OPENCL_LIBRARIES_OPTION="${CMAKE_OPENCL_LIBRARIES_OPTION} -lpthread"
+              fi
+          fi
+          cmake .. -G Ninja \
+                -DCMAKE_BUILD_TYPE=${{ matrix.build-type }} \
+                -DCMAKE_CACHE_OPTIONS="-DCMAKE_C_COMPILER_LAUNCHER=sccache -DCMAKE_CXX_COMPILER_LAUNCHER=sccache" \
+                -DCL_INCLUDE_DIR='${{ github.workspace }}'/OpenCL-Headers \
+                -DCL_LIB_DIR='${{ github.workspace }}'/OpenCL-ICD-Loader/build \
+                -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_FILE} \
+                -DCMAKE_RUNTIME_OUTPUT_DIRECTORY=./bin \
+                -DOPENCL_LIBRARIES="${CMAKE_OPENCL_LIBRARIES_OPTION}" \
+                -DUSE_CL_EXPERIMENTAL=ON \
+                -DGL_IS_SUPPORTED=${{ matrix.gl }} \
+                -DVULKAN_IS_SUPPORTED=ON \
+                -DVULKAN_INCLUDE_DIR='${{ github.workspace }}'/Vulkan-Headers/include/ \
+                -DVULKAN_LIB_DIR='${{ github.workspace }}'/Vulkan-Loader/build/loader/ \
+                "${CMAKE_CONFIG_ARGS_ANDROID}"
+          cmake --build . --parallel
   formatcheck:
     name: Check code format
     runs-on: ubuntu-22.04

CMakeLists.txt

@@ -7,7 +7,7 @@ project(CLConform${CONFORMANCE_SUFFIX})
 
 set(CMAKE_C_STANDARD 99)
 set(CMAKE_C_STANDARD_REQUIRED ON)
-set(CMAKE_CXX_STANDARD 11)
+set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
 
 add_definitions(-DCL_TARGET_OPENCL_VERSION=300)
@@ -100,7 +100,11 @@ if(CMAKE_COMPILER_IS_GNUCC OR "${CMAKE_CXX_COMPILER_ID}" MATCHES "(Apple)?Clang"
     add_cxx_flag_if_supported(-Wmisleading-indentation)
     add_cxx_flag_if_supported(-Wunused-function)
     add_cxx_flag_if_supported(-Wunused-variable)
-    add_cxx_flag_if_supported(-Werror)
+    if(CMAKE_VERSION VERSION_GREATER_EQUAL "3.24")
+      set(CMAKE_COMPILE_WARNING_AS_ERROR ON)
+    else()
+      add_cxx_flag_if_supported(-Werror)
+    endif()
     if(NOT CMAKE_BUILD_TYPE MATCHES "Release|RelWithDebInfo|MinSizeRel")
         # Enable more warnings if not doing a release build.
         add_cxx_flag_if_supported(-Wall)

README.md

@@ -14,6 +14,9 @@ Compiling the CTS requires the following CMake configuration options to be set:
 * `CL_INCLUDE_DIR` Points to the unified
   [OpenCL-Headers](https://github.com/KhronosGroup/OpenCL-Headers).
 * `CL_LIB_DIR` Directory containing the OpenCL library to build against.
+* `SPIRV_TOOLS_DIR` Directory containing the `spirv-as` and `spirv-val` binaries
+   to be used in the CTS build process. Alternatively, the location to these binaries
+   can be provided via the `PATH` variable.
 * `OPENCL_LIBRARIES` Name of the OpenCL library to link.
 
 It is advised that the [OpenCL ICD-Loader](https://github.com/KhronosGroup/OpenCL-ICD-Loader)
@@ -29,16 +32,26 @@ a build, and compile.
 git clone https://github.com/KhronosGroup/OpenCL-CTS.git
 git clone https://github.com/KhronosGroup/OpenCL-Headers.git
 git clone https://github.com/KhronosGroup/OpenCL-ICD-Loader.git
+git clone https://github.com/KhronosGroup/SPIRV-Tools.git
+git clone https://github.com/KhronosGroup/SPIRV-Headers.git SPIRV-Tools/external/spirv-headers
+git clone https://github.com/google/effcee.git SPIRV-Tools/external/effcee
+git clone https://github.com/google/re2.git SPIRV-Tools/external/re2
+
 
 mkdir OpenCL-ICD-Loader/build
 cmake -S OpenCL-ICD-Loader -B OpenCL-ICD-Loader/build \
       -DOPENCL_ICD_LOADER_HEADERS_DIR=$PWD/OpenCL-Headers
 cmake --build ./OpenCL-ICD-Loader/build --config Release
 
+mkdir SPIRV-Tools/build
+cmake -S SPIRV-Tools -B SPIRV-Tools/build -DSPIRV_SKIP_TESTS=ON
+cmake --build SPIRV-Tools/build --config Release
+
 mkdir OpenCL-CTS/build
 cmake -S OpenCL-CTS -B OpenCL-CTS/build \
       -DCL_INCLUDE_DIR=$PWD/OpenCL-Headers \
       -DCL_LIB_DIR=$PWD/OpenCL-ICD-Loader/build \
+      -DSPIRV_TOOLS_DIR=$PWD/SPIRV-Tools/build/tools/ \
       -DOPENCL_LIBRARIES=OpenCL
 cmake --build OpenCL-CTS/build --config Release
 ```

presubmit.sh

@@ -1,112 +0,0 @@
-#!/usr/bin/env bash
-
-set -e
-
-export TOP=$(pwd)
-
-TOOLCHAIN_PREFIX_arm=arm-linux-gnueabihf
-TOOLCHAIN_PREFIX_aarch64=aarch64-linux-gnu
-
-if [[ ${JOB_ARCHITECTURE} == android-* ]]; then
-    TOOLCHAIN_FILE=${ANDROID_NDK}/build/cmake/android.toolchain.cmake
-    CMAKE_CONFIG_ARGS_ANDROID="-DCMAKE_ANDROID_ARCH_ABI=${ANDROID_ARCH_ABI}"
-else
-    TOOLCHAIN_FILE=${TOP}/toolchain.cmake
-    touch ${TOOLCHAIN_FILE}
-fi
-
-BUILD_OPENGL_TEST="OFF"
-BUILD_VULKAN_TEST="ON"
-
-cmake --version
-echo
-
-# Prepare toolchain if needed
-if [[ ${JOB_ARCHITECTURE} != android-* ]]; then
-    if [[ ${JOB_ARCHITECTURE} != "" && ${RUNNER_OS} != "Windows" ]]; then
-        TOOLCHAIN_PREFIX_VAR=TOOLCHAIN_PREFIX_${JOB_ARCHITECTURE}
-        TOOLCHAIN_PREFIX=${!TOOLCHAIN_PREFIX_VAR}
-
-        echo "SET(CMAKE_SYSTEM_NAME Linux)" >> ${TOOLCHAIN_FILE}
-        echo "SET(CMAKE_SYSTEM_PROCESSOR ${JOB_ARCHITECTURE})" >> ${TOOLCHAIN_FILE}
-        echo "SET(CMAKE_C_COMPILER   ${TOOLCHAIN_PREFIX}-gcc)" >> ${TOOLCHAIN_FILE}
-        echo "SET(CMAKE_CXX_COMPILER ${TOOLCHAIN_PREFIX}-g++)" >> ${TOOLCHAIN_FILE}
-        echo "SET(CMAKE_FIND_ROOT_PATH_MODE_PROGRAM NEVER)" >> ${TOOLCHAIN_FILE}
-        echo "SET(CMAKE_FIND_ROOT_PATH_MODE_LIBRARY ONLY)" >> ${TOOLCHAIN_FILE}
-        echo "SET(CMAKE_FIND_ROOT_PATH_MODE_INCLUDE ONLY)" >> ${TOOLCHAIN_FILE}
-    fi
-fi
-
-if [[ ( ${JOB_ARCHITECTURE} == "" && ${JOB_ENABLE_GL} == "1" ) ]]; then
-    BUILD_OPENGL_TEST="ON"
-fi
-
-if [[ ${JOB_ENABLE_DEBUG} == 1 ]]; then
-    BUILD_CONFIG="Debug"
-else
-    BUILD_CONFIG="Release"
-fi
-
-#Vulkan Headers
-git clone https://github.com/KhronosGroup/Vulkan-Headers.git
-
-# Get and build loader
-git clone https://github.com/KhronosGroup/OpenCL-ICD-Loader.git
-cd ${TOP}/OpenCL-ICD-Loader
-mkdir build
-cd build
-cmake .. -G Ninja \
-      -DCMAKE_BUILD_TYPE=Release \
-      -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_FILE} \
-      -DOPENCL_ICD_LOADER_HEADERS_DIR=${TOP}/OpenCL-Headers/ \
-      "${CMAKE_CONFIG_ARGS_ANDROID}"
-cmake --build . --parallel
-
-#Vulkan Loader
-if [[ ${JOB_ARCHITECTURE} != android-* ]]; then
-    # Building the Vulkan loader is not supported on Android,
-    # instead, the loader is shipped as part of the operating system
-    cd ${TOP}
-    git clone https://github.com/KhronosGroup/Vulkan-Loader.git
-    cd Vulkan-Loader
-    mkdir build
-    cd build
-    python3 ../scripts/update_deps.py
-    cmake .. -G Ninja \
-          -DCMAKE_BUILD_TYPE=Release \
-          -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_FILE} \
-          -DBUILD_WSI_XLIB_SUPPORT=OFF \
-          -DBUILD_WSI_XCB_SUPPORT=OFF \
-          -DBUILD_WSI_WAYLAND_SUPPORT=OFF \
-          -C helper.cmake ..
-    cmake --build . --parallel
-fi
-
-# Build CTS
-cd ${TOP}
-ls -l
-mkdir build
-cd build
-if [[ ${RUNNER_OS} == "Windows" ]]; then
-    CMAKE_OPENCL_LIBRARIES_OPTION="OpenCL"
-else
-    CMAKE_OPENCL_LIBRARIES_OPTION="-lOpenCL"
-    if [[ ${JOB_ARCHITECTURE} != android-* ]]; then
-        CMAKE_OPENCL_LIBRARIES_OPTION="${CMAKE_OPENCL_LIBRARIES_OPTION} -lpthread"
-    fi
-fi
-cmake .. -G Ninja \
-      -DCMAKE_BUILD_TYPE="${BUILD_CONFIG}" \
-      -DCMAKE_CACHE_OPTIONS="-DCMAKE_C_COMPILER_LAUNCHER=sccache -DCMAKE_CXX_COMPILER_LAUNCHER=sccache" \
-      -DCL_INCLUDE_DIR=${TOP}/OpenCL-Headers \
-      -DCL_LIB_DIR=${TOP}/OpenCL-ICD-Loader/build \
-      -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_FILE} \
-      -DCMAKE_RUNTIME_OUTPUT_DIRECTORY=./bin \
-      -DOPENCL_LIBRARIES="${CMAKE_OPENCL_LIBRARIES_OPTION}" \
-      -DUSE_CL_EXPERIMENTAL=ON \
-      -DGL_IS_SUPPORTED=${BUILD_OPENGL_TEST} \
-      -DVULKAN_IS_SUPPORTED=${BUILD_VULKAN_TEST} \
-      -DVULKAN_INCLUDE_DIR=${TOP}/Vulkan-Headers/include/ \
-      -DVULKAN_LIB_DIR=${TOP}/Vulkan-Loader/build/loader/ \
-      "${CMAKE_CONFIG_ARGS_ANDROID}"
-cmake --build . --parallel

test_common/gl/setup.h

@@ -13,12 +13,13 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 //
-#ifndef _setup_h
-#define _setup_h
+#ifndef _gl_setup_h
+#define _gl_setup_h
 
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
+#include <vector>
 #include "gl_headers.h"
 #ifdef __APPLE__
 #include <OpenCL/opencl.h>
@@ -36,13 +37,11 @@ class GLEnvironment
         GLEnvironment() {}
         virtual ~GLEnvironment() {}
 
-         virtual int Init( int *argc, char **argv, int use_opengl_32 ) = 0;
+        virtual int Init(int *argc, char **argv, int use_opengl_32) = 0;
         virtual cl_context CreateCLContext( void ) = 0;
         virtual int SupportsCLGLInterop( cl_device_type device_type) = 0;
 
-        static GLEnvironment *    Instance( void );
-
-
+        static GLEnvironment *Instance(void);
 };
 
-#endif // _setup_h
+#endif // _gl_setup_h

test_common/gl/setup_osx.cpp

@@ -1,6 +1,6 @@
 //
-// Copyright (c) 2017 The Khronos Group Inc.
-// 
+// Copyright (c) 2024 The Khronos Group Inc.
+//
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
@@ -25,65 +25,74 @@ private:
 public:
     OSXGLEnvironment()
     {
-        mCGLContext = NULL;
         mIsGlutInit = false;
+        mCGLContext = NULL;
+        mShareGroup = NULL;
+        mPlatform = NULL;
     }
 
-  virtual int Init( int *argc, char **argv, int use_opengl_32 )
+    int Init(int *argc, char **argv, int use_opengl_32) override
+    {
+        if (!use_opengl_32)
         {
-      if (!use_opengl_32) {
-          if (!mIsGlutInit)
-          {
-              // Create a GLUT window to render into
-              glutInit(argc, argv);
-              glutInitWindowSize(512, 512);
-              glutInitDisplayMode(GLUT_RGB | GLUT_DOUBLE);
-              glutCreateWindow("OpenCL <-> OpenGL Test");
-              mIsGlutInit = true;
-          }
-      }
-
-      else {
-
-        CGLPixelFormatAttribute attribs[] = {
-          kCGLPFAOpenGLProfile, (CGLPixelFormatAttribute)kCGLOGLPVersion_3_2_Core,
-          kCGLPFAAllowOfflineRenderers,
-          kCGLPFANoRecovery,
-          kCGLPFAAccelerated,
-          kCGLPFADoubleBuffer,
-          (CGLPixelFormatAttribute)0
-        };
-
-        CGLError err;
-        CGLPixelFormatObj pix;
-        GLint npix;
-        err = CGLChoosePixelFormat (attribs, &pix, &npix);
-        if(err != kCGLNoError)
-          {
-            log_error("Failed to choose pixel format\n");
-            return -1;
-          }
-        err = CGLCreateContext(pix, NULL, &mCGLContext);
-        if(err != kCGLNoError)
-          {
-            log_error("Failed to create GL context\n");
-            return -1;
-          }
-        CGLSetCurrentContext(mCGLContext);
-      }
-
-            return 0;
+            if (!mIsGlutInit)
+            {
+                // Create a GLUT window to render into
+                glutInit(argc, argv);
+                glutInitWindowSize(512, 512);
+                glutInitDisplayMode(GLUT_RGB | GLUT_DOUBLE);
+                glutCreateWindow("OpenCL <-> OpenGL Test");
+                mIsGlutInit = true;
+            }
         }
 
-        virtual cl_context CreateCLContext( void )
+        else
+        {
+
+            CGLPixelFormatAttribute attribs[] = {
+                kCGLPFAOpenGLProfile,
+                (CGLPixelFormatAttribute)kCGLOGLPVersion_3_2_Core,
+                kCGLPFAAllowOfflineRenderers,
+                kCGLPFANoRecovery,
+                kCGLPFAAccelerated,
+                kCGLPFADoubleBuffer,
+                (CGLPixelFormatAttribute)0
+            };
+
+            CGLError err;
+            CGLPixelFormatObj pix;
+            GLint npix;
+            err = CGLChoosePixelFormat(attribs, &pix, &npix);
+            if (err != kCGLNoError)
+            {
+                log_error("Failed to choose pixel format\n");
+                return -1;
+            }
+            err = CGLCreateContext(pix, NULL, &mCGLContext);
+            if (err != kCGLNoError)
+            {
+                log_error("Failed to create GL context\n");
+                return -1;
+            }
+            CGLSetCurrentContext(mCGLContext);
+        }
+
+        return 0;
+    }
+
+    cl_context CreateCLContext(void) override
     {
       int error;
 
       if( mCGLContext == NULL )
         mCGLContext = CGLGetCurrentContext();
 
-      CGLShareGroupObj share_group = CGLGetShareGroup(mCGLContext);
-      cl_context_properties properties[] = { CL_CONTEXT_PROPERTY_USE_CGL_SHAREGROUP_APPLE, (cl_context_properties)share_group, 0 };
+      mShareGroup = CGLGetShareGroup(mCGLContext);
+      cl_context_properties properties[] = {
+          CL_CONTEXT_PLATFORM, (cl_context_properties)mPlatform,
+          CL_CONTEXT_PROPERTY_USE_CGL_SHAREGROUP_APPLE,
+          (cl_context_properties)mShareGroup, 0
+      };
       cl_context context = clCreateContext(properties, 0, 0, 0, 0, &error);
       if (error) {
         print_error(error, "clCreateContext failed");
@@ -108,16 +117,24 @@ public:
       return context;
     }
 
-    virtual int SupportsCLGLInterop( cl_device_type device_type )
+    int SupportsCLGLInterop(cl_device_type device_type) override
     {
       int found_valid_device = 0;
       cl_device_id devices[64];
       cl_uint num_of_devices;
       int error;
-      error = clGetDeviceIDs(NULL, device_type, 64, devices, &num_of_devices);
+      error = clGetPlatformIDs(1, &mPlatform, NULL);
+      if (error)
+      {
+          print_error(error, "clGetPlatformIDs failed");
+          return 0;
+      }
+
+      error =
+          clGetDeviceIDs(mPlatform, device_type, 64, devices, &num_of_devices);
       if (error) {
         print_error(error, "clGetDeviceIDs failed");
-        return -1;
+        return 0;
       }
 
       for (int i=0; i<(int)num_of_devices; i++) {
@@ -131,13 +148,11 @@ public:
             return found_valid_device;
     }
 
-        virtual ~OSXGLEnvironment()
-        {
-            CGLDestroyContext( mCGLContext );
-        }
-
-        CGLContextObj mCGLContext;
+    virtual ~OSXGLEnvironment() { CGLDestroyContext(mCGLContext); }
 
+    CGLContextObj mCGLContext;
+    CGLShareGroupObj mShareGroup;
+    cl_platform_id mPlatform;
 };
 
 GLEnvironment * GLEnvironment::Instance( void )

test_common/gl/setup_win32.cpp

@@ -1,6 +1,6 @@
 //
-// Copyright (c) 2017 The Khronos Group Inc.
-// 
+// Copyright (c) 2024 The Khronos Group Inc.
+//
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
@@ -42,14 +42,19 @@ private:
     cl_platform_id m_platform;
     bool m_is_glut_init;
 
+    HGLRC m_hGLRC;
+    HDC m_hDC;
+
 public:
     WGLEnvironment()
     {
         m_device_count = 0;
         m_platform = 0;
         m_is_glut_init = false;
+        m_hGLRC = 0;
+        m_hDC = 0;
     }
-    virtual int Init( int *argc, char **argv, int use_opengl_32 )
+    int Init(int *argc, char **argv, int use_opengl_32) override
     {
         if (!m_is_glut_init)
         {
@@ -64,21 +69,25 @@ public:
         return 0;
     }
 
-    virtual cl_context CreateCLContext( void )
+    cl_context CreateCLContext(void) override
     {
-        HGLRC hGLRC = wglGetCurrentContext();
-        HDC hDC = wglGetCurrentDC();
+        m_hGLRC = wglGetCurrentContext();
+        m_hDC = wglGetCurrentDC();
         cl_context_properties properties[] = {
-            CL_CONTEXT_PLATFORM, (cl_context_properties) m_platform,
-            CL_GL_CONTEXT_KHR,   (cl_context_properties) hGLRC,
-            CL_WGL_HDC_KHR,      (cl_context_properties) hDC,
+            CL_CONTEXT_PLATFORM,
+            (cl_context_properties)m_platform,
+            CL_GL_CONTEXT_KHR,
+            (cl_context_properties)m_hGLRC,
+            CL_WGL_HDC_KHR,
+            (cl_context_properties)m_hDC,
             0
         };
         cl_device_id devices[MAX_DEVICES];
         size_t dev_size;
         cl_int status;
 
-        if (!hGLRC || !hDC) {
+        if (!m_hGLRC || !m_hDC)
+        {
             print_error(CL_INVALID_CONTEXT, "No GL context bound");
             return 0;
         }
@@ -155,7 +164,7 @@ public:
         return clCreateContext(properties, 1, &ctxDevice, NULL, NULL, &status);
     }
 
-    virtual int SupportsCLGLInterop( cl_device_type device_type )
+    int SupportsCLGLInterop(cl_device_type device_type) override
     {
         cl_device_id devices[MAX_DEVICES];
         cl_uint num_of_devices;

test_common/gl/setup_x11.cpp

@@ -28,13 +28,21 @@ private:
     cl_uint m_device_count;
     bool m_glut_init;
 
+    cl_platform_id m_platform;
+    GLXContext m_context;
+    Display *m_dpy;
+
 public:
     X11GLEnvironment()
     {
         m_device_count = 0;
         m_glut_init = false;
+        m_platform = 0;
+        m_context = 0;
+        m_dpy = nullptr;
     }
-    virtual int Init( int *argc, char **argv, int use_opencl_32 )
+
+    int Init(int *argc, char **argv, int use_opencl_32) override
     {
          // Create a GLUT window to render into
          if (!m_glut_init)
@@ -49,19 +57,24 @@ public:
         return 0;
     }
 
-    virtual cl_context CreateCLContext( void )
+    cl_context CreateCLContext(void) override
     {
-        GLXContext context = glXGetCurrentContext();
-        Display *dpy = glXGetCurrentDisplay();
+        m_context = glXGetCurrentContext();
+        m_dpy = glXGetCurrentDisplay();
 
         cl_context_properties properties[] = {
-            CL_GL_CONTEXT_KHR,  (cl_context_properties) context,
-            CL_GLX_DISPLAY_KHR, (cl_context_properties) dpy,
+            CL_CONTEXT_PLATFORM,
+            (cl_context_properties)m_platform,
+            CL_GL_CONTEXT_KHR,
+            (cl_context_properties)m_context,
+            CL_GLX_DISPLAY_KHR,
+            (cl_context_properties)m_dpy,
             0
         };
         cl_int status;
 
-        if (!context || !dpy) {
+        if (!m_context || !m_dpy)
+        {
             print_error(CL_INVALID_CONTEXT, "No GL context bound");
             return 0;
         }
@@ -69,19 +82,19 @@ public:
         return clCreateContext(properties, 1, m_devices, NULL, NULL, &status);
     }
 
-    virtual int SupportsCLGLInterop( cl_device_type device_type )
+    int SupportsCLGLInterop(cl_device_type device_type) override
     {
         int found_valid_device = 0;
-        cl_platform_id platform;
         cl_device_id devices[64];
         cl_uint num_of_devices;
         int error;
-        error = clGetPlatformIDs(1, &platform, NULL);
+        error = clGetPlatformIDs(1, &m_platform, NULL);
         if (error) {
             print_error(error, "clGetPlatformIDs failed");
             return -1;
         }
-        error = clGetDeviceIDs(platform, device_type, 64, devices, &num_of_devices);
+        error = clGetDeviceIDs(m_platform, device_type, 64, devices,
+                               &num_of_devices);
         // If this platform doesn't have any of the requested device_type (namely GPUs) then return 0
         if (error == CL_DEVICE_NOT_FOUND)
           return 0;

test_common/harness/errorHelpers.cpp

@@ -213,6 +213,12 @@ const char *GetChannelTypeName(cl_channel_type type)
         case CL_UNORM_INT24: return "CL_UNORM_INT24";
         case CL_UNSIGNED_INT_RAW10_EXT: return "CL_UNSIGNED_INT_RAW10_EXT";
         case CL_UNSIGNED_INT_RAW12_EXT: return "CL_UNSIGNED_INT_RAW12_EXT";
+        case CL_UNSIGNED_INT10X6_EXT: return "CL_UNSIGNED_INT10X6_EXT";
+        case CL_UNSIGNED_INT12X4_EXT: return "CL_UNSIGNED_INT12X4_EXT";
+        case CL_UNSIGNED_INT14X2_EXT: return "CL_UNSIGNED_INT14X2_EXT";
+        case CL_UNORM_INT10X6_EXT: return "CL_UNORM_INT10X6_EXT";
+        case CL_UNORM_INT12X4_EXT: return "CL_UNORM_INT12X4_EXT";
+        case CL_UNORM_INT14X2_EXT: return "CL_UNORM_INT14X2_EXT";
         default: return NULL;
     }
 }
@@ -238,10 +244,16 @@ int IsChannelTypeSupported(cl_channel_type type)
         case CL_UNSIGNED_INT16:
         case CL_UNSIGNED_INT32:
         case CL_HALF_FLOAT:
-        case CL_FLOAT: return 1;
+        case CL_FLOAT:
 #ifdef CL_SFIXED14_APPLE
-        case CL_SFIXED14_APPLE: return 1;
+        case CL_SFIXED14_APPLE:
 #endif
+        case CL_UNSIGNED_INT10X6_EXT:
+        case CL_UNSIGNED_INT12X4_EXT:
+        case CL_UNSIGNED_INT14X2_EXT:
+        case CL_UNORM_INT10X6_EXT:
+        case CL_UNORM_INT12X4_EXT:
+        case CL_UNORM_INT14X2_EXT: return 1;
         default: return 0;
     }
 }

test_common/harness/imageHelpers.cpp

@@ -97,7 +97,13 @@ uint32_t get_channel_data_type_size(cl_channel_type channelType)
         case CL_UNSIGNED_INT32: return sizeof(cl_int);
 
         case CL_UNORM_SHORT_565:
-        case CL_UNORM_SHORT_555: return 2;
+        case CL_UNORM_SHORT_555:
+        case CL_UNSIGNED_INT10X6_EXT:
+        case CL_UNSIGNED_INT12X4_EXT:
+        case CL_UNSIGNED_INT14X2_EXT:
+        case CL_UNORM_INT10X6_EXT:
+        case CL_UNORM_INT12X4_EXT:
+        case CL_UNORM_INT14X2_EXT: return 2;
 
         case CL_UNORM_INT_101010:
         case CL_UNORM_INT_101010_2:
@@ -186,6 +192,12 @@ cl_channel_type get_channel_type_from_name(const char *name)
 #ifdef CL_SFIXED14_APPLE
         { CL_SFIXED14_APPLE, "CL_SFIXED14_APPLE" }
 #endif
+        { CL_UNSIGNED_INT10X6_EXT, "CL_UNSIGNED_INT10X6_EXT" },
+        { CL_UNSIGNED_INT12X4_EXT, "CL_UNSIGNED_INT12X4_EXT" },
+        { CL_UNSIGNED_INT14X2_EXT, "CL_UNSIGNED_INT14X2_EXT" },
+        { CL_UNORM_INT10X6_EXT, "CL_UNORM_INT10X6_EXT" },
+        { CL_UNORM_INT12X4_EXT, "CL_UNORM_INT12X4_EXT" },
+        { CL_UNORM_INT14X2_EXT, "CL_UNORM_INT14X2_EXT" },
     };
     for (size_t i = 0; i < sizeof(typeNames) / sizeof(typeNames[0]); i++)
     {
@@ -278,6 +290,12 @@ uint32_t get_pixel_size(const cl_image_format *format)
 #ifdef CL_SFIXED14_APPLE
         case CL_SFIXED14_APPLE:
 #endif
+        case CL_UNSIGNED_INT10X6_EXT:
+        case CL_UNSIGNED_INT12X4_EXT:
+        case CL_UNSIGNED_INT14X2_EXT:
+        case CL_UNORM_INT10X6_EXT:
+        case CL_UNORM_INT12X4_EXT:
+        case CL_UNORM_INT14X2_EXT:
             return get_format_channel_count(format) * sizeof(cl_ushort);
 
         case CL_SIGNED_INT32:
@@ -940,6 +958,9 @@ float get_max_absolute_error(const cl_image_format *format,
 #endif
         case CL_UNORM_SHORT_555:
         case CL_UNORM_SHORT_565: return 1.0f / 31.0f;
+        case CL_UNORM_INT10X6_EXT: return 1.0f / 1023.0f;
+        case CL_UNORM_INT12X4_EXT: return 1.0f / 4095.0f;
+        case CL_UNORM_INT14X2_EXT: return 1.0f / 16383.0f;
         default: return 0.0f;
     }
 }
@@ -968,6 +989,9 @@ float get_max_relative_error(const cl_image_format *format,
         case CL_UNORM_INT_101010:
         case CL_UNORM_INT_101010_2:
         case CL_UNORM_INT_2_101010_EXT:
+        case CL_UNORM_INT10X6_EXT:
+        case CL_UNORM_INT12X4_EXT:
+        case CL_UNORM_INT14X2_EXT:
             // Maximum sampling error for round to zero normalization based on
             // multiplication by reciprocal (using reciprocal generated in
             // round to +inf mode, so that 1.0 matches spec)
@@ -1053,7 +1077,15 @@ size_t get_format_max_int(const cl_image_format *format)
 
         case CL_UNORM_INT_101010:
         case CL_UNORM_INT_101010_2:
-        case CL_UNORM_INT_2_101010_EXT: return 1023;
+        case CL_UNORM_INT_2_101010_EXT:
+        case CL_UNSIGNED_INT10X6_EXT:
+        case CL_UNORM_INT10X6_EXT: return 1023;
+
+        case CL_UNSIGNED_INT12X4_EXT:
+        case CL_UNORM_INT12X4_EXT: return 4095;
+
+        case CL_UNSIGNED_INT14X2_EXT:
+        case CL_UNORM_INT14X2_EXT: return 16383;
 
         case CL_HALF_FLOAT: return 1 << 10;
 
@@ -1087,7 +1119,13 @@ int get_format_min_int(const cl_image_format *format)
         case CL_UNORM_SHORT_555:
         case CL_UNORM_INT_101010:
         case CL_UNORM_INT_101010_2:
-        case CL_UNORM_INT_2_101010_EXT: return 0;
+        case CL_UNORM_INT_2_101010_EXT:
+        case CL_UNSIGNED_INT10X6_EXT:
+        case CL_UNSIGNED_INT12X4_EXT:
+        case CL_UNSIGNED_INT14X2_EXT:
+        case CL_UNORM_INT10X6_EXT:
+        case CL_UNORM_INT12X4_EXT:
+        case CL_UNORM_INT14X2_EXT: return 0;
 
         case CL_HALF_FLOAT: return -(1 << 10);
 
@@ -1521,6 +1559,30 @@ void read_image_pixel_float(void *imageData, image_descriptor *imageInfo, int x,
             break;
         }
 
+        case CL_UNORM_INT10X6_EXT: {
+            cl_ushort *dPtr = (cl_ushort *)ptr;
+            for (i = 0; i < channelCount; i++)
+                tempData[i] =
+                    CLAMP_FLOAT((float)((dPtr[i] >> 6) & 0x3ff) / 1023.0f);
+            break;
+        }
+
+        case CL_UNORM_INT12X4_EXT: {
+            cl_ushort *dPtr = (cl_ushort *)ptr;
+            for (i = 0; i < channelCount; i++)
+                tempData[i] =
+                    CLAMP_FLOAT((float)((dPtr[i] >> 4) & 0xfff) / 4095.0f);
+            break;
+        }
+
+        case CL_UNORM_INT14X2_EXT: {
+            cl_ushort *dPtr = (cl_ushort *)ptr;
+            for (i = 0; i < channelCount; i++)
+                tempData[i] =
+                    CLAMP_FLOAT((float)((dPtr[i] >> 2) & 0x3fff) / 16383.0f);
+            break;
+        }
+
         case CL_FLOAT: {
             float *dPtr = (float *)ptr;
             for (i = 0; i < channelCount; i++) tempData[i] = (float)dPtr[i];
@@ -2634,6 +2696,24 @@ void pack_image_pixel(unsigned int *srcVector,
                 ptr[i] = (unsigned int)srcVector[i];
             break;
         }
+        case CL_UNSIGNED_INT10X6_EXT: {
+            unsigned short *ptr = (unsigned short *)outData;
+            for (unsigned int i = 0; i < channelCount; i++)
+                ptr[i] = (unsigned short)SATURATE(srcVector[i], 0, 1023) << 6;
+            break;
+        }
+        case CL_UNSIGNED_INT12X4_EXT: {
+            unsigned short *ptr = (unsigned short *)outData;
+            for (unsigned int i = 0; i < channelCount; i++)
+                ptr[i] = (unsigned short)SATURATE(srcVector[i], 0, 4095) << 4;
+            break;
+        }
+        case CL_UNSIGNED_INT14X2_EXT: {
+            unsigned short *ptr = (unsigned short *)outData;
+            for (unsigned int i = 0; i < channelCount; i++)
+                ptr[i] = (unsigned short)SATURATE(srcVector[i], 0, 16383) << 2;
+            break;
+        }
         default: break;
     }
 }
@@ -2809,6 +2889,27 @@ void pack_image_pixel(float *srcVector, const cl_image_format *imageFormat,
                 | (((unsigned int)NORMALIZE(srcVector[3], 1023.f) & 1023) << 0);
             break;
         }
+        case CL_UNORM_INT10X6_EXT: {
+            cl_ushort *ptr = (cl_ushort *)outData;
+            for (unsigned int i = 0; i < channelCount; i++)
+                ptr[i] = ((cl_ushort)NORMALIZE(srcVector[i], 1023.f) & 1023)
+                    << 6;
+            break;
+        }
+        case CL_UNORM_INT12X4_EXT: {
+            cl_ushort *ptr = (cl_ushort *)outData;
+            for (unsigned int i = 0; i < channelCount; i++)
+                ptr[i] = ((cl_ushort)NORMALIZE(srcVector[i], 4095.f) & 4095)
+                    << 4;
+            break;
+        }
+        case CL_UNORM_INT14X2_EXT: {
+            cl_ushort *ptr = (cl_ushort *)outData;
+            for (unsigned int i = 0; i < channelCount; i++)
+                ptr[i] = ((cl_ushort)NORMALIZE(srcVector[i], 16383.f) & 16383)
+                    << 2;
+            break;
+        }
         case CL_SIGNED_INT8: {
             cl_char *ptr = (cl_char *)outData;
             for (unsigned int i = 0; i < channelCount; i++)
@@ -2852,6 +2953,27 @@ void pack_image_pixel(float *srcVector, const cl_image_format *imageFormat,
                     CL_UINT_MAX);
             break;
         }
+        case CL_UNSIGNED_INT10X6_EXT: {
+            cl_ushort *ptr = (cl_ushort *)outData;
+            for (unsigned int i = 0; i < channelCount; i++)
+                ptr[i] = ((cl_ushort)NORMALIZE(srcVector[i], 1023.f) & 1023)
+                    << 6;
+            break;
+        }
+        case CL_UNSIGNED_INT12X4_EXT: {
+            cl_ushort *ptr = (cl_ushort *)outData;
+            for (unsigned int i = 0; i < channelCount; i++)
+                ptr[i] = ((cl_ushort)NORMALIZE(srcVector[i], 4095.f) & 4095)
+                    << 4;
+            break;
+        }
+        case CL_UNSIGNED_INT14X2_EXT: {
+            cl_ushort *ptr = (cl_ushort *)outData;
+            for (unsigned int i = 0; i < channelCount; i++)
+                ptr[i] = ((cl_ushort)NORMALIZE(srcVector[i], 16383.f) & 16383)
+                    << 2;
+            break;
+        }
 #ifdef CL_SFIXED14_APPLE
         case CL_SFIXED14_APPLE: {
             cl_ushort *ptr = (cl_ushort *)outData;
@@ -2999,6 +3121,33 @@ void pack_image_pixel_error(const float *srcVector,
 
             break;
         }
+        case CL_UNORM_INT10X6_EXT: {
+            const cl_ushort *ptr = (const cl_ushort *)results;
+
+            for (unsigned int i = 0; i < channelCount; i++)
+                errors[i] =
+                    (ptr[i] >> 6) - NORMALIZE_UNROUNDED(srcVector[i], 1023.f);
+
+            break;
+        }
+        case CL_UNORM_INT12X4_EXT: {
+            const cl_ushort *ptr = (const cl_ushort *)results;
+
+            for (unsigned int i = 0; i < channelCount; i++)
+                errors[i] =
+                    (ptr[i] >> 4) - NORMALIZE_UNROUNDED(srcVector[i], 4095.f);
+
+            break;
+        }
+        case CL_UNORM_INT14X2_EXT: {
+            const cl_ushort *ptr = (const cl_ushort *)results;
+
+            for (unsigned int i = 0; i < channelCount; i++)
+                errors[i] =
+                    (ptr[i] >> 2) - NORMALIZE_UNROUNDED(srcVector[i], 16383.f);
+
+            break;
+        }
         case CL_SIGNED_INT8: {
             const cl_char *ptr = (const cl_char *)results;
 
@@ -3050,6 +3199,29 @@ void pack_image_pixel_error(const float *srcVector,
                         CL_UINT_MAX));
             break;
         }
+        case CL_UNSIGNED_INT10X6_EXT: {
+            cl_ushort *ptr = (cl_ushort *)results;
+            for (unsigned int i = 0; i < channelCount; i++)
+                errors[i] = static_cast<float>(
+                    (cl_int)ptr[i] - (*((cl_int *)(&srcVector[i])) & 0xffc0));
+            break;
+        }
+        case CL_UNSIGNED_INT12X4_EXT: {
+            cl_ushort *ptr = (cl_ushort *)results;
+            for (unsigned int i = 0; i < channelCount; i++)
+                errors[i] = static_cast<float>(
+                    (cl_int)ptr[i] - (*((cl_int *)(&srcVector[i])) & 0xfff0));
+            break;
+        }
+        case CL_UNSIGNED_INT14X2_EXT: {
+            cl_ushort *ptr = (cl_ushort *)results;
+            for (unsigned int i = 0; i < channelCount; i++)
+                errors[i] = static_cast<float>(
+                    (cl_int)ptr[i]
+                    - (cl_int)CONVERT_UINT(srcVector[i], 16383.f,
+                                           CL_USHRT_MAX));
+            break;
+        }
 #ifdef CL_SFIXED14_APPLE
         case CL_SFIXED14_APPLE: {
             const cl_ushort *ptr = (const cl_ushort *)results;

test_common/harness/imageHelpers.h

@@ -126,6 +126,7 @@ typedef struct
     const cl_image_format *format;
     cl_mem buffer;
     cl_mem_object_type type;
+    cl_mem_flags mem_flags;
     cl_uint num_mip_levels;
 } image_descriptor;
 
@@ -447,6 +448,27 @@ void read_image_pixel(void *imageData, image_descriptor *imageInfo, int x,
             tempData[0] = (T)(hi_val | lo_val);
             break;
         }
+        case CL_UNSIGNED_INT10X6_EXT: {
+            cl_short *dPtr = (cl_short *)ptr;
+            const size_t channelCount = get_format_channel_count(format);
+            for (i = 0; i < channelCount; i++)
+                tempData[i] = (dPtr[i] >> 6) & 0x3ff;
+            break;
+        }
+        case CL_UNSIGNED_INT12X4_EXT: {
+            cl_short *dPtr = (cl_short *)ptr;
+            const size_t channelCount = get_format_channel_count(format);
+            for (i = 0; i < channelCount; i++)
+                tempData[i] = (dPtr[i] >> 4) & 0xfff;
+            break;
+        }
+        case CL_UNSIGNED_INT14X2_EXT: {
+            cl_short *dPtr = (cl_short *)ptr;
+            const size_t channelCount = get_format_channel_count(format);
+            for (i = 0; i < channelCount; i++)
+                tempData[i] = (dPtr[i] >> 2) & 0x3fff;
+            break;
+        }
     }
 
 

test_common/harness/kernelHelpers.cpp

@@ -1328,6 +1328,12 @@ size_t get_pixel_bytes(const cl_image_format *fmt)
         case CL_SIGNED_INT8:
         case CL_UNSIGNED_INT8: return chanCount;
 
+        case CL_UNSIGNED_INT10X6_EXT:
+        case CL_UNSIGNED_INT12X4_EXT:
+        case CL_UNSIGNED_INT14X2_EXT:
+        case CL_UNORM_INT10X6_EXT:
+        case CL_UNORM_INT12X4_EXT:
+        case CL_UNORM_INT14X2_EXT:
         case CL_SNORM_INT16:
         case CL_UNORM_INT16:
         case CL_HALF_FLOAT:

test_common/harness/typeWrappers.h

@@ -377,4 +377,24 @@ public:
     size_t getSize() const { return allocsize; };
 };
 
+// scope guard helper to ensure proper releasing of sub devices
+struct SubDevicesScopeGuarded
+{
+    SubDevicesScopeGuarded(const cl_int dev_count)
+    {
+        sub_devices.resize(dev_count);
+    }
+    ~SubDevicesScopeGuarded()
+    {
+        for (auto &device : sub_devices)
+        {
+            cl_int err = clReleaseDevice(device);
+            if (err != CL_SUCCESS)
+                log_error("\n Releasing sub-device failed \n");
+        }
+    }
+
+    std::vector<cl_device_id> sub_devices;
+};
+
 #endif // _typeWrappers_h

test_conformance/allocations/main.cpp

@@ -26,6 +26,8 @@ typedef long long unsigned llu;
 
 #define REDUCTION_PERCENTAGE_DEFAULT 50
 
+#define BYTES_PER_WORK_ITEM 2048ULL
+
 int g_repetition_count = 1;
 int g_reduction_percentage = REDUCTION_PERCENTAGE_DEFAULT;
 int g_write_allocations = 1;
@@ -125,7 +127,7 @@ int doTest(cl_device_id device, cl_context context, cl_command_queue queue,
     int number_of_mems_used;
     cl_ulong max_individual_allocation_size = g_max_individual_allocation_size;
     cl_ulong global_mem_size = g_global_mem_size;
-    unsigned int number_of_work_items = 8192 * 32;
+    unsigned int number_of_work_items;
     const bool allocate_image =
         (alloc_type != BUFFER) && (alloc_type != BUFFER_NON_BLOCKING);
 
@@ -183,12 +185,16 @@ int doTest(cl_device_id device, cl_context context, cl_command_queue queue,
                  g_reduction_percentage);
         g_max_size = (size_t)((double)g_max_size
                               * (double)g_reduction_percentage / 100.0);
-        number_of_work_items = 8192 * 2;
     }
 
     // Round to nearest MB.
     g_max_size &= (size_t)(0xFFFFFFFFFF00000ULL);
 
+    // Scales the number of work-items to keep the amount of bytes processed
+    // per work-item the same.
+    number_of_work_items =
+        std::max(g_max_size / BYTES_PER_WORK_ITEM, 8192ULL * 2ULL);
+
     log_info("** Target allocation size (rounded to nearest MB) is: %llu bytes "
              "(%gMB).\n",
              llu(g_max_size), toMB(g_max_size));

test_conformance/api/CMakeLists.txt

@@ -4,6 +4,7 @@ set(${MODULE_NAME}_SOURCES
          main.cpp
          negative_platform.cpp
          negative_queue.cpp
+         negative_enqueue_map_image.cpp
          test_api_consistency.cpp
          test_bool.cpp
          test_retain.cpp

test_conformance/api/negative_enqueue_map_image.cpp

@@ -0,0 +1,191 @@
+//
+// Copyright (c) 2024 The Khronos Group Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//    http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#include "testBase.h"
+#include "harness/clImageHelper.h"
+
+#include <array>
+#include <vector>
+#include <memory>
+
+static constexpr cl_mem_object_type image_types[] = {
+    CL_MEM_OBJECT_IMAGE2D, CL_MEM_OBJECT_IMAGE3D, CL_MEM_OBJECT_IMAGE2D_ARRAY,
+    CL_MEM_OBJECT_IMAGE1D, CL_MEM_OBJECT_IMAGE1D_ARRAY
+};
+
+REGISTER_TEST(negative_enqueue_map_image)
+{
+    constexpr size_t image_dim = 32;
+
+    REQUIRE_EXTENSION("cl_ext_immutable_memory_objects");
+
+    static constexpr cl_mem_flags mem_flags[]{
+        CL_MEM_IMMUTABLE_EXT | CL_MEM_USE_HOST_PTR,
+        CL_MEM_IMMUTABLE_EXT | CL_MEM_COPY_HOST_PTR,
+        CL_MEM_IMMUTABLE_EXT | CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR
+    };
+
+    static constexpr const char *mem_flags_string[]{
+        "CL_MEM_IMMUTABLE_EXT | CL_MEM_USE_HOST_PTR",
+        "CL_MEM_IMMUTABLE_EXT | CL_MEM_COPY_HOST_PTR",
+        "CL_MEM_IMMUTABLE_EXT | CL_MEM_COPY_HOST_PTR | "
+        "CL_MEM_ALLOC_HOST_PTR"
+    };
+
+    static_assert(ARRAY_SIZE(mem_flags) == ARRAY_SIZE(mem_flags_string),
+                  "mem_flags and mem_flags_string must be of the same size");
+
+    using CLUCharPtr = std::unique_ptr<cl_uchar, decltype(&free)>;
+
+    for (size_t index = 0; index < ARRAY_SIZE(mem_flags); ++index)
+    {
+        cl_mem_flags mem_flag = mem_flags[index];
+
+        log_info("Testing memory flag: %s\n", mem_flags_string[index]);
+        for (cl_mem_object_type image_type : image_types)
+        {
+            // find supported image formats
+            cl_uint num_formats = 0;
+
+            cl_int error = clGetSupportedImageFormats(
+                context, mem_flag, image_type, 0, nullptr, &num_formats);
+            test_error(error,
+                       "clGetSupportedImageFormats failed to return supported "
+                       "formats");
+
+            std::vector<cl_image_format> formats(num_formats);
+            error = clGetSupportedImageFormats(context, mem_flag, image_type,
+                                               num_formats, formats.data(),
+                                               nullptr);
+            test_error(error,
+                       "clGetSupportedImageFormats failed to return supported "
+                       "formats");
+
+            clMemWrapper image;
+            for (cl_image_format &fmt : formats)
+            {
+                log_info("Testing %s %s\n",
+                         GetChannelOrderName(fmt.image_channel_order),
+                         GetChannelTypeName(fmt.image_channel_data_type));
+
+                RandomSeed seed(gRandomSeed);
+                size_t origin[3] = { 0, 0, 0 };
+                size_t region[3] = { image_dim, image_dim, image_dim };
+                switch (image_type)
+                {
+                    case CL_MEM_OBJECT_IMAGE1D: {
+                        const size_t pixel_size = get_pixel_size(&fmt);
+                        const size_t image_size =
+                            image_dim * pixel_size * sizeof(cl_uchar);
+                        CLUCharPtr imgptr{ static_cast<cl_uchar *>(
+                                               create_random_data(kUChar, seed,
+                                                                  image_size)),
+                                           free };
+                        image =
+                            create_image_1d(context, mem_flag, &fmt, image_dim,
+                                            0, imgptr.get(), nullptr, &error);
+                        region[1] = 1;
+                        region[2] = 1;
+                        break;
+                    }
+                    case CL_MEM_OBJECT_IMAGE2D: {
+                        const size_t pixel_size = get_pixel_size(&fmt);
+                        const size_t image_size = image_dim * image_dim
+                            * pixel_size * sizeof(cl_uchar);
+                        CLUCharPtr imgptr{ static_cast<cl_uchar *>(
+                                               create_random_data(kUChar, seed,
+                                                                  image_size)),
+                                           free };
+                        image =
+                            create_image_2d(context, mem_flag, &fmt, image_dim,
+                                            image_dim, 0, imgptr.get(), &error);
+                        region[2] = 1;
+                        break;
+                    }
+                    case CL_MEM_OBJECT_IMAGE3D: {
+                        const size_t pixel_size = get_pixel_size(&fmt);
+                        const size_t image_size = image_dim * image_dim
+                            * image_dim * pixel_size * sizeof(cl_uchar);
+                        CLUCharPtr imgptr{ static_cast<cl_uchar *>(
+                                               create_random_data(kUChar, seed,
+                                                                  image_size)),
+                                           free };
+                        image = create_image_3d(context, mem_flag, &fmt,
+                                                image_dim, image_dim, image_dim,
+                                                0, 0, imgptr.get(), &error);
+                        break;
+                    }
+                    case CL_MEM_OBJECT_IMAGE1D_ARRAY: {
+                        const size_t pixel_size = get_pixel_size(&fmt);
+                        const size_t image_size = image_dim * image_dim
+                            * pixel_size * sizeof(cl_uchar);
+                        CLUCharPtr imgptr{ static_cast<cl_uchar *>(
+                                               create_random_data(kUChar, seed,
+                                                                  image_size)),
+                                           free };
+                        image = create_image_1d_array(context, mem_flag, &fmt,
+                                                      image_dim, image_dim, 0,
+                                                      0, imgptr.get(), &error);
+                        region[1] = 1;
+                        region[2] = 1;
+                        break;
+                    }
+                    case CL_MEM_OBJECT_IMAGE2D_ARRAY: {
+                        const size_t pixel_size = get_pixel_size(&fmt);
+                        const size_t image_size = image_dim * image_dim
+                            * image_dim * pixel_size * sizeof(cl_uchar);
+                        CLUCharPtr imgptr{ static_cast<cl_uchar *>(
+                                               create_random_data(kUChar, seed,
+                                                                  image_size)),
+                                           free };
+                        image = create_image_2d_array(
+                            context, mem_flag, &fmt, image_dim, image_dim,
+                            image_dim, 0, 0, imgptr.get(), &error);
+                        region[2] = 1;
+                        break;
+                    }
+                }
+                test_error(error, "Failed to create image");
+
+                void *map = clEnqueueMapImage(
+                    queue, image, CL_TRUE, CL_MAP_WRITE, origin, region,
+                    nullptr, nullptr, 0, nullptr, nullptr, &error);
+
+                constexpr const char *write_err_msg =
+                    "clEnqueueMapImage should return CL_INVALID_OPERATION "
+                    "when: \"image has been created with CL_MEM_IMMUTABLE_EXT "
+                    "and CL_MAP_WRITE is set in map_flags\"";
+                test_assert_error(map == nullptr, write_err_msg);
+                test_failure_error_ret(error, CL_INVALID_OPERATION,
+                                       write_err_msg, TEST_FAIL);
+
+                map = clEnqueueMapImage(queue, image, CL_TRUE,
+                                        CL_MAP_WRITE_INVALIDATE_REGION, origin,
+                                        region, nullptr, nullptr, 0, nullptr,
+                                        nullptr, &error);
+
+                constexpr const char *write_invalidate_err_msg =
+                    "clEnqueueMapImage should return CL_INVALID_OPERATION "
+                    "when: \"image has been created with CL_MEM_IMMUTABLE_EXT "
+                    "and CL_MAP_WRITE_INVALIDATE_REGION is set in map_flags\"";
+                test_assert_error(map == nullptr, write_invalidate_err_msg);
+                test_failure_error_ret(error, CL_INVALID_OPERATION,
+                                       write_invalidate_err_msg, TEST_FAIL);
+            }
+        }
+    }
+
+    return TEST_PASS;
+}

test_conformance/api/testBase.h

@@ -27,24 +27,4 @@
 #include <sys/types.h>
 #include <sys/stat.h>
 
-// scope guard helper to ensure proper releasing of sub devices
-struct SubDevicesScopeGuarded
-{
-    SubDevicesScopeGuarded(const cl_int dev_count)
-    {
-        sub_devices.resize(dev_count);
-    }
-    ~SubDevicesScopeGuarded()
-    {
-        for (auto &device : sub_devices)
-        {
-            cl_int err = clReleaseDevice(device);
-            if (err != CL_SUCCESS)
-                log_error("\n Releasing sub-device failed \n");
-        }
-    }
-
-    std::vector<cl_device_id> sub_devices;
-};
-
 #endif // _testBase_h

test_conformance/api/test_api_consistency.cpp

@@ -1243,13 +1243,6 @@ REGISTER_TEST(consistency_requirements_fp16)
     }
     else
     {
-        error = clGetDeviceInfo(device, CL_DEVICE_HALF_FP_CONFIG, sizeof(value),
-                                &value, nullptr);
-        test_failure_error(
-            error, CL_INVALID_VALUE,
-            "cl_khr_fp16 is not available; CL_DEVICE_HALF_FP_CONFIG must fail "
-            "with CL_INVALID_VALUE");
-
         error = clGetDeviceInfo(device, CL_DEVICE_PREFERRED_VECTOR_WIDTH_HALF,
                                 sizeof(value), &value, nullptr);
         test_error(

test_conformance/api/test_kernel_local_memory_size.cpp

@@ -153,6 +153,8 @@ REGISTER_TEST(kernel_local_memory_size)
                       "kernel local mem size failed");
 
 
+    program.reset();
+    kernel.reset();
     // Check memory needed to execute empty kernel with __local parameter with
     // setKernelArg
     if (create_single_kernel_helper(context, &program, &kernel, 1,
@@ -225,6 +227,8 @@ REGISTER_TEST(kernel_local_memory_size)
                       "kernel local mem size failed");
 
 
+    program.reset();
+    kernel.reset();
     // Check memory needed to execute kernel with __local variable and __local
     // parameter with setKernelArg
     if (create_single_kernel_helper(context, &program, &kernel, 1,

test_conformance/api/test_kernels.cpp

@@ -16,6 +16,7 @@
 #include "testBase.h"
 #include "harness/typeWrappers.h"
 #include "harness/conversions.h"
+#include <vector>
 
 const char *sample_single_test_kernel[] = {
 "__kernel void sample_test(__global float *src, __global int *dst)\n"
@@ -49,6 +50,17 @@ const char *sample_const_test_kernel[] = {
 "\n"
 "}\n" };
 
+const char *sample_image_test_kernel[] = {
+    "__kernel void sample_image_test(__read_only image2d_t src, __write_only "
+    "image2d_t dst)\n"
+    "{\n"
+    "    int2 coord = (int2)(get_global_id(0), get_global_id(1));\n"
+    "    uint4 value = read_imageui(src, coord);\n"
+    "    write_imageui(dst, coord, value);\n"
+    "\n"
+    "}\n"
+};
+
 const char *sample_const_global_test_kernel[] = {
 "__constant int addFactor = 1024;\n"
 "__kernel void sample_test(__global int *src1, __global int *dst)\n"
@@ -631,3 +643,64 @@ REGISTER_TEST(kernel_global_constant)
 
     return 0;
 }
+
+REGISTER_TEST(negative_set_immutable_memory_to_writeable_kernel_arg)
+{
+    REQUIRE_EXTENSION("cl_ext_immutable_memory_objects");
+
+    cl_int error = CL_SUCCESS;
+    clProgramWrapper program;
+    clKernelWrapper kernels[2];
+    clMemWrapper image, buffer;
+    const char *test_kernels[2] = { sample_const_test_kernel[0],
+                                    sample_image_test_kernel[0] };
+    constexpr cl_image_format formats = { CL_RGBA, CL_UNSIGNED_INT8 };
+    constexpr size_t size_dim = 128;
+
+    // Setup the test
+    error = create_single_kernel_helper(context, &program, nullptr, 2,
+                                        test_kernels, nullptr);
+    test_error(error, "Unable to build test program");
+
+    kernels[0] = clCreateKernel(program, "sample_test", &error);
+    test_error(error, "Unable to get sample_test kernel for built program");
+
+    kernels[1] = clCreateKernel(program, "sample_image_test", &error);
+    test_error(error,
+               "Unable to get sample_image_test kernel for built program");
+
+    std::vector<cl_uchar> mem_data(size_dim * size_dim);
+    buffer = clCreateBuffer(context, CL_MEM_IMMUTABLE_EXT | CL_MEM_USE_HOST_PTR,
+                            sizeof(cl_int) * size_dim, mem_data.data(), &error);
+    test_error(error, "clCreateBuffer failed");
+
+    image = create_image_2d(context, CL_MEM_IMMUTABLE_EXT | CL_MEM_USE_HOST_PTR,
+                            &formats, size_dim, size_dim, 0, mem_data.data(),
+                            &error);
+    test_error(error, "create_image_2d failed");
+
+    // Run the test
+    error = clSetKernelArg(kernels[0], 0, sizeof(buffer), &buffer);
+    test_error(error, "clSetKernelArg failed");
+
+    error = clSetKernelArg(kernels[0], 2, sizeof(buffer), &buffer);
+    test_failure_error_ret(error, CL_INVALID_ARG_VALUE,
+                           "clSetKernelArg is supposed to fail "
+                           "with CL_INVALID_ARG_VALUE when a buffer is "
+                           "created with CL_MEM_IMMUTABLE_EXT is "
+                           "passed to a non-constant kernel argument",
+                           TEST_FAIL);
+
+    error = clSetKernelArg(kernels[1], 0, sizeof(image), &image);
+    test_error(error, "clSetKernelArg failed");
+
+    error = clSetKernelArg(kernels[1], 1, sizeof(image), &image);
+    test_failure_error_ret(error, CL_INVALID_ARG_VALUE,
+                           "clSetKernelArg is supposed to fail "
+                           "with CL_INVALID_ARG_VALUE when an image is "
+                           "created with CL_MEM_IMMUTABLE_EXT is "
+                           "passed to a write_only kernel argument",
+                           TEST_FAIL);
+
+    return TEST_PASS;
+}

test_conformance/api/test_queries.cpp

@@ -796,6 +796,60 @@ REGISTER_TEST(get_device_info)
     return 0;
 }
 
+REGISTER_TEST(get_device_info_comparability)
+{
+    int error = CL_SUCCESS;
+
+    // comparability test for CL_DEVICE_PLATFORM
+    // 1. find platform related to device without using query
+    cl_uint num_platforms = 0;
+    error = clGetPlatformIDs(16, nullptr, &num_platforms);
+    test_error(error, "clGetPlatformIDs failed");
+
+    std::vector<cl_platform_id> platforms(num_platforms);
+
+    error = clGetPlatformIDs(num_platforms, platforms.data(), &num_platforms);
+    test_error(error, "clGetPlatformIDs failed");
+
+    cl_uint num_devices = 0;
+    cl_platform_id comp_platform = nullptr;
+    for (int p = 0; p < (int)num_platforms && comp_platform == nullptr; p++)
+    {
+        error = clGetDeviceIDs(platforms[p], CL_DEVICE_TYPE_ALL, 0, nullptr,
+                               &num_devices);
+        if (error != CL_SUCCESS || num_devices == 0) continue;
+
+        std::vector<cl_device_id> devices(num_devices);
+        error = clGetDeviceIDs(platforms[p], CL_DEVICE_TYPE_ALL, num_devices,
+                               devices.data(), nullptr);
+        if (error != CL_SUCCESS) continue;
+
+        // find correct device
+        for (auto did : devices)
+        {
+            if (did == device)
+            {
+                comp_platform = platforms[p];
+                break;
+            }
+        }
+    }
+
+    test_error_fail(comp_platform == nullptr,
+                    "Test failed to locate platform for comparison!");
+
+    // 2. compare platforms found with and without using query
+    cl_platform_id plat = nullptr;
+    error =
+        clGetDeviceInfo(device, CL_DEVICE_PLATFORM, sizeof(plat), &plat, NULL);
+    test_error(error, "clGetDeviceInfo failed");
+
+    test_assert_error(plat == comp_platform,
+                      "Unexpected result returned by clGetDeviceInfo for "
+                      "CL_DEVICE_PLATFORM query");
+
+    return TEST_PASS;
+}
 
 static const char *sample_compile_size[2] = {
     "__kernel void sample_test(__global int *src, __global int *dst)\n"

test_conformance/api/test_queue.cpp

@@ -23,7 +23,8 @@ REGISTER_TEST(queue_flush_on_release)
     cl_int err;
 
     // Create a command queue
-    cl_command_queue cmd_queue = clCreateCommandQueue(context, device, 0, &err);
+    clCommandQueueWrapper cmd_queue =
+        clCreateCommandQueue(context, device, 0, &err);
     test_error(err, "Could not create command queue");
 
     // Create a kernel
@@ -42,7 +43,7 @@ REGISTER_TEST(queue_flush_on_release)
     test_error(err, "Could not enqueue kernel");
 
     // Release the queue
-    err = clReleaseCommandQueue(cmd_queue);
+    cmd_queue.reset();
 
     // Wait for kernel to execute since the queue must flush on release
     bool success = poll_until(2000, 50, [&event]() {
@@ -64,11 +65,13 @@ REGISTER_TEST(multi_queue_flush_on_release)
     cl_int err;
 
     // Create A command queue
-    cl_command_queue queue_A = clCreateCommandQueue(context, device, 0, &err);
+    clCommandQueueWrapper queue_A =
+        clCreateCommandQueue(context, device, 0, &err);
     test_error(err, "Could not create command queue A");
 
     // Create B command queue
-    cl_command_queue queue_B = clCreateCommandQueue(context, device, 0, &err);
+    clCommandQueueWrapper queue_B =
+        clCreateCommandQueue(context, device, 0, &err);
     test_error(err, "Could not create command queue B");
 
     // Create a kernel
@@ -96,8 +99,7 @@ REGISTER_TEST(multi_queue_flush_on_release)
 
     // Release queue_A, which performs an implicit flush to issue any previously
     // queued OpenCL commands
-    err = clReleaseCommandQueue(queue_A);
-    test_error(err, "clReleaseCommandQueue failed");
+    queue_A.reset();
 
     err = clFlush(queue_B);
     test_error(err, "clFlush failed");

test_conformance/api/test_zero_sized_enqueue.cpp

@@ -1,6 +1,6 @@
 //
-// Copyright (c) 2017 The Khronos Group Inc.
-// 
+// Copyright (c) 2017-2025 The Khronos Group Inc.
+//
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
@@ -14,18 +14,14 @@
 // limitations under the License.
 //
 #include "testBase.h"
-#include "harness/typeWrappers.h"
-#include "harness/conversions.h"
 
-const char* zero_sized_enqueue_test_kernel[] = {
-    "__kernel void foo_kernel(__global int *dst)\n"
-    "{\n"
-    "    int  tid = get_global_id(0);\n"
-    "\n"
-    "    dst[tid] = 1;\n"
-    "\n"
-    "}\n"
-};
+const char* zero_sized_enqueue_test_kernel = R"(
+__kernel void foo_kernel(__global int *dst)
+{
+    int  tid = get_global_id(0);
+    dst[tid] = 1;
+}
+)";
 
 const int bufSize = 128;
 
@@ -81,15 +77,20 @@ int test_zero_sized_enqueue_helper(cl_device_id device, cl_context context,
     output_stream =
         clCreateBuffer(context, CL_MEM_READ_WRITE | CL_MEM_ALLOC_HOST_PTR,
                        bufSize * sizeof(int), NULL, &error);
+    test_error(error, "clCreateBuffer failed.");
 
     // Initialise output buffer.
     int output_buffer_data = 0;
     error = clEnqueueFillBuffer(queue, output_stream, &output_buffer_data,
                                 sizeof(int), 0, sizeof(int) * bufSize, 0, NULL,
                                 NULL);
+    test_error(error, "clEnqueueFillBuffer failed.");
 
     /* Create a kernel to test with */
-    if( create_single_kernel_helper( context, &program, &kernel, 1, zero_sized_enqueue_test_kernel, "foo_kernel" ) != 0 )
+    if (create_single_kernel_helper(context, &program, &kernel, 1,
+                                    &zero_sized_enqueue_test_kernel,
+                                    "foo_kernel")
+        != 0)
     {
         return -1;
     }
@@ -98,6 +99,16 @@ int test_zero_sized_enqueue_helper(cl_device_id device, cl_context context,
     test_error( error, "clSetKernelArg failed." );
 
     // Simple API return code tests for 1D, 2D and 3D zero sized ND range.
+    error = test_zero_sized_enqueue_and_test_output_buffer(
+        queue, kernel, output_stream, 1, nullptr);
+    test_error(error, "1D null sized kernel enqueue failed.");
+    error = test_zero_sized_enqueue_and_test_output_buffer(
+        queue, kernel, output_stream, 2, nullptr);
+    test_error(error, "2D null sized kernel enqueue failed.");
+    error = test_zero_sized_enqueue_and_test_output_buffer(
+        queue, kernel, output_stream, 3, nullptr);
+    test_error(error, "3D null sized kernel enqueue failed.");
+
     error = test_zero_sized_enqueue_and_test_output_buffer(
         queue, kernel, output_stream, 1, &ndrange1);
     test_error( error, "1D zero sized kernel enqueue failed." );
@@ -156,6 +167,17 @@ int test_zero_sized_enqueue_helper(cl_device_id device, cl_context context,
         return -1;
     }
 
+    cl_command_type cmdtype;
+    error = clGetEventInfo(ev, CL_EVENT_COMMAND_TYPE, sizeof(cmdtype), &cmdtype,
+                           NULL);
+    test_error(error, "Failed to get event command type.");
+    if (cmdtype != CL_COMMAND_NDRANGE_KERNEL)
+    {
+        log_error(
+            "ERROR: incorrect zero sized kernel enqueue event command type.\n");
+        return -1;
+    }
+
     cl_int sta;
     error = clGetEventInfo(ev, CL_EVENT_COMMAND_EXECUTION_STATUS, sizeof(cl_int), &sta, NULL);
     test_error( error, "Failed to get event status.");

test_conformance/basic/CMakeLists.txt

@@ -71,6 +71,6 @@ if(APPLE)
     list(APPEND ${MODULE_NAME}_SOURCES test_queue_priority.cpp)
 endif(APPLE)
 
-set_gnulike_module_compile_flags("-Wno-sign-compare -Wno-format")
+set_gnulike_module_compile_flags("-Wno-sign-compare")
 
 include(../CMakeCommon.txt)

test_conformance/basic/test_arrayreadwrite.cpp

@@ -20,31 +20,42 @@
 #include <string.h>
 #include <sys/types.h>
 #include <sys/stat.h>
+#include <vector>
 
 #include "testBase.h"
 
-REGISTER_TEST(arrayreadwrite)
+static int test_arrayreadwrite_impl(cl_device_id device, cl_context context,
+                                    cl_command_queue queue, int num_elements,
+                                    cl_mem_flags flags)
 {
-    cl_uint                *inptr, *outptr;
-    cl_mem              streams[1];
+    clMemWrapper buffer;
     int                 num_tries = 400;
     num_elements = 1024 * 1024 * 4;
-    int                 i, j, err;
-    MTdata              d;
+    MTdataHolder d(gRandomSeed);
 
-    inptr = (cl_uint*)malloc(num_elements*sizeof(cl_uint));
-    outptr = (cl_uint*)malloc(num_elements*sizeof(cl_uint));
+    std::vector<cl_uint> reference_vals(num_elements);
+    std::vector<cl_uint> inptr(num_elements);
+    std::vector<cl_uint> outptr(num_elements);
 
     // randomize data
-    d = init_genrand( gRandomSeed );
-    for (i=0; i<num_elements; i++)
+    for (int i = 0; i < num_elements; i++)
+    {
         inptr[i] = (cl_uint)(genrand_int32(d) & 0x7FFFFFFF);
+        reference_vals[i] = (cl_uint)(genrand_int32(d) & 0x7FFFFFFF);
+    }
 
-    streams[0] = clCreateBuffer(context, CL_MEM_READ_WRITE,
-                                sizeof(cl_uint) * num_elements, NULL, &err);
+    void* host_ptr = nullptr;
+    if ((flags & CL_MEM_USE_HOST_PTR) || (flags & CL_MEM_COPY_HOST_PTR))
+    {
+        host_ptr = inptr.data();
+    }
+
+    cl_int err = CL_SUCCESS;
+    buffer = clCreateBuffer(context, flags, sizeof(cl_uint) * num_elements,
+                            host_ptr, &err);
     test_error(err, "clCreateBuffer failed");
 
-    for (i=0; i<num_tries; i++)
+    for (int i = 0; i < num_tries; i++)
     {
         int        offset;
         int        cb;
@@ -58,15 +69,19 @@ REGISTER_TEST(arrayreadwrite)
         if (cb > (num_elements - offset))
             cb = num_elements - offset;
 
-        err = clEnqueueWriteBuffer(queue, streams[0], CL_TRUE, offset*sizeof(cl_uint), sizeof(cl_uint)*cb,&inptr[offset], 0, NULL, NULL);
+        err = clEnqueueWriteBuffer(
+            queue, buffer, CL_TRUE, offset * sizeof(cl_uint),
+            sizeof(cl_uint) * cb, &reference_vals[offset], 0, nullptr, nullptr);
         test_error(err, "clEnqueueWriteBuffer failed");
 
-        err = clEnqueueReadBuffer( queue, streams[0], CL_TRUE, offset*sizeof(cl_uint), cb*sizeof(cl_uint), &outptr[offset], 0, NULL, NULL );
+        err = clEnqueueReadBuffer(
+            queue, buffer, CL_TRUE, offset * sizeof(cl_uint),
+            cb * sizeof(cl_uint), &outptr[offset], 0, nullptr, nullptr);
         test_error(err, "clEnqueueReadBuffer failed");
 
-        for (j=offset; j<offset+cb; j++)
+        for (int j = offset; j < offset + cb; j++)
         {
-            if (inptr[j] != outptr[j])
+            if (reference_vals[j] != outptr[j])
             {
                 log_error("ARRAY read, write test failed\n");
                 err = -1;
@@ -78,13 +93,15 @@ REGISTER_TEST(arrayreadwrite)
             break;
     }
 
-    free_mtdata(d);
-    clReleaseMemObject(streams[0]);
-    free(inptr);
-    free(outptr);
-
     if (!err)
         log_info("ARRAY read, write test passed\n");
 
     return err;
 }
+
+
+REGISTER_TEST(arrayreadwrite)
+{
+    return test_arrayreadwrite_impl(device, context, queue, num_elements,
+                                    CL_MEM_READ_WRITE);
+}

test_conformance/basic/test_bufferreadwriterect.cpp

@@ -21,6 +21,8 @@
 #include <sys/types.h>
 #include <sys/stat.h>
 
+#include <cinttypes>
+
 #include "testBase.h"
 
 #define CL_EXIT_ERROR(cmd,format,...)                \
@@ -340,8 +342,21 @@ void CL_CALLBACK mem_obj_destructor_callback( cl_mem, void *data )
     free( data );
 }
 
-// This is the main test function for the conformance test.
-REGISTER_TEST(bufferreadwriterect)
+using test_fn = int (*)(size_t, size_t[3], size_t[3], size_t, size_t[3],
+                        size_t[3]);
+struct TestFunctions
+{
+    test_fn copy;
+    test_fn read;
+    test_fn write;
+};
+
+static int test_bufferreadwriterect_impl(cl_device_id device,
+                                         cl_context context,
+                                         cl_command_queue queue,
+                                         int num_elements,
+                                         cl_map_flags buffer_flags,
+                                         const TestFunctions& test_functions)
 {
     gQueue = queue;
     cl_int err;
@@ -352,7 +367,8 @@ REGISTER_TEST(bufferreadwriterect)
     // Compute a maximum buffer size based on the number of test images and the device maximum.
     cl_ulong max_mem_alloc_size = 0;
     CL_EXIT_ERROR(clGetDeviceInfo(device, CL_DEVICE_MAX_MEM_ALLOC_SIZE, sizeof(cl_ulong), &max_mem_alloc_size, NULL),"Could not get device info");
-    log_info("CL_DEVICE_MAX_MEM_ALLOC_SIZE = %llu bytes.\n", max_mem_alloc_size);
+    log_info("CL_DEVICE_MAX_MEM_ALLOC_SIZE = %" PRIu64 " bytes.\n",
+             max_mem_alloc_size);
 
     // Confirm that the maximum allocation size is not zero.
     if (max_mem_alloc_size == 0) {
@@ -390,7 +406,7 @@ REGISTER_TEST(bufferreadwriterect)
         // Check to see if adequately sized buffers were found.
         if (tries >= max_tries) {
             log_error("Error: Could not find random buffer sized less than "
-                      "%llu bytes in %zu tries.\n",
+                      "%" PRIu64 " bytes in %zu tries.\n",
                       max_mem_alloc_size, max_tries);
             return -1;
         }
@@ -431,7 +447,8 @@ REGISTER_TEST(bufferreadwriterect)
         memcpy(backing[i], verify[i], size_bytes);
 
         // Create the CL buffer.
-        buffer[i] = clCreateBuffer (context, CL_MEM_USE_HOST_PTR | CL_MEM_READ_WRITE, size_bytes, backing[i], &err);
+        buffer[i] =
+            clCreateBuffer(context, buffer_flags, size_bytes, backing[i], &err);
         CL_EXIT_ERROR(err,"clCreateBuffer failed for buffer %u", i);
 
         // Make sure buffer is cleaned up appropriately if we encounter an error in the rest of the calls.
@@ -496,7 +513,8 @@ REGISTER_TEST(bufferreadwriterect)
                          doffset[0], doffset[1], doffset[2], sregion[0],
                          sregion[1], sregion[2],
                          sregion[0] * sregion[1] * sregion[2]);
-                if ((err = copy_region(src, soffset, sregion, dst, doffset, dregion)))
+                if ((err = test_functions.copy(src, soffset, sregion, dst,
+                                               doffset, dregion)))
                     return err;
                 break;
             case 1:
@@ -506,7 +524,8 @@ REGISTER_TEST(bufferreadwriterect)
                          doffset[0], doffset[1], doffset[2], sregion[0],
                          sregion[1], sregion[2],
                          sregion[0] * sregion[1] * sregion[2]);
-                if ((err = read_verify_region(src, soffset, sregion, dst, doffset, dregion)))
+                if ((err = test_functions.read(src, soffset, sregion, dst,
+                                               doffset, dregion)))
                     return err;
                 break;
             case 2:
@@ -516,7 +535,8 @@ REGISTER_TEST(bufferreadwriterect)
                          doffset[0], doffset[1], doffset[2], sregion[0],
                          sregion[1], sregion[2],
                          sregion[0] * sregion[1] * sregion[2]);
-                if ((err = write_region(src, soffset, sregion, dst, doffset, dregion)))
+                if ((err = test_functions.write(src, soffset, sregion, dst,
+                                                doffset, dregion)))
                     return err;
                 break;
         }
@@ -559,3 +579,15 @@ REGISTER_TEST(bufferreadwriterect)
 
     return err;
 }
+
+// This is the main test function for the conformance test.
+REGISTER_TEST(bufferreadwriterect)
+{
+    TestFunctions test_functions;
+    test_functions.copy = copy_region;
+    test_functions.read = read_verify_region;
+    test_functions.write = write_region;
+    return test_bufferreadwriterect_impl(
+        device, context, queue, num_elements,
+        CL_MEM_USE_HOST_PTR | CL_MEM_READ_WRITE, test_functions);
+}

test_conformance/basic/test_constant.cpp

@@ -22,6 +22,7 @@
 #include <sys/stat.h>
 
 #include <algorithm>
+#include <cinttypes>
 #include <vector>
 
 #include "testBase.h"
@@ -116,7 +117,8 @@ REGISTER_TEST(constant)
     err = clGetDeviceInfo(device, CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE,
                           sizeof(maxSize), &maxSize, 0);
     test_error(err, "Unable to get max constant buffer size");
-    log_info("Device reports CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE %llu bytes.\n",
+    log_info("Device reports CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE %" PRIu64
+             " bytes.\n",
              maxSize);
 
     // Limit test buffer size to 1/4 of CL_DEVICE_GLOBAL_MEM_SIZE

test_conformance/basic/test_enqueue_map.cpp

@@ -29,6 +29,9 @@ const cl_mem_flags flag_set[] = {
   CL_MEM_ALLOC_HOST_PTR | CL_MEM_COPY_HOST_PTR,
   CL_MEM_USE_HOST_PTR,
   CL_MEM_COPY_HOST_PTR,
+  CL_MEM_USE_HOST_PTR | CL_MEM_IMMUTABLE_EXT,
+  CL_MEM_COPY_HOST_PTR | CL_MEM_IMMUTABLE_EXT,
+  CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR | CL_MEM_IMMUTABLE_EXT,
   0
 };
 
@@ -37,6 +40,9 @@ const char *flag_set_names[] = {
   "CL_MEM_ALLOC_HOST_PTR | CL_MEM_COPY_HOST_PTR",
   "CL_MEM_USE_HOST_PTR",
   "CL_MEM_COPY_HOST_PTR",
+  "CL_MEM_USE_HOST_PTR | CL_MEM_IMMUTABLE_EXT",
+  "CL_MEM_COPY_HOST_PTR | CL_MEM_IMMUTABLE_EXT",
+  "CL_MEM_COPY_HOST_PTR | CL_MEM_ALLOC_HOST_PTR | CL_MEM_IMMUTABLE_EXT",
   "0"
 };
 // clang-format on
@@ -44,7 +50,7 @@ const char *flag_set_names[] = {
 REGISTER_TEST(enqueue_map_buffer)
 {
     int error;
-    const size_t bufferSize = 256 * 256;
+    constexpr size_t bufferSize = 256 * 256;
     MTdataHolder d{ gRandomSeed };
     BufferOwningPtr<cl_char> hostPtrData{ malloc(bufferSize) };
     BufferOwningPtr<cl_char> referenceData{ malloc(bufferSize) };
@@ -57,18 +63,28 @@ REGISTER_TEST(enqueue_map_buffer)
         log_info("Testing with cl_mem_flags src: %s\n",
                  flag_set_names[src_flag_id]);
 
+        if ((flag_set[src_flag_id] & CL_MEM_IMMUTABLE_EXT)
+            && !is_extension_available(device,
+                                       "cl_ext_immutable_memory_objects"))
+        {
+            log_info("Device does not support CL_MEM_IMMUTABLE_EXT. "
+                     "Skipping the memory flag.\n");
+            continue;
+        }
+
         generate_random_data(kChar, (unsigned int)bufferSize, d, hostPtrData);
         memcpy(referenceData, hostPtrData, bufferSize);
 
         void *hostPtr = nullptr;
         cl_mem_flags flags = flag_set[src_flag_id];
+        const bool is_immutable_buffer = flags & CL_MEM_IMMUTABLE_EXT;
         bool hasHostPtr =
             (flags & CL_MEM_USE_HOST_PTR) || (flags & CL_MEM_COPY_HOST_PTR);
         if (hasHostPtr) hostPtr = hostPtrData;
         memObject = clCreateBuffer(context, flags, bufferSize, hostPtr, &error);
         test_error(error, "Unable to create testing buffer");
 
-        if (!hasHostPtr)
+        if (!hasHostPtr && !is_immutable_buffer)
         {
             error =
                 clEnqueueWriteBuffer(queue, memObject, CL_TRUE, 0, bufferSize,
@@ -86,7 +102,18 @@ REGISTER_TEST(enqueue_map_buffer)
             cl_char *mappedRegion = (cl_char *)clEnqueueMapBuffer(
                 queue, memObject, CL_TRUE, CL_MAP_READ | CL_MAP_WRITE, offset,
                 length, 0, NULL, NULL, &error);
-            if (error != CL_SUCCESS)
+
+            // Mapping should fail if the buffer is immutable
+            if (is_immutable_buffer)
+            {
+                test_failure_error_ret(
+                    error, CL_INVALID_OPERATION,
+                    "clEnqueueMapBuffer call was expected to fail "
+                    "with CL_INVALID_OPERATION",
+                    TEST_FAIL);
+                continue;
+            }
+            else if (error != CL_SUCCESS)
             {
                 print_error(error, "clEnqueueMapBuffer call failed");
                 log_error("\tOffset: %d  Length: %d\n", (int)offset,
@@ -122,6 +149,11 @@ REGISTER_TEST(enqueue_map_buffer)
                                     finalData, 0, NULL, NULL);
         test_error(error, "Unable to read results");
 
+        if (is_immutable_buffer && !hasHostPtr)
+        {
+            continue;
+        }
+
         for (size_t q = 0; q < bufferSize; q++)
         {
             if (referenceData[q] != finalData[q])
@@ -140,9 +172,10 @@ REGISTER_TEST(enqueue_map_buffer)
 REGISTER_TEST(enqueue_map_image)
 {
     int error;
-    cl_image_format format = { CL_RGBA, CL_UNSIGNED_INT32 };
-    const size_t imageSize = 256;
-    const size_t imageDataSize = imageSize * imageSize * 4 * sizeof(cl_uint);
+    constexpr cl_image_format format = { CL_RGBA, CL_UNSIGNED_INT32 };
+    constexpr size_t imageSize = 256;
+    constexpr size_t imageDataSize =
+        imageSize * imageSize * 4 * sizeof(cl_uint);
 
     PASSIVE_REQUIRE_IMAGE_SUPPORT(device)
 
@@ -158,20 +191,30 @@ REGISTER_TEST(enqueue_map_image)
         log_info("Testing with cl_mem_flags src: %s\n",
                  flag_set_names[src_flag_id]);
 
+        if ((flag_set[src_flag_id] & CL_MEM_IMMUTABLE_EXT)
+            && !is_extension_available(device,
+                                       "cl_ext_immutable_memory_objects"))
+        {
+            log_info("Device does not support CL_MEM_IMMUTABLE_EXT. "
+                     "Skipping the memory flag.\n");
+            continue;
+        }
+
         generate_random_data(kUInt, (unsigned int)(imageSize * imageSize * 4),
                              d, hostPtrData);
         memcpy(referenceData, hostPtrData, imageDataSize);
 
         cl_mem_flags flags = flag_set[src_flag_id];
+        bool is_immutable_image = flags & CL_MEM_IMMUTABLE_EXT;
         bool hasHostPtr =
             (flags & CL_MEM_USE_HOST_PTR) || (flags & CL_MEM_COPY_HOST_PTR);
         void *hostPtr = nullptr;
         if (hasHostPtr) hostPtr = hostPtrData;
-        memObject = create_image_2d(context, CL_MEM_READ_WRITE | flags, &format,
-                                    imageSize, imageSize, 0, hostPtr, &error);
+        memObject = create_image_2d(context, flags, &format, imageSize,
+                                    imageSize, 0, hostPtr, &error);
         test_error(error, "Unable to create testing buffer");
 
-        if (!hasHostPtr)
+        if (!hasHostPtr && !is_immutable_image)
         {
             size_t write_origin[3] = { 0, 0, 0 },
                    write_region[3] = { imageSize, imageSize, 1 };
@@ -198,7 +241,17 @@ REGISTER_TEST(enqueue_map_image)
             cl_uint *mappedRegion = (cl_uint *)clEnqueueMapImage(
                 queue, memObject, CL_TRUE, CL_MAP_READ | CL_MAP_WRITE, offset,
                 region, &rowPitch, NULL, 0, NULL, NULL, &error);
-            if (error != CL_SUCCESS)
+
+            if (is_immutable_image)
+            {
+                test_failure_error_ret(
+                    error, CL_INVALID_OPERATION,
+                    "clEnqueueMapImage call was expected to fail "
+                    "with CL_INVALID_OPERATION",
+                    TEST_FAIL);
+                continue;
+            }
+            else if (error != CL_SUCCESS)
             {
                 print_error(error, "clEnqueueMapImage call failed");
                 log_error("\tOffset: %d,%d  Region: %d,%d\n", (int)offset[0],
@@ -245,6 +298,11 @@ REGISTER_TEST(enqueue_map_image)
                                    finalRegion, 0, 0, finalData, 0, NULL, NULL);
         test_error(error, "Unable to read results");
 
+        if (is_immutable_image && !hasHostPtr)
+        {
+            continue;
+        }
+
         for (size_t q = 0; q < imageSize * imageSize * 4; q++)
         {
             if (referenceData[q] != finalData[q])

test_conformance/basic/test_imagecopy.cpp

@@ -1,6 +1,6 @@
 //
 // Copyright (c) 2017 The Khronos Group Inc.
-// 
+//
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
@@ -20,213 +20,217 @@
 #include <string.h>
 #include <sys/types.h>
 #include <sys/stat.h>
+#include <memory>
 
 #include "testBase.h"
 
-static unsigned char *
-generate_rgba8_image(int w, int h, MTdata d)
+static std::unique_ptr<unsigned char[]> generate_rgba8_image(int w, int h,
+                                                             MTdata d)
 {
-    unsigned char   *ptr = (unsigned char*)malloc(w * h * 4);
-    int             i;
+    std::unique_ptr<unsigned char[]> ptr{ new unsigned char[w * h * 4] };
 
-    for (i=0; i<w*h*4; i++)
+    for (int i = 0; i < w * h * 4; i++)
         ptr[i] = (unsigned char)genrand_int32(d);
 
     return ptr;
 }
 
-static int
-verify_rgba8_image(unsigned char *image, unsigned char *outptr, int w, int h)
+static int verify_rgba8_image(const unsigned char *image,
+                              const unsigned char *outptr, int w, int h)
 {
-    int     i;
+    int i;
 
-    for (i=0; i<w*h*4; i++)
+    for (i = 0; i < w * h * 4; i++)
     {
-        if (outptr[i] != image[i])
-            return -1;
+        if (outptr[i] != image[i]) return -1;
     }
 
     return 0;
 }
 
 
-static unsigned short *
-generate_rgba16_image(int w, int h, MTdata d)
+static std::unique_ptr<unsigned short[]> generate_rgba16_image(int w, int h,
+                                                               MTdata d)
 {
-    unsigned short    *ptr = (unsigned short *)malloc(w * h * 4 * sizeof(unsigned short));
-    int             i;
+    std::unique_ptr<unsigned short[]> ptr{ new unsigned short[w * h * 4] };
 
-    for (i=0; i<w*h*4; i++)
+    for (int i = 0; i < w * h * 4; i++)
         ptr[i] = (unsigned short)genrand_int32(d);
 
     return ptr;
 }
 
-static int
-verify_rgba16_image(unsigned short *image, unsigned short *outptr, int w, int h)
+static int verify_rgba16_image(const unsigned short *image,
+                               const unsigned short *outptr, int w, int h)
 {
-    int     i;
+    int i;
 
-    for (i=0; i<w*h*4; i++)
+    for (i = 0; i < w * h * 4; i++)
     {
-        if (outptr[i] != image[i])
-            return -1;
+        if (outptr[i] != image[i]) return -1;
     }
 
     return 0;
 }
 
 
-static float *
-generate_rgbafp_image(int w, int h, MTdata d)
+static std::unique_ptr<float[]> generate_rgbafp_image(int w, int h, MTdata d)
 {
-    float   *ptr = (float*)malloc(w * h * 4 * sizeof(float));
-    int     i;
+    std::unique_ptr<float[]> ptr{ new float[w * h * 4] };
 
-    for (i=0; i<w*h*4; i++)
+    for (int i = 0; i < w * h * 4; i++)
         ptr[i] = get_random_float(-0x40000000, 0x40000000, d);
 
     return ptr;
 }
 
-static int
-verify_rgbafp_image(float *image, float *outptr, int w, int h)
+static int verify_rgbafp_image(const float *image, const float *outptr, int w,
+                               int h)
 {
-    int     i;
+    int i;
 
-    for (i=0; i<w*h*4; i++)
+    for (i = 0; i < w * h * 4; i++)
     {
-        if (outptr[i] != image[i])
-            return -1;
+        if (outptr[i] != image[i]) return -1;
     }
 
     return 0;
 }
 
+static constexpr cl_image_format image_formats[] = { { CL_RGBA, CL_UNORM_INT8 },
+                                                     { CL_RGBA,
+                                                       CL_UNORM_INT16 },
+                                                     { CL_RGBA, CL_FLOAT } };
 
-REGISTER_TEST(imagecopy)
+static int test_imagecopy_impl(cl_device_id device, cl_context context,
+                               cl_command_queue queue, int num_elements,
+                               cl_mem_flags src_image_flags)
 {
-    cl_image_format    img_format;
-    unsigned char    *rgba8_inptr, *rgba8_outptr;
-    unsigned short    *rgba16_inptr, *rgba16_outptr;
-    float            *rgbafp_inptr, *rgbafp_outptr;
-    clMemWrapper            streams[6];
-    int                img_width = 512;
-    int                img_height = 512;
-    int                i, err;
-    MTdata          d;
+    constexpr size_t image_formats_count = ARRAY_SIZE(image_formats);
+    std::unique_ptr<unsigned char[]> rgba8_inptr, rgba8_outptr;
+    std::unique_ptr<unsigned short[]> rgba16_inptr, rgba16_outptr;
+    std::unique_ptr<float[]> rgbafp_inptr, rgbafp_outptr;
+    clMemWrapper streams[6];
+    int img_width = 512;
+    int img_height = 512;
+    int i, err;
+    MTdataHolder d(gRandomSeed);
 
-    PASSIVE_REQUIRE_IMAGE_SUPPORT( device )
+    rgba8_inptr = generate_rgba8_image(img_width, img_height, d);
+    rgba16_inptr = generate_rgba16_image(img_width, img_height, d);
+    rgbafp_inptr = generate_rgbafp_image(img_width, img_height, d);
 
-    d = init_genrand( gRandomSeed );
-    rgba8_inptr = (unsigned char *)generate_rgba8_image(img_width, img_height, d);
-    rgba16_inptr = (unsigned short *)generate_rgba16_image(img_width, img_height, d);
-    rgbafp_inptr = (float *)generate_rgbafp_image(img_width, img_height, d);
-    free_mtdata(d); d = NULL;
+    rgba8_outptr.reset(new unsigned char[4 * img_width * img_height]);
+    rgba16_outptr.reset(new unsigned short[4 * img_width * img_height]);
+    rgbafp_outptr.reset(new float[4 * img_width * img_height]);
 
-    rgba8_outptr = (unsigned char*)malloc(sizeof(unsigned char) * 4 * img_width * img_height);
-    rgba16_outptr = (unsigned short*)malloc(sizeof(unsigned short) * 4 * img_width * img_height);
-    rgbafp_outptr = (float*)malloc(sizeof(float) * 4 * img_width * img_height);
-
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_UNORM_INT8;
-    streams[0] = create_image_2d(context, CL_MEM_READ_WRITE, &img_format,
-                                 img_width, img_height, 0, NULL, &err);
-    test_error(err, "create_image_2d failed");
-    streams[1] = create_image_2d(context, CL_MEM_READ_WRITE, &img_format,
-                                 img_width, img_height, 0, NULL, &err);
-    test_error(err, "create_image_2d failed");
-
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_UNORM_INT16;
-    streams[2] = create_image_2d(context, CL_MEM_READ_WRITE, &img_format,
-                                 img_width, img_height, 0, NULL, &err);
-    test_error(err, "create_image_2d failed");
-    streams[3] = create_image_2d(context, CL_MEM_READ_WRITE, &img_format,
-                                 img_width, img_height, 0, NULL, &err);
-    test_error(err, "create_image_2d failed");
-
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_FLOAT;
-    streams[4] = create_image_2d(context, CL_MEM_READ_WRITE, &img_format,
-                                 img_width, img_height, 0, NULL, &err);
-    test_error(err, "create_image_2d failed");
-    streams[5] = create_image_2d(context, CL_MEM_READ_WRITE, &img_format,
-                                 img_width, img_height, 0, NULL, &err);
-    test_error(err, "create_image_2d failed");
-
-    for (i=0; i<3; i++)
+    for (size_t index = 0; index < image_formats_count; ++index)
     {
-        void    *p, *outp;
-        int        x, y, delta_w = img_width/8, delta_h = img_height/16;
+        void *ptr = nullptr;
+        if (src_image_flags & CL_MEM_USE_HOST_PTR
+            || src_image_flags & CL_MEM_COPY_HOST_PTR)
+
+        {
+            switch (index)
+            {
+                case 0: ptr = rgba8_inptr.get(); break;
+                case 1: ptr = rgba16_inptr.get(); break;
+                case 2: ptr = rgbafp_inptr.get(); break;
+                default: break;
+            }
+        }
+        streams[index * 2] =
+            create_image_2d(context, src_image_flags, &image_formats[index],
+                            img_width, img_height, 0, ptr, &err);
+        test_error(err, "create_image_2d failed");
+
+        streams[index * 2 + 1] =
+            create_image_2d(context, CL_MEM_READ_WRITE, &image_formats[index],
+                            img_width, img_height, 0, nullptr, &err);
+        test_error(err, "create_image_2d failed");
+    }
+
+    for (i = 0; i < 3; i++)
+    {
+        void *p, *outp;
+        int x, y, delta_w = img_width / 8, delta_h = img_height / 16;
 
         switch (i)
         {
             case 0:
-                p = (void *)rgba8_inptr;
-                outp = (void *)rgba8_outptr;
-        log_info("Testing CL_RGBA CL_UNORM_INT8\n");
+                p = rgba8_inptr.get();
+                outp = rgba8_outptr.get();
+                log_info("Testing CL_RGBA CL_UNORM_INT8\n");
                 break;
             case 1:
-                p = (void *)rgba16_inptr;
-                outp = (void *)rgba16_outptr;
-        log_info("Testing CL_RGBA CL_UNORM_INT16\n");
+                p = rgba16_inptr.get();
+                outp = rgba16_outptr.get();
+                log_info("Testing CL_RGBA CL_UNORM_INT16\n");
                 break;
             case 2:
-                p = (void *)rgbafp_inptr;
-                outp = (void *)rgbafp_outptr;
-        log_info("Testing CL_RGBA CL_FLOAT\n");
+                p = rgbafp_inptr.get();
+                outp = rgbafp_outptr.get();
+                log_info("Testing CL_RGBA CL_FLOAT\n");
                 break;
         }
 
-        size_t origin[3] = {0,0,0}, region[3] = {img_width, img_height, 1};
-        err = clEnqueueWriteImage(queue, streams[i*2], CL_TRUE, origin, region, 0, 0, p, 0, NULL, NULL);
-        test_error(err, "create_image_2d failed");
+        size_t origin[3] = { 0, 0, 0 },
+               region[3] = { img_width, img_height, 1 };
+        if (!(src_image_flags & CL_MEM_USE_HOST_PTR
+              || src_image_flags & CL_MEM_COPY_HOST_PTR))
+        {
+            err = clEnqueueWriteImage(queue, streams[i * 2], CL_TRUE, origin,
+                                      region, 0, 0, p, 0, nullptr, nullptr);
+            test_error(err, "create_image_2d failed");
+        }
 
         int copy_number = 0;
-        for (y=0; y<img_height; y+=delta_h)
+        for (y = 0; y < img_height; y += delta_h)
         {
-            for (x=0; x<img_width; x+=delta_w)
+            for (x = 0; x < img_width; x += delta_w)
             {
-        copy_number++;
-        size_t copy_origin[3] = {x,y,0}, copy_region[3]={delta_w, delta_h, 1};
-        err = clEnqueueCopyImage(queue, streams[i*2], streams[i*2+1],
-                                 copy_origin, copy_origin, copy_region,
-                                 0, NULL, NULL);
-        if (err) {
-          log_error("Copy %d (origin [%d, %d], size [%d, %d], image size [%d x %d]) Failed\n", copy_number, x, y, delta_w, delta_h, img_width, img_height);
-        }
-        test_error(err, "clEnqueueCopyImage failed");
+                copy_number++;
+                size_t copy_origin[3] = { x, y, 0 },
+                       copy_region[3] = { delta_w, delta_h, 1 };
+                err = clEnqueueCopyImage(
+                    queue, streams[i * 2], streams[i * 2 + 1], copy_origin,
+                    copy_origin, copy_region, 0, NULL, NULL);
+                if (err)
+                {
+                    log_error("Copy %d (origin [%d, %d], size [%d, %d], image "
+                              "size [%d x %d]) Failed\n",
+                              copy_number, x, y, delta_w, delta_h, img_width,
+                              img_height);
+                }
+                test_error(err, "clEnqueueCopyImage failed");
             }
         }
 
-        err = clEnqueueReadImage(queue, streams[i*2+1], CL_TRUE, origin, region, 0, 0, outp, 0, NULL, NULL);
+        err = clEnqueueReadImage(queue, streams[i * 2 + 1], CL_TRUE, origin,
+                                 region, 0, 0, outp, 0, NULL, NULL);
         test_error(err, "clEnqueueReadImage failed");
 
         switch (i)
         {
             case 0:
-                err = verify_rgba8_image(rgba8_inptr, rgba8_outptr, img_width, img_height);
+                err = verify_rgba8_image(rgba8_inptr.get(), rgba8_outptr.get(),
+                                         img_width, img_height);
                 break;
             case 1:
-                err = verify_rgba16_image(rgba16_inptr, rgba16_outptr, img_width, img_height);
+                err =
+                    verify_rgba16_image(rgba16_inptr.get(), rgba16_outptr.get(),
+                                        img_width, img_height);
                 break;
             case 2:
-                err = verify_rgbafp_image(rgbafp_inptr, rgbafp_outptr, img_width, img_height);
+                err =
+                    verify_rgbafp_image(rgbafp_inptr.get(), rgbafp_outptr.get(),
+                                        img_width, img_height);
                 break;
         }
 
-        if (err)
-            break;
+        if (err) break;
     }
 
-  free(rgba8_inptr);
-  free(rgba16_inptr);
-  free(rgbafp_inptr);
-  free(rgba8_outptr);
-  free(rgba16_outptr);
-  free(rgbafp_outptr);
-
     if (err)
         log_error("IMAGE copy test failed\n");
     else
@@ -234,3 +238,11 @@ REGISTER_TEST(imagecopy)
 
     return err;
 }
+
+REGISTER_TEST(imagecopy)
+{
+    PASSIVE_REQUIRE_IMAGE_SUPPORT(device);
+
+    return test_imagecopy_impl(device, context, queue, num_elements,
+                               CL_MEM_READ_WRITE);
+}

test_conformance/basic/test_imagecopy3d.cpp

@@ -1,6 +1,6 @@
 //
 // Copyright (c) 2017 The Khronos Group Inc.
-// 
+//
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
@@ -20,213 +20,227 @@
 #include <string.h>
 #include <sys/types.h>
 #include <sys/stat.h>
+#include <memory>
 
 #include "testBase.h"
 
-static unsigned char *
+static std::unique_ptr<unsigned char[]>
 generate_uint8_image(unsigned num_elements, MTdata d)
 {
-    unsigned char *ptr = (unsigned char*)malloc(num_elements);
-    unsigned i;
+    std::unique_ptr<unsigned char[]> ptr{ new unsigned char[num_elements] };
 
-    for (i=0; i<num_elements; i++)
+    for (unsigned i = 0; i < num_elements; i++)
         ptr[i] = (unsigned char)genrand_int32(d);
 
     return ptr;
 }
 
-static int
-verify_uint8_image(unsigned char *image, unsigned char *outptr, unsigned num_elements)
+static int verify_uint8_image(const unsigned char *image,
+                              const unsigned char *outptr,
+                              unsigned num_elements)
 {
     unsigned i;
 
-    for (i=0; i<num_elements; i++)
+    for (i = 0; i < num_elements; i++)
     {
-        if (outptr[i] != image[i])
-            return -1;
+        if (outptr[i] != image[i]) return -1;
     }
 
     return 0;
 }
 
 
-static unsigned short *
+static std::unique_ptr<unsigned short[]>
 generate_uint16_image(unsigned num_elements, MTdata d)
 {
-    unsigned short    *ptr = (unsigned short *)malloc(num_elements * sizeof(unsigned short));
-    unsigned i;
+    std::unique_ptr<unsigned short[]> ptr{ new unsigned short[num_elements] };
 
-    for (i=0; i<num_elements; i++)
+    for (unsigned i = 0; i < num_elements; i++)
         ptr[i] = (unsigned short)genrand_int32(d);
 
     return ptr;
 }
 
-static int
-verify_uint16_image(unsigned short *image, unsigned short *outptr, unsigned num_elements)
+static int verify_uint16_image(const unsigned short *image,
+                               const unsigned short *outptr,
+                               unsigned num_elements)
 {
     unsigned i;
 
-    for (i=0; i<num_elements; i++)
+    for (i = 0; i < num_elements; i++)
     {
-        if (outptr[i] != image[i])
-            return -1;
+        if (outptr[i] != image[i]) return -1;
     }
 
     return 0;
 }
 
 
-static float *
-generate_float_image(unsigned num_elements, MTdata d)
+static std::unique_ptr<float[]> generate_float_image(unsigned num_elements,
+                                                     MTdata d)
 {
-    float   *ptr = (float*)malloc(num_elements * sizeof(float));
-    unsigned i;
+    std::unique_ptr<float[]> ptr{ new float[num_elements] };
 
-    for (i=0; i<num_elements; i++)
+    for (unsigned i = 0; i < num_elements; i++)
         ptr[i] = get_random_float(-0x40000000, 0x40000000, d);
 
     return ptr;
 }
 
-static int
-verify_float_image(float *image, float *outptr, unsigned num_elements)
+static int verify_float_image(const float *image, const float *outptr,
+                              unsigned num_elements)
 {
     unsigned i;
 
-    for (i=0; i<num_elements; i++)
+    for (i = 0; i < num_elements; i++)
     {
-        if (outptr[i] != image[i])
-            return -1;
+        if (outptr[i] != image[i]) return -1;
     }
 
     return 0;
 }
 
+static constexpr cl_image_format image_formats[] = { { CL_RGBA, CL_UNORM_INT8 },
+                                                     { CL_RGBA,
+                                                       CL_UNORM_INT16 },
+                                                     { CL_RGBA, CL_FLOAT } };
 
-REGISTER_TEST(imagecopy3d)
+static int test_imagecopy3d_impl(cl_device_id device, cl_context context,
+                                 cl_command_queue queue,
+                                 int num_elements_ignored,
+                                 cl_mem_flags src_image_flags)
 {
-    cl_image_format    img_format;
-    unsigned char    *rgba8_inptr, *rgba8_outptr;
-    unsigned short *rgba16_inptr, *rgba16_outptr;
-    float *rgbafp_inptr, *rgbafp_outptr;
+    constexpr size_t image_formats_count = ARRAY_SIZE(image_formats);
+    std::unique_ptr<unsigned char[]> rgba8_inptr, rgba8_outptr;
+    std::unique_ptr<unsigned short[]> rgba16_inptr, rgba16_outptr;
+    std::unique_ptr<float[]> rgbafp_inptr, rgbafp_outptr;
     clMemWrapper streams[6];
     int img_width = 128;
     int img_height = 128;
     int img_depth = 64;
     int i;
-    cl_int        err;
-    unsigned num_elems = img_width * img_height * img_depth * 4;
-    MTdata      d;
+    cl_int err;
+    unsigned num_elements = img_width * img_height * img_depth * 4;
+    MTdataHolder d(gRandomSeed);
 
-    PASSIVE_REQUIRE_3D_IMAGE_SUPPORT( device )
+    rgba8_inptr = generate_uint8_image(num_elements, d);
+    rgba16_inptr = generate_uint16_image(num_elements, d);
+    rgbafp_inptr = generate_float_image(num_elements, d);
 
-    d = init_genrand( gRandomSeed );
-    rgba8_inptr = (unsigned char *)generate_uint8_image(num_elems, d);
-    rgba16_inptr = (unsigned short *)generate_uint16_image(num_elems, d);
-    rgbafp_inptr = (float *)generate_float_image(num_elems, d);
-    free_mtdata(d); d = NULL;
+    rgba8_outptr.reset(new unsigned char[num_elements]);
+    rgba16_outptr.reset(new unsigned short[num_elements]);
+    rgbafp_outptr.reset(new float[num_elements]);
 
-    rgba8_outptr = (unsigned char *)malloc(sizeof(unsigned char) * num_elems);
-    rgba16_outptr =
-        (unsigned short *)malloc(sizeof(unsigned short) * num_elems);
-    rgbafp_outptr = (float *)malloc(sizeof(float) * num_elems);
-
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_UNORM_INT8;
-    streams[0] = create_image_3d(context, CL_MEM_READ_ONLY, &img_format, img_width, img_height, img_depth, 0, 0, NULL, &err);
-    test_error(err, "create_image_3d failed");
-    streams[1] = create_image_3d(context, CL_MEM_READ_ONLY, &img_format, img_width, img_height, img_depth, 0, 0, NULL, &err);
-    test_error(err, "create_image_3d failed");
-
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_UNORM_INT16;
-    streams[2] = create_image_3d(context, CL_MEM_READ_ONLY, &img_format, img_width, img_height, img_depth, 0, 0, NULL, &err);
-    test_error(err, "create_image_3d failed");
-    streams[3] = create_image_3d(context, CL_MEM_READ_ONLY, &img_format, img_width, img_height, img_depth, 0, 0, NULL, &err);
-    test_error(err, "create_image_3d failed");
-
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_FLOAT;
-    streams[4] = create_image_3d(context, CL_MEM_READ_ONLY, &img_format, img_width, img_height, img_depth, 0, 0, NULL, &err);
-    test_error(err, "create_image_3d failed");
-    streams[5] = create_image_3d(context, CL_MEM_READ_ONLY, &img_format, img_width, img_height, img_depth, 0, 0, NULL, &err);
-    test_error(err, "create_image_3d failed");
-
-    for (i=0; i<3; i++)
+    for (size_t index = 0; index < image_formats_count; ++index)
     {
-        void    *p, *outp;
-        int        x, y, z, delta_w = img_width/8, delta_h = img_height/16, delta_d = img_depth/4;
+        void *ptr = nullptr;
+        if (src_image_flags & CL_MEM_USE_HOST_PTR
+            || src_image_flags & CL_MEM_COPY_HOST_PTR)
+        {
+            switch (index)
+            {
+                case 0: ptr = rgba8_inptr.get(); break;
+                case 1: ptr = rgba16_inptr.get(); break;
+                case 2: ptr = rgbafp_inptr.get(); break;
+                default: break;
+            }
+        }
+        streams[index * 2] =
+            create_image_3d(context, src_image_flags, &image_formats[index],
+                            img_width, img_height, img_depth, 0, 0, ptr, &err);
+        test_error(err, "create_image_3d failed");
+
+        streams[index * 2 + 1] = create_image_3d(
+            context, CL_MEM_READ_ONLY, &image_formats[index], img_width,
+            img_height, img_depth, 0, 0, nullptr, &err);
+        test_error(err, "create_image_3d failed");
+    }
+
+    for (i = 0; i < image_formats_count; i++)
+    {
+        void *p, *outp;
+        int x, y, z, delta_w = img_width / 8, delta_h = img_height / 16,
+                     delta_d = img_depth / 4;
 
         switch (i)
         {
             case 0:
-                p = (void *)rgba8_inptr;
-                outp = (void *)rgba8_outptr;
+                p = rgba8_inptr.get();
+                outp = rgba8_outptr.get();
                 break;
             case 1:
-                p = (void *)rgba16_inptr;
-                outp = (void *)rgba16_outptr;
+                p = rgba16_inptr.get();
+                outp = rgba16_outptr.get();
                 break;
             case 2:
-                p = (void *)rgbafp_inptr;
-                outp = (void *)rgbafp_outptr;
+                p = rgbafp_inptr.get();
+                outp = rgbafp_outptr.get();
                 break;
         }
 
-        size_t origin[3]={0,0,0}, region[3]={img_width, img_height, img_depth};
-        err = clEnqueueWriteImage(queue, streams[i*2], CL_TRUE, origin, region, 0, 0, p, 0, NULL, NULL);
-        test_error(err, "clEnqueueWriteImage failed");
-
-        for (z=0; z<img_depth; z+=delta_d)
+        size_t origin[3] = { 0, 0, 0 },
+               region[3] = { img_width, img_height, img_depth };
+        if (!(src_image_flags & CL_MEM_USE_HOST_PTR
+              || src_image_flags & CL_MEM_COPY_HOST_PTR))
         {
-            for (y=0; y<img_height; y+=delta_h)
-            {
-                for (x=0; x<img_width; x+=delta_w)
-                {
-                  origin[0] = x; origin[1] = y; origin[2] = z;
-                  region[0] = delta_w; region[1] = delta_h; region[2] = delta_d;
+            err = clEnqueueWriteImage(queue, streams[i * 2], CL_TRUE, origin,
+                                      region, 0, 0, p, 0, nullptr, nullptr);
+            test_error(err, "clEnqueueWriteImage failed");
+        }
 
-                  err = clEnqueueCopyImage(queue, streams[i*2], streams[i*2+1], origin, origin, region, 0, NULL, NULL);
-                  test_error(err, "clEnqueueCopyImage failed");
+        for (z = 0; z < img_depth; z += delta_d)
+        {
+            for (y = 0; y < img_height; y += delta_h)
+            {
+                for (x = 0; x < img_width; x += delta_w)
+                {
+                    origin[0] = x;
+                    origin[1] = y;
+                    origin[2] = z;
+                    region[0] = delta_w;
+                    region[1] = delta_h;
+                    region[2] = delta_d;
+
+                    err = clEnqueueCopyImage(queue, streams[i * 2],
+                                             streams[i * 2 + 1], origin, origin,
+                                             region, 0, NULL, NULL);
+                    test_error(err, "clEnqueueCopyImage failed");
                 }
             }
         }
 
-        origin[0] = 0; origin[1] = 0; origin[2] = 0;
-        region[0] = img_width; region[1] = img_height; region[2] = img_depth;
-        err = clEnqueueReadImage(queue, streams[i*2+1], CL_TRUE, origin, region, 0, 0, outp, 0, NULL, NULL);
+        origin[0] = 0;
+        origin[1] = 0;
+        origin[2] = 0;
+        region[0] = img_width;
+        region[1] = img_height;
+        region[2] = img_depth;
+        err = clEnqueueReadImage(queue, streams[i * 2 + 1], CL_TRUE, origin,
+                                 region, 0, 0, outp, 0, NULL, NULL);
         test_error(err, "clEnqueueReadImage failed");
 
         switch (i)
         {
             case 0:
-                err = verify_uint8_image(rgba8_inptr, rgba8_outptr, num_elems);
+                err = verify_uint8_image(rgba8_inptr.get(), rgba8_outptr.get(),
+                                         num_elements);
                 if (err) log_error("Failed uint8\n");
                 break;
             case 1:
-                err =
-                    verify_uint16_image(rgba16_inptr, rgba16_outptr, num_elems);
+                err = verify_uint16_image(rgba16_inptr.get(),
+                                          rgba16_outptr.get(), num_elements);
                 if (err) log_error("Failed uint16\n");
                 break;
             case 2:
-                err =
-                    verify_float_image(rgbafp_inptr, rgbafp_outptr, num_elems);
+                err = verify_float_image(rgbafp_inptr.get(),
+                                         rgbafp_outptr.get(), num_elements);
                 if (err) log_error("Failed float\n");
                 break;
         }
 
-        if (err)
-            break;
+        if (err) break;
     }
 
-  free(rgba8_inptr);
-  free(rgba16_inptr);
-  free(rgbafp_inptr);
-  free(rgba8_outptr);
-  free(rgba16_outptr);
-  free(rgbafp_outptr);
-
     if (err)
         log_error("IMAGE3D copy test failed\n");
     else
@@ -234,3 +248,11 @@ REGISTER_TEST(imagecopy3d)
 
     return err;
 }
+
+REGISTER_TEST(imagecopy3d)
+{
+    PASSIVE_REQUIRE_3D_IMAGE_SUPPORT(device);
+
+    return test_imagecopy3d_impl(device, context, queue, num_elements,
+                                 CL_MEM_READ_WRITE);
+}

test_conformance/basic/test_imagereadwrite.cpp

@@ -15,6 +15,7 @@
 //
 #include "harness/compat.h"
 
+#include <memory>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -23,16 +24,15 @@
 
 #include "testBase.h"
 
-static unsigned char *
-generate_rgba8_image(int w, int h, MTdata d)
+static std::unique_ptr<unsigned char[]> generate_rgba8_image(int w, int h,
+                                                             MTdata d)
 {
-  unsigned char   *ptr = (unsigned char*)malloc(w * h * 4);
-  int             i;
+    std::unique_ptr<unsigned char[]> ptr{ new unsigned char[w * h * 4] };
 
-  for (i=0; i<w*h*4; i++)
-    ptr[i] = (unsigned char)genrand_int32(d);
+    for (int i = 0; i < w * h * 4; i++)
+        ptr[i] = (unsigned char)genrand_int32(d);
 
-  return ptr;
+    return ptr;
 }
 
 static void
@@ -73,8 +73,8 @@ update_image_from_image(void *out, void *in, int x, int y, int w, int h, int img
     }
 }
 
-static int
-verify_rgba8_image(unsigned char *image, unsigned char *outptr, int w, int h)
+static int verify_rgba8_image(const unsigned char *image,
+                              const unsigned char *outptr, int w, int h)
 {
   int     i;
 
@@ -91,16 +91,15 @@ verify_rgba8_image(unsigned char *image, unsigned char *outptr, int w, int h)
 }
 
 
-static unsigned short *
-generate_rgba16_image(int w, int h, MTdata d)
+static std::unique_ptr<unsigned short[]> generate_rgba16_image(int w, int h,
+                                                               MTdata d)
 {
-  unsigned short    *ptr = (unsigned short*)malloc(w * h * 4 * sizeof(unsigned short));
-  int             i;
+    std::unique_ptr<unsigned short[]> ptr{ new unsigned short[w * h * 4] };
 
-  for (i=0; i<w*h*4; i++)
-    ptr[i] = (unsigned short)genrand_int32(d);
+    for (int i = 0; i < w * h * 4; i++)
+        ptr[i] = (unsigned short)genrand_int32(d);
 
-  return ptr;
+    return ptr;
 }
 
 static void
@@ -121,8 +120,8 @@ update_rgba16_image(unsigned short *p, int x, int y, int w, int h, int img_width
     }
 }
 
-static int
-verify_rgba16_image(unsigned short *image, unsigned short *outptr, int w, int h)
+static int verify_rgba16_image(const unsigned short *image,
+                               const unsigned short *outptr, int w, int h)
 {
   int     i;
 
@@ -139,16 +138,14 @@ verify_rgba16_image(unsigned short *image, unsigned short *outptr, int w, int h)
 }
 
 
-static float *
-generate_rgbafp_image(int w, int h, MTdata d)
+static std::unique_ptr<float[]> generate_rgbafp_image(int w, int h, MTdata d)
 {
-  float   *ptr = (float*)malloc(w * h * 4 * sizeof(float));
-  int     i;
+    std::unique_ptr<float[]> ptr{ new float[w * h * 4] };
 
-  for (i=0; i<w*h*4; i++)
-    ptr[i] = get_random_float(-0x40000000, 0x40000000, d);
+    for (int i = 0; i < w * h * 4; i++)
+        ptr[i] = get_random_float(-0x40000000, 0x40000000, d);
 
-  return ptr;
+    return ptr;
 }
 
 static void
@@ -169,8 +166,8 @@ update_rgbafp_image(float *p, int x, int y, int w, int h, int img_width, MTdata
     }
 }
 
-static int
-verify_rgbafp_image(float *image, float *outptr, int w, int h)
+static int verify_rgbafp_image(const float *image, const float *outptr, int w,
+                               int h)
 {
   int     i;
 
@@ -186,59 +183,52 @@ verify_rgbafp_image(float *image, float *outptr, int w, int h)
   return 0;
 }
 
+static constexpr cl_image_format image_formats[] = { { CL_RGBA, CL_UNORM_INT8 },
+                                                     { CL_RGBA,
+                                                       CL_UNORM_INT16 },
+                                                     { CL_RGBA, CL_FLOAT } };
 
 REGISTER_TEST(imagereadwrite)
 {
-    cl_image_format    img_format;
-    unsigned char    *rgba8_inptr, *rgba8_outptr;
-    unsigned short    *rgba16_inptr, *rgba16_outptr;
-    float            *rgbafp_inptr, *rgbafp_outptr;
+    constexpr size_t image_formats_count = ARRAY_SIZE(image_formats);
+    std::unique_ptr<unsigned char[]> rgba8_inptr, rgba8_outptr;
+    std::unique_ptr<unsigned short[]> rgba16_inptr, rgba16_outptr;
+    std::unique_ptr<float[]> rgbafp_inptr, rgbafp_outptr;
     clMemWrapper            streams[3];
     int                img_width = 512;
     int                img_height = 512;
     int                num_tries = 200;
     int                i, j, err;
-    MTdata          d;
+    MTdataHolder d(gRandomSeed);
 
     PASSIVE_REQUIRE_IMAGE_SUPPORT( device )
 
-    d = init_genrand( gRandomSeed );
-    rgba8_inptr = (unsigned char *)generate_rgba8_image(img_width, img_height, d);
-    rgba16_inptr = (unsigned short *)generate_rgba16_image(img_width, img_height, d);
-    rgbafp_inptr = (float *)generate_rgbafp_image(img_width, img_height, d);
+    rgba8_inptr = generate_rgba8_image(img_width, img_height, d);
+    rgba16_inptr = generate_rgba16_image(img_width, img_height, d);
+    rgbafp_inptr = generate_rgbafp_image(img_width, img_height, d);
 
-    rgba8_outptr = (unsigned char*)malloc(sizeof(unsigned char) * 4 * img_width * img_height);
-    rgba16_outptr = (unsigned short*)malloc(sizeof(unsigned short) * 4 * img_width * img_height);
-    rgbafp_outptr = (float*)malloc(sizeof(float) * 4 * img_width * img_height);
+    rgba8_outptr.reset(new unsigned char[4 * img_width * img_height]);
+    rgba16_outptr.reset(new unsigned short[4 * img_width * img_height]);
+    rgbafp_outptr.reset(new float[4 * img_width * img_height]);
 
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_UNORM_INT8;
-    streams[0] = create_image_2d(context, CL_MEM_READ_WRITE, &img_format,
-                                 img_width, img_height, 0, NULL, &err);
-    test_error(err, "create_image_2d failed");
-
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_UNORM_INT16;
-    streams[1] = create_image_2d(context, CL_MEM_READ_WRITE, &img_format,
-                                 img_width, img_height, 0, NULL, &err);
-    test_error(err, "create_image_2d failed");
-
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_FLOAT;
-    streams[2] = create_image_2d(context, CL_MEM_READ_WRITE, &img_format,
-                                 img_width, img_height, 0, NULL, &err);
-    test_error(err, "create_image_2d failed");
+    for (size_t index = 0; index < image_formats_count; ++index)
+    {
+        streams[index] =
+            create_image_2d(context, CL_MEM_READ_WRITE, &image_formats[index],
+                            img_width, img_height, 0, NULL, &err);
+        test_error(err, "create_image_2d failed");
+    }
 
     for (i=0; i<3; i++)
     {
         void    *p;
 
         if (i == 0)
-            p = (void *)rgba8_inptr;
+            p = rgba8_inptr.get();
         else if (i == 1)
-            p = (void *)rgba16_inptr;
+            p = rgba16_inptr.get();
         else
-            p = (void *)rgbafp_inptr;
+            p = rgbafp_inptr.get();
         size_t origin[3] = {0,0,0}, region[3] = {img_width, img_height, 1};
         err = clEnqueueWriteImage(queue, streams[i], CL_TRUE,
                               origin, region, 0, 0,
@@ -250,7 +240,7 @@ REGISTER_TEST(imagereadwrite)
         }
     }
 
-    for (i=0,j=0; i<num_tries*3; i++,j++)
+    for (i = 0, j = 0; i < num_tries * image_formats_count; i++, j++)
     {
         int        x = (int)get_random_float(0, img_width, d);
         int        y = (int)get_random_float(0, img_height, d);
@@ -260,10 +250,12 @@ REGISTER_TEST(imagereadwrite)
         int     set_input_pitch = (int)(genrand_int32(d) & 0x01);
         int     packed_update = (int)(genrand_int32(d) & 0x01);
         void    *p, *outp;
+        std::unique_ptr<unsigned char[]> p_rgba8;
+        std::unique_ptr<unsigned short[]> p_rgba16;
+        std::unique_ptr<float[]> p_rgbaf;
         int        elem_size;
 
-        if (j == 3)
-            j = 0;
+        if (j == image_formats_count) j = 0;
 
         switch (j)
         {
@@ -272,45 +264,57 @@ REGISTER_TEST(imagereadwrite)
                 elem_size = 4;
                 if(packed_update)
                 {
-                    p = generate_rgba8_image(w, h, d);
-                    update_image_from_image(rgba8_inptr, p, x, y, w, h, img_width, elem_size);
+                    p_rgba8 = generate_rgba8_image(w, h, d);
+                    p = p_rgba8.get();
+                    update_image_from_image(rgba8_inptr.get(), p, x, y, w, h,
+                                            img_width, elem_size);
                 }
                 else
                 {
-                    update_rgba8_image(rgba8_inptr, x, y, w, h, img_width, d);
-                    p = (void *)(rgba8_inptr + ((y * img_width + x) * 4));
+                    update_rgba8_image(rgba8_inptr.get(), x, y, w, h, img_width,
+                                       d);
+                    p = static_cast<void *>(rgba8_inptr.get()
+                                            + ((y * img_width + x) * 4));
                 }
-                outp = (void *)rgba8_outptr;
+                outp = static_cast<void *>(rgba8_outptr.get());
                 break;
             case 1:
                 //if ((w<=8) || (h<=8)) continue;
                 elem_size = 2*4;
                 if(packed_update)
                 {
-                    p = generate_rgba16_image(w, h, d);
-                    update_image_from_image(rgba16_inptr, p, x, y, w, h, img_width, elem_size);
+                    p_rgba16 = generate_rgba16_image(w, h, d);
+                    p = p_rgba16.get();
+                    update_image_from_image(rgba16_inptr.get(), p, x, y, w, h,
+                                            img_width, elem_size);
                 }
                 else
                 {
-                    update_rgba16_image(rgba16_inptr, x, y, w, h, img_width, d);
-                    p = (void *)(rgba16_inptr + ((y * img_width + x) * 4));
+                    update_rgba16_image(rgba16_inptr.get(), x, y, w, h,
+                                        img_width, d);
+                    p = static_cast<void *>(rgba16_inptr.get()
+                                            + ((y * img_width + x) * 4));
                 }
-                outp = (void *)rgba16_outptr;
+                outp = static_cast<void *>(rgba16_outptr.get());
                 break;
             case 2:
                 //if ((w<=8) || (h<=8)) continue;
                 elem_size = 4*4;
                 if(packed_update)
                 {
-                    p = generate_rgbafp_image(w, h, d);
-                    update_image_from_image(rgbafp_inptr, p, x, y, w, h, img_width, elem_size);
+                    p_rgbaf = generate_rgbafp_image(w, h, d);
+                    p = p_rgbaf.get();
+                    update_image_from_image(rgbafp_inptr.get(), p, x, y, w, h,
+                                            img_width, elem_size);
                 }
                 else
                 {
-                    update_rgbafp_image(rgbafp_inptr, x, y, w, h, img_width, d);
-                    p = (void *)(rgbafp_inptr + ((y * img_width + x) * 4));
+                    update_rgbafp_image(rgbafp_inptr.get(), x, y, w, h,
+                                        img_width, d);
+                    p = static_cast<void *>(rgbafp_inptr.get()
+                                            + ((y * img_width + x) * 4));
                 }
-                outp = (void *)rgbafp_outptr;
+                outp = static_cast<void *>(rgbafp_outptr.get());
                 break;
             default:
                 log_error("ERROR Invalid j = %d\n", j);
@@ -358,8 +362,7 @@ REGISTER_TEST(imagereadwrite)
 
         if(packed_update)
         {
-            free(p);
-            p = NULL;
+            p = nullptr;
         }
 
         memset(outp, 0x7, img_width*img_height*elem_size);
@@ -379,7 +382,8 @@ REGISTER_TEST(imagereadwrite)
         switch (j)
         {
             case 0:
-                err = verify_rgba8_image(rgba8_inptr, rgba8_outptr, img_width, img_height);
+                err = verify_rgba8_image(rgba8_inptr.get(), rgba8_outptr.get(),
+                                         img_width, img_height);
                 if (err)
                 {
                     log_error("x=%d y=%d w=%d h=%d, pitch=%d, try=%d\n", x, y, w, h, (int)input_pitch, (int)i);
@@ -387,7 +391,9 @@ REGISTER_TEST(imagereadwrite)
                 }
                 break;
             case 1:
-                err = verify_rgba16_image(rgba16_inptr, rgba16_outptr, img_width, img_height);
+                err =
+                    verify_rgba16_image(rgba16_inptr.get(), rgba16_outptr.get(),
+                                        img_width, img_height);
                 if (err)
                 {
                     log_error("x=%d y=%d w=%d h=%d, pitch=%d, try=%d\n", x, y, w, h, (int)input_pitch, (int)i);
@@ -395,7 +401,9 @@ REGISTER_TEST(imagereadwrite)
                 }
                 break;
             case 2:
-                err = verify_rgbafp_image(rgbafp_inptr, rgbafp_outptr, img_width, img_height);
+                err =
+                    verify_rgbafp_image(rgbafp_inptr.get(), rgbafp_outptr.get(),
+                                        img_width, img_height);
                 if (err)
                 {
                     log_error("x=%d y=%d w=%d h=%d, pitch=%d, try=%d\n", x, y, w, h, (int)input_pitch, (int)i);
@@ -407,14 +415,6 @@ REGISTER_TEST(imagereadwrite)
         if (err) break;
     }
 
-    free_mtdata(d);
-    free(rgba8_inptr);
-    free(rgba16_inptr);
-    free(rgbafp_inptr);
-    free(rgba8_outptr);
-    free(rgba16_outptr);
-    free(rgbafp_outptr);
-
     if (!err)
         log_info("IMAGE read, write test passed\n");
 

test_conformance/basic/test_imagereadwrite3d.cpp

@@ -15,6 +15,7 @@
 //
 #include "harness/compat.h"
 
+#include <memory>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -23,14 +24,13 @@
 
 #include "testBase.h"
 
-static unsigned char *
+static std::unique_ptr<unsigned char[]>
 generate_rgba8_image(int w, int h, int d, MTdata mtData)
 {
-    unsigned char   *ptr = (unsigned char*)malloc(w * h * d *4);
-    int             i;
+    std::unique_ptr<unsigned char[]> ptr{ new unsigned char[w * h * d * 4] };
 
-    for (i=0; i<w*h*d*4; i++)
-            ptr[i] = (unsigned char)genrand_int32(mtData);
+    for (int i = 0; i < w * h * d * 4; i++)
+        ptr[i] = (unsigned char)genrand_int32(mtData);
 
     return ptr;
 }
@@ -77,8 +77,8 @@ update_image_from_image(void *out, void *in, int x, int y, int z, int w, int h,
         }
 }
 
-static int
-verify_rgba8_image(unsigned char *image, unsigned char *outptr, int w, int h, int d)
+static int verify_rgba8_image(const unsigned char *image,
+                              const unsigned char *outptr, int w, int h, int d)
 {
     int     i;
 
@@ -95,14 +95,13 @@ verify_rgba8_image(unsigned char *image, unsigned char *outptr, int w, int h, in
 }
 
 
-static unsigned short *
+static std::unique_ptr<unsigned short[]>
 generate_rgba16_image(int w, int h, int d, MTdata mtData)
 {
-    unsigned short    *ptr = (unsigned short*)malloc(w * h * d * 4 * sizeof(unsigned short));
-    int             i;
+    std::unique_ptr<unsigned short[]> ptr{ new unsigned short[w * h * d * 4] };
 
-    for (i=0; i<w*h*d*4; i++)
-            ptr[i] = (unsigned short)genrand_int32(mtData);
+    for (int i = 0; i < w * h * d * 4; i++)
+        ptr[i] = (unsigned short)genrand_int32(mtData);
 
     return ptr;
 }
@@ -127,8 +126,9 @@ update_rgba16_image(unsigned short *p, int x, int y, int z, int w, int h, int d,
     }
 }
 
-static int
-verify_rgba16_image(unsigned short *image, unsigned short *outptr, int w, int h, int d)
+static int verify_rgba16_image(const unsigned short *image,
+                               const unsigned short *outptr, int w, int h,
+                               int d)
 {
     int     i;
 
@@ -145,14 +145,13 @@ verify_rgba16_image(unsigned short *image, unsigned short *outptr, int w, int h,
 }
 
 
-static float *
-generate_rgbafp_image(int w, int h, int d, MTdata mtData)
+static std::unique_ptr<float[]> generate_rgbafp_image(int w, int h, int d,
+                                                      MTdata mtData)
 {
-    float   *ptr = (float*)malloc(w * h * d *4 * sizeof(float));
-    int     i;
+    std::unique_ptr<float[]> ptr{ new float[w * h * d * 4] };
 
-    for (i=0; i<w*h*d*4; i++)
-            ptr[i] = get_random_float(-0x40000000, 0x40000000, mtData);
+    for (int i = 0; i < w * h * d * 4; i++)
+        ptr[i] = get_random_float(-0x40000000, 0x40000000, mtData);
 
     return ptr;
 }
@@ -177,8 +176,8 @@ update_rgbafp_image(float *p, int x, int y, int z, int w, int h, int d, int img_
         }
 }
 
-static int
-verify_rgbafp_image(float *image, float *outptr, int w, int h, int d)
+static int verify_rgbafp_image(const float *image, const float *outptr, int w,
+                               int h, int d)
 {
     int     i;
 
@@ -194,13 +193,17 @@ verify_rgbafp_image(float *image, float *outptr, int w, int h, int d)
     return 0;
 }
 
+static constexpr cl_image_format image_formats[] = { { CL_RGBA, CL_UNORM_INT8 },
+                                                     { CL_RGBA,
+                                                       CL_UNORM_INT16 },
+                                                     { CL_RGBA, CL_FLOAT } };
 
 REGISTER_TEST(imagereadwrite3d)
 {
-    cl_image_format    img_format;
-    unsigned char    *rgba8_inptr, *rgba8_outptr;
-    unsigned short *rgba16_inptr, *rgba16_outptr;
-    float            *rgbafp_inptr, *rgbafp_outptr;
+    constexpr size_t image_formats_count = ARRAY_SIZE(image_formats);
+    std::unique_ptr<unsigned char[]> rgba8_inptr, rgba8_outptr;
+    std::unique_ptr<unsigned short[]> rgba16_inptr, rgba16_outptr;
+    std::unique_ptr<float[]> rgbafp_inptr, rgbafp_outptr;
     clMemWrapper    streams[3];
     int       img_width = 64;
     int       img_height = 64;
@@ -208,44 +211,41 @@ REGISTER_TEST(imagereadwrite3d)
     int       img_slice = img_width * img_height;
     int       num_tries = 30;
     int       i, j, err;
-    MTdata      mtData;
+    MTdataHolder mtData(gRandomSeed);
 
     PASSIVE_REQUIRE_3D_IMAGE_SUPPORT( device )
 
-    mtData = init_genrand( gRandomSeed );
-    rgba8_inptr = (unsigned char *)generate_rgba8_image(img_width, img_height, img_depth, mtData);
-    rgba16_inptr = (unsigned short *)generate_rgba16_image(img_width, img_height, img_depth, mtData);
-    rgbafp_inptr = (float *)generate_rgbafp_image(img_width, img_height, img_depth, mtData);
+    rgba8_inptr =
+        generate_rgba8_image(img_width, img_height, img_depth, mtData);
+    rgba16_inptr =
+        generate_rgba16_image(img_width, img_height, img_depth, mtData);
+    rgbafp_inptr =
+        generate_rgbafp_image(img_width, img_height, img_depth, mtData);
 
-    rgba8_outptr = (unsigned char*)malloc(sizeof(unsigned char) * 4 * img_width * img_height * img_depth);
-    rgba16_outptr = (unsigned short*)malloc(sizeof(unsigned short) * 4 * img_width * img_height * img_depth);
-    rgbafp_outptr = (float*)malloc(sizeof(float) * 4 * img_width * img_height * img_depth);
+    rgba8_outptr.reset(
+        new unsigned char[4 * img_width * img_height * img_depth]);
+    rgba16_outptr.reset(
+        new unsigned short[4 * img_width * img_height * img_depth]);
+    rgbafp_outptr.reset(new float[4 * img_width * img_height * img_depth]);
 
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_UNORM_INT8;
-    streams[0] = create_image_3d(context, CL_MEM_READ_ONLY, &img_format, img_width, img_height, img_depth, 0, 0, NULL, &err);
-    test_error(err, "create_image_3d failed");
+    for (size_t index = 0; index < image_formats_count; ++index)
+    {
+        streams[index] = create_image_3d(
+            context, CL_MEM_READ_ONLY, &image_formats[index], img_width,
+            img_height, img_depth, 0, 0, nullptr, &err);
+        test_error(err, "create_image_3d failed");
+    }
 
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_UNORM_INT16;
-    streams[1] = create_image_3d(context, CL_MEM_READ_ONLY, &img_format, img_width, img_height, img_depth, 0, 0, NULL, &err);
-    test_error(err, "create_image_3d failed");
-
-    img_format.image_channel_order = CL_RGBA;
-    img_format.image_channel_data_type = CL_FLOAT;
-    streams[2] = create_image_3d(context, CL_MEM_READ_ONLY, &img_format, img_width, img_height, img_depth, 0, 0, NULL, &err);
-    test_error(err, "create_image_3d failed");
-
-    for (i=0; i<3; i++)
+    for (i = 0; i < image_formats_count; i++)
     {
         void    *p;
 
         if (i == 0)
-            p = (void *)rgba8_inptr;
+            p = rgba8_inptr.get();
         else if (i == 1)
-            p = (void *)rgba16_inptr;
+            p = rgba16_inptr.get();
         else
-            p = (void *)rgbafp_inptr;
+            p = rgbafp_inptr.get();
 
         size_t origin[3] = {0,0,0}, region[3] = {img_width, img_height, img_depth};
         err = clEnqueueWriteImage(queue, streams[i], CL_TRUE,
@@ -255,7 +255,7 @@ REGISTER_TEST(imagereadwrite3d)
         test_error(err, "clEnqueueWriteImage failed");
     }
 
-    for (i=0,j=0; i<num_tries*3; i++,j++)
+    for (i = 0, j = 0; i < num_tries * image_formats_count; i++, j++)
     {
         int        x = (int)get_random_float(0, (float)img_width - 1, mtData);
         int        y = (int)get_random_float(0, (float)img_height - 1, mtData);
@@ -267,10 +267,12 @@ REGISTER_TEST(imagereadwrite3d)
         int     set_input_pitch = (int)(genrand_int32(mtData) & 0x01);
         int     packed_update = (int)(genrand_int32(mtData) & 0x01);
         void    *p, *outp;
+        std::unique_ptr<unsigned char[]> p_rgba8;
+        std::unique_ptr<unsigned short[]> p_rgba16;
+        std::unique_ptr<float[]> p_rgbaf;
         int        elem_size;
 
-        if (j == 3)
-            j = 0;
+        if (j == image_formats_count) j = 0;
 
         // packed: the source image for the write is a whole image                                                                                                                                                                                                                                                      .
         // unpacked: the source image for the write is a subset within a larger image
@@ -280,43 +282,64 @@ REGISTER_TEST(imagereadwrite3d)
                 elem_size = 4;
                 if(packed_update)
                 {
-                    p = generate_rgba8_image(w, h, d, mtData);
-                    update_image_from_image(rgba8_inptr, p, x, y, z, w, h, d, img_width, img_height, img_depth, elem_size);
+                    p_rgba8 = generate_rgba8_image(w, h, d, mtData);
+                    p = p_rgba8.get();
+                    update_image_from_image(rgba8_inptr.get(), p, x, y, z, w, h,
+                                            d, img_width, img_height, img_depth,
+                                            elem_size);
                 }
                 else
                 {
-                    update_rgba8_image(rgba8_inptr, x, y, z, w, h, d, img_width, img_height, img_depth, mtData);
-                    p = (void *)(rgba8_inptr + ((z * img_slice + y * img_width + x) * 4));
+                    update_rgba8_image(rgba8_inptr.get(), x, y, z, w, h, d,
+                                       img_width, img_height, img_depth,
+                                       mtData);
+                    p = static_cast<void *>(
+                        rgba8_inptr.get()
+                        + ((z * img_slice + y * img_width + x) * 4));
                 }
-                outp = (void *)rgba8_outptr;
+                outp = static_cast<void *>(rgba8_outptr.get());
                 break;
             case 1:
                 elem_size = 2*4;
                 if(packed_update)
                 {
-                    p = generate_rgba16_image(w, h, d, mtData);
-                    update_image_from_image(rgba16_inptr, p, x, y, z, w, h, d, img_width, img_height, img_depth, elem_size);
+                    p_rgba16 = generate_rgba16_image(w, h, d, mtData);
+                    p = p_rgba16.get();
+                    update_image_from_image(rgba16_inptr.get(), p, x, y, z, w,
+                                            h, d, img_width, img_height,
+                                            img_depth, elem_size);
                 }
                 else
                 {
-                    update_rgba16_image(rgba16_inptr, x, y, z, w, h, d, img_width, img_height, img_depth, mtData);
-                    p = (void *)(rgba16_inptr + ((z * img_slice + y * img_width + x) * 4));
+                    update_rgba16_image(rgba16_inptr.get(), x, y, z, w, h, d,
+                                        img_width, img_height, img_depth,
+                                        mtData);
+                    p = static_cast<void *>(
+                        rgba16_inptr.get()
+                        + ((z * img_slice + y * img_width + x) * 4));
                 }
-                outp = (void *)rgba16_outptr;
+                outp = static_cast<void *>(rgba16_outptr.get());
                 break;
             case 2:
                 elem_size = 4*4;
                 if(packed_update)
                 {
-                    p = generate_rgbafp_image(w, h, d, mtData);
-                    update_image_from_image(rgbafp_inptr, p, x, y, z, w, h, d, img_width, img_height, img_depth, elem_size);
+                    p_rgbaf = generate_rgbafp_image(w, h, d, mtData);
+                    p = p_rgbaf.get();
+                    update_image_from_image(rgbafp_inptr.get(), p, x, y, z, w,
+                                            h, d, img_width, img_height,
+                                            img_depth, elem_size);
                 }
                 else
                 {
-                    update_rgbafp_image(rgbafp_inptr, x, y, z, w, h, d, img_width, img_height, img_depth, mtData);
-                    p = (void *)(rgbafp_inptr + ((z * img_slice + y * img_width + x) * 4));
+                    update_rgbafp_image(rgbafp_inptr.get(), x, y, z, w, h, d,
+                                        img_width, img_height, img_depth,
+                                        mtData);
+                    p = static_cast<void *>(
+                        rgbafp_inptr.get()
+                        + ((z * img_slice + y * img_width + x) * 4));
                 }
-                outp = (void *)rgbafp_outptr;
+                outp = static_cast<void *>(rgbafp_outptr.get());
                 break;
             default:
                 log_error("ERROR Invalid j = %d\n", j);
@@ -360,8 +383,7 @@ REGISTER_TEST(imagereadwrite3d)
 
         if(packed_update)
         {
-            free(p);
-            p = NULL;
+            p = nullptr;
         }
 
         memset(outp, 0x7, img_width*img_height*img_depth*elem_size);
@@ -375,7 +397,8 @@ REGISTER_TEST(imagereadwrite3d)
         switch (j)
         {
             case 0:
-                err = verify_rgba8_image(rgba8_inptr, rgba8_outptr, img_width, img_height, img_depth);
+                err = verify_rgba8_image(rgba8_inptr.get(), rgba8_outptr.get(),
+                                         img_width, img_height, img_depth);
                 if (err)
                 {
                     log_error("x=%d y=%d z=%d w=%d h=%d d=%d pitch=%d, slice_pitch=%d, try=%d\n", x, y, z, w, h, d, (int)input_pitch, (int)input_slice_pitch, (int)i);
@@ -383,7 +406,9 @@ REGISTER_TEST(imagereadwrite3d)
                 }
                 break;
             case 1:
-                err = verify_rgba16_image(rgba16_inptr, rgba16_outptr, img_width, img_height, img_depth);
+                err =
+                    verify_rgba16_image(rgba16_inptr.get(), rgba16_outptr.get(),
+                                        img_width, img_height, img_depth);
                 if (err)
                 {
                     log_error("x=%d y=%d z=%d w=%d h=%d d=%d pitch=%d, slice_pitch=%d, try=%d\n", x, y, z, w, h, d, (int)input_pitch, (int)input_slice_pitch, (int)i);
@@ -391,7 +416,9 @@ REGISTER_TEST(imagereadwrite3d)
                 }
                 break;
             case 2:
-                err = verify_rgbafp_image(rgbafp_inptr, rgbafp_outptr, img_width, img_height, img_depth);
+                err =
+                    verify_rgbafp_image(rgbafp_inptr.get(), rgbafp_outptr.get(),
+                                        img_width, img_height, img_depth);
                 if (err)
                 {
                     log_error("x=%d y=%d z=%d w=%d h=%d d=%d pitch=%d, slice_pitch=%d, try=%d\n", x, y, z, w, h, d, (int)input_pitch, (int)input_slice_pitch, (int)i);
@@ -404,14 +431,6 @@ REGISTER_TEST(imagereadwrite3d)
             break;
     }
 
-    free_mtdata(mtData);
-    free(rgba8_inptr);
-    free(rgba16_inptr);
-    free(rgbafp_inptr);
-    free(rgba8_outptr);
-    free(rgba16_outptr);
-    free(rgbafp_outptr);
-
     if (!err)
         log_info("IMAGE read, write test passed\n");
 

test_conformance/basic/test_intmath.cpp

@@ -14,6 +14,7 @@
 // limitations under the License.
 //
 
+#include <cinttypes>
 #include <functional>
 #include <string>
 #include <vector>
@@ -182,13 +183,13 @@ static int test_intmath(cl_device_id device, cl_context context,
             if (r != output[i])
             {
                 log_error("\n\nverification failed at index %d\n", i);
-                log_error("-> inputs: %llu, %llu, %llu\n",
-                          static_cast<cl_uint>(inputA[i]),
-                          static_cast<cl_uint>(inputB[i]),
-                          static_cast<cl_uint>(inputC[i]));
-                log_error("-> expected %llu, got %llu\n\n",
-                          static_cast<cl_uint>(r),
-                          static_cast<cl_uint>(output[i]));
+                log_error("-> inputs: %" PRIu64 "%" PRIu64 "%" PRIu64 "\n",
+                          static_cast<cl_ulong>(inputA[i]),
+                          static_cast<cl_ulong>(inputB[i]),
+                          static_cast<cl_ulong>(inputC[i]));
+                log_error("-> expected %" PRIu64 "%" PRIu64 "\n\n",
+                          static_cast<cl_ulong>(r),
+                          static_cast<cl_ulong>(output[i]));
                 return TEST_FAIL;
             }
         }

test_conformance/basic/test_kernel_memory_alignment.cpp

@@ -17,6 +17,8 @@
 #include <unistd.h>
 #endif
 
+#include <cinttypes>
+
 #include "harness/conversions.h"
 #include "harness/typeWrappers.h"
 #include "harness/errorHelpers.h"
@@ -315,9 +317,14 @@ static int test_kernel_memory_alignment(cl_device_id device, cl_context context,
                 for (int i = 0; i < 6; i++) {
                     if ((results_data[i] & alignments[i]) != 0) {
                         total_errors++;
-                        log_error("\tVector size %d failed: 0x%llx is not properly aligned.\n", 1 << i, results_data[i]);
+                        log_error("\tVector size %d failed: 0x%" PRIx64
+                                  " is not properly aligned.\n",
+                                  1 << i, results_data[i]);
                     } else {
-                        if (DEBUG) log_info("\tVector size %d passed: 0x%llx is properly aligned.\n", 1 << i, results_data[i]);
+                        if (DEBUG)
+                            log_info("\tVector size %d passed: 0x%" PRIx64
+                                     " is properly aligned.\n",
+                                     1 << i, results_data[i]);
                     }
                 }
             }
@@ -328,9 +335,14 @@ static int test_kernel_memory_alignment(cl_device_id device, cl_context context,
                 for (int i = 0; i < 6; i++) {
                     if ((results_data_no_long[i] & alignments[i]) != 0) {
                         total_errors++;
-                        log_error("\tVector size %d failed: 0x%llx is not properly aligned.\n", 1 << i, results_data_no_long[i]);
+                        log_error("\tVector size %d failed: 0x%x is not "
+                                  "properly aligned.\n",
+                                  1 << i, results_data_no_long[i]);
                     } else {
-                        if (DEBUG) log_info("\tVector size %d passed: 0x%llx is properly aligned.\n", 1 << i, results_data_no_long[i]);
+                        if (DEBUG)
+                            log_info("\tVector size %d passed: 0x%x is "
+                                     "properly aligned.\n",
+                                     1 << i, results_data_no_long[i]);
                     }
                 }
             }
@@ -383,9 +395,14 @@ static int test_kernel_memory_alignment(cl_device_id device, cl_context context,
                 for (int i = 0; i < 5; i++) {
                     if ((results_data[i] & alignments[i]) != 0) {
                         total_errors++;
-                        log_error("\tVector size %d failed: 0x%llx is not properly aligned.\n", 1 << i, results_data[i]);
+                        log_error("\tVector size %d failed: 0x%" PRIx64
+                                  " is not properly aligned.\n",
+                                  1 << i, results_data[i]);
                     } else {
-                        if (DEBUG) log_info("\tVector size %d passed: 0x%llx is properly aligned.\n", 1 << i, results_data[i]);
+                        if (DEBUG)
+                            log_info("\tVector size %d passed: 0x%" PRIx64
+                                     " is properly aligned.\n",
+                                     1 << i, results_data[i]);
                     }
                 }
             }
@@ -396,9 +413,14 @@ static int test_kernel_memory_alignment(cl_device_id device, cl_context context,
                 for (int i = 0; i < 5; i++) {
                     if ((results_data_no_long[i] & alignments[i]) != 0) {
                         total_errors++;
-                        log_error("\tVector size %d failed: 0x%llx is not properly aligned.\n", 1 << i, results_data_no_long[i]);
+                        log_error("\tVector size %d failed: 0x%x is not "
+                                  "properly aligned.\n",
+                                  1 << i, results_data_no_long[i]);
                     } else {
-                        if (DEBUG) log_info("\tVector size %d passed: 0x%llx is properly aligned.\n", 1 << i, results_data_no_long[i]);
+                        if (DEBUG)
+                            log_info("\tVector size %d passed: 0x%x is "
+                                     "properly aligned.\n",
+                                     1 << i, results_data_no_long[i]);
                     }
                 }
             }
@@ -455,9 +477,14 @@ static int test_kernel_memory_alignment(cl_device_id device, cl_context context,
                 for (int i = 0; i < 5; i++) {
                     if ((results_data[i] & alignments[i]) != 0) {
                         total_errors++;
-                        log_error("\tVector size %d failed: 0x%llx is not properly aligned.\n", 1 << i, results_data[i]);
+                        log_error("\tVector size %d failed: 0x%" PRIx64
+                                  " is not properly aligned.\n",
+                                  1 << i, results_data[i]);
                     } else {
-                        if (DEBUG) log_info("\tVector size %d passed: 0x%llx is properly aligned.\n", 1 << i, results_data[i]);
+                        if (DEBUG)
+                            log_info("\tVector size %d passed: 0x%" PRIx64
+                                     " is properly aligned.\n",
+                                     1 << i, results_data[i]);
                     }
                 }
             }
@@ -468,9 +495,14 @@ static int test_kernel_memory_alignment(cl_device_id device, cl_context context,
                 for (int i = 0; i < 5; i++) {
                     if ((results_data_no_long[i] & alignments[i]) != 0) {
                         total_errors++;
-                        log_error("\tVector size %d failed: 0x%llx is not properly aligned.\n", 1 << i, results_data_no_long[i]);
+                        log_error("\tVector size %d failed: 0x%x is not "
+                                  "properly aligned.\n",
+                                  1 << i, results_data_no_long[i]);
                     } else {
-                        if (DEBUG) log_info("\tVector size %d passed: 0x%llx is properly aligned.\n", 1 << i, results_data_no_long[i]);
+                        if (DEBUG)
+                            log_info("\tVector size %d passed: 0x%x is "
+                                     "properly aligned.\n",
+                                     1 << i, results_data_no_long[i]);
                     }
                 }
             }

test_conformance/basic/test_numeric_constants.cpp

@@ -15,6 +15,8 @@
 //
 #include "testBase.h"
 
+#include <cinttypes>
+
 #define TEST_VALUE_POSITIVE( string_name, name, value ) \
 { \
 if (name < value) { \
@@ -528,7 +530,14 @@ const char *kernel_constant_double_limits[] = {
 };
 
 #define TEST_FLOAT_ASSERTION( a, msg, f ) if( !( a ) ) { log_error( "ERROR: Float constant failed requirement: %s (bitwise value is 0x%8.8x)\n", msg, *( (uint32_t *)&f ) ); return -1; }
-#define TEST_DOUBLE_ASSERTION( a, msg, f ) if( !( a ) ) { log_error( "ERROR: Double constant failed requirement: %s (bitwise value is 0x%16.16llx)\n", msg, *( (uint64_t *)&f ) ); return -1; }
+#define TEST_DOUBLE_ASSERTION(a, msg, f)                                       \
+    if (!(a))                                                                  \
+    {                                                                          \
+        log_error("ERROR: Double constant failed requirement: %s (bitwise "    \
+                  "value is 0x%16.16" PRIx64 ")\n",                            \
+                  msg, *((uint64_t *)&f));                                     \
+        return -1;                                                             \
+    }
 
 REGISTER_TEST(kernel_limit_constants)
 {

test_conformance/basic/test_sizeof.cpp

@@ -21,6 +21,8 @@
 #include <sys/types.h>
 #include <sys/stat.h>
 
+#include <cinttypes>
+
 #include "testBase.h"
 
 cl_int get_type_size( cl_context context, cl_command_queue queue, const char *type, cl_ulong *size, cl_device_id device  )
@@ -161,12 +163,16 @@ REGISTER_TEST(sizeof)
             return err;
         if( test != scalar_table[i].size )
         {
-            log_error( "\nFAILED: Type %s has size %lld, but expected size %lld!\n", scalar_table[i].name, test, scalar_table[i].size );
+            log_error("\nFAILED: Type %s has size %" PRId64
+                      ", but expected size %" PRId64 "!\n",
+                      scalar_table[i].name, test, scalar_table[i].size);
             return -1;
         }
         if( test != scalar_table[i].cl_size )
         {
-            log_error( "\nFAILED: Type %s has size %lld, but cl_ size is %lld!\n", scalar_table[i].name, test, scalar_table[i].cl_size );
+            log_error("\nFAILED: Type %s has size %" PRId64
+                      ", but cl_ size is %" PRId64 "!\n",
+                      scalar_table[i].name, test, scalar_table[i].cl_size);
             return -2;
         }
         log_info( "%16s", scalar_table[i].name );
@@ -196,16 +202,16 @@ REGISTER_TEST(sizeof)
                 return err;
             if( test != j * vector_table[i].size )
             {
-                log_error(
-                    "\nFAILED: Type %s has size %lld, but expected size %zu!\n",
-                    name, test, j * vector_table[i].size);
+                log_error("\nFAILED: Type %s has size %" PRId64
+                          ", but expected size %" PRIu64 "!\n",
+                          name, test, j * vector_table[i].size);
                 return -1;
             }
             if( test != j * vector_table[i].cl_size )
             {
-                log_error(
-                    "\nFAILED: Type %s has size %lld, but cl_ size is %zu!\n",
-                    name, test, j * vector_table[i].cl_size);
+                log_error("\nFAILED: Type %s has size %" PRId64
+                          ", but cl_ size is %" PRIu64 "!\n",
+                          name, test, j * vector_table[i].cl_size);
                 return -2;
             }
             log_info( "%16s", name );
@@ -222,7 +228,9 @@ REGISTER_TEST(sizeof)
             return err;
         if( test != ptr_size )
         {
-            log_error( "\nFAILED: Type %s has size %lld, but expected size %u!\n", ptr_table[i], test, ptr_size );
+            log_error("\nFAILED: Type %s has size %" PRId64
+                      ", but expected size %u!\n",
+                      ptr_table[i], test, ptr_size);
             return -1;
         }
         log_info( "%16s", ptr_table[i] );
@@ -235,12 +243,16 @@ REGISTER_TEST(sizeof)
         return err;
     if( test < ptr_size )
     {
-        log_error( "\nFAILED: intptr_t has size %lld, but must be at least %u!\n", test, ptr_size );
+        log_error("\nFAILED: intptr_t has size %" PRId64
+                  ", but must be at least %u!\n",
+                  test, ptr_size);
         return -1;
     }
     if( ! IsPowerOfTwo( test ) )
     {
-        log_error( "\nFAILED: sizeof(intptr_t) is %lld, but must be a power of two!\n", test );
+        log_error("\nFAILED: sizeof(intptr_t) is %" PRId64
+                  ", but must be a power of two!\n",
+                  test);
         return -2;
     }
     log_info( "%16s", "intptr_t" );
@@ -252,12 +264,16 @@ REGISTER_TEST(sizeof)
         return err;
     if( test < ptr_size )
     {
-        log_error( "\nFAILED: uintptr_t has size %lld, but must be at least %u!\n", test, ptr_size );
+        log_error("\nFAILED: uintptr_t has size %" PRId64
+                  ", but must be at least %u!\n",
+                  test, ptr_size);
         return -1;
     }
     if( ! IsPowerOfTwo( test ) )
     {
-        log_error( "\nFAILED: sizeof(uintptr_t) is %lld, but must be a power of two!\n", test );
+        log_error("\nFAILED: sizeof(uintptr_t) is %" PRId64
+                  ", but must be a power of two!\n",
+                  test);
         return -2;
     }
     log_info( "%16s\n", "uintptr_t" );
@@ -293,7 +309,9 @@ REGISTER_TEST(sizeof)
             return err;
         if( ! IsPowerOfTwo( test ) )
         {
-            log_error( "\nFAILED: Type %s has size %lld, which is not a power of two (section 6.1.5)!\n", other_types[i], test );
+            log_error("\nFAILED: Type %s has size %" PRId64
+                      ", which is not a power of two (section 6.1.5)!\n",
+                      other_types[i], test);
             return -1;
         }
         log_info( "%16s", other_types[i] );
@@ -311,7 +329,8 @@ REGISTER_TEST(sizeof)
             return err;
         if( test != 8 )
         {
-            log_error( "\nFAILED: double has size %lld, but must be 8!\n", test );
+            log_error("\nFAILED: double has size %" PRId64 ", but must be 8!\n",
+                      test);
             return -1;
         }
         log_info( "%16s", "double" );
@@ -328,7 +347,8 @@ REGISTER_TEST(sizeof)
                 return err;
             if( test != 8*j )
             {
-                log_error("\nFAILED: %s has size %lld, but must be %zu!\n",
+                log_error("\nFAILED: %s has size %" PRId64
+                          ", but must be %zu!\n",
                           name, test, 8 * j);
                 return -1;
             }
@@ -347,7 +367,8 @@ REGISTER_TEST(sizeof)
             return err;
         if( test != 2 )
         {
-            log_error( "\nFAILED: half has size %lld, but must be 2!\n", test );
+            log_error("\nFAILED: half has size %" PRId64 ", but must be 2!\n",
+                      test);
             return -1;
         }
         log_info( "%16s", "half" );
@@ -364,7 +385,8 @@ REGISTER_TEST(sizeof)
                 return err;
             if( test != 2*j )
             {
-                log_error("\nFAILED: %s has size %lld, but must be %zu!\n",
+                log_error("\nFAILED: %s has size %" PRId64
+                          ", but must be %zu!\n",
                           name, test, 2 * j);
                 return -1;
             }

test_conformance/basic/test_vector_swizzle.cpp

@@ -605,7 +605,7 @@ template <typename T, size_t N>
 static int test_vectype(const char* type_name, cl_device_id device,
                         cl_context context, cl_command_queue queue)
 {
-    log_info("    testing type %s%d\n", type_name, N);
+    log_info("    testing type %s%zu\n", type_name, N);
 
     cl_int error = CL_SUCCESS;
     int result = TEST_PASS;

test_conformance/buffers/test_buffer_copy.cpp

@@ -15,6 +15,7 @@
 //
 #include "harness/compat.h"
 
+#include <memory>
 #include <stdio.h>
 #include <stdlib.h>
 #include <string.h>
@@ -36,210 +37,246 @@ static int verify_copy_buffer(int *inptr, int *outptr, int n)
     return 0;
 }
 
+using alignedOwningPtr = std::unique_ptr<cl_int[], decltype(&align_free)>;
 
 static int test_copy( cl_command_queue queue, cl_context context, int num_elements, MTdata d )
 {
-    cl_mem  buffers[2];
-    cl_int  *int_input_ptr, *int_output_ptr;
-    cl_int  err;
-    int     i;
-    int     src_flag_id, dst_flag_id;
-    int     errors = 0;
+    clMemWrapper buffers[2];
+    cl_int err = CL_SUCCESS;
 
     size_t  min_alignment = get_min_alignment(context);
 
-    int_input_ptr = (cl_int*) align_malloc(sizeof(cl_int) * num_elements, min_alignment);
-    int_output_ptr = (cl_int*)align_malloc(sizeof(cl_int) * num_elements, min_alignment);
+    alignedOwningPtr invalid_ptr{
+        (cl_int *)align_malloc(sizeof(cl_int) * num_elements, min_alignment),
+        align_free
+    };
+    if (!invalid_ptr)
+    {
+        log_error(" unable to allocate %zu bytes of memory\n",
+                  sizeof(cl_int) * num_elements);
+        return TEST_FAIL;
+    }
+    alignedOwningPtr out_ptr{ (cl_int *)align_malloc(
+                                  sizeof(cl_int) * num_elements, min_alignment),
+                              align_free };
+    if (!out_ptr)
+    {
+        log_error(" unable to allocate %zu bytes of memory\n",
+                  sizeof(cl_int) * num_elements);
+        return TEST_FAIL;
+    }
+    alignedOwningPtr reference_ptr{
+        (cl_int *)align_malloc(sizeof(cl_int) * num_elements, min_alignment),
+        align_free
+    };
+    if (!reference_ptr)
+    {
+        log_error(" unable to allocate %zu bytes of memory\n",
+                  sizeof(cl_int) * num_elements);
+        return TEST_FAIL;
+    }
 
-    for (src_flag_id=0; src_flag_id < NUM_FLAGS; src_flag_id++) {
-        for (dst_flag_id=0; dst_flag_id < NUM_FLAGS; dst_flag_id++) {
+    for (int src_flag_id = 0; src_flag_id < NUM_FLAGS; src_flag_id++)
+    {
+        for (int dst_flag_id = 0; dst_flag_id < NUM_FLAGS; dst_flag_id++)
+        {
             log_info("Testing with cl_mem_flags src: %s dst: %s\n", flag_set_names[src_flag_id], flag_set_names[dst_flag_id]);
 
-            for (i=0; i<num_elements; i++){
-                int_input_ptr[i] = (int)genrand_int32( d );
-                int_output_ptr[i] = 0xdeaddead; // seed with incorrect data
+            for (int i = 0; i < num_elements; i++)
+            {
+                invalid_ptr[i] = static_cast<int>(0xdeaddead);
+                out_ptr[i] = static_cast<int>(0xdeadbeef);
+                reference_ptr[i] = (int)genrand_int32(d);
             }
 
+
             if ((flag_set[src_flag_id] & CL_MEM_USE_HOST_PTR) || (flag_set[src_flag_id] & CL_MEM_COPY_HOST_PTR))
-                buffers[0] = clCreateBuffer(context, flag_set[src_flag_id],  sizeof(cl_int) * num_elements, int_input_ptr, &err);
+                buffers[0] = clCreateBuffer(context, flag_set[src_flag_id],
+                                            sizeof(cl_int) * num_elements,
+                                            reference_ptr.get(), &err);
             else
-                buffers[0] = clCreateBuffer(context, flag_set[src_flag_id],  sizeof(cl_int) * num_elements, NULL, &err);
+                buffers[0] = clCreateBuffer(context, flag_set[src_flag_id],
+                                            sizeof(cl_int) * num_elements,
+                                            nullptr, &err);
             if ( err != CL_SUCCESS ){
-                print_error(err, " clCreateBuffer failed\n" );
-                align_free( (void *)int_input_ptr );
-                align_free( (void *)int_output_ptr );
-                return -1;
+                print_error(err, "clCreateBuffer failed\n");
+                return TEST_FAIL;
             }
 
             if ((flag_set[dst_flag_id] & CL_MEM_USE_HOST_PTR) || (flag_set[dst_flag_id] & CL_MEM_COPY_HOST_PTR))
-                buffers[1] = clCreateBuffer(context, flag_set[dst_flag_id],  sizeof(cl_int) * num_elements, int_output_ptr, &err);
+                buffers[1] = clCreateBuffer(context, flag_set[dst_flag_id],
+                                            sizeof(cl_int) * num_elements,
+                                            invalid_ptr.get(), &err);
             else
-                buffers[1] = clCreateBuffer(context, flag_set[dst_flag_id],  sizeof(cl_int) * num_elements, NULL, &err);
+                buffers[1] = clCreateBuffer(context, flag_set[dst_flag_id],
+                                            sizeof(cl_int) * num_elements,
+                                            nullptr, &err);
             if ( err != CL_SUCCESS ){
-                print_error(err, " clCreateBuffer failed\n" );
-                clReleaseMemObject( buffers[0] );
-                align_free( (void *)int_input_ptr );
-                align_free( (void *)int_output_ptr );
-                return -1;
+                print_error(err, "clCreateBuffer failed\n");
+                return TEST_FAIL;
             }
 
             if (!(flag_set[src_flag_id] & CL_MEM_USE_HOST_PTR) && !(flag_set[src_flag_id] & CL_MEM_COPY_HOST_PTR)) {
-                err = clEnqueueWriteBuffer(queue, buffers[0], CL_TRUE, 0, sizeof(cl_int)*num_elements, (void *)int_input_ptr, 0, NULL, NULL);
+                err = clEnqueueWriteBuffer(queue, buffers[0], CL_TRUE, 0,
+                                           sizeof(cl_int) * num_elements,
+                                           reference_ptr.get(), 0, nullptr,
+                                           nullptr);
                 if ( err != CL_SUCCESS ){
-                    print_error( err, "clEnqueueWriteBuffer failed" );
-                    clReleaseMemObject( buffers[0] );
-                    clReleaseMemObject( buffers[1] );
-                    align_free( (void *)int_output_ptr );
-                    align_free( (void *)int_input_ptr );
-                    return -1;
+                    print_error(err, "clEnqueueWriteBuffer failed\n");
+                    return TEST_FAIL;
                 }
             }
 
-            err = clEnqueueCopyBuffer(queue, buffers[0], buffers[1], 0, 0, sizeof(cl_int)*num_elements, 0, NULL, NULL);
+            err = clEnqueueCopyBuffer(queue, buffers[0], buffers[1], 0, 0,
+                                      sizeof(cl_int) * num_elements, 0, nullptr,
+                                      nullptr);
             if ( err != CL_SUCCESS ){
-                print_error( err, "clCopyArray failed" );
-                clReleaseMemObject( buffers[0] );
-                clReleaseMemObject( buffers[1] );
-                align_free( (void *)int_output_ptr );
-                align_free( (void *)int_input_ptr );
-                return -1;
+                print_error(err, "clCopyArray failed\n");
+                return TEST_FAIL;
             }
 
-            err = clEnqueueReadBuffer( queue, buffers[1], true, 0, sizeof(int)*num_elements, (void *)int_output_ptr, 0, NULL, NULL );
+            err = clEnqueueReadBuffer(queue, buffers[1], true, 0,
+                                      sizeof(int) * num_elements, out_ptr.get(),
+                                      0, nullptr, nullptr);
             if ( err != CL_SUCCESS ){
-                print_error( err, "clEnqueueReadBuffer failed" );
-                clReleaseMemObject( buffers[0] );
-                clReleaseMemObject( buffers[1] );
-                align_free( (void *)int_output_ptr );
-                align_free( (void *)int_input_ptr );
-                return -1;
+                print_error(err, "clEnqueueReadBuffer failed\n");
+                return TEST_FAIL;
             }
 
-            if ( verify_copy_buffer(int_input_ptr, int_output_ptr, num_elements) ){
+            if (verify_copy_buffer(reference_ptr.get(), out_ptr.get(),
+                                   num_elements))
+            {
                 log_error( " test failed\n" );
-                errors++;
+                return TEST_FAIL;
             }
             else{
                 log_info( " test passed\n" );
             }
-    // cleanup
-            clReleaseMemObject( buffers[0] );
-            clReleaseMemObject( buffers[1] );
         } // dst flags
-    }  // src flags
-    // cleanup
-    align_free( (void *)int_output_ptr );
-    align_free( (void *)int_input_ptr );
+    } // src flags
 
-    return errors;
+    return TEST_PASS;
 
 }   // end test_copy()
 
 
 static int testPartialCopy( cl_command_queue queue, cl_context context, int num_elements, cl_uint srcStart, cl_uint dstStart, int size, MTdata d )
 {
-    cl_mem  buffers[2];
-    int     *inptr, *outptr;
-    cl_int  err;
-    int     i;
-    int     src_flag_id, dst_flag_id;
-    int     errors = 0;
+    clMemWrapper buffers[2];
+    cl_int err = CL_SUCCESS;
 
     size_t  min_alignment = get_min_alignment(context);
 
-    inptr = (int *)align_malloc( sizeof(int) * num_elements, min_alignment);
-    if ( ! inptr ){
-        log_error( " unable to allocate %d bytes of memory\n", (int)sizeof(int) * num_elements );
-        return -1;
+    alignedOwningPtr invalid_ptr{
+        (cl_int *)align_malloc(sizeof(cl_int) * num_elements, min_alignment),
+        align_free
+    };
+    if (!invalid_ptr)
+    {
+        log_error(" unable to allocate %zu bytes of memory\n",
+                  sizeof(cl_int) * num_elements);
+        return TEST_FAIL;
     }
-    outptr = (int *)align_malloc( sizeof(int) * num_elements, min_alignment);
-    if ( ! outptr ){
-        log_error( " unable to allocate %d bytes of memory\n", (int)sizeof(int) * num_elements );
-        align_free( (void *)inptr );
-        return -1;
+    alignedOwningPtr out_ptr{ (cl_int *)align_malloc(
+                                  sizeof(cl_int) * num_elements, min_alignment),
+                              align_free };
+    if (!out_ptr)
+    {
+        log_error(" unable to allocate %zu bytes of memory\n",
+                  sizeof(cl_int) * num_elements);
+        return TEST_FAIL;
+    }
+    alignedOwningPtr reference_ptr{
+        (cl_int *)align_malloc(sizeof(cl_int) * num_elements, min_alignment),
+        align_free
+    };
+    if (!reference_ptr)
+    {
+        log_error(" unable to allocate %zu bytes of memory\n",
+                  sizeof(cl_int) * num_elements);
+        return TEST_FAIL;
     }
 
-    for (src_flag_id=0; src_flag_id < NUM_FLAGS; src_flag_id++) {
-        for (dst_flag_id=0; dst_flag_id < NUM_FLAGS; dst_flag_id++) {
+    for (int src_flag_id = 0; src_flag_id < NUM_FLAGS; src_flag_id++)
+    {
+        for (int dst_flag_id = 0; dst_flag_id < NUM_FLAGS; dst_flag_id++)
+        {
             log_info("Testing with cl_mem_flags src: %s dst: %s\n", flag_set_names[src_flag_id], flag_set_names[dst_flag_id]);
 
-            for (i=0; i<num_elements; i++){
-                inptr[i] = (int)genrand_int32( d );
-                outptr[i] = (int)0xdeaddead;    // seed with incorrect data
+            for (int i = 0; i < num_elements; i++)
+            {
+                invalid_ptr[i] = static_cast<int>(0xdeaddead);
+                out_ptr[i] = static_cast<int>(0xdeadbeef);
+                reference_ptr[i] = (int)genrand_int32(d);
             }
 
             if ((flag_set[src_flag_id] & CL_MEM_USE_HOST_PTR) || (flag_set[src_flag_id] & CL_MEM_COPY_HOST_PTR))
-                buffers[0] = clCreateBuffer(context, flag_set[src_flag_id],  sizeof(cl_int) * num_elements, inptr, &err);
+                buffers[0] = clCreateBuffer(context, flag_set[src_flag_id],
+                                            sizeof(cl_int) * num_elements,
+                                            reference_ptr.get(), &err);
             else
-                buffers[0] = clCreateBuffer(context, flag_set[src_flag_id],  sizeof(cl_int) * num_elements, NULL, &err);
+                buffers[0] = clCreateBuffer(context, flag_set[src_flag_id],
+                                            sizeof(cl_int) * num_elements,
+                                            nullptr, &err);
             if ( err != CL_SUCCESS ){
-                print_error(err, " clCreateBuffer failed\n" )
-                align_free( (void *)outptr );
-                align_free( (void *)inptr );
-                return -1;
+                print_error(err, "clCreateBuffer failed\n");
+                return TEST_FAIL;
             }
 
             if ((flag_set[dst_flag_id] & CL_MEM_USE_HOST_PTR) || (flag_set[dst_flag_id] & CL_MEM_COPY_HOST_PTR))
-                buffers[1] = clCreateBuffer(context, flag_set[dst_flag_id],  sizeof(cl_int) * num_elements, outptr, &err);
+                buffers[1] = clCreateBuffer(context, flag_set[dst_flag_id],
+                                            sizeof(cl_int) * num_elements,
+                                            invalid_ptr.get(), &err);
             else
-                buffers[1] = clCreateBuffer(context, flag_set[dst_flag_id],  sizeof(cl_int) * num_elements, NULL, &err);
+                buffers[1] = clCreateBuffer(context, flag_set[dst_flag_id],
+                                            sizeof(cl_int) * num_elements,
+                                            nullptr, &err);
             if ( err != CL_SUCCESS ){
-                print_error(err, " clCreateBuffer failed\n" );
-                clReleaseMemObject( buffers[0] );
-                align_free( (void *)outptr );
-                align_free( (void *)inptr );
-                return -1;
+                print_error(err, "clCreateBuffer failed\n");
+                return TEST_FAIL;
             }
 
             if (!(flag_set[src_flag_id] & CL_MEM_USE_HOST_PTR) && !(flag_set[src_flag_id] & CL_MEM_COPY_HOST_PTR)){
-                err = clEnqueueWriteBuffer(queue, buffers[0], CL_TRUE, 0, sizeof(cl_int)*num_elements, (void *)inptr, 0, NULL, NULL);
+                err = clEnqueueWriteBuffer(queue, buffers[0], CL_TRUE, 0,
+                                           sizeof(cl_int) * num_elements,
+                                           reference_ptr.get(), 0, nullptr,
+                                           nullptr);
                 if ( err != CL_SUCCESS ){
-                    print_error( err, "clEnqueueWriteBuffer failed" );
-                    clReleaseMemObject( buffers[1] );
-                    clReleaseMemObject( buffers[0] );
-                    align_free( (void *)outptr );
-                    align_free( (void *)inptr );
-                    return -1;
+                    print_error(err, "clEnqueueWriteBuffer failed\n");
+                    return TEST_FAIL;
                 }
             }
 
-            err = clEnqueueCopyBuffer(queue, buffers[0], buffers[1], srcStart*sizeof(cl_int), dstStart*sizeof(cl_int), sizeof(cl_int)*size, 0, NULL, NULL);
+            err = clEnqueueCopyBuffer(
+                queue, buffers[0], buffers[1], srcStart * sizeof(cl_int),
+                dstStart * sizeof(cl_int), sizeof(cl_int) * size, 0, nullptr,
+                nullptr);
             if ( err != CL_SUCCESS){
-                print_error( err, "clEnqueueCopyBuffer failed" );
-                clReleaseMemObject( buffers[1] );
-                clReleaseMemObject( buffers[0] );
-                align_free( (void *)outptr );
-                align_free( (void *)inptr );
-                return -1;
+                print_error(err, "clEnqueueCopyBuffer failed\n");
+                return TEST_FAIL;
             }
 
-            err = clEnqueueReadBuffer( queue, buffers[1], true, 0, sizeof(int)*num_elements, (void *)outptr, 0, NULL, NULL );
+            err = clEnqueueReadBuffer(queue, buffers[1], true, 0,
+                                      sizeof(int) * num_elements, out_ptr.get(),
+                                      0, nullptr, nullptr);
             if ( err != CL_SUCCESS){
-                print_error( err, "clEnqueueReadBuffer failed" );
-                clReleaseMemObject( buffers[1] );
-                clReleaseMemObject( buffers[0] );
-                align_free( (void *)outptr );
-                align_free( (void *)inptr );
-                return -1;
+                print_error(err, "clEnqueueReadBuffer failed\n");
+                return TEST_FAIL;
             }
 
-            if ( verify_copy_buffer(inptr + srcStart, outptr + dstStart, size) ){
+            if (verify_copy_buffer(reference_ptr.get() + srcStart,
+                                   out_ptr.get() + dstStart, size))
+            {
                 log_error("buffer_COPY test failed\n");
-                errors++;
+                return TEST_FAIL;
             }
             else{
                 log_info("buffer_COPY test passed\n");
             }
-    // cleanup
-            clReleaseMemObject( buffers[1] );
-            clReleaseMemObject( buffers[0] );
         } // dst mem flags
     } // src mem flags
-    // cleanup
-    align_free( (void *)outptr );
-    align_free( (void *)inptr );
 
-    return errors;
+    return TEST_PASS;
 
 }   // end testPartialCopy()
 
@@ -252,15 +289,19 @@ REGISTER_TEST(buffer_copy)
 
     // test the preset size
     log_info( "set size: %d: ", num_elements );
-    if (test_copy( queue, context, num_elements, d ))
+    if (test_copy(queue, context, num_elements, d) != TEST_PASS)
+    {
         err++;
+    }
 
     // now test random sizes
     for ( i = 0; i < 8; i++ ){
         size = (int)get_random_float(2.f,131072.f, d);
         log_info( "random size: %d: ", size );
-        if (test_copy( queue, context, size, d ))
+        if (test_copy(queue, context, size, d) != TEST_PASS)
+        {
             err++;
+        }
     }
 
     free_mtdata(d);
@@ -283,8 +324,12 @@ REGISTER_TEST(buffer_partial_copy)
         size = (int)get_random_float( 8.f, (float)(num_elements - srcStart), d );
         dstStart = (cl_uint)get_random_float( 0.f, (float)(num_elements - size), d );
         log_info( "random partial copy from %d to %d, size: %d: ", (int)srcStart, (int)dstStart, size );
-        if (testPartialCopy( queue, context, num_elements, srcStart, dstStart, size, d ))
+        if (testPartialCopy(queue, context, num_elements, srcStart, dstStart,
+                            size, d)
+            != TEST_PASS)
+        {
             err++;
+        }
     }
 
     free_mtdata(d);

test_conformance/c11_atomics/common.cpp

@@ -58,30 +58,28 @@ cl_uint AtomicTypeInfo::Size(cl_device_id device)
 {
   switch(_type)
   {
-  case TYPE_ATOMIC_INT:
-  case TYPE_ATOMIC_UINT:
-  case TYPE_ATOMIC_FLOAT:
-  case TYPE_ATOMIC_FLAG:
-    return sizeof(cl_int);
-  case TYPE_ATOMIC_LONG:
-  case TYPE_ATOMIC_ULONG:
-  case TYPE_ATOMIC_DOUBLE:
-    return sizeof(cl_long);
-  case TYPE_ATOMIC_INTPTR_T:
-  case TYPE_ATOMIC_UINTPTR_T:
-  case TYPE_ATOMIC_SIZE_T:
-  case TYPE_ATOMIC_PTRDIFF_T:
-    {
-      int error;
-      cl_uint addressBits = 0;
+      case TYPE_ATOMIC_HALF: return sizeof(cl_half);
+      case TYPE_ATOMIC_INT:
+      case TYPE_ATOMIC_UINT:
+      case TYPE_ATOMIC_FLOAT:
+      case TYPE_ATOMIC_FLAG: return sizeof(cl_int);
+      case TYPE_ATOMIC_LONG:
+      case TYPE_ATOMIC_ULONG:
+      case TYPE_ATOMIC_DOUBLE: return sizeof(cl_long);
+      case TYPE_ATOMIC_INTPTR_T:
+      case TYPE_ATOMIC_UINTPTR_T:
+      case TYPE_ATOMIC_SIZE_T:
+      case TYPE_ATOMIC_PTRDIFF_T: {
+          int error;
+          cl_uint addressBits = 0;
 
-      error = clGetDeviceInfo(device, CL_DEVICE_ADDRESS_BITS, sizeof(addressBits), &addressBits, 0);
-      test_error_ret(error, "clGetDeviceInfo", 0);
+          error = clGetDeviceInfo(device, CL_DEVICE_ADDRESS_BITS,
+                                  sizeof(addressBits), &addressBits, 0);
+          test_error_ret(error, "clGetDeviceInfo", 0);
 
-      return addressBits/8;
-    }
-  default:
-    return 0;
+          return addressBits / 8;
+      }
+      default: return 0;
   }
 }
 
@@ -93,6 +91,7 @@ const char *AtomicTypeInfo::AtomicTypeName()
     return "atomic_int";
   case TYPE_ATOMIC_UINT:
     return "atomic_uint";
+  case TYPE_ATOMIC_HALF: return "atomic_half";
   case TYPE_ATOMIC_FLOAT:
     return "atomic_float";
   case TYPE_ATOMIC_FLAG:
@@ -124,6 +123,7 @@ const char *AtomicTypeInfo::RegularTypeName()
     return "int";
   case TYPE_ATOMIC_UINT:
     return "uint";
+  case TYPE_ATOMIC_HALF: return "half";
   case TYPE_ATOMIC_FLOAT:
     return "float";
   case TYPE_ATOMIC_FLAG:
@@ -163,29 +163,30 @@ int AtomicTypeInfo::IsSupported(cl_device_id device)
 {
   switch(_type)
   {
-  case TYPE_ATOMIC_INT:
-  case TYPE_ATOMIC_UINT:
-  case TYPE_ATOMIC_FLOAT:
-  case TYPE_ATOMIC_FLAG:
-    return 1;
-  case TYPE_ATOMIC_LONG:
-  case TYPE_ATOMIC_ULONG:
-    return is_extension_available(device, "cl_khr_int64_base_atomics") &&
-      is_extension_available(device, "cl_khr_int64_extended_atomics");
-  case TYPE_ATOMIC_DOUBLE:
-    return is_extension_available(device, "cl_khr_int64_base_atomics") &&
-      is_extension_available(device, "cl_khr_int64_extended_atomics") &&
-      is_extension_available(device, "cl_khr_fp64");
-  case TYPE_ATOMIC_INTPTR_T:
-  case TYPE_ATOMIC_UINTPTR_T:
-  case TYPE_ATOMIC_SIZE_T:
-  case TYPE_ATOMIC_PTRDIFF_T:
-    if(Size(device) == 4)
-      return 1;
-    return is_extension_available(device, "cl_khr_int64_base_atomics") &&
-      is_extension_available(device, "cl_khr_int64_extended_atomics");
-  default:
-    return 0;
+      case TYPE_ATOMIC_HALF:
+          return is_extension_available(device, "cl_khr_fp16");
+      case TYPE_ATOMIC_INT:
+      case TYPE_ATOMIC_UINT:
+      case TYPE_ATOMIC_FLOAT:
+      case TYPE_ATOMIC_FLAG: return 1;
+      case TYPE_ATOMIC_LONG:
+      case TYPE_ATOMIC_ULONG:
+          return is_extension_available(device, "cl_khr_int64_base_atomics")
+              && is_extension_available(device,
+                                        "cl_khr_int64_extended_atomics");
+      case TYPE_ATOMIC_DOUBLE:
+          return is_extension_available(device, "cl_khr_int64_base_atomics")
+              && is_extension_available(device, "cl_khr_int64_extended_atomics")
+              && is_extension_available(device, "cl_khr_fp64");
+      case TYPE_ATOMIC_INTPTR_T:
+      case TYPE_ATOMIC_UINTPTR_T:
+      case TYPE_ATOMIC_SIZE_T:
+      case TYPE_ATOMIC_PTRDIFF_T:
+          if (Size(device) == 4) return 1;
+          return is_extension_available(device, "cl_khr_int64_base_atomics")
+              && is_extension_available(device,
+                                        "cl_khr_int64_extended_atomics");
+      default: return 0;
   }
 }
 

test_conformance/c11_atomics/common.h

@@ -22,6 +22,8 @@
 
 #include "host_atomics.h"
 
+#include "CL/cl_half.h"
+
 #include <vector>
 #include <sstream>
 
@@ -38,6 +40,7 @@ enum TExplicitAtomicType
     TYPE_ATOMIC_UINT,
     TYPE_ATOMIC_LONG,
     TYPE_ATOMIC_ULONG,
+    TYPE_ATOMIC_HALF,
     TYPE_ATOMIC_FLOAT,
     TYPE_ATOMIC_DOUBLE,
     TYPE_ATOMIC_INTPTR_T,
@@ -71,6 +74,9 @@ extern int
     gMaxDeviceThreads; // maximum number of threads executed on OCL device
 extern cl_device_atomic_capabilities gAtomicMemCap,
     gAtomicFenceCap; // atomic memory and fence capabilities for this device
+extern cl_half_rounding_mode gHalfRoundingMode;
+extern bool gFloatAtomicsSupported;
+extern cl_device_fp_atomic_capabilities_ext gHalfAtomicCaps;
 
 extern const char *
 get_memory_order_type_name(TExplicitMemoryOrderType orderType);
@@ -240,13 +246,13 @@ public:
         int error = 0;
         if (_maxDeviceThreads > 0 && !UseSVM())
         {
-            LocalMemory(true);
+            SetLocalMemory(true);
             EXECUTE_TEST(
                 error, ExecuteForEachDeclarationType(deviceID, context, queue));
         }
         if (_maxDeviceThreads + MaxHostThreads() > 0)
         {
-            LocalMemory(false);
+            SetLocalMemory(false);
             EXECUTE_TEST(
                 error, ExecuteForEachDeclarationType(deviceID, context, queue));
         }
@@ -401,7 +407,7 @@ public:
     bool UseSVM() { return _useSVM; }
     void StartValue(HostDataType startValue) { _startValue = startValue; }
     HostDataType StartValue() { return _startValue; }
-    void LocalMemory(bool local) { _localMemory = local; }
+    void SetLocalMemory(bool local) { _localMemory = local; }
     bool LocalMemory() { return _localMemory; }
     void DeclaredInProgram(bool declaredInProgram)
     {
@@ -781,6 +787,8 @@ CBasicTest<HostAtomicType, HostDataType>::PragmaHeader(cl_device_id deviceID)
     }
     if (_dataType == TYPE_ATOMIC_DOUBLE)
         pragma += "#pragma OPENCL EXTENSION cl_khr_fp64 : enable\n";
+    if (_dataType == TYPE_ATOMIC_HALF)
+        pragma += "#pragma OPENCL EXTENSION cl_khr_fp16 : enable\n";
     return pragma;
 }
 
@@ -875,7 +883,14 @@ CBasicTest<HostAtomicType, HostDataType>::ProgramHeader(cl_uint maxNumDestItems)
         header += std::string("__global volatile ") + aTypeName + " destMemory["
             + ss.str() + "] = {\n";
         ss.str("");
-        ss << _startValue;
+
+        if (CBasicTest<HostAtomicType, HostDataType>::DataType()._type
+            != TYPE_ATOMIC_HALF)
+            ss << _startValue;
+        else
+            ss << static_cast<HostDataType>(
+                cl_half_to_float(static_cast<cl_half>(_startValue)));
+
         for (cl_uint i = 0; i < maxNumDestItems; i++)
         {
             if (aTypeName == "atomic_flag")

test_conformance/c11_atomics/host_atomics.h

@@ -41,6 +41,7 @@ enum TExplicitMemoryOrderType
 #define HOST_ATOMIC_UINT        unsigned long
 #define HOST_ATOMIC_LONG        unsigned long long
 #define HOST_ATOMIC_ULONG       unsigned long long
+#define HOST_ATOMIC_HALF unsigned short
 #define HOST_ATOMIC_FLOAT       float
 #define HOST_ATOMIC_DOUBLE      double
 #else
@@ -48,6 +49,7 @@ enum TExplicitMemoryOrderType
 #define HOST_ATOMIC_UINT        cl_uint
 #define HOST_ATOMIC_LONG        cl_long
 #define HOST_ATOMIC_ULONG       cl_ulong
+#define HOST_ATOMIC_HALF cl_half
 #define HOST_ATOMIC_FLOAT       cl_float
 #define HOST_ATOMIC_DOUBLE      cl_double
 #endif
@@ -69,6 +71,7 @@ enum TExplicitMemoryOrderType
 #define HOST_UINT               cl_uint
 #define HOST_LONG               cl_long
 #define HOST_ULONG              cl_ulong
+#define HOST_HALF cl_half
 #define HOST_FLOAT              cl_float
 #define HOST_DOUBLE             cl_double
 
@@ -120,9 +123,12 @@ CorrespondingType host_atomic_exchange(volatile AtomicType *a, CorrespondingType
                                        TExplicitMemoryOrderType order)
 {
 #if defined( _MSC_VER ) || (defined( __INTEL_COMPILER ) && defined(WIN32))
-  return InterlockedExchange(a, c);
+    if (sizeof(CorrespondingType) == 2)
+        return InterlockedExchange16(reinterpret_cast<volatile SHORT *>(a), c);
+    else
+        return InterlockedExchange(reinterpret_cast<volatile LONG *>(a), c);
 #elif defined(__GNUC__)
-  return __sync_lock_test_and_set(a, c);
+    return __sync_lock_test_and_set(a, c);
 #else
   log_info("Host function not implemented: atomic_exchange\n");
   return 0;
@@ -158,7 +164,10 @@ CorrespondingType host_atomic_load(volatile AtomicType *a,
                                    TExplicitMemoryOrderType order)
 {
 #if defined( _MSC_VER ) || (defined( __INTEL_COMPILER ) && defined(WIN32))
-  return InterlockedExchangeAdd(a, 0);
+    if (sizeof(CorrespondingType) == 2)
+        auto prev = InterlockedOr16(reinterpret_cast<volatile SHORT *>(a), 0);
+    else
+        return InterlockedExchangeAdd(reinterpret_cast<volatile LONG *>(a), 0);
 #elif defined(__GNUC__)
   return __sync_add_and_fetch(a, 0);
 #else

test_conformance/c11_atomics/main.cpp

@@ -1,6 +1,6 @@
 //
-// Copyright (c) 2017 The Khronos Group Inc.
-// 
+// Copyright (c) 2024 The Khronos Group Inc.
+//
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
 // You may obtain a copy of the License at
@@ -14,8 +14,11 @@
 // limitations under the License.
 //
 #include "harness/testHarness.h"
+#include "harness/deviceInfo.h"
+#include "harness/kernelHelpers.h"
 #include <iostream>
 #include <string>
+#include "CL/cl_half.h"
 
 bool gHost = false; // flag for testing native host threads (test verification)
 bool gOldAPI = false; // flag for testing with old API (OpenCL 1.2) - test verification
@@ -28,6 +31,9 @@ int gInternalIterations = 10000; // internal test iterations for atomic operatio
 int gMaxDeviceThreads = 1024; // maximum number of threads executed on OCL device
 cl_device_atomic_capabilities gAtomicMemCap,
     gAtomicFenceCap; // atomic memory and fence capabilities for this device
+cl_half_rounding_mode gHalfRoundingMode = CL_HALF_RTE;
+bool gFloatAtomicsSupported = false;
+cl_device_fp_atomic_capabilities_ext gHalfAtomicCaps = 0;
 
 test_status InitCL(cl_device_id device) {
     auto version = get_device_cl_version(device);
@@ -123,6 +129,34 @@ test_status InitCL(cl_device_id device) {
             | CL_DEVICE_ATOMIC_SCOPE_ALL_DEVICES;
     }
 
+    if (is_extension_available(device, "cl_ext_float_atomics"))
+    {
+        gFloatAtomicsSupported = true;
+        if (is_extension_available(device, "cl_khr_fp16"))
+        {
+            cl_int error = clGetDeviceInfo(
+                device, CL_DEVICE_HALF_FP_ATOMIC_CAPABILITIES_EXT,
+                sizeof(gHalfAtomicCaps), &gHalfAtomicCaps, nullptr);
+            test_error_ret(error, "clGetDeviceInfo failed!", TEST_FAIL);
+
+            const cl_device_fp_config fpConfigHalf =
+                get_default_rounding_mode(device, CL_DEVICE_HALF_FP_CONFIG);
+            if ((fpConfigHalf & CL_FP_ROUND_TO_NEAREST) != 0)
+            {
+                gHalfRoundingMode = CL_HALF_RTE;
+            }
+            else if ((fpConfigHalf & CL_FP_ROUND_TO_ZERO) != 0)
+            {
+                gHalfRoundingMode = CL_HALF_RTZ;
+            }
+            else
+            {
+                log_error("Error while acquiring half rounding mode\n");
+                return TEST_FAIL;
+            }
+        }
+    }
+
     return TEST_PASS;
 }
 

test_conformance/c11_atomics/test_atomics.cpp

@@ -1,5 +1,5 @@
 //
-// Copyright (c) 2017 The Khronos Group Inc.
+// Copyright (c) 2024 The Khronos Group Inc.
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
 // you may not use this file except in compliance with the License.
@@ -33,6 +33,7 @@ public:
     using CBasicTestMemOrderScope<HostAtomicType,
                                   HostDataType>::MemoryOrderScopeStr;
     using CBasicTest<HostAtomicType, HostDataType>::CheckCapabilities;
+    using CBasicTestMemOrderScope<HostAtomicType, HostDataType>::LocalMemory;
     CBasicTestStore(TExplicitAtomicType dataType, bool useSVM)
         : CBasicTestMemOrderScope<HostAtomicType, HostDataType>(dataType,
                                                                 useSVM)
@@ -54,6 +55,21 @@ public:
             == TEST_SKIPPED_ITSELF)
             return 0; // skip test - not applicable
 
+        if (CBasicTestMemOrderScope<HostAtomicType, HostDataType>::DataType()
+                ._type
+            == TYPE_ATOMIC_HALF)
+        {
+            if (LocalMemory()
+                && (gHalfAtomicCaps & CL_DEVICE_LOCAL_FP_ATOMIC_LOAD_STORE_EXT)
+                    == 0)
+                return 0; // skip test - not applicable
+
+            if (!LocalMemory()
+                && (gHalfAtomicCaps & CL_DEVICE_GLOBAL_FP_ATOMIC_LOAD_STORE_EXT)
+                    == 0)
+                return 0;
+        }
+
         return CBasicTestMemOrderScope<
             HostAtomicType, HostDataType>::ExecuteSingleTest(deviceID, context,
                                                              queue);
@@ -75,7 +91,13 @@ public:
                                HostDataType *startRefValues,
                                cl_uint whichDestValue)
     {
-        expected = (HostDataType)whichDestValue;
+        if (CBasicTestMemOrderScope<HostAtomicType, HostDataType>::DataType()
+                ._type
+            != TYPE_ATOMIC_HALF)
+            expected = (HostDataType)whichDestValue;
+        else
+            expected = cl_half_from_float(static_cast<float>(whichDestValue),
+                                          gHalfRoundingMode);
         return true;
     }
 };
@@ -109,6 +131,15 @@ static int test_atomic_store_generic(cl_device_id deviceID, cl_context context,
         TYPE_ATOMIC_DOUBLE, useSVM);
     EXECUTE_TEST(error,
                  test_double.Execute(deviceID, context, queue, num_elements));
+
+    if (gFloatAtomicsSupported)
+    {
+        CBasicTestStore<HOST_ATOMIC_HALF, HOST_HALF> test_half(TYPE_ATOMIC_HALF,
+                                                               useSVM);
+        EXECUTE_TEST(error,
+                     test_half.Execute(deviceID, context, queue, num_elements));
+    }
+
     if (AtomicTypeInfo(TYPE_ATOMIC_SIZE_T).Size(deviceID) == 4)
     {
         CBasicTestStore<HOST_ATOMIC_INTPTR_T32, HOST_INTPTR_T32> test_intptr_t(
@@ -297,6 +328,7 @@ public:
                                   HostDataType>::MemoryOrderScopeStr;
     using CBasicTestMemOrderScope<HostAtomicType, HostDataType>::MemoryScopeStr;
     using CBasicTest<HostAtomicType, HostDataType>::CheckCapabilities;
+    using CBasicTestMemOrderScope<HostAtomicType, HostDataType>::LocalMemory;
     CBasicTestLoad(TExplicitAtomicType dataType, bool useSVM)
         : CBasicTestMemOrderScope<HostAtomicType, HostDataType>(dataType,
                                                                 useSVM)
@@ -318,6 +350,21 @@ public:
             == TEST_SKIPPED_ITSELF)
             return 0; // skip test - not applicable
 
+        if (CBasicTestMemOrderScope<HostAtomicType, HostDataType>::DataType()
+                ._type
+            == TYPE_ATOMIC_HALF)
+        {
+            if (LocalMemory()
+                && (gHalfAtomicCaps & CL_DEVICE_LOCAL_FP_ATOMIC_LOAD_STORE_EXT)
+                    == 0)
+                return 0; // skip test - not applicable
+
+            if (!LocalMemory()
+                && (gHalfAtomicCaps & CL_DEVICE_GLOBAL_FP_ATOMIC_LOAD_STORE_EXT)
+                    == 0)
+                return 0;
+        }
+
         return CBasicTestMemOrderScope<
             HostAtomicType, HostDataType>::ExecuteSingleTest(deviceID, context,
                                                              queue);
@@ -351,7 +398,13 @@ public:
                                HostDataType *startRefValues,
                                cl_uint whichDestValue)
     {
-        expected = (HostDataType)whichDestValue;
+        if (CBasicTestMemOrderScope<HostAtomicType, HostDataType>::DataType()
+                ._type
+            != TYPE_ATOMIC_HALF)
+            expected = (HostDataType)whichDestValue;
+        else
+            expected = cl_half_from_float(static_cast<float>(whichDestValue),
+                                          gHalfRoundingMode);
         return true;
     }
     virtual bool VerifyRefs(bool &correct, cl_uint threadCount,
@@ -361,11 +414,25 @@ public:
         correct = true;
         for (cl_uint i = 0; i < threadCount; i++)
         {
-            if (refValues[i] != (HostDataType)i)
+            if constexpr (std::is_same<HostDataType, cl_half>::value)
             {
-                log_error("Invalid value for thread %u\n", (cl_uint)i);
-                correct = false;
-                return true;
+                HostDataType test = cl_half_from_float(static_cast<float>(i),
+                                                       gHalfRoundingMode);
+                if (refValues[i] != test)
+                {
+                    log_error("Invalid value for thread %u\n", (cl_uint)i);
+                    correct = false;
+                    return true;
+                }
+            }
+            else
+            {
+                if (refValues[i] != (HostDataType)i)
+                {
+                    log_error("Invalid value for thread %u\n", (cl_uint)i);
+                    correct = false;
+                    return true;
+                }
             }
         }
         return true;
@@ -400,6 +467,15 @@ static int test_atomic_load_generic(cl_device_id deviceID, cl_context context,
         TYPE_ATOMIC_DOUBLE, useSVM);
     EXECUTE_TEST(error,
                  test_double.Execute(deviceID, context, queue, num_elements));
+
+    if (gFloatAtomicsSupported)
+    {
+        CBasicTestLoad<HOST_ATOMIC_HALF, HOST_HALF> test_half(TYPE_ATOMIC_HALF,
+                                                              useSVM);
+        EXECUTE_TEST(error,
+                     test_half.Execute(deviceID, context, queue, num_elements));
+    }
+
     if (AtomicTypeInfo(TYPE_ATOMIC_SIZE_T).Size(deviceID) == 4)
     {
         CBasicTestLoad<HOST_ATOMIC_INTPTR_T32, HOST_INTPTR_T32> test_intptr_t(
@@ -469,11 +545,36 @@ public:
                                   HostDataType>::MemoryOrderScopeStr;
     using CBasicTestMemOrderScope<HostAtomicType, HostDataType>::Iterations;
     using CBasicTestMemOrderScope<HostAtomicType, HostDataType>::IterationsStr;
+    using CBasicTestMemOrderScope<HostAtomicType, HostDataType>::LocalMemory;
     CBasicTestExchange(TExplicitAtomicType dataType, bool useSVM)
         : CBasicTestMemOrderScope<HostAtomicType, HostDataType>(dataType,
                                                                 useSVM)
     {
-        StartValue(123456);
+        if constexpr (std::is_same_v<HostDataType, HOST_ATOMIC_HALF>)
+            StartValue(cl_half_from_float(static_cast<float>(1234),
+                                          gHalfRoundingMode));
+        else
+            StartValue(123456);
+    }
+    virtual int ExecuteSingleTest(cl_device_id deviceID, cl_context context,
+                                  cl_command_queue queue)
+    {
+        if constexpr (std::is_same_v<HostDataType, HOST_ATOMIC_HALF>)
+        {
+            if (LocalMemory()
+                && (gHalfAtomicCaps & CL_DEVICE_LOCAL_FP_ATOMIC_LOAD_STORE_EXT)
+                    == 0)
+                return 0; // skip test - not applicable
+
+            if (!LocalMemory()
+                && (gHalfAtomicCaps & CL_DEVICE_GLOBAL_FP_ATOMIC_LOAD_STORE_EXT)
+                    == 0)
+                return 0;
+        }
+
+        return CBasicTestMemOrderScope<
+            HostAtomicType, HostDataType>::ExecuteSingleTest(deviceID, context,
+                                                             queue);
     }
     virtual std::string ProgramCore()
     {
@@ -515,17 +616,35 @@ public:
         /* Any repeated value is treated as an error */
         std::vector<bool> tidFound(threadCount);
         bool startValueFound = false;
-        cl_uint i;
+        cl_uint startVal = StartValue();
 
-        for (i = 0; i <= threadCount; i++)
+        if constexpr (std::is_same_v<HostDataType, HOST_ATOMIC_HALF>)
+            startVal = static_cast<cl_uint>(
+                cl_half_to_float(static_cast<cl_half>(StartValue())));
+
+        for (cl_uint i = 0; i <= threadCount; i++)
         {
-            cl_uint value;
+            cl_uint value = 0;
             if (i == threadCount)
-                value = (cl_uint)finalValues[0]; // additional value from atomic
+            {
+                if constexpr (!std::is_same_v<HostDataType, HOST_ATOMIC_HALF>)
+                    value =
+                        (cl_uint)finalValues[0]; // additional value from atomic
                                                  // variable (last written)
+                else
+                    value =
+                        cl_half_to_float(static_cast<cl_half>(finalValues[0]));
+            }
             else
-                value = (cl_uint)refValues[i];
-            if (value == (cl_uint)StartValue())
+            {
+                if constexpr (!std::is_same_v<HostDataType, HOST_ATOMIC_HALF>)
+                    value = (cl_uint)refValues[i];
+                else
+                    value =
+                        cl_half_to_float(static_cast<cl_half>(refValues[i]));
+            }
+
+            if (value == startVal)
             {
                 // Special initial value
                 if (startValueFound)
@@ -590,6 +709,13 @@ static int test_atomic_exchange_generic(cl_device_id deviceID,
         TYPE_ATOMIC_DOUBLE, useSVM);
     EXECUTE_TEST(error,
                  test_double.Execute(deviceID, context, queue, num_elements));
+    if (gFloatAtomicsSupported)
+    {
+        CBasicTestExchange<HOST_ATOMIC_HALF, HOST_HALF> test_half(
+            TYPE_ATOMIC_HALF, useSVM);
+        EXECUTE_TEST(error,
+                     test_half.Execute(deviceID, context, queue, num_elements));
+    }
     if (AtomicTypeInfo(TYPE_ATOMIC_SIZE_T).Size(deviceID) == 4)
     {
         CBasicTestExchange<HOST_ATOMIC_INTPTR_T32, HOST_INTPTR_T32>

test_conformance/common/vulkan_wrapper/CMakeLists.txt

@@ -6,6 +6,7 @@ set(VULKAN_WRAPPER_SOURCES
 )
 
 # needed by Vulkan wrapper to compile
+set(CMAKE_COMPILE_WARNING_AS_ERROR OFF)
 add_cxx_flag_if_supported(-Wmisleading-indentation)
 add_cxx_flag_if_supported(-Wno-narrowing)
 add_cxx_flag_if_supported(-Wno-format)

test_conformance/common/vulkan_wrapper/opencl_vulkan_wrapper.cpp

@@ -428,38 +428,68 @@ size_t GetElementNBytes(const cl_image_format *format)
     return result;
 }
 
-cl_int get2DImageDimensions(const VkImageCreateInfo *VulkanImageCreateInfo,
-                            cl_image_format *img_fmt, size_t totalImageSize,
-                            size_t &width, size_t &height)
+cl_int getImageDimensions(const VkImageCreateInfo *VulkanImageCreateInfo,
+                          cl_image_format *img_fmt, cl_image_desc *img_desc,
+                          VkExtent3D max_ext, const VkSubresourceLayout *layout)
 {
-    cl_int result = CL_SUCCESS;
-    if (totalImageSize == 0)
+    test_assert_error(
+        VulkanImageCreateInfo != nullptr,
+        "getImageDimensions: invalid VulkanImageCreateInfo pointer!");
+    test_assert_error(img_fmt != nullptr,
+                      "getImageDimensions: invalid img_fmt pointer!");
+    test_assert_error(img_desc != nullptr,
+                      "getImageDimensions: invalid img_desc pointer!");
+
+    img_desc->image_depth = VulkanImageCreateInfo->extent.depth;
+    img_desc->image_width = VulkanImageCreateInfo->extent.width;
+    img_desc->image_height = VulkanImageCreateInfo->extent.height;
+
+    if (layout != nullptr)
     {
-        result = CL_INVALID_VALUE;
+        size_t element_size = GetElementNBytes(img_fmt);
+        size_t row_pitch = element_size * VulkanImageCreateInfo->extent.width;
+        row_pitch = row_pitch < layout->rowPitch ? layout->rowPitch : row_pitch;
+        img_desc->image_row_pitch = row_pitch;
+
+        size_t slice_pitch = row_pitch * VulkanImageCreateInfo->extent.height;
+        slice_pitch =
+            slice_pitch < layout->depthPitch ? layout->depthPitch : slice_pitch;
+        img_desc->image_slice_pitch = slice_pitch;
     }
-    size_t element_size = GetElementNBytes(img_fmt);
-    size_t row_pitch = element_size * VulkanImageCreateInfo->extent.width;
-    row_pitch = row_pitch % 64 == 0 ? row_pitch : ((row_pitch / 64) + 1) * 64;
 
-    width = row_pitch / element_size;
-    height = totalImageSize / row_pitch;
+    switch (img_desc->image_type)
+    {
+        case CL_MEM_OBJECT_IMAGE3D:
+            test_assert_error(img_desc->image_depth >= 1
+                                  && img_desc->image_depth <= max_ext.depth,
+                              "getImageDimensions: invalid image depth!");
+        case CL_MEM_OBJECT_IMAGE2D:
+            test_assert_error(img_desc->image_height >= 1
+                                  && img_desc->image_height <= max_ext.height,
+                              "getImageDimensions: invalid image height!");
+        case CL_MEM_OBJECT_IMAGE1D:
+            test_assert_error(img_desc->image_width >= 1
+                                  && img_desc->image_width <= max_ext.width,
+                              "getImageDimensions: invalid image width!");
+    }
 
-    return result;
+    return CL_SUCCESS;
 }
 
 cl_int
-getCLImageInfoFromVkImageInfo(const VkImageCreateInfo *VulkanImageCreateInfo,
-                              size_t totalImageSize, cl_image_format *img_fmt,
-                              cl_image_desc *img_desc)
+getCLImageInfoFromVkImageInfo(const cl_device_id device,
+                              const VkImageCreateInfo *VulkanImageCreateInfo,
+                              cl_image_format *img_fmt, cl_image_desc *img_desc,
+                              const VkSubresourceLayout *layout)
 {
-    cl_int result = CL_SUCCESS;
+    cl_int error = CL_SUCCESS;
 
     cl_image_format clImgFormat = { 0 };
-    result =
+    error =
         getCLFormatFromVkFormat(VulkanImageCreateInfo->format, &clImgFormat);
-    if (CL_SUCCESS != result)
+    if (CL_SUCCESS != error)
     {
-        return result;
+        return error;
     }
     memcpy(img_fmt, &clImgFormat, sizeof(cl_image_format));
 
@@ -469,25 +499,57 @@ getCLImageInfoFromVkImageInfo(const VkImageCreateInfo *VulkanImageCreateInfo,
         return CL_INVALID_VALUE;
     }
 
-    result =
-        get2DImageDimensions(VulkanImageCreateInfo, img_fmt, totalImageSize,
-                             img_desc->image_width, img_desc->image_height);
-    if (CL_SUCCESS != result)
+    VkExtent3D max_ext = { 0, 0, 0 };
+
+    size_t width = 0, height = 0, depth = 0;
+    if (img_desc->image_type == CL_MEM_OBJECT_IMAGE3D)
     {
-        throw std::runtime_error("get2DImageDimensions failed!!!");
+        error = clGetDeviceInfo(device, CL_DEVICE_IMAGE3D_MAX_WIDTH,
+                                sizeof(width), &width, NULL);
+        test_error(error, "Unable to get CL_DEVICE_IMAGE3D_MAX_WIDTH");
+        error = clGetDeviceInfo(device, CL_DEVICE_IMAGE3D_MAX_HEIGHT,
+                                sizeof(height), &height, NULL);
+        test_error(error, "Unable to get CL_DEVICE_IMAGE3D_MAX_HEIGHT");
+        error = clGetDeviceInfo(device, CL_DEVICE_IMAGE3D_MAX_DEPTH,
+                                sizeof(depth), &depth, NULL);
+        test_error(error, "Unable to get CL_DEVICE_IMAGE3D_MAX_DEPTH");
+    }
+    else
+    {
+        error = clGetDeviceInfo(device, CL_DEVICE_IMAGE2D_MAX_WIDTH,
+                                sizeof(width), &width, NULL);
+        test_error(error, "Unable to get CL_DEVICE_IMAGE2D_MAX_WIDTH");
+        error = clGetDeviceInfo(device, CL_DEVICE_IMAGE2D_MAX_HEIGHT,
+                                sizeof(height), &height, NULL);
+        test_error(error, "Unable to get CL_DEVICE_IMAGE2D_MAX_HEIGHT");
+    }
+
+    max_ext.depth = depth;
+    max_ext.height = height;
+    max_ext.width = width;
+
+    // If image_row_pitch is zero and the image is created from an external
+    // memory handle, then the image row pitch is implementation-defined
+    img_desc->image_row_pitch = 0;
+    // If image_slice_pitch is zero and the image is created from an external
+    // memory handle, then the image slice pitch is implementation-defined
+    img_desc->image_slice_pitch = 0;
+
+    error = getImageDimensions(VulkanImageCreateInfo, img_fmt, img_desc,
+                               max_ext, layout);
+    if (CL_SUCCESS != error)
+    {
+        throw std::runtime_error("getImageDimensions failed!!!");
     }
 
     img_desc->image_depth =
         static_cast<size_t>(VulkanImageCreateInfo->extent.depth);
     img_desc->image_array_size = 0;
-    img_desc->image_row_pitch = 0; // Row pitch set to zero as host_ptr is NULL
-    img_desc->image_slice_pitch =
-        img_desc->image_row_pitch * img_desc->image_height;
     img_desc->num_mip_levels = 0;
     img_desc->num_samples = 0;
     img_desc->buffer = NULL;
 
-    return result;
+    return error;
 }
 
 cl_int check_external_memory_handle_type(
@@ -529,8 +591,8 @@ cl_int check_external_memory_handle_type(
     return CL_INVALID_VALUE;
 }
 
-cl_int check_external_semaphore_handle_type(
-    cl_device_id deviceID,
+void check_external_semaphore_handle_type(
+    cl_device_id device,
     cl_external_semaphore_handle_type_khr requiredHandleType,
     cl_device_info queryParamName)
 {
@@ -540,8 +602,8 @@ cl_int check_external_semaphore_handle_type(
     cl_int errNum = CL_SUCCESS;
 
     errNum =
-        clGetDeviceInfo(deviceID, queryParamName, 0, NULL, &handle_type_size);
-    test_error(errNum, "clGetDeviceInfo failed");
+        clGetDeviceInfo(device, queryParamName, 0, NULL, &handle_type_size);
+    ASSERT_SUCCESS(errNum, "clGetDeviceInfo");
 
     if (handle_type_size == 0)
     {
@@ -549,31 +611,37 @@ cl_int check_external_semaphore_handle_type(
                   queryParamName == CL_DEVICE_SEMAPHORE_IMPORT_HANDLE_TYPES_KHR
                       ? "importing"
                       : "exporting");
-        return CL_INVALID_VALUE;
+
+        throw std::runtime_error("");
     }
 
     handle_type =
         (cl_external_semaphore_handle_type_khr *)malloc(handle_type_size);
 
-    errNum = clGetDeviceInfo(deviceID, queryParamName, handle_type_size,
+    errNum = clGetDeviceInfo(device, queryParamName, handle_type_size,
                              handle_type, NULL);
+    ASSERT_SUCCESS(errNum, "clGetDeviceInfo");
 
-    test_error(
-        errNum,
-        "Unable to query supported device semaphore handle types list\n");
-
+    bool found = false;
     for (i = 0; i < handle_type_size; i++)
     {
         if (requiredHandleType == handle_type[i])
         {
-            return CL_SUCCESS;
+            found = true;
+            break;
         }
     }
-    log_error("cl_khr_external_semaphore extension is missing support for %d\n",
-              requiredHandleType);
 
-    return CL_INVALID_VALUE;
+    if (!found)
+    {
+        log_error("cl_khr_external_semaphore extension is missing support for "
+                  "handle type %d\n",
+                  requiredHandleType);
+
+        throw std::runtime_error("");
+    }
 }
+
 clExternalMemory::clExternalMemory() {}
 
 clExternalMemory::clExternalMemory(const clExternalMemory &externalMemory)
@@ -705,13 +773,17 @@ clExternalMemoryImage::clExternalMemoryImage(
     std::vector<cl_mem_properties> extMemProperties1;
     cl_device_id devList[] = { deviceId, NULL };
 
-    VulkanImageTiling vulkanImageTiling =
-        vkClExternalMemoryHandleTilingAssumption(
-            deviceId, externalMemoryHandleType, &errcode_ret);
+    auto vulkanImageTiling = vkClExternalMemoryHandleTilingAssumption(
+        deviceId, externalMemoryHandleType, &errcode_ret);
     if (CL_SUCCESS != errcode_ret)
     {
         throw std::runtime_error("Failed to query OpenCL tiling mode");
     }
+    if (vulkanImageTiling == std::nullopt)
+    {
+        throw std::runtime_error(
+            "Could not find image tiling supported by both Vulkan and OpenCL");
+    }
 
 #ifdef _WIN32
     if (!is_extension_available(devList[0], "cl_khr_external_memory_win32"))
@@ -796,7 +868,7 @@ clExternalMemoryImage::clExternalMemoryImage(
         image2D.getVkImageCreateInfo();
 
     errcode_ret = getCLImageInfoFromVkImageInfo(
-        &VulkanImageCreateInfo, image2D.getSize(), &img_format, &image_desc);
+        deviceId, &VulkanImageCreateInfo, &img_format, &image_desc);
     if (CL_SUCCESS != errcode_ret)
     {
         throw std::runtime_error("getCLImageInfoFromVkImageInfo failed\n");
@@ -855,9 +927,15 @@ clExternalImportableSemaphore::clExternalImportableSemaphore(
     cl_device_id deviceId)
     : m_deviceSemaphore(semaphore)
 {
-
     cl_int err = 0;
     cl_device_id devList[] = { deviceId, NULL };
+    cl_external_semaphore_handle_type_khr clSemaphoreHandleType =
+        getCLSemaphoreTypeFromVulkanType(externalSemaphoreHandleType);
+
+    check_external_semaphore_handle_type(
+        deviceId, clSemaphoreHandleType,
+        CL_DEVICE_SEMAPHORE_IMPORT_HANDLE_TYPES_KHR);
+
     m_externalHandleType = externalSemaphoreHandleType;
     m_externalSemaphore = nullptr;
     m_device = deviceId;
@@ -875,8 +953,6 @@ clExternalImportableSemaphore::clExternalImportableSemaphore(
             ASSERT(0);
 #else
             fd = (int)semaphore.getHandle(externalSemaphoreHandleType);
-            err = check_external_semaphore_handle_type(
-                devList[0], CL_SEMAPHORE_HANDLE_OPAQUE_FD_KHR);
             sema_props.push_back(
                 (cl_semaphore_properties_khr)CL_SEMAPHORE_HANDLE_OPAQUE_FD_KHR);
             sema_props.push_back((cl_semaphore_properties_khr)fd);
@@ -888,8 +964,6 @@ clExternalImportableSemaphore::clExternalImportableSemaphore(
             ASSERT(0);
 #else
             handle = semaphore.getHandle(externalSemaphoreHandleType);
-            err = check_external_semaphore_handle_type(
-                devList[0], CL_SEMAPHORE_HANDLE_OPAQUE_WIN32_KHR);
             sema_props.push_back((cl_semaphore_properties_khr)
                                      CL_SEMAPHORE_HANDLE_OPAQUE_WIN32_KHR);
             sema_props.push_back((cl_semaphore_properties_khr)handle);
@@ -903,8 +977,6 @@ clExternalImportableSemaphore::clExternalImportableSemaphore(
             const std::wstring &name = semaphore.getName();
             if (name.size())
             {
-                err = check_external_semaphore_handle_type(
-                    devList[0], CL_SEMAPHORE_HANDLE_OPAQUE_WIN32_NAME_KHR);
                 sema_props.push_back(
                     (cl_semaphore_properties_khr)
                         CL_SEMAPHORE_HANDLE_OPAQUE_WIN32_NAME_KHR);
@@ -924,21 +996,17 @@ clExternalImportableSemaphore::clExternalImportableSemaphore(
             ASSERT(0);
 #else
             handle = semaphore.getHandle(externalSemaphoreHandleType);
-            err = check_external_semaphore_handle_type(
-                devList[0], CL_SEMAPHORE_HANDLE_OPAQUE_WIN32_KMT_KHR);
             sema_props.push_back((cl_semaphore_properties_khr)
                                      CL_SEMAPHORE_HANDLE_OPAQUE_WIN32_KMT_KHR);
             sema_props.push_back((cl_semaphore_properties_khr)handle);
 #endif
             break;
         case VULKAN_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD:
-            err = check_external_semaphore_handle_type(
-                devList[0], CL_SEMAPHORE_HANDLE_SYNC_FD_KHR);
-
             sema_props.push_back(static_cast<cl_semaphore_properties_khr>(
                 CL_SEMAPHORE_HANDLE_SYNC_FD_KHR));
             sema_props.push_back(static_cast<cl_semaphore_properties_khr>(-1));
             break;
+
         default:
             log_error("Unsupported external memory handle type\n");
             ASSERT(0);
@@ -1013,11 +1081,15 @@ clExternalExportableSemaphore::clExternalExportableSemaphore(
     cl_device_id deviceId)
     : m_deviceSemaphore(semaphore)
 {
-
     cl_int err = 0;
     cl_device_id devList[] = { deviceId, NULL };
     cl_external_semaphore_handle_type_khr clSemaphoreHandleType =
         getCLSemaphoreTypeFromVulkanType(externalSemaphoreHandleType);
+
+    check_external_semaphore_handle_type(
+        deviceId, clSemaphoreHandleType,
+        CL_DEVICE_SEMAPHORE_EXPORT_HANDLE_TYPES_KHR);
+
     m_externalHandleType = externalSemaphoreHandleType;
     m_externalSemaphore = nullptr;
     m_device = deviceId;
@@ -1191,6 +1263,10 @@ cl_external_memory_handle_type_khr vkToOpenCLExternalMemoryHandleType(
 {
     switch (vkExternalMemoryHandleType)
     {
+        default:
+        case VULKAN_EXTERNAL_MEMORY_HANDLE_TYPE_NONE:
+            log_error("Unexpected external memory handle type\n");
+            return 0;
         case VULKAN_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD:
             return CL_EXTERNAL_MEMORY_HANDLE_OPAQUE_FD_KHR;
         case VULKAN_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_NT:
@@ -1198,17 +1274,17 @@ cl_external_memory_handle_type_khr vkToOpenCLExternalMemoryHandleType(
         case VULKAN_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_KMT:
         case VULKAN_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_NT_KMT:
             return CL_EXTERNAL_MEMORY_HANDLE_OPAQUE_WIN32_KMT_KHR;
-        case VULKAN_EXTERNAL_MEMORY_HANDLE_TYPE_NONE: return 0;
+        case VULKAN_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_WIN32_NT_NAME:
+            return CL_EXTERNAL_MEMORY_HANDLE_OPAQUE_WIN32_NAME_KHR;
     }
     return 0;
 }
 
-VulkanImageTiling vkClExternalMemoryHandleTilingAssumption(
+std::optional<VulkanImageTiling> vkClExternalMemoryHandleTilingAssumption(
     cl_device_id deviceId,
     VulkanExternalMemoryHandleType vkExternalMemoryHandleType, int *error_ret)
 {
     size_t size = 0;
-    VulkanImageTiling mode = VULKAN_IMAGE_TILING_OPTIMAL;
 
     assert(error_ret
            != nullptr); // errcode_ret is not optional, it must be checked
@@ -1219,12 +1295,12 @@ VulkanImageTiling vkClExternalMemoryHandleTilingAssumption(
         0, nullptr, &size);
     if (*error_ret != CL_SUCCESS)
     {
-        return mode;
+        return std::nullopt;
     }
 
     if (size == 0)
     {
-        return mode;
+        return std::nullopt;
     }
 
     std::vector<cl_external_memory_handle_type_khr> assume_linear_types(
@@ -1236,7 +1312,7 @@ VulkanImageTiling vkClExternalMemoryHandleTilingAssumption(
         size, assume_linear_types.data(), nullptr);
     if (*error_ret != CL_SUCCESS)
     {
-        return mode;
+        return std::nullopt;
     }
 
     if (std::find(
@@ -1244,8 +1320,8 @@ VulkanImageTiling vkClExternalMemoryHandleTilingAssumption(
             vkToOpenCLExternalMemoryHandleType(vkExternalMemoryHandleType))
         != assume_linear_types.end())
     {
-        mode = VULKAN_IMAGE_TILING_LINEAR;
+        return VULKAN_IMAGE_TILING_LINEAR;
     }
 
-    return mode;
+    return std::nullopt;
 }

test_conformance/common/vulkan_wrapper/opencl_vulkan_wrapper.hpp

@@ -26,22 +26,7 @@
 #include <OpenCL/cl_ext.h>
 #endif
 
-#define CREATE_OPENCL_SEMAPHORE(clSemaphore, vkSemaphore, ctx, handleType,     \
-                                devIdx, createExportable)                      \
-    if (!(createExportable                                                     \
-          && (check_external_semaphore_handle_type(                            \
-                  devIdx, getCLSemaphoreTypeFromVulkanType(handleType),        \
-                  CL_DEVICE_SEMAPHORE_EXPORT_HANDLE_TYPES_KHR)                 \
-              == CL_SUCCESS)))                                                 \
-    {                                                                          \
-        clSemaphore = new clExternalImportableSemaphore(vkSemaphore, ctx,      \
-                                                        handleType, devIdx);   \
-    }                                                                          \
-    else                                                                       \
-    {                                                                          \
-        clSemaphore = new clExternalExportableSemaphore(vkSemaphore, ctx,      \
-                                                        handleType, devIdx);   \
-    }
+#include <optional>
 
 typedef cl_semaphore_khr (*pfnclCreateSemaphoreWithPropertiesKHR)(
     cl_context context, cl_semaphore_properties_khr *sema_props,
@@ -86,16 +71,17 @@ extern pfnclEnqueueReleaseExternalMemObjectsKHR
 extern pfnclReleaseSemaphoreKHR clReleaseSemaphoreKHRptr;
 extern pfnclReImportSemaphoreSyncFdKHR pfnclReImportSemaphoreSyncFdKHRptr;
 
-cl_int getCLImageInfoFromVkImageInfo(const VkImageCreateInfo *, size_t,
-                                     cl_image_format *, cl_image_desc *);
+cl_int
+getCLImageInfoFromVkImageInfo(const cl_device_id, const VkImageCreateInfo *,
+                              cl_image_format *, cl_image_desc *,
+                              const VkSubresourceLayout *layout = nullptr);
 cl_int check_external_memory_handle_type(
     cl_device_id deviceID,
     cl_external_memory_handle_type_khr requiredHandleType);
-cl_int check_external_semaphore_handle_type(
-    cl_device_id deviceID,
+void check_external_semaphore_handle_type(
+    cl_device_id device,
     cl_external_semaphore_handle_type_khr requiredHandleType,
-    cl_device_info queryParamName =
-        CL_DEVICE_SEMAPHORE_IMPORT_HANDLE_TYPES_KHR);
+    cl_device_info queryParamName);
 cl_int setMaxImageDimensions(cl_device_id deviceID, size_t &width,
                              size_t &height);
 
@@ -187,8 +173,8 @@ public:
 extern void init_cl_vk_ext(cl_platform_id, cl_uint num_devices,
                            cl_device_id *deviceIds);
 
-VulkanImageTiling vkClExternalMemoryHandleTilingAssumption(
+std::optional<VulkanImageTiling> vkClExternalMemoryHandleTilingAssumption(
     cl_device_id deviceId,
     VulkanExternalMemoryHandleType vkExternalMemoryHandleType, int *error_ret);
 
-#endif // _opencl_vulkan_wrapper_hpp_
+#endif // _opencl_vulkan_wrapper_hpp_

test_conformance/common/vulkan_wrapper/vulkan_api_list.hpp

@@ -102,6 +102,8 @@
     VK_FUNC_DECL(vkImportSemaphoreFdKHR)                                       \
     VK_FUNC_DECL(vkGetPhysicalDeviceExternalSemaphorePropertiesKHR)            \
     VK_FUNC_DECL(vkGetImageSubresourceLayout)                                  \
+    VK_FUNC_DECL(vkCreateDebugUtilsMessengerEXT)                               \
+    VK_FUNC_DECL(vkDestroyDebugUtilsMessengerEXT)                              \
     VK_FUNC_DECL(vkGetPhysicalDeviceExternalBufferProperties)
 #define VK_WINDOWS_FUNC_LIST                                                   \
     VK_FUNC_DECL(vkGetMemoryWin32HandleKHR)                                    \
@@ -203,6 +205,8 @@
 #define vkGetSemaphoreWin32HandleKHR _vkGetSemaphoreWin32HandleKHR
 #define vkImportSemaphoreWin32HandleKHR _vkImportSemaphoreWin32HandleKHR
 #define vkGetImageSubresourceLayout _vkGetImageSubresourceLayout
+#define vkCreateDebugUtilsMessengerEXT _vkCreateDebugUtilsMessengerEXT
+#define vkDestroyDebugUtilsMessengerEXT _vkDestroyDebugUtilsMessengerEXT
 #define vkGetPhysicalDeviceExternalBufferProperties                            \
     _vkGetPhysicalDeviceExternalBufferProperties
 

test_conformance/common/vulkan_wrapper/vulkan_utility.cpp

@@ -32,13 +32,13 @@
 #define BUFFERSIZE 3000
 
 
-const VulkanInstance &getVulkanInstance()
+const VulkanInstance &getVulkanInstance(bool useValidationLayers)
 {
-    static VulkanInstance instance;
+    static VulkanInstance instance(useValidationLayers);
     return instance;
 }
 
-const VulkanPhysicalDevice &getVulkanPhysicalDevice()
+const VulkanPhysicalDevice &getVulkanPhysicalDevice(bool useValidationLayers)
 {
     size_t pdIdx = 0;
     cl_int errNum = 0;
@@ -47,7 +47,7 @@ const VulkanPhysicalDevice &getVulkanPhysicalDevice()
     cl_uint num_devices = 0;
     cl_uint device_no = 0;
     const size_t bufsize = BUFFERSIZE;
-    const VulkanInstance &instance = getVulkanInstance();
+    const VulkanInstance &instance = getVulkanInstance(useValidationLayers);
     const VulkanPhysicalDeviceList &physicalDeviceList =
         instance.getPhysicalDeviceList();
 
@@ -112,12 +112,13 @@ const VulkanPhysicalDevice &getVulkanPhysicalDevice()
 }
 
 const VulkanPhysicalDevice &
-getAssociatedVulkanPhysicalDevice(cl_device_id deviceId)
+getAssociatedVulkanPhysicalDevice(cl_device_id deviceId,
+                                  bool useValidationLayers)
 {
     size_t pdIdx;
     cl_int errNum = 0;
     cl_uchar uuid[CL_UUID_SIZE_KHR];
-    const VulkanInstance &instance = getVulkanInstance();
+    const VulkanInstance &instance = getVulkanInstance(useValidationLayers);
     const VulkanPhysicalDeviceList &physicalDeviceList =
         instance.getPhysicalDeviceList();
 
@@ -188,10 +189,10 @@ getVulkanMemoryType(const VulkanDevice &device,
     return memoryTypeList[mtIdx];
 }
 
-bool checkVkSupport()
+bool checkVkSupport(bool useValidationLayers)
 {
     bool result = true;
-    const VulkanInstance &instance = getVulkanInstance();
+    const VulkanInstance &instance = getVulkanInstance(useValidationLayers);
     const VulkanPhysicalDeviceList &physicalDeviceList =
         instance.getPhysicalDeviceList();
     if (physicalDeviceList() == NULL)
@@ -711,6 +712,7 @@ operator<<(std::ostream &os,
 {
     switch (externalMemoryHandleType)
     {
+        default:
         case VULKAN_EXTERNAL_MEMORY_HANDLE_TYPE_NONE: return os << "None";
         case VULKAN_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD:
             return os << "Opaque file descriptor";
@@ -731,6 +733,7 @@ operator<<(std::ostream &os,
 {
     switch (externalSemaphoreHandleType)
     {
+        default:
         case VULKAN_EXTERNAL_SEMAPHORE_HANDLE_TYPE_NONE: return os << "None";
         case VULKAN_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD:
             return os << "Opaque file descriptor";

test_conformance/common/vulkan_wrapper/vulkan_utility.hpp

@@ -30,10 +30,12 @@
 #define ROUND_UP(n, multiple)                                                  \
     (((n) + (multiple)-1) - ((((n) + (multiple)-1)) % (multiple)))
 
-const VulkanInstance& getVulkanInstance();
-const VulkanPhysicalDevice& getVulkanPhysicalDevice();
+const VulkanInstance& getVulkanInstance(bool useValidationLayers = false);
 const VulkanPhysicalDevice&
-getAssociatedVulkanPhysicalDevice(cl_device_id deviceId);
+getVulkanPhysicalDevice(bool useValidationLayers = false);
+const VulkanPhysicalDevice&
+getAssociatedVulkanPhysicalDevice(cl_device_id deviceId,
+                                  bool useValidationLayers = false);
 const VulkanQueueFamily& getVulkanQueueFamily(
     const VulkanPhysicalDevice& physicalDevice = getVulkanPhysicalDevice(),
     uint32_t queueFlags = VULKAN_QUEUE_FLAG_GRAPHICS
@@ -41,7 +43,7 @@ const VulkanQueueFamily& getVulkanQueueFamily(
 const VulkanMemoryType&
 getVulkanMemoryType(const VulkanDevice& device,
                     VulkanMemoryTypeProperty memoryTypeProperty);
-bool checkVkSupport();
+bool checkVkSupport(bool useValidationLayers = false);
 const VulkanQueueFamilyList& getEmptyVulkanQueueFamilyList();
 const VulkanDescriptorSetLayoutList& getEmptyVulkanDescriptorSetLayoutList();
 const VulkanQueueFamilyToQueueCountMap&

test_conformance/common/vulkan_wrapper/vulkan_wrapper.cpp

@@ -48,16 +48,38 @@ VK_WINDOWS_FUNC_LIST
 
 #define CHECK_VK(call)                                                         \
     if (call != VK_SUCCESS) return call;
+
+static VKAPI_ATTR VkBool32 VKAPI_CALL logCallback(
+    VkDebugUtilsMessageSeverityFlagBitsEXT messageSeverity,
+    VkDebugUtilsMessageTypeFlagsEXT messageType,
+    const VkDebugUtilsMessengerCallbackDataEXT *pCallbackData, void *pUserData)
+{
+    switch (messageSeverity)
+    {
+        case VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT:
+            log_error("Vulkan validation layer: %s\n", pCallbackData->pMessage);
+            break;
+        default:
+            log_info("Vulkan validation layer: %s\n", pCallbackData->pMessage);
+            break;
+    }
+
+    return VK_FALSE;
+}
+
 ///////////////////////////////////
 // VulkanInstance implementation //
 ///////////////////////////////////
 
 VulkanInstance::VulkanInstance(const VulkanInstance &instance)
     : m_vkInstance(instance.m_vkInstance),
-      m_physicalDeviceList(instance.m_physicalDeviceList)
+      m_physicalDeviceList(instance.m_physicalDeviceList),
+      m_useValidationLayers(instance.m_useValidationLayers),
+      m_validationLayers(instance.m_validationLayers)
 {}
 
-VulkanInstance::VulkanInstance(): m_vkInstance(VK_NULL_HANDLE)
+VulkanInstance::VulkanInstance(bool useValidationLayers)
+    : m_vkInstance(VK_NULL_HANDLE), m_useValidationLayers(useValidationLayers)
 {
 #if defined(__linux__) && !defined(__ANDROID__)
     char *glibcVersion = strdup(gnu_get_libc_version());
@@ -130,6 +152,35 @@ VulkanInstance::VulkanInstance(): m_vkInstance(VK_NULL_HANDLE)
     VK_GET_NULL_INSTANCE_PROC_ADDR(vkCreateInstance);
 #undef VK_GET_NULL_INSTANCE_PROC_ADDR
 
+    if (m_useValidationLayers)
+    {
+        uint32_t layerCount = 0;
+        vkEnumerateInstanceLayerProperties(&layerCount, nullptr);
+
+        std::vector<VkLayerProperties> layers(layerCount);
+        vkEnumerateInstanceLayerProperties(&layerCount, layers.data());
+
+        for (auto it = m_validationLayers.begin();
+             it != m_validationLayers.end();)
+        {
+            bool found = false;
+            for (const auto &layerProps : layers)
+                if (strcmp(*it, layerProps.layerName) == 0)
+                {
+                    found = true;
+                    break;
+                }
+            if (!found)
+            {
+                log_info("Vulkan layer not found: %s\n", *it);
+                it = m_validationLayers.erase(it);
+            }
+            else
+                ++it;
+        }
+        m_useValidationLayers = !m_validationLayers.empty();
+    }
+
     VkApplicationInfo vkApplicationInfo = {};
     vkApplicationInfo.sType = VK_STRUCTURE_TYPE_APPLICATION_INFO;
     vkApplicationInfo.pNext = NULL;
@@ -147,6 +198,9 @@ VulkanInstance::VulkanInstance(): m_vkInstance(VK_NULL_HANDLE)
     enabledExtensionNameList.push_back(
         VK_KHR_EXTERNAL_SEMAPHORE_CAPABILITIES_EXTENSION_NAME);
 
+    if (m_useValidationLayers)
+        enabledExtensionNameList.push_back(VK_EXT_DEBUG_UTILS_EXTENSION_NAME);
+
     std::vector<VkExtensionProperties> vkExtensionPropertiesList(
         instanceExtensionPropertiesCount);
     vkEnumerateInstanceExtensionProperties(NULL,
@@ -174,18 +228,51 @@ VulkanInstance::VulkanInstance(): m_vkInstance(VK_NULL_HANDLE)
 
     VkInstanceCreateInfo vkInstanceCreateInfo = {};
     vkInstanceCreateInfo.sType = VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO;
-    vkInstanceCreateInfo.pNext = NULL;
     vkInstanceCreateInfo.flags = 0;
     vkInstanceCreateInfo.pApplicationInfo = &vkApplicationInfo;
-    vkInstanceCreateInfo.enabledLayerCount = 0;
     vkInstanceCreateInfo.ppEnabledLayerNames = NULL;
     vkInstanceCreateInfo.enabledExtensionCount =
         (uint32_t)enabledExtensionNameList.size();
     vkInstanceCreateInfo.ppEnabledExtensionNames =
         enabledExtensionNameList.data();
+    vkInstanceCreateInfo.enabledLayerCount = 0;
+    vkInstanceCreateInfo.pNext = NULL;
+
+    VkDebugUtilsMessengerCreateInfoEXT debugCreateInfo{};
+    if (m_useValidationLayers)
+    {
+        vkInstanceCreateInfo.enabledLayerCount =
+            static_cast<uint32_t>(m_validationLayers.size());
+        vkInstanceCreateInfo.ppEnabledLayerNames = m_validationLayers.data();
+
+        debugCreateInfo.messageType =
+            VK_DEBUG_UTILS_MESSAGE_TYPE_GENERAL_BIT_EXT
+            | VK_DEBUG_UTILS_MESSAGE_TYPE_VALIDATION_BIT_EXT
+            | VK_DEBUG_UTILS_MESSAGE_TYPE_PERFORMANCE_BIT_EXT;
+
+        debugCreateInfo.sType =
+            VK_STRUCTURE_TYPE_DEBUG_UTILS_MESSENGER_CREATE_INFO_EXT;
+        debugCreateInfo.messageSeverity =
+            VK_DEBUG_UTILS_MESSAGE_SEVERITY_WARNING_BIT_EXT
+            | VK_DEBUG_UTILS_MESSAGE_SEVERITY_ERROR_BIT_EXT;
+        debugCreateInfo.pfnUserCallback = logCallback;
+
+        vkInstanceCreateInfo.pNext =
+            (VkDebugUtilsMessengerCreateInfoEXT *)&debugCreateInfo;
+    }
 
     vkCreateInstance(&vkInstanceCreateInfo, NULL, &m_vkInstance);
 
+    if (m_useValidationLayers)
+    {
+        _vkCreateDebugUtilsMessengerEXT =
+            (PFN_vkCreateDebugUtilsMessengerEXT)vkGetInstanceProcAddr(
+                m_vkInstance, "vkCreateDebugUtilsMessengerEXT");
+        if (_vkCreateDebugUtilsMessengerEXT != nullptr)
+            vkCreateDebugUtilsMessengerEXT(m_vkInstance, &debugCreateInfo,
+                                           nullptr, &m_debugMessenger);
+    }
+
 #define VK_FUNC_DECL(name)                                                     \
     _##name = (PFN_##name)vkGetInstanceProcAddr(m_vkInstance, #name);          \
     // ASSERT_NEQ((unsigned long long)name, 0ULL) << "Couldn't obtain address
@@ -228,6 +315,17 @@ VulkanInstance::~VulkanInstance()
             m_physicalDeviceList[pdIdx];
         delete &physicalDevice;
     }
+
+    _vkDestroyDebugUtilsMessengerEXT =
+        (PFN_vkDestroyDebugUtilsMessengerEXT)vkGetInstanceProcAddr(
+            m_vkInstance, "vkDestroyDebugUtilsMessengerEXT");
+
+    if (_vkDestroyDebugUtilsMessengerEXT != nullptr)
+    {
+        vkDestroyDebugUtilsMessengerEXT(m_vkInstance, m_debugMessenger,
+                                        nullptr);
+    }
+
     if (m_vkInstance)
     {
         vkDestroyInstance(m_vkInstance, NULL);
@@ -1741,7 +1839,13 @@ VulkanImage::VulkanImage(
         vkImageCreateInfo.pNext = &vkExternalMemoryImageCreateInfo;
     }
 
-    vkCreateImage(m_device, &vkImageCreateInfo, NULL, &m_vkImage);
+    VkResult vkStatus =
+        vkCreateImage(m_device, &vkImageCreateInfo, NULL, &m_vkImage);
+    if (vkStatus != VK_SUCCESS)
+    {
+        throw std::runtime_error("Error: Failed create image.");
+    }
+
     VulkanImageCreateInfo = vkImageCreateInfo;
 
     VkMemoryDedicatedRequirements vkMemoryDedicatedRequirements = {};
@@ -1870,7 +1974,7 @@ VulkanExtent3D VulkanImage2D::getExtent3D(uint32_t mipLevel) const
     return VulkanExtent3D(width, height, depth);
 }
 
-VkSubresourceLayout VulkanImage2D::getSubresourceLayout() const
+VkSubresourceLayout VulkanImage::getSubresourceLayout() const
 {
     VkImageSubresource subresource = { VK_IMAGE_ASPECT_COLOR_BIT, 0, 0 };
 

test_conformance/common/vulkan_wrapper/vulkan_wrapper.hpp

@@ -25,14 +25,21 @@
 #include <cassert>
 
 class VulkanInstance {
-    friend const VulkanInstance &getVulkanInstance();
+    friend const VulkanInstance &getVulkanInstance(bool useValidationLayers);
 
 protected:
     VkInstance m_vkInstance;
     VulkanPhysicalDeviceList m_physicalDeviceList;
+    VkDebugUtilsMessengerEXT m_debugMessenger;
+    bool m_useValidationLayers;
+    std::vector<const char *> m_validationLayers = {
+        "VK_LAYER_KHRONOS_validation",
+    };
 
-    VulkanInstance();
     VulkanInstance(const VulkanInstance &);
+
+public:
+    VulkanInstance(bool useValidationLayers = false);
     virtual ~VulkanInstance();
 
 public:
@@ -478,12 +485,14 @@ public:
         VulkanExternalMemoryHandleType externalMemoryHandleType =
             VULKAN_EXTERNAL_MEMORY_HANDLE_TYPE_NONE,
         VulkanImageCreateFlag imageCreateFlags = VULKAN_IMAGE_CREATE_FLAG_NONE,
-        VulkanImageTiling imageTiling = VULKAN_IMAGE_TILING_OPTIMAL,
+        VulkanImageTiling imageTiling = VULKAN_IMAGE_TILING_LINEAR,
         VulkanImageUsage imageUsage =
             VULKAN_IMAGE_USAGE_SAMPLED_STORAGE_TRANSFER_SRC_DST,
         VulkanSharingMode sharingMode = VULKAN_SHARING_MODE_EXCLUSIVE);
     virtual ~VulkanImage();
     virtual VulkanExtent3D getExtent3D(uint32_t mipLevel = 0) const;
+    virtual VkSubresourceLayout getSubresourceLayout() const;
+
     VulkanFormat getFormat() const;
     uint32_t getNumMipLevels() const;
     uint32_t getNumLayers() const;
@@ -553,7 +562,6 @@ public:
         VulkanSharingMode sharingMode = VULKAN_SHARING_MODE_EXCLUSIVE);
     virtual ~VulkanImage2D();
     virtual VulkanExtent3D getExtent3D(uint32_t mipLevel = 0) const;
-    virtual VkSubresourceLayout getSubresourceLayout() const;
 
     VulkanImage2D(const VulkanImage2D &image2D);
 };

test_conformance/compiler/testBase.h

@@ -45,27 +45,4 @@
     }
 
 
-// scope guard helper to ensure proper releasing of sub devices
-struct SubDevicesScopeGuarded
-{
-    SubDevicesScopeGuarded(const cl_int dev_count)
-    {
-        sub_devices.resize(dev_count);
-    }
-    ~SubDevicesScopeGuarded()
-    {
-        for (auto &device : sub_devices)
-        {
-            cl_int err = clReleaseDevice(device);
-            if (err != CL_SUCCESS)
-                log_error("\n Releasing sub-device failed \n");
-        }
-    }
-
-    std::vector<cl_device_id> sub_devices;
-};
-
 #endif // _testBase_h
-
-
-

test_conformance/compiler/test_build_helpers.cpp

@@ -55,7 +55,6 @@ const char *sample_kernel_code_two_line[] = {
 "\n"
 "}\n" };
 
-
 const char *sample_kernel_code_bad_multi_line[] = {
 "__kernel void sample_test(__global float *src, __global int *dst)",
 "{",
@@ -87,6 +86,35 @@ __kernel void sample_test_C(__global float *src, __global int *dst)
 }
 )";
 
+const char *sample_multi_kernel_code_AB_with_macro = R"(
+__kernel void sample_test_A(__global float *src, __global int *dst)
+{
+    size_t tid = get_global_id(0);
+    dst[tid] = (int)src[tid];
+}
+#ifdef USE_SAMPLE_TEST_B
+__kernel void sample_test_B(__global float *src, __global int *dst)
+{
+    size_t tid = get_global_id(0);
+    dst[tid] = (int)src[tid];
+}
+#endif
+)";
+
+const char *sample_multi_kernel_code_CD_with_macro = R"(
+__kernel void sample_test_C(__global float *src, __global int *dst)
+{
+    size_t tid = get_global_id(0);
+    dst[tid] = (int)src[tid];
+}
+#ifdef USE_SAMPLE_TEST_D
+__kernel void sample_test_D(__global float *src, __global int *dst)
+{
+    size_t tid = get_global_id(0);
+    dst[tid] = (int)src[tid];
+}
+#endif
+)";
 
 REGISTER_TEST(load_program_source)
 {
@@ -550,6 +578,128 @@ REGISTER_TEST(get_program_info_kernel_names)
     return CL_SUCCESS;
 }
 
+REGISTER_TEST(get_linked_program_info_kernel_names)
+{
+    int error = CL_SUCCESS;
+    size_t total_kernels = 0;
+    size_t kernel_names_len = 0;
+
+    clProgramWrapper program_AB = clCreateProgramWithSource(
+        context, 1, &sample_multi_kernel_code_AB_with_macro, nullptr, &error);
+    test_error(error, "clCreateProgramWithSource failed");
+
+    clProgramWrapper program_CD = clCreateProgramWithSource(
+        context, 1, &sample_multi_kernel_code_CD_with_macro, nullptr, &error);
+    test_error(error, "clCreateProgramWithSource failed");
+
+    clProgramWrapper program = nullptr;
+
+    // 1) Compile and link the programs with the preprocessor macro undefined.
+    //    Query CL_PROGRAM_NUM_KERNELS and check that the correct number is
+    //    returned. Query CL_PROGRAM_KERNEL_NAMES and check that the right
+    //    kernel names are returned.
+    {
+        error =
+            clCompileProgram(program_AB, 1, &device, nullptr, 0, 0, 0, 0, 0);
+        test_error(error, "clCompileProgram failed");
+
+        error =
+            clCompileProgram(program_CD, 1, &device, nullptr, 0, 0, 0, 0, 0);
+        test_error(error, "clCompileProgram failed");
+
+        cl_program progs[] = { program_AB, program_CD };
+        program =
+            clLinkProgram(context, 1, &device, "", 2, progs, 0, 0, &error);
+        test_error(error, "clLinkProgram failed");
+
+        error = clGetProgramInfo(program, CL_PROGRAM_NUM_KERNELS,
+                                 sizeof(size_t), &total_kernels, nullptr);
+        test_error(error, "clGetProgramInfo failed");
+
+        test_assert_error(total_kernels == 2,
+                          "Unexpected clGetProgramInfo result");
+
+        error = clGetProgramInfo(program, CL_PROGRAM_KERNEL_NAMES, 0, nullptr,
+                                 &kernel_names_len);
+        test_error(error, "clGetProgramInfo failed");
+
+        const size_t len = kernel_names_len + 1;
+        std::vector<char> kernel_names(len, '\0');
+        error =
+            clGetProgramInfo(program, CL_PROGRAM_KERNEL_NAMES, kernel_names_len,
+                             kernel_names.data(), &kernel_names_len);
+        test_error(error, "Unable to get kernel names list.");
+        std::string program_names = kernel_names.data();
+
+        std::vector<std::string> expected_names = { "sample_test_A",
+                                                    "sample_test_C" };
+        for (const auto &name : expected_names)
+        {
+            test_assert_error(program_names.find(name) != std::string::npos,
+                              "Unexpected kernel name");
+        }
+
+        std::vector<std::string> unexpected_names = { "sample_test_B",
+                                                      "sample_test_D" };
+        for (const auto &name : unexpected_names)
+        {
+            test_assert_error(program_names.find(name) == std::string::npos,
+                              "Unexpected kernel name");
+        }
+    }
+
+    // 2) Compile and link the programs with the preprocessor macro defined.
+    //    Query CL_PROGRAM_NUM_KERNELS and check that the correct number is
+    //    returned. Query CL_PROGRAM_KERNEL_NAMES and check that the right
+    //    kernel names are returned.
+    {
+        const char *build_options_B = "-DUSE_SAMPLE_TEST_B";
+        error = clCompileProgram(program_AB, 1, &device, build_options_B, 0, 0,
+                                 0, 0, 0);
+        test_error(error, "clCompileProgram failed");
+
+        const char *build_options_D = "-DUSE_SAMPLE_TEST_D";
+        error = clCompileProgram(program_CD, 1, &device, build_options_D, 0, 0,
+                                 0, 0, 0);
+        test_error(error, "clCompileProgram failed");
+
+        cl_program progs[] = { program_AB, program_CD };
+        program =
+            clLinkProgram(context, 1, &device, "", 2, progs, 0, 0, &error);
+        test_error(error, "clLinkProgram failed");
+
+        error = clGetProgramInfo(program, CL_PROGRAM_NUM_KERNELS,
+                                 sizeof(size_t), &total_kernels, nullptr);
+        test_error(error, "clGetProgramInfo failed");
+
+        test_assert_error(total_kernels == 4,
+                          "Unexpected clGetProgramInfo result");
+
+        error = clGetProgramInfo(program, CL_PROGRAM_KERNEL_NAMES, 0, nullptr,
+                                 &kernel_names_len);
+        test_error(error, "clGetProgramInfo failed");
+
+        std::vector<std::string> expected_names = {
+            "sample_test_A", "sample_test_B", "sample_test_C", "sample_test_D"
+        };
+
+        const size_t len = kernel_names_len + 1;
+        std::vector<char> kernel_names(len, '\0');
+        error =
+            clGetProgramInfo(program, CL_PROGRAM_KERNEL_NAMES, kernel_names_len,
+                             kernel_names.data(), &kernel_names_len);
+        test_error(error, "Could not find expected kernel name");
+
+        std::string program_names = kernel_names.data();
+        for (const auto &name : expected_names)
+        {
+            test_assert_error(program_names.find(name) != std::string::npos,
+                              "Unexpected kernel name");
+        }
+    }
+    return TEST_PASS;
+}
+
 REGISTER_TEST(get_program_info_mult_devices)
 {
     size_t size = 0;
@@ -641,7 +791,7 @@ REGISTER_TEST(get_program_info_mult_devices)
     if (program == nullptr)
     {
         log_error("ERROR: Unable to create reference program!\n");
-        return -1;
+        return TEST_FAIL;
     }
 
     err = clGetProgramInfo(program, CL_PROGRAM_NUM_DEVICES, sizeof(num_devices),
@@ -670,7 +820,7 @@ REGISTER_TEST(get_program_info_mult_devices)
             }
         }
         test_error_fail(
-            !found, "Unexpected result returned by CL_CONTEXT_DEVICES query");
+            !found, "Unexpected result returned by CL_PROGRAM_DEVICES query");
     }
 
     return TEST_PASS;

test_conformance/compiler/test_compiler_defines_for_extensions.cpp

@@ -94,7 +94,8 @@ const char *known_extensions[] = {
     "cl_khr_external_memory_dma_buf",
     "cl_khr_command_buffer",
     "cl_khr_command_buffer_mutable_dispatch",
-    "cl_khr_command_buffer_multi_device"
+    "cl_khr_command_buffer_multi_device",
+    "cl_khr_external_memory_android_hardware_buffer"
 };
 // clang-format on
 

test_conformance/compiler/test_opencl_c_versions.cpp

@@ -56,9 +56,9 @@ static int test_CL_DEVICE_OPENCL_C_VERSION(cl_device_id device,
     // For OpenCL 2.x, the minimum required OpenCL C version is OpenCL C 2.0.
     // For other OpenCL versions, the minimum required OpenCL C version is
     // the same as the API version.
-    const Version min_clc_version = api_version == Version(3, 0)
-        ? Version(1, 2)
-        : api_version >= Version(2, 0) ? Version(2, 0) : api_version;
+    const Version min_clc_version = api_version == Version(3, 0) ? Version(1, 2)
+        : api_version >= Version(2, 0)                           ? Version(2, 0)
+                                                                 : api_version;
     if (clc_version < min_clc_version)
     {
         log_error("The minimum required OpenCL C version for API version %s is "
@@ -88,10 +88,10 @@ static int test_CL_DEVICE_OPENCL_C_VERSION(cl_device_id device,
         if (clc_version >= testcase.version)
         {
             clProgramWrapper program;
-            cl_int error =
-                create_single_kernel_helper_create_program_for_device(
-                    context, device, &program, 1, &test_kernel,
-                    testcase.buildOptions);
+            clKernelWrapper kernel;
+            cl_int error = create_single_kernel_helper(
+                context, &program, &kernel, 1, &test_kernel, "test",
+                testcase.buildOptions);
             test_error(error, "Unable to build program!");
 
             log_info("    successfully built program with build options '%s'\n",
@@ -152,9 +152,10 @@ static int test_CL_DEVICE_OPENCL_C_ALL_VERSIONS(cl_device_id device,
             buildOptions += std::to_string(minor);
 
             clProgramWrapper program;
-            error = create_single_kernel_helper_create_program_for_device(
-                context, device, &program, 1, &test_kernel,
-                buildOptions.c_str());
+            clKernelWrapper kernel;
+            error = create_single_kernel_helper(context, &program, &kernel, 1,
+                                                &test_kernel, "test",
+                                                buildOptions.c_str());
             test_error(error, "Unable to build program!");
 
             log_info("    successfully built program with build options '%s'\n",

test_conformance/conversions/CMakeLists.txt

@@ -12,6 +12,6 @@ if("${CLConform_TARGET_ARCH}" STREQUAL "ARM" OR "${CLConform_TARGET_ARCH}" STREQ
     list(APPEND ${MODULE_NAME}_SOURCES fplib.cpp)
 endif()
 
-set_gnulike_module_compile_flags("-Wno-sign-compare -Wno-format")
+set_gnulike_module_compile_flags("-Wno-sign-compare")
 
 include(../CMakeCommon.txt)

test_conformance/conversions/basic_test_conversions.cpp

@@ -68,12 +68,6 @@ cl_context gContext = NULL;
 cl_command_queue gQueue = NULL;
 int gStartTestNumber = -1;
 int gEndTestNumber = 0;
-#if defined(__APPLE__)
-int gTimeResults = 1;
-#else
-int gTimeResults = 0;
-#endif
-int gReportAverageTimes = 0;
 void *gIn = NULL;
 void *gRef = NULL;
 void *gAllowZ = NULL;
@@ -103,8 +97,6 @@ const char **argList = NULL;
 int argCount = 0;
 
 
-double SubtractTime(uint64_t endTime, uint64_t startTime);
-
 cl_half_rounding_mode DataInitInfo::halfRoundingMode = CL_HALF_RTE;
 cl_half_rounding_mode ConversionsTest::defaultHalfRoundingMode = CL_HALF_RTE;
 
@@ -318,7 +310,7 @@ int CalcRefValsPat<InType, OutType, InFP, OutFP>::check_result(void *test,
 {
     const cl_uchar *a = (const cl_uchar *)gAllowZ;
 
-    if (is_half<OutType, OutFP>())
+    if constexpr (is_half<OutType, OutFP>())
     {
         const cl_half *t = (const cl_half *)test;
         const cl_half *c = (const cl_half *)gRef;
@@ -335,7 +327,7 @@ int CalcRefValsPat<InType, OutType, InFP, OutFP>::check_result(void *test,
                 return i + 1;
             }
     }
-    else if (std::is_integral<OutType>::value)
+    else if constexpr (std::is_integral<OutType>::value)
     { // char/uchar/short/ushort/half/int/uint/long/ulong
         const OutType *t = (const OutType *)test;
         const OutType *c = (const OutType *)gRef;
@@ -350,7 +342,7 @@ int CalcRefValsPat<InType, OutType, InFP, OutFP>::check_result(void *test,
                 return i + 1;
             }
     }
-    else if (std::is_same<OutType, cl_float>::value)
+    else if constexpr (std::is_same<OutType, cl_float>::value)
     {
         // cast to integral - from original test
         const cl_uint *t = (const cl_uint *)test;
@@ -904,61 +896,6 @@ int ConversionsTest::DoTest(Type outType, Type inType, SaturationMode sat,
 
     log_info("done.\n");
 
-    if (gTimeResults)
-    {
-        // Kick off tests for the various vector lengths
-        for (vectorSize = gMinVectorSize; vectorSize < gMaxVectorSize;
-             vectorSize++)
-        {
-            size_t workItemCount = blockCount / vectorSizes[vectorSize];
-            if (vectorSizes[vectorSize] * gTypeSizes[outType] < 4)
-                workItemCount /=
-                    4 / (vectorSizes[vectorSize] * gTypeSizes[outType]);
-
-            double sum = 0.0;
-            double bestTime = INFINITY;
-            cl_uint k;
-            for (k = 0; k < PERF_LOOP_COUNT; k++)
-            {
-                uint64_t startTime = conv_test::GetTime();
-                if ((error = conv_test::RunKernel(
-                         writeInputBufferInfo.calcInfo[vectorSize]->kernel,
-                         gInBuffer, gOutBuffers[vectorSize], workItemCount)))
-                {
-                    gFailCount++;
-                    return error;
-                }
-
-                // Make sure OpenCL is done
-                if ((error = clFinish(gQueue)))
-                {
-                    vlog_error("Error %d at clFinish\n", error);
-                    return error;
-                }
-
-                uint64_t endTime = conv_test::GetTime();
-                double time = SubtractTime(endTime, startTime);
-                sum += time;
-                if (time < bestTime) bestTime = time;
-            }
-
-            if (gReportAverageTimes) bestTime = sum / PERF_LOOP_COUNT;
-            double clocksPerOp = bestTime * (double)gDeviceFrequency
-                * gComputeDevices * gSimdSize * 1e6
-                / (workItemCount * vectorSizes[vectorSize]);
-            if (0 == vectorSize)
-                vlog_perf(clocksPerOp, LOWER_IS_BETTER, "clocks / element",
-                          "implicit convert %s -> %s", gTypeNames[inType],
-                          gTypeNames[outType]);
-            else
-                vlog_perf(clocksPerOp, LOWER_IS_BETTER, "clocks / element",
-                          "convert_%s%s%s%s( %s%s )", gTypeNames[outType],
-                          sizeNames[vectorSize], gSaturationNames[sat],
-                          gRoundingModeNames[round], gTypeNames[inType],
-                          sizeNames[vectorSize]);
-        }
-    }
-
     if (gWimpyMode)
         vlog("\tWimp pass");
     else
@@ -978,33 +915,6 @@ int ConversionsTest::DoTest(Type outType, Type inType, SaturationMode sat,
 void memset_pattern4(void *dest, const void *src_pattern, size_t bytes);
 #endif
 
-#if defined(_MSC_VER)
-/* function is defined in "compat.h" */
-#else
-double SubtractTime(uint64_t endTime, uint64_t startTime)
-{
-    uint64_t diff = endTime - startTime;
-    static double conversion = 0.0;
-
-    if (0.0 == conversion)
-    {
-#if defined(__APPLE__)
-        mach_timebase_info_data_t info = { 0, 0 };
-        kern_return_t err = mach_timebase_info(&info);
-        if (0 == err)
-            conversion = 1e-9 * (double)info.numer / (double)info.denom;
-#else
-        // This function consumes output from GetTime() above, and converts the
-        // time to secionds.
-#warning need accurate ticks to seconds conversion factor here. Times are invalid.
-#endif
-    }
-
-    // strictly speaking we should also be subtracting out timer latency here
-    return conversion * (double)diff;
-}
-#endif
-
 void MapResultValuesComplete(const std::unique_ptr<CalcRefValsBase> &ptr);
 
 void CL_CALLBACK CalcReferenceValuesComplete(cl_event e, cl_int status,
@@ -1352,20 +1262,6 @@ cl_int PrepareReference(cl_uint job_id, cl_uint thread_id, void *p)
     return CL_SUCCESS;
 }
 
-uint64_t GetTime(void)
-{
-#if defined(__APPLE__)
-    return mach_absolute_time();
-#elif defined(_MSC_VER)
-    return ReadTime();
-#else
-    // mach_absolute_time is a high precision timer with precision < 1
-    // microsecond.
-#warning need accurate clock here.  Times are invalid.
-    return 0;
-#endif
-}
-
 // Note: not called reentrantly
 void WriteInputBufferComplete(void *data)
 {

test_conformance/conversions/basic_test_conversions.h

@@ -86,8 +86,6 @@ extern int gSkipTesting;
 extern int gMinVectorSize;
 extern int gMaxVectorSize;
 extern int gForceFTZ;
-extern int gTimeResults;
-extern int gReportAverageTimes;
 extern int gStartTestNumber;
 extern int gEndTestNumber;
 extern int gIsRTZ;

test_conformance/conversions/test_conversions.cpp

@@ -226,8 +226,6 @@ static int ParseArgs(int argc, const char **argv)
                         gForceFTZ ^= 1;
                         gForceHalfFTZ ^= 1;
                         break;
-                    case 't': gTimeResults ^= 1; break;
-                    case 'a': gReportAverageTimes ^= 1; break;
                     case '1':
                         if (arg[1] == '6')
                         {

test_conformance/extensions/CMakeLists.txt

@@ -8,6 +8,7 @@ add_subdirectory( cl_khr_dx9_media_sharing )
 add_subdirectory( cl_khr_external_memory_dma_buf )
 add_subdirectory( cl_khr_semaphore )
 add_subdirectory( cl_khr_kernel_clock )
+add_subdirectory( cl_ext_buffer_device_address )
 if(VULKAN_IS_SUPPORTED)
     add_subdirectory( cl_khr_external_semaphore )
 endif()

test_conformance/extensions/cl_ext_buffer_device_address/CMakeLists.txt

@@ -0,0 +1,7 @@
+set(MODULE_NAME CL_EXT_BUFFER_DEVICE_ADDRESS)
+
+set(${MODULE_NAME}_SOURCES
+    main.cpp buffer_device_address.cpp
+)
+
+include(../../CMakeCommon.txt)

test_conformance/extensions/cl_ext_buffer_device_address/buffer_device_address.cpp

@@ -0,0 +1,437 @@
+// Copyright (c) 2024 The Khronos Group Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//    http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+
+#include "harness/typeWrappers.h"
+#include <cinttypes>
+
+#define BUF_SIZE 1024
+#define BUF_SIZE_STR "1024"
+
+namespace {
+
+static const char *program_source =
+    R"raw(
+    // A kernel that gets the device-seen address of the buffer.
+    __kernel void get_addr (__global int *buffer,
+                          __global ulong* addr) {
+      for (int i = 0; i < BUF_SIZE; ++i)
+      buffer[i] += 1;
+      *addr = (ulong)buffer;
+    }
+
+    // A kernel that accesses another buffer indirectly.
+    __kernel void indirect_access (__global long* in_addr,
+                                 __global int* out) {
+      *out = **(int __global* __global*)in_addr;
+    }
+
+    // A kernel that gets passed a pointer to a middle of a buffer,
+    // with the data _before_ the passed pointer. Tests the property
+    // of sub-buffers to synchronize the whole parent buffer when
+    // using the CL_MEM_BUFFER_DEVICE_ADDRESS flag.
+    __kernel void ptr_arith (__global int* in_addr,
+                           __global int* out) {
+      *out = *(in_addr - 2);
+    }
+    )raw";
+
+class BufferDeviceAddressTest {
+
+public:
+    BufferDeviceAddressTest(cl_device_id device, cl_context context,
+                            cl_command_queue queue,
+                            cl_mem_properties address_type)
+        : device(device), context(context), queue(queue),
+          address_type(address_type)
+    {}
+
+    bool Skip()
+    {
+        cl_int error = 0;
+
+        cl_mem_properties buf_props[] = { address_type, CL_TRUE, 0 };
+        clMemWrapper TempBuffer = clCreateBufferWithProperties(
+            context, buf_props, CL_MEM_READ_WRITE,
+            (size_t)BUF_SIZE * sizeof(cl_int), nullptr, &error);
+        return (error != CL_SUCCESS);
+    }
+
+    cl_int RunTest()
+    {
+        cl_int error;
+
+        clProgramWrapper program;
+        clKernelWrapper get_addr_kernel, indirect_access_kernel,
+            ptr_arith_kernel;
+        clMemWrapper dev_addr_buffer, buffer_long, buffer_int,
+            dev_addr_no_host_buffer;
+
+        error = create_single_kernel_helper(context, &program, &get_addr_kernel,
+                                            1, &program_source, "get_addr",
+                                            "-DBUF_SIZE=" BUF_SIZE_STR);
+        test_error(error, "Failed to create program with source\n");
+
+        indirect_access_kernel =
+            clCreateKernel(program, "indirect_access", &error);
+        test_error(error, "Failed to create kernel indirect_access\n");
+
+        ptr_arith_kernel = clCreateKernel(program, "ptr_arith", &error);
+        test_error(error, "Failed to create kernel ptr_arith\n");
+
+        buffer_long = clCreateBuffer(context, CL_MEM_READ_WRITE,
+                                     sizeof(cl_ulong), nullptr, &error);
+        test_error(error, "clCreateBuffer failed\n");
+
+        buffer_int = clCreateBuffer(context, CL_MEM_READ_WRITE, sizeof(cl_int),
+                                    nullptr, &error);
+        test_error(error, "clCreateBuffer failed\n");
+
+        // Test a buffer with hostptr copied data
+        cl_mem_properties buf_props[] = { address_type, CL_TRUE, 0 };
+        dev_addr_buffer = clCreateBufferWithProperties(
+            context, buf_props, CL_MEM_READ_WRITE | CL_MEM_COPY_HOST_PTR,
+            sizeof(cl_int) * BUF_SIZE, BufferHost, &error);
+        test_error(error, "clCreateBuffer with device address 1 failed\n");
+
+        if (test_buffer(dev_addr_buffer, buffer_long, get_addr_kernel)
+            != TEST_PASS)
+            test_fail("test_buffer_host failed\n");
+
+        // Test a buffer which doesn't have any hostptr associated with it.
+        dev_addr_no_host_buffer = clCreateBufferWithProperties(
+            context, buf_props, CL_MEM_READ_WRITE, sizeof(cl_int) * BUF_SIZE,
+            nullptr, &error);
+        test_error(error, "clCreateBuffer with device address 2 failed\n");
+
+        if (test_buffer(dev_addr_no_host_buffer, buffer_long, get_addr_kernel)
+            != TEST_PASS)
+            test_fail("test_buffer_no_host failed\n");
+
+        // Test a buffer passed indirectly
+        if (test_indirect_buffer(dev_addr_buffer, buffer_long, buffer_int,
+                                 indirect_access_kernel)
+            != TEST_PASS)
+            test_fail("test_indirect_buffer failed\n");
+
+        if (test_set_kernel_arg(dev_addr_buffer, buffer_int, ptr_arith_kernel)
+            != TEST_PASS)
+            test_fail("test_set_kernel_arg failed\n");
+
+        if (test_svm_buffer() == TEST_FAIL)
+            test_fail("test_svm_buffer failed\n");
+
+        return TEST_PASS;
+    }
+
+private:
+    int BufferHost[BUF_SIZE];
+    size_t global_size_one[3] = { 1, 1, 1 };
+    cl_device_id device;
+    cl_context context;
+    cl_command_queue queue;
+    cl_mem_properties address_type;
+
+    int check_device_address_from_api(cl_mem buf,
+                                      cl_mem_device_address_ext &Addr)
+    {
+        Addr = 0;
+        cl_int error = clGetMemObjectInfo(buf, CL_MEM_DEVICE_ADDRESS_EXT,
+                                          sizeof(Addr), &Addr, NULL);
+        test_error(error,
+                   "clGetMemObjectInfo(CL_MEM_DEVICE_ADDRESS_EXT) failed\n");
+        if (Addr == 0)
+        {
+            print_error(error,
+                        "clGetMemObjectInfo(CL_MEM_DEVICE_ADDRESS_EXT) "
+                        "returned 0 as address\n");
+            return CL_INVALID_VALUE;
+        }
+        return CL_SUCCESS;
+    }
+
+    int test_svm_buffer()
+    {
+        clSVMWrapper svm_buffer;
+        clMemWrapper buffer;
+        cl_int error = 0;
+
+        cl_device_svm_capabilities svm_caps = 0;
+        error = clGetDeviceInfo(device, CL_DEVICE_SVM_CAPABILITIES,
+                                sizeof(svm_caps), &svm_caps, NULL);
+        if (error != CL_SUCCESS)
+        {
+            print_missing_feature(error,
+                                  "Unable to get SVM capabilities, "
+                                  "skipping");
+            return TEST_SKIP;
+        }
+        if ((svm_caps & CL_DEVICE_SVM_COARSE_GRAIN_BUFFER) == 0)
+        {
+            print_missing_feature(error,
+                                  "Device doesn't support "
+                                  "CL_DEVICE_SVM_COARSE_"
+                                  "GRAIN_BUFFER, skipping");
+            return TEST_SKIP;
+        }
+
+        svm_buffer =
+            clSVMWrapper(context, sizeof(cl_int) * BUF_SIZE,
+                         CL_DEVICE_SVM_COARSE_GRAIN_BUFFER | CL_MEM_READ_WRITE);
+        if (svm_buffer() == nullptr)
+        {
+            test_error(CL_OUT_OF_RESOURCES, "SVM allocation failed");
+        }
+
+        cl_mem_properties buf_props[] = { address_type, CL_TRUE, 0 };
+        buffer = clCreateBufferWithProperties(
+            context, buf_props, CL_MEM_READ_WRITE | CL_MEM_USE_HOST_PTR,
+            sizeof(cl_int) * BUF_SIZE, svm_buffer(), &error);
+        test_error(error, "clCreateBuffer with device address 1 failed\n");
+
+        cl_mem_device_address_ext Addr = 0;
+        error = clGetMemObjectInfo(buffer, CL_MEM_DEVICE_ADDRESS_EXT,
+                                   sizeof(Addr), &Addr, NULL);
+        test_error(error,
+                   "clGetMemObjectInfo(CL_MEM_DEVICE_ADDRESS_EXT) failed\n");
+
+        if ((void *)Addr != svm_buffer())
+        {
+            test_fail("clGetMemObjectInfo(CL_MEM_DEVICE_ADDRESS_EXT) "
+                      "returned different address than clSVMAlloc\n");
+        }
+        return TEST_PASS;
+    }
+
+
+    int test_buffer(clMemWrapper &dev_addr_buffer, clMemWrapper &plain_buffer,
+                    clKernelWrapper &get_addr_kernel)
+    {
+        cl_int error = 0;
+        cl_ulong DeviceAddrFromKernel = 0;
+        cl_mem_device_address_ext DeviceAddrFromAPI = 0;
+
+        for (int i = 0; i < BUF_SIZE; ++i)
+        {
+            BufferHost[i] = i;
+        }
+
+        error =
+            check_device_address_from_api(dev_addr_buffer, DeviceAddrFromAPI);
+        test_error_fail(error,
+                        "device address buffer does not have device address");
+
+        error = clEnqueueWriteBuffer(queue, dev_addr_buffer,
+                                     CL_FALSE, // block
+                                     0, BUF_SIZE * sizeof(cl_int), BufferHost,
+                                     0, NULL, NULL);
+        test_error_fail(error,
+                        "clEnqueueWriteBuffer of dev_addr_buffer failed\n");
+
+        error = clSetKernelArg(get_addr_kernel, 0, sizeof(cl_mem),
+                               &dev_addr_buffer);
+        test_error_fail(error, "clSetKernelArg 0 failed\n");
+        error =
+            clSetKernelArg(get_addr_kernel, 1, sizeof(cl_mem), &plain_buffer);
+        test_error_fail(error, "clSetKernelArg 1 failed\n");
+
+        error = clEnqueueNDRangeKernel(queue, get_addr_kernel, 1, NULL,
+                                       global_size_one, NULL, 0, NULL, NULL);
+        test_error_fail(error, "clNDRangeKernel failed\n");
+
+        error = clEnqueueReadBuffer(queue, dev_addr_buffer, CL_FALSE, 0,
+                                    sizeof(cl_int) * BUF_SIZE, BufferHost, 0,
+                                    NULL, NULL);
+        test_error_fail(error,
+                        "clEnqueueReadBuffer of dev_addr_buffer failed\n");
+
+        error = clEnqueueReadBuffer(queue, plain_buffer, CL_FALSE, 0,
+                                    sizeof(cl_ulong), &DeviceAddrFromKernel, 0,
+                                    NULL, NULL);
+        test_error_fail(error, "clEnqueueReadBuffer of buffer failed\n");
+
+        error = clFinish(queue);
+        test_error_fail(error, "clFinish failed\n");
+
+        for (int i = 0; i < BUF_SIZE; ++i)
+        {
+            if (BufferHost[i] != (i + 1))
+            {
+                test_fail("BufferHost[%i] expected "
+                          "to be: %i, but is: %i\n",
+                          i, i + 1, BufferHost[i]);
+            }
+        }
+
+        if (DeviceAddrFromAPI != DeviceAddrFromKernel)
+        {
+            test_fail("DeviceAddrFromAPI(%" PRIu64
+                      ") != DeviceAddrFromKernel(%" PRIu64 ")\n",
+                      (uint64_t)DeviceAddrFromAPI,
+                      (uint64_t)DeviceAddrFromKernel);
+        }
+        return TEST_PASS;
+    }
+
+    int test_indirect_buffer(clMemWrapper &dev_addr_buffer,
+                             clMemWrapper &buffer_in_long,
+                             clMemWrapper &buffer_out_int,
+                             clKernelWrapper &ind_access_kernel)
+    {
+        cl_int error = 0;
+        cl_mem_device_address_ext DeviceAddrFromAPI = 0;
+
+        int DataIn = 0x12348765;
+        int DataOut = -1;
+
+        // A devaddr buffer with the payload data.
+        error = clEnqueueWriteBuffer(queue, dev_addr_buffer,
+                                     CL_TRUE, // block
+                                     0, sizeof(cl_int), &DataIn, 0, NULL, NULL);
+        test_error_fail(error,
+                        "clEnqueueWriteBuffer of dev_addr_buffer failed\n");
+
+        error =
+            check_device_address_from_api(dev_addr_buffer, DeviceAddrFromAPI);
+        test_error_fail(error,
+                        "device address buffer does not have device address")
+
+            // A basic buffer used to pass the other buffer's address.
+            error = clEnqueueWriteBuffer(queue, buffer_in_long,
+                                         CL_TRUE, // block
+                                         0, sizeof(cl_long), &DeviceAddrFromAPI,
+                                         0, NULL, NULL);
+        test_error_fail(error,
+                        "clEnqueueWriteBuffer of dev_addr_buffer failed\n");
+
+        error = clSetKernelArg(ind_access_kernel, 0, sizeof(cl_mem),
+                               &buffer_in_long);
+        test_error_fail(error, "clSetKernelArg 0 failed\n");
+        error = clSetKernelArg(ind_access_kernel, 1, sizeof(cl_mem),
+                               &buffer_out_int);
+        test_error_fail(error, "clSetKernelArg 1 failed\n");
+
+        error = clSetKernelExecInfo(ind_access_kernel,
+                                    CL_KERNEL_EXEC_INFO_DEVICE_PTRS_EXT,
+                                    sizeof(void *), &DeviceAddrFromAPI);
+        test_error_fail(error,
+                        "Setting indirect access for "
+                        "device ptrs failed!\n");
+
+        error = clEnqueueNDRangeKernel(queue, ind_access_kernel, 1, NULL,
+                                       global_size_one, NULL, 0, NULL, NULL);
+        test_error_fail(error, "clNDRangeKernel failed\n");
+
+        error = clEnqueueReadBuffer(queue, buffer_out_int, CL_FALSE, 0,
+                                    sizeof(cl_int), &DataOut, 0, NULL, NULL);
+        test_error_fail(error, "clEnqueueReadBuffer of buffer failed\n");
+
+        error = clFinish(queue);
+        test_error_fail(error, "clFinish failed\n");
+
+        for (int i = 0; i < BUF_SIZE; ++i)
+        {
+            if (BufferHost[i] != i + 1)
+            {
+                test_fail("PinnedBufferHost[%i] expected "
+                          "to be: %i, but is: %i\n",
+                          i, i + 1, BufferHost[i]);
+            }
+        }
+
+        if (DataIn != DataOut)
+        {
+            test_fail("Passing data via indirect buffers failed. "
+                      "Got: %i expected: %i\n",
+                      DataOut, DataIn);
+        }
+        return TEST_PASS;
+    }
+
+    int test_set_kernel_arg(clMemWrapper &dev_addr_buffer,
+                            clMemWrapper &buffer_out_int,
+                            clKernelWrapper &ptr_arith_kernel)
+    {
+        cl_int error = 0;
+        cl_mem_device_address_ext DeviceAddrFromAPI = 0;
+        int DataOut = -1;
+        int DataIn = 0x12348765;
+
+        clSetKernelArgDevicePointerEXT_fn clSetKernelArgDevicePointer =
+            (clSetKernelArgDevicePointerEXT_fn)
+                clGetExtensionFunctionAddressForPlatform(
+                    getPlatformFromDevice(device),
+                    "clSetKernelArgDevicePointerEXT");
+        if (clSetKernelArgDevicePointer == nullptr)
+            test_error_fail(
+                error, "Cannot get address of clSetKernelArgDevicePointerEXT");
+
+        error = clEnqueueWriteBuffer(queue, dev_addr_buffer,
+                                     CL_FALSE, // block
+                                     0, sizeof(cl_int), &DataIn, 0, NULL, NULL);
+        test_error_fail(error,
+                        "clEnqueueWriteBuffer of dev_addr_buffer failed\n");
+
+        error =
+            check_device_address_from_api(dev_addr_buffer, DeviceAddrFromAPI);
+        test_error_fail(error, "dev_addr_buffer does not have device address");
+
+        cl_mem_device_address_ext DeviceAddrFromAPIP2 =
+            (cl_mem_device_address_ext)(((cl_uint *)DeviceAddrFromAPI) + 2);
+        error = clSetKernelArgDevicePointer(ptr_arith_kernel, 0,
+                                            DeviceAddrFromAPIP2);
+        test_error_fail(error, "clSetKernelArgDevicePointer failed\n");
+        error = clSetKernelArg(ptr_arith_kernel, 1, sizeof(cl_mem),
+                               &buffer_out_int);
+        test_error_fail(error, "clSetKernelArg 1 failed\n");
+
+        error = clEnqueueNDRangeKernel(queue, ptr_arith_kernel, 1, NULL,
+                                       global_size_one, NULL, 0, NULL, NULL);
+        test_error_fail(error, "clNDRangeKernel failed\n");
+
+        error = clEnqueueReadBuffer(queue, buffer_out_int, CL_FALSE, 0,
+                                    sizeof(cl_int), &DataOut, 0, NULL, NULL);
+        test_error_fail(error, "clEnqueueReadBuffer of buffer failed\n");
+
+        error = clFinish(queue);
+        test_error_fail(error, "clFinish failed\n");
+
+        if (DataIn != DataOut)
+        {
+            test_fail("Negative offsetting from passed in pointer failed: "
+                      "got: %i expected: %i",
+                      DataOut, DataIn);
+        }
+        return TEST_PASS;
+    }
+};
+
+}
+
+REGISTER_TEST(private_address)
+{
+    BufferDeviceAddressTest test_fixture = BufferDeviceAddressTest(
+        device, context, queue, CL_MEM_DEVICE_PRIVATE_ADDRESS_EXT);
+
+    if (test_fixture.Skip())
+    {
+        log_info("Test fixture skip\n");
+        return TEST_SKIPPED_ITSELF;
+    }
+
+    cl_int error = test_fixture.RunTest();
+    test_error_ret(error, "Test Failed", TEST_FAIL);
+
+    return TEST_PASS;
+}

test_conformance/extensions/cl_ext_buffer_device_address/main.cpp

@@ -0,0 +1,58 @@
+// Copyright (c) 2024 The Khronos Group Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//    http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+
+#include "harness/testHarness.h"
+#include "harness/deviceInfo.h"
+
+test_status InitCL(cl_device_id device)
+{
+    auto version = get_device_cl_version(device);
+    auto expected_min_version = Version(3, 0);
+
+    if (version < expected_min_version)
+    {
+        version_expected_info("Test", "OpenCL",
+                              expected_min_version.to_string().c_str(),
+                              version.to_string().c_str());
+        return TEST_SKIP;
+    }
+
+    if (!is_extension_available(device, "cl_ext_buffer_device_address"))
+    {
+        log_info("The device does not support the "
+                 "cl_ext_buffer_device_address extension.\n");
+        return TEST_SKIPPED_ITSELF;
+    }
+
+    cl_version ext_version =
+        get_extension_version(device, "cl_ext_buffer_device_address");
+    if (ext_version != CL_MAKE_VERSION(1, 0, 2))
+    {
+        log_info("The test is written against cl_ext_buffer_device_address "
+                 "extension version 1.0.2, device supports version: %u.%u.%u\n",
+                 CL_VERSION_MAJOR(ext_version), CL_VERSION_MINOR(ext_version),
+                 CL_VERSION_PATCH(ext_version));
+        return TEST_SKIPPED_ITSELF;
+    }
+
+    return TEST_PASS;
+}
+
+int main(int argc, const char *argv[])
+{
+    return runTestHarnessWithCheck(
+        argc, argv, test_registry::getInstance().num_tests(),
+        test_registry::getInstance().definitions(), false, 0, InitCL);
+}

test_conformance/extensions/cl_ext_cxx_for_opencl/cxx_for_opencl_ext.cpp

@@ -13,8 +13,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 //
-#include "procs.h"
+#include <CL/cl.h>
 
+#include "harness/typeWrappers.h"
 
 int test_cxx_for_opencl(cl_device_id device, cl_context context,
                         cl_command_queue queue)
@@ -90,8 +91,7 @@ int test_cxx_for_opencl(cl_device_id device, cl_context context,
     return TEST_PASS;
 }
 
-int test_cxx_for_opencl_ext(cl_device_id device, cl_context context,
-                            cl_command_queue queue, int)
+REGISTER_TEST_VERSION(cxx_for_opencl_ext, Version(2, 0))
 {
     if (!is_extension_available(device, "cl_ext_cxx_for_opencl"))
     {

test_conformance/extensions/cl_ext_cxx_for_opencl/cxx_for_opencl_ver.cpp

@@ -13,8 +13,9 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 //
-#include "procs.h"
+#include <CL/cl.h>
 
+#include "harness/typeWrappers.h"
 
 int test_cxx_for_opencl_version(cl_device_id device, cl_context context,
                                 cl_command_queue queue)
@@ -88,8 +89,7 @@ int test_cxx_for_opencl_version(cl_device_id device, cl_context context,
     return TEST_PASS;
 }
 
-int test_cxx_for_opencl_ver(cl_device_id device, cl_context context,
-                            cl_command_queue queue, int)
+REGISTER_TEST_VERSION(cxx_for_opencl_ver, Version(2, 0))
 {
     if (!is_extension_available(device, "cl_ext_cxx_for_opencl"))
     {

test_conformance/extensions/cl_ext_cxx_for_opencl/main.cpp

@@ -14,15 +14,11 @@
 // limitations under the License.
 //
 
-#include "procs.h"
-
-test_definition test_list[] = {
-    ADD_TEST_VERSION(cxx_for_opencl_ext, Version(2, 0)),
-    ADD_TEST_VERSION(cxx_for_opencl_ver, Version(2, 0))
-};
+#include "harness/testHarness.h"
 
 int main(int argc, const char *argv[])
 {
-    return runTestHarnessWithCheck(argc, argv, ARRAY_SIZE(test_list), test_list,
-                                   false, 0, nullptr);
+    return runTestHarnessWithCheck(
+        argc, argv, test_registry::getInstance().num_tests(),
+        test_registry::getInstance().definitions(), false, 0, nullptr);
 }

test_conformance/extensions/cl_ext_cxx_for_opencl/procs.h

@@ -1,26 +0,0 @@
-//
-// Copyright (c) 2021 The Khronos Group Inc.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//    http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-//
-#ifndef _procs_h
-#define _procs_h
-
-#include "harness/typeWrappers.h"
-
-extern int test_cxx_for_opencl_ext(cl_device_id device, cl_context context,
-                                   cl_command_queue queue, int);
-extern int test_cxx_for_opencl_ver(cl_device_id device, cl_context context,
-                                   cl_command_queue queue, int);
-
-#endif /*_procs_h*/

test_conformance/extensions/cl_khr_command_buffer/CMakeLists.txt

@@ -3,6 +3,7 @@ set(MODULE_NAME CL_KHR_COMMAND_BUFFER)
 set(${MODULE_NAME}_SOURCES
     main.cpp
     basic_command_buffer.cpp
+    basic_command_buffer_tests.cpp
     svm_command_basic.cpp
     command_buffer_printf.cpp
     command_buffer_get_command_buffer_info.cpp

test_conformance/extensions/cl_khr_command_buffer/basic_command_buffer.cpp

@@ -14,7 +14,6 @@
 // limitations under the License.
 //
 #include "basic_command_buffer.h"
-#include "procs.h"
 
 #include <algorithm>
 #include <cstring>
@@ -186,324 +185,253 @@ cl_int BasicCommandBufferTest::SetUp(int elements)
     return CL_SUCCESS;
 }
 
-namespace {
-
-// Test that the CL_COMMAND_BUFFER_FLAGS_KHR bitfield is parsed correctly when
-// multiple flags are set.
-struct MultiFlagCreationTest : public BasicCommandBufferTest
+cl_int MultiFlagCreationTest::Run()
 {
-    using BasicCommandBufferTest::BasicCommandBufferTest;
+    cl_command_buffer_properties_khr flags = 0;
+    cl_int error = CL_SUCCESS;
 
-    cl_int Run() override
+    // First try to find multiple flags that are supported by the driver and
+    // device.
+    if (simultaneous_use_support)
     {
-        cl_command_buffer_properties_khr flags = 0;
-        cl_int error = CL_SUCCESS;
-
-        // First try to find multiple flags that are supported by the driver and
-        // device.
-        if (simultaneous_use_support)
-        {
-            flags |= CL_COMMAND_BUFFER_SIMULTANEOUS_USE_KHR;
-        }
-
-        if (is_extension_available(
-                device, CL_KHR_COMMAND_BUFFER_MULTI_DEVICE_EXTENSION_NAME))
-        {
-            flags |= CL_COMMAND_BUFFER_DEVICE_SIDE_SYNC_KHR;
-        }
-
-        if (is_extension_available(
-                device, CL_KHR_COMMAND_BUFFER_MUTABLE_DISPATCH_EXTENSION_NAME))
-        {
-            flags |= CL_COMMAND_BUFFER_MUTABLE_KHR;
-        }
-
-        cl_command_buffer_properties_khr props[] = {
-            CL_COMMAND_BUFFER_FLAGS_KHR, flags, 0
-        };
-
-        command_buffer = clCreateCommandBufferKHR(1, &queue, props, &error);
-        test_error(error, "clCreateCommandBufferKHR failed");
-
-        return CL_SUCCESS;
+        flags |= CL_COMMAND_BUFFER_SIMULTANEOUS_USE_KHR;
     }
+
+    if (is_extension_available(
+            device, CL_KHR_COMMAND_BUFFER_MULTI_DEVICE_EXTENSION_NAME))
+    {
+        flags |= CL_COMMAND_BUFFER_DEVICE_SIDE_SYNC_KHR;
+    }
+
+    if (is_extension_available(
+            device, CL_KHR_COMMAND_BUFFER_MUTABLE_DISPATCH_EXTENSION_NAME))
+    {
+        flags |= CL_COMMAND_BUFFER_MUTABLE_KHR;
+    }
+
+    cl_command_buffer_properties_khr props[] = { CL_COMMAND_BUFFER_FLAGS_KHR,
+                                                 flags, 0 };
+
+    command_buffer = clCreateCommandBufferKHR(1, &queue, props, &error);
+    test_error(error, "clCreateCommandBufferKHR failed");
+
+    return CL_SUCCESS;
 };
 
-// Test enqueuing a command-buffer containing a single NDRange command once
-struct BasicEnqueueTest : public BasicCommandBufferTest
+cl_int BasicEnqueueTest::Run()
 {
-    using BasicCommandBufferTest::BasicCommandBufferTest;
 
-    cl_int Run() override
+    cl_int error = clCommandNDRangeKernelKHR(
+        command_buffer, nullptr, nullptr, kernel, 1, nullptr, &num_elements,
+        nullptr, 0, nullptr, nullptr, nullptr);
+    test_error(error, "clCommandNDRangeKernelKHR failed");
+
+    error = clFinalizeCommandBufferKHR(command_buffer);
+    test_error(error, "clFinalizeCommandBufferKHR failed");
+
+    const cl_int pattern = 42;
+    error = clEnqueueFillBuffer(queue, in_mem, &pattern, sizeof(cl_int), 0,
+                                data_size(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueFillBuffer failed");
+
+    error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0, nullptr,
+                                      nullptr);
+    test_error(error, "clEnqueueCommandBufferKHR failed");
+
+    std::vector<cl_int> output_data_1(num_elements);
+    error = clEnqueueReadBuffer(queue, out_mem, CL_TRUE, 0, data_size(),
+                                output_data_1.data(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueReadBuffer failed");
+
+    for (size_t i = 0; i < num_elements; i++)
     {
-        cl_int error = clCommandNDRangeKernelKHR(
+        CHECK_VERIFICATION_ERROR(pattern, output_data_1[i], i);
+    }
+
+    const cl_int new_pattern = 12;
+    error = clEnqueueFillBuffer(queue, in_mem, &new_pattern, sizeof(cl_int), 0,
+                                data_size(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueFillBuffer failed");
+
+    error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0, nullptr,
+                                      nullptr);
+    test_error(error, "clEnqueueCommandBufferKHR failed");
+
+    std::vector<cl_int> output_data_2(num_elements);
+    error = clEnqueueReadBuffer(queue, out_mem, CL_TRUE, 0, data_size(),
+                                output_data_2.data(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueReadBuffer failed");
+
+    for (size_t i = 0; i < num_elements; i++)
+    {
+        CHECK_VERIFICATION_ERROR(new_pattern, output_data_2[i], i);
+    }
+
+    return CL_SUCCESS;
+};
+
+cl_int MixedCommandsTest::Run()
+{
+    cl_int error;
+    const size_t iterations = 4;
+    clMemWrapper result_mem =
+        clCreateBuffer(context, CL_MEM_READ_WRITE, sizeof(cl_int) * iterations,
+                       nullptr, &error);
+    test_error(error, "clCreateBuffer failed");
+
+    const cl_int pattern_base = 42;
+    for (size_t i = 0; i < iterations; i++)
+    {
+        const cl_int pattern = pattern_base + i;
+        cl_int error = clCommandFillBufferKHR(
+            command_buffer, nullptr, nullptr, in_mem, &pattern, sizeof(cl_int),
+            0, data_size(), 0, nullptr, nullptr, nullptr);
+        test_error(error, "clCommandFillBufferKHR failed");
+
+        error = clCommandNDRangeKernelKHR(
             command_buffer, nullptr, nullptr, kernel, 1, nullptr, &num_elements,
             nullptr, 0, nullptr, nullptr, nullptr);
         test_error(error, "clCommandNDRangeKernelKHR failed");
 
-        error = clFinalizeCommandBufferKHR(command_buffer);
-        test_error(error, "clFinalizeCommandBufferKHR failed");
-
-        const cl_int pattern = 42;
-        error = clEnqueueFillBuffer(queue, in_mem, &pattern, sizeof(cl_int), 0,
-                                    data_size(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueFillBuffer failed");
-
-        error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0,
-                                          nullptr, nullptr);
-        test_error(error, "clEnqueueCommandBufferKHR failed");
-
-        std::vector<cl_int> output_data_1(num_elements);
-        error = clEnqueueReadBuffer(queue, out_mem, CL_TRUE, 0, data_size(),
-                                    output_data_1.data(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueReadBuffer failed");
-
-        for (size_t i = 0; i < num_elements; i++)
-        {
-            CHECK_VERIFICATION_ERROR(pattern, output_data_1[i], i);
-        }
-
-        const cl_int new_pattern = 12;
-        error = clEnqueueFillBuffer(queue, in_mem, &new_pattern, sizeof(cl_int),
-                                    0, data_size(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueFillBuffer failed");
-
-        error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0,
-                                          nullptr, nullptr);
-        test_error(error, "clEnqueueCommandBufferKHR failed");
-
-        std::vector<cl_int> output_data_2(num_elements);
-        error = clEnqueueReadBuffer(queue, out_mem, CL_TRUE, 0, data_size(),
-                                    output_data_2.data(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueReadBuffer failed");
-
-        for (size_t i = 0; i < num_elements; i++)
-        {
-            CHECK_VERIFICATION_ERROR(new_pattern, output_data_2[i], i);
-        }
-
-        return CL_SUCCESS;
+        const size_t result_offset = i * sizeof(cl_int);
+        error = clCommandCopyBufferKHR(
+            command_buffer, nullptr, nullptr, out_mem, result_mem, 0,
+            result_offset, sizeof(cl_int), 0, nullptr, nullptr, nullptr);
+        test_error(error, "clCommandCopyBufferKHR failed");
     }
-};
 
-// Test enqueuing a command-buffer containing multiple command, including
-// operations other than NDRange kernel execution.
-struct MixedCommandsTest : public BasicCommandBufferTest
-{
-    using BasicCommandBufferTest::BasicCommandBufferTest;
+    error = clFinalizeCommandBufferKHR(command_buffer);
+    test_error(error, "clFinalizeCommandBufferKHR failed");
 
-    cl_int Run() override
+    error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0, nullptr,
+                                      nullptr);
+    test_error(error, "clEnqueueCommandBufferKHR failed");
+
+    std::vector<cl_int> result_data(num_elements);
+    error = clEnqueueReadBuffer(queue, result_mem, CL_TRUE, 0,
+                                iterations * sizeof(cl_int), result_data.data(),
+                                0, nullptr, nullptr);
+    test_error(error, "clEnqueueReadBuffer failed");
+
+    for (size_t i = 0; i < iterations; i++)
     {
-        cl_int error;
-        const size_t iterations = 4;
-        clMemWrapper result_mem =
-            clCreateBuffer(context, CL_MEM_READ_WRITE,
-                           sizeof(cl_int) * iterations, nullptr, &error);
-        test_error(error, "clCreateBuffer failed");
-
-        const cl_int pattern_base = 42;
-        for (size_t i = 0; i < iterations; i++)
-        {
-            const cl_int pattern = pattern_base + i;
-            cl_int error = clCommandFillBufferKHR(
-                command_buffer, nullptr, nullptr, in_mem, &pattern,
-                sizeof(cl_int), 0, data_size(), 0, nullptr, nullptr, nullptr);
-            test_error(error, "clCommandFillBufferKHR failed");
-
-            error = clCommandNDRangeKernelKHR(
-                command_buffer, nullptr, nullptr, kernel, 1, nullptr,
-                &num_elements, nullptr, 0, nullptr, nullptr, nullptr);
-            test_error(error, "clCommandNDRangeKernelKHR failed");
-
-            const size_t result_offset = i * sizeof(cl_int);
-            error = clCommandCopyBufferKHR(
-                command_buffer, nullptr, nullptr, out_mem, result_mem, 0,
-                result_offset, sizeof(cl_int), 0, nullptr, nullptr, nullptr);
-            test_error(error, "clCommandCopyBufferKHR failed");
-        }
-
-        error = clFinalizeCommandBufferKHR(command_buffer);
-        test_error(error, "clFinalizeCommandBufferKHR failed");
-
-        error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0,
-                                          nullptr, nullptr);
-        test_error(error, "clEnqueueCommandBufferKHR failed");
-
-        std::vector<cl_int> result_data(num_elements);
-        error = clEnqueueReadBuffer(queue, result_mem, CL_TRUE, 0,
-                                    iterations * sizeof(cl_int),
-                                    result_data.data(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueReadBuffer failed");
-
-        for (size_t i = 0; i < iterations; i++)
-        {
-            const cl_int ref = pattern_base + i;
-            CHECK_VERIFICATION_ERROR(ref, result_data[i], i);
-        }
-
-        return CL_SUCCESS;
-    }
-};
-
-// Test flushing the command-queue between command-buffer enqueues
-struct ExplicitFlushTest : public BasicCommandBufferTest
-{
-    using BasicCommandBufferTest::BasicCommandBufferTest;
-
-    cl_int Run() override
-    {
-        cl_int error = clCommandNDRangeKernelKHR(
-            command_buffer, nullptr, nullptr, kernel, 1, nullptr, &num_elements,
-            nullptr, 0, nullptr, nullptr, nullptr);
-        test_error(error, "clCommandNDRangeKernelKHR failed");
-
-        error = clFinalizeCommandBufferKHR(command_buffer);
-        test_error(error, "clFinalizeCommandBufferKHR failed");
-
-        const cl_int pattern_A = 42;
-        error = clEnqueueFillBuffer(queue, in_mem, &pattern_A, sizeof(cl_int),
-                                    0, data_size(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueFillBuffer failed");
-
-        error = clFlush(queue);
-        test_error(error, "clFlush failed");
-
-        error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0,
-                                          nullptr, nullptr);
-        test_error(error, "clEnqueueCommandBufferKHR failed");
-
-        std::vector<cl_int> output_data_A(num_elements);
-        error = clEnqueueReadBuffer(queue, out_mem, CL_FALSE, 0, data_size(),
-                                    output_data_A.data(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueReadBuffer failed");
-
-        const cl_int pattern_B = 0xA;
-        error = clEnqueueFillBuffer(queue, in_mem, &pattern_B, sizeof(cl_int),
-                                    0, data_size(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueFillBuffer failed");
-
-        error = clFlush(queue);
-        test_error(error, "clFlush failed");
-
-        error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0,
-                                          nullptr, nullptr);
-        test_error(error, "clEnqueueCommandBufferKHR failed");
-
-        error = clFlush(queue);
-        test_error(error, "clFlush failed");
-
-        std::vector<cl_int> output_data_B(num_elements);
-        error = clEnqueueReadBuffer(queue, out_mem, CL_FALSE, 0, data_size(),
-                                    output_data_B.data(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueReadBuffer failed");
-
-        error = clFinish(queue);
-        test_error(error, "clFinish failed");
-
-        for (size_t i = 0; i < num_elements; i++)
-        {
-            CHECK_VERIFICATION_ERROR(pattern_A, output_data_A[i], i);
-
-            CHECK_VERIFICATION_ERROR(pattern_B, output_data_B[i], i);
-        }
-        return CL_SUCCESS;
+        const cl_int ref = pattern_base + i;
+        CHECK_VERIFICATION_ERROR(ref, result_data[i], i);
     }
 
-    bool Skip() override
-    {
-        return BasicCommandBufferTest::Skip() || !simultaneous_use_support;
-    }
-};
-
-// Test enqueueing a command-buffer twice separated by another enqueue operation
-struct InterleavedEnqueueTest : public BasicCommandBufferTest
-{
-    using BasicCommandBufferTest::BasicCommandBufferTest;
-
-    cl_int Run() override
-    {
-        cl_int error = clCommandNDRangeKernelKHR(
-            command_buffer, nullptr, nullptr, kernel, 1, nullptr, &num_elements,
-            nullptr, 0, nullptr, nullptr, nullptr);
-        test_error(error, "clCommandNDRangeKernelKHR failed");
-
-        error = clFinalizeCommandBufferKHR(command_buffer);
-        test_error(error, "clFinalizeCommandBufferKHR failed");
-
-        cl_int pattern = 42;
-        error = clEnqueueFillBuffer(queue, in_mem, &pattern, sizeof(cl_int), 0,
-                                    data_size(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueFillBuffer failed");
-
-        error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0,
-                                          nullptr, nullptr);
-        test_error(error, "clEnqueueCommandBufferKHR failed");
-
-        pattern = 0xABCD;
-        error = clEnqueueFillBuffer(queue, in_mem, &pattern, sizeof(cl_int), 0,
-                                    data_size(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueFillBuffer failed");
-
-        error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0,
-                                          nullptr, nullptr);
-        test_error(error, "clEnqueueCommandBufferKHR failed");
-
-        error = clEnqueueCopyBuffer(queue, in_mem, out_mem, 0, 0, data_size(),
-                                    0, nullptr, nullptr);
-        test_error(error, "clEnqueueCopyBuffer failed");
-
-        std::vector<cl_int> output_data(num_elements);
-        error = clEnqueueReadBuffer(queue, out_mem, CL_TRUE, 0, data_size(),
-                                    output_data.data(), 0, nullptr, nullptr);
-        test_error(error, "clEnqueueReadBuffer failed");
-
-        for (size_t i = 0; i < num_elements; i++)
-        {
-            CHECK_VERIFICATION_ERROR(pattern, output_data[i], i);
-        }
-
-        return CL_SUCCESS;
-    }
-
-    bool Skip() override
-    {
-        return BasicCommandBufferTest::Skip() || !simultaneous_use_support;
-    }
-};
-
-} // anonymous namespace
-
-int test_multi_flag_creation(cl_device_id device, cl_context context,
-                             cl_command_queue queue, int num_elements)
-{
-    return MakeAndRunTest<MultiFlagCreationTest>(device, context, queue,
-                                                 num_elements);
+    return CL_SUCCESS;
 }
 
-int test_single_ndrange(cl_device_id device, cl_context context,
-                        cl_command_queue queue, int num_elements)
+cl_int ExplicitFlushTest::Run()
 {
-    return MakeAndRunTest<BasicEnqueueTest>(device, context, queue,
-                                            num_elements);
+    cl_int error = clCommandNDRangeKernelKHR(
+        command_buffer, nullptr, nullptr, kernel, 1, nullptr, &num_elements,
+        nullptr, 0, nullptr, nullptr, nullptr);
+    test_error(error, "clCommandNDRangeKernelKHR failed");
+
+    error = clFinalizeCommandBufferKHR(command_buffer);
+    test_error(error, "clFinalizeCommandBufferKHR failed");
+
+    const cl_int pattern_A = 42;
+    error = clEnqueueFillBuffer(queue, in_mem, &pattern_A, sizeof(cl_int), 0,
+                                data_size(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueFillBuffer failed");
+
+    error = clFlush(queue);
+    test_error(error, "clFlush failed");
+
+    error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0, nullptr,
+                                      nullptr);
+    test_error(error, "clEnqueueCommandBufferKHR failed");
+
+    std::vector<cl_int> output_data_A(num_elements);
+    error = clEnqueueReadBuffer(queue, out_mem, CL_FALSE, 0, data_size(),
+                                output_data_A.data(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueReadBuffer failed");
+
+    const cl_int pattern_B = 0xA;
+    error = clEnqueueFillBuffer(queue, in_mem, &pattern_B, sizeof(cl_int), 0,
+                                data_size(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueFillBuffer failed");
+
+    error = clFlush(queue);
+    test_error(error, "clFlush failed");
+
+    error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0, nullptr,
+                                      nullptr);
+    test_error(error, "clEnqueueCommandBufferKHR failed");
+
+    error = clFlush(queue);
+    test_error(error, "clFlush failed");
+
+    std::vector<cl_int> output_data_B(num_elements);
+    error = clEnqueueReadBuffer(queue, out_mem, CL_FALSE, 0, data_size(),
+                                output_data_B.data(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueReadBuffer failed");
+
+    error = clFinish(queue);
+    test_error(error, "clFinish failed");
+
+    for (size_t i = 0; i < num_elements; i++)
+    {
+        CHECK_VERIFICATION_ERROR(pattern_A, output_data_A[i], i);
+
+        CHECK_VERIFICATION_ERROR(pattern_B, output_data_B[i], i);
+    }
+    return CL_SUCCESS;
 }
 
-int test_interleaved_enqueue(cl_device_id device, cl_context context,
-                             cl_command_queue queue, int num_elements)
+bool ExplicitFlushTest::Skip()
 {
-    return MakeAndRunTest<InterleavedEnqueueTest>(device, context, queue,
-                                                  num_elements);
+    return BasicCommandBufferTest::Skip() || !simultaneous_use_support;
 }
 
-int test_mixed_commands(cl_device_id device, cl_context context,
-                        cl_command_queue queue, int num_elements)
+cl_int InterleavedEnqueueTest::Run()
 {
-    return MakeAndRunTest<MixedCommandsTest>(device, context, queue,
-                                             num_elements);
+    cl_int error = clCommandNDRangeKernelKHR(
+        command_buffer, nullptr, nullptr, kernel, 1, nullptr, &num_elements,
+        nullptr, 0, nullptr, nullptr, nullptr);
+    test_error(error, "clCommandNDRangeKernelKHR failed");
+
+    error = clFinalizeCommandBufferKHR(command_buffer);
+    test_error(error, "clFinalizeCommandBufferKHR failed");
+
+    cl_int pattern = 42;
+    error = clEnqueueFillBuffer(queue, in_mem, &pattern, sizeof(cl_int), 0,
+                                data_size(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueFillBuffer failed");
+
+    error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0, nullptr,
+                                      nullptr);
+    test_error(error, "clEnqueueCommandBufferKHR failed");
+
+    pattern = 0xABCD;
+    error = clEnqueueFillBuffer(queue, in_mem, &pattern, sizeof(cl_int), 0,
+                                data_size(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueFillBuffer failed");
+
+    error = clEnqueueCommandBufferKHR(0, nullptr, command_buffer, 0, nullptr,
+                                      nullptr);
+    test_error(error, "clEnqueueCommandBufferKHR failed");
+
+    error = clEnqueueCopyBuffer(queue, in_mem, out_mem, 0, 0, data_size(), 0,
+                                nullptr, nullptr);
+    test_error(error, "clEnqueueCopyBuffer failed");
+
+    std::vector<cl_int> output_data(num_elements);
+    error = clEnqueueReadBuffer(queue, out_mem, CL_TRUE, 0, data_size(),
+                                output_data.data(), 0, nullptr, nullptr);
+    test_error(error, "clEnqueueReadBuffer failed");
+
+    for (size_t i = 0; i < num_elements; i++)
+    {
+        CHECK_VERIFICATION_ERROR(pattern, output_data[i], i);
+    }
+
+    return CL_SUCCESS;
 }
 
-int test_explicit_flush(cl_device_id device, cl_context context,
-                        cl_command_queue queue, int num_elements)
+bool InterleavedEnqueueTest::Skip()
 {
-    return MakeAndRunTest<ExplicitFlushTest>(device, context, queue,
-                                             num_elements);
+    return BasicCommandBufferTest::Skip() || !simultaneous_use_support;
 }

test_conformance/extensions/cl_khr_command_buffer/basic_command_buffer.h

@@ -84,6 +84,49 @@ protected:
     clCommandBufferWrapper command_buffer;
 };
 
+// Test that CL_COMMAND_BUFFER_FLAGS_KHR bitfield is parsed correctly when
+// multiple flags are set.
+struct MultiFlagCreationTest : public BasicCommandBufferTest
+{
+    using BasicCommandBufferTest::BasicCommandBufferTest;
+
+    cl_int Run() override;
+};
+
+// Test enqueuing a command-buffer containing a single NDRange command once
+struct BasicEnqueueTest : public BasicCommandBufferTest
+{
+    using BasicCommandBufferTest::BasicCommandBufferTest;
+
+    cl_int Run() override;
+};
+
+// Test enqueuing a command-buffer containing multiple command, including
+// operations other than NDRange kernel execution.
+struct MixedCommandsTest : public BasicCommandBufferTest
+{
+    using BasicCommandBufferTest::BasicCommandBufferTest;
+
+    cl_int Run() override;
+};
+
+// Test flushing the command-queue between command-buffer enqueues
+struct ExplicitFlushTest : public BasicCommandBufferTest
+{
+    using BasicCommandBufferTest::BasicCommandBufferTest;
+
+    cl_int Run() override;
+    bool Skip() override;
+};
+
+// Test enqueueing a command-buffer twice separated by another enqueue operation
+struct InterleavedEnqueueTest : public BasicCommandBufferTest
+{
+    using BasicCommandBufferTest::BasicCommandBufferTest;
+
+    cl_int Run() override;
+    bool Skip() override;
+};
 
 template <class T>
 int MakeAndRunTest(cl_device_id device, cl_context context,

test_conformance/extensions/cl_khr_command_buffer/basic_command_buffer_tests.cpp

@@ -0,0 +1,46 @@
+//
+// Copyright (c) 2025 The Khronos Group Inc.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//    http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+#include "basic_command_buffer.h"
+
+REGISTER_TEST(multi_flag_creation)
+{
+    return MakeAndRunTest<MultiFlagCreationTest>(device, context, queue,
+                                                 num_elements);
+}
+
+REGISTER_TEST(single_ndrange)
+{
+    return MakeAndRunTest<BasicEnqueueTest>(device, context, queue,
+                                            num_elements);
+}
+
+REGISTER_TEST(interleaved_enqueue)
+{
+    return MakeAndRunTest<InterleavedEnqueueTest>(device, context, queue,
+                                                  num_elements);
+}
+
+REGISTER_TEST(mixed_commands)
+{
+    return MakeAndRunTest<MixedCommandsTest>(device, context, queue,
+                                             num_elements);
+}
+
+REGISTER_TEST(explicit_flush)
+{
+    return MakeAndRunTest<ExplicitFlushTest>(device, context, queue,
+                                             num_elements);
+}

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/main.cpp

@@ -12,43 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 //
-#include "procs.h"
 #include "harness/testHarness.h"
 
-test_definition test_list[] = {
-    ADD_TEST(mutable_command_info_device_query),
-    ADD_TEST(mutable_command_info_buffer),
-    ADD_TEST(mutable_command_properties_array),
-    ADD_TEST(mutable_command_kernel),
-    ADD_TEST(mutable_command_dimensions),
-    ADD_TEST(mutable_command_info_type),
-    ADD_TEST(mutable_command_info_queue),
-    ADD_TEST(mutable_command_info_global_work_offset),
-    ADD_TEST(mutable_command_info_local_work_size),
-    ADD_TEST(mutable_command_info_global_work_size),
-    ADD_TEST(mutable_command_full_dispatch),
-    ADD_TEST(mutable_command_overwrite_update),
-    ADD_TEST(mutable_command_multiple_dispatches),
-    ADD_TEST(mutable_command_iterative_arg_update),
-    ADD_TEST(mutable_dispatch_image_1d_arguments),
-    ADD_TEST(mutable_dispatch_image_2d_arguments),
-    ADD_TEST(mutable_dispatch_out_of_order),
-    ADD_TEST(mutable_dispatch_simultaneous_out_of_order),
-    ADD_TEST(mutable_dispatch_simultaneous_in_order),
-    ADD_TEST(mutable_dispatch_simultaneous_cross_queue),
-    ADD_TEST(mutable_dispatch_global_size),
-    ADD_TEST(mutable_dispatch_local_size),
-    ADD_TEST(mutable_dispatch_global_offset),
-    ADD_TEST(mutable_dispatch_svm_arguments),
-    ADD_TEST(mutable_dispatch_local_arguments),
-    ADD_TEST(mutable_dispatch_global_arguments),
-    ADD_TEST(mutable_dispatch_pod_arguments),
-    ADD_TEST(mutable_dispatch_null_arguments),
-    ADD_TEST(command_buffer_with_no_additional_work_groups),
-    ADD_TEST(ndrange_with_no_additional_work_groups),
-    ADD_TEST(ndrange_command_buffer_with_no_additional_work_groups),
-};
-
 int main(int argc, const char *argv[])
 {
     // A device may report the required properties of a queue that
@@ -57,7 +22,9 @@ int main(int argc, const char *argv[])
     // for this in the tests themselves, rather than here, where we have a
     // device to query.
     const cl_command_queue_properties queue_properties = 0;
-    return runTestHarnessWithCheck(argc, argv, ARRAY_SIZE(test_list), test_list,
+    return runTestHarnessWithCheck(argc, argv,
+                                   test_registry::getInstance().num_tests(),
+                                   test_registry::getInstance().definitions(),
                                    false, queue_properties, nullptr);
     return 0;
 }

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_arguments.cpp

@@ -826,44 +826,31 @@ struct MutableDispatchSVMArguments : public MutableDispatchArgumentsTest
 
 }
 
-int test_mutable_dispatch_local_arguments(cl_device_id device,
-                                          cl_context context,
-                                          cl_command_queue queue,
-                                          int num_elements)
+REGISTER_TEST(mutable_dispatch_local_arguments)
 {
     return MakeAndRunTest<MutableDispatchLocalArguments>(device, context, queue,
                                                          num_elements);
 }
 
-int test_mutable_dispatch_global_arguments(cl_device_id device,
-                                           cl_context context,
-                                           cl_command_queue queue,
-                                           int num_elements)
+REGISTER_TEST(mutable_dispatch_global_arguments)
 {
     return MakeAndRunTest<MutableDispatchGlobalArguments>(device, context,
                                                           queue, num_elements);
 }
 
-int test_mutable_dispatch_pod_arguments(cl_device_id device, cl_context context,
-                                        cl_command_queue queue,
-                                        int num_elements)
+REGISTER_TEST(mutable_dispatch_pod_arguments)
 {
     return MakeAndRunTest<MutableDispatchPODArguments>(device, context, queue,
                                                        num_elements);
 }
 
-int test_mutable_dispatch_null_arguments(cl_device_id device,
-                                         cl_context context,
-                                         cl_command_queue queue,
-                                         int num_elements)
+REGISTER_TEST(mutable_dispatch_null_arguments)
 {
     return MakeAndRunTest<MutableDispatchNullArguments>(device, context, queue,
                                                         num_elements);
 }
 
-int test_mutable_dispatch_svm_arguments(cl_device_id device, cl_context context,
-                                        cl_command_queue queue,
-                                        int num_elements)
+REGISTER_TEST(mutable_dispatch_svm_arguments)
 {
     return MakeAndRunTest<MutableDispatchSVMArguments>(device, context, queue,
                                                        num_elements);

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_full_dispatch.cpp

@@ -492,8 +492,7 @@ struct MutableCommandFullDispatch : InfoMutableCommandBufferTest
 
 }
 
-int test_mutable_command_full_dispatch(cl_device_id device, cl_context context,
-                                       cl_command_queue queue, int num_elements)
+REGISTER_TEST(mutable_command_full_dispatch)
 {
     return MakeAndRunTest<MutableCommandFullDispatch>(device, context, queue,
                                                       num_elements);

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_global_offset.cpp

@@ -160,9 +160,7 @@ struct MutableDispatchGlobalOffset : InfoMutableCommandBufferTest
     cl_mutable_command_khr command = nullptr;
 };
 
-int test_mutable_dispatch_global_offset(cl_device_id device, cl_context context,
-                                        cl_command_queue queue,
-                                        int num_elements)
+REGISTER_TEST(mutable_dispatch_global_offset)
 {
 
     return MakeAndRunTest<MutableDispatchGlobalOffset>(device, context, queue,

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_global_size.cpp

@@ -159,8 +159,7 @@ struct MutableDispatchGlobalSize : public InfoMutableCommandBufferTest
     cl_mutable_command_khr command = nullptr;
 };
 
-int test_mutable_dispatch_global_size(cl_device_id device, cl_context context,
-                                      cl_command_queue queue, int num_elements)
+REGISTER_TEST(mutable_dispatch_global_size)
 {
     return MakeAndRunTest<MutableDispatchGlobalSize>(device, context, queue,
                                                      num_elements);

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_image_arguments.cpp

@@ -412,19 +412,13 @@ struct MutableDispatchImage2DArguments : public BasicMutableCommandBufferTest
     cl_mutable_command_khr command = nullptr;
 };
 
-int test_mutable_dispatch_image_1d_arguments(cl_device_id device,
-                                             cl_context context,
-                                             cl_command_queue queue,
-                                             int num_elements)
+REGISTER_TEST(mutable_dispatch_image_1d_arguments)
 {
     return MakeAndRunTest<MutableDispatchImage1DArguments>(device, context,
                                                            queue, num_elements);
 }
 
-int test_mutable_dispatch_image_2d_arguments(cl_device_id device,
-                                             cl_context context,
-                                             cl_command_queue queue,
-                                             int num_elements)
+REGISTER_TEST(mutable_dispatch_image_2d_arguments)
 {
     return MakeAndRunTest<MutableDispatchImage2DArguments>(device, context,
                                                            queue, num_elements);

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_info.cpp

@@ -16,7 +16,6 @@
 
 #include <extensionHelpers.h>
 #include "typeWrappers.h"
-#include "procs.h"
 #include "testHarness.h"
 #include <vector>
 #include <iostream>
@@ -441,76 +440,56 @@ struct InfoLocalWorkSize : public InfoMutableCommandBufferTest
     size_t test_local_work_size = 0;
 };
 
-int test_mutable_command_info_device_query(cl_device_id device,
-                                           cl_context context,
-                                           cl_command_queue queue,
-                                           int num_elements)
+REGISTER_TEST(mutable_command_info_device_query)
 {
     return MakeAndRunTest<InfoDeviceQuery>(device, context, queue,
                                            num_elements);
 }
 
-int test_mutable_command_info_buffer(cl_device_id device, cl_context context,
-                                     cl_command_queue queue, int num_elements)
+REGISTER_TEST(mutable_command_info_buffer)
 {
     return MakeAndRunTest<InfoBuffer>(device, context, queue, num_elements);
 }
 
-int test_mutable_command_properties_array(cl_device_id device,
-                                          cl_context context,
-                                          cl_command_queue queue,
-                                          int num_elements)
+REGISTER_TEST(mutable_command_properties_array)
 {
     return MakeAndRunTest<PropertiesArray>(device, context, queue,
                                            num_elements);
 }
 
-int test_mutable_command_kernel(cl_device_id device, cl_context context,
-                                cl_command_queue queue, int num_elements)
+REGISTER_TEST(mutable_command_kernel)
 {
     return MakeAndRunTest<Kernel>(device, context, queue, num_elements);
 }
 
-int test_mutable_command_dimensions(cl_device_id device, cl_context context,
-                                    cl_command_queue queue, int num_elements)
+REGISTER_TEST(mutable_command_dimensions)
 {
     return MakeAndRunTest<Dimensions>(device, context, queue, num_elements);
 }
 
-int test_mutable_command_info_type(cl_device_id device, cl_context context,
-                                   cl_command_queue queue, int num_elements)
+REGISTER_TEST(mutable_command_info_type)
 {
     return MakeAndRunTest<InfoType>(device, context, queue, num_elements);
 }
 
-int test_mutable_command_info_queue(cl_device_id device, cl_context context,
-                                    cl_command_queue queue, int num_elements)
+REGISTER_TEST(mutable_command_info_queue)
 {
     return MakeAndRunTest<InfoQueue>(device, context, queue, num_elements);
 }
 
-int test_mutable_command_info_global_work_offset(cl_device_id device,
-                                                 cl_context context,
-                                                 cl_command_queue queue,
-                                                 int num_elements)
+REGISTER_TEST(mutable_command_info_global_work_offset)
 {
     return MakeAndRunTest<InfoGlobalWorkOffset>(device, context, queue,
                                                 num_elements);
 }
 
-int test_mutable_command_info_global_work_size(cl_device_id device,
-                                               cl_context context,
-                                               cl_command_queue queue,
-                                               int num_elements)
+REGISTER_TEST(mutable_command_info_global_work_size)
 {
     return MakeAndRunTest<InfoGlobalWorkSize>(device, context, queue,
                                               num_elements);
 }
 
-int test_mutable_command_info_local_work_size(cl_device_id device,
-                                              cl_context context,
-                                              cl_command_queue queue,
-                                              int num_elements)
+REGISTER_TEST(mutable_command_info_local_work_size)
 {
     return MakeAndRunTest<InfoLocalWorkSize>(device, context, queue,
                                              num_elements);

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_iterative_arg_update.cpp

@@ -212,10 +212,7 @@ struct IterativeArgUpdateDispatch : BasicMutableCommandBufferTest
 
 }
 
-int test_mutable_command_iterative_arg_update(cl_device_id device,
-                                              cl_context context,
-                                              cl_command_queue queue,
-                                              int num_elements)
+REGISTER_TEST(mutable_command_iterative_arg_update)
 {
     return MakeAndRunTest<IterativeArgUpdateDispatch>(device, context, queue,
                                                       num_elements);

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_local_size.cpp

@@ -16,7 +16,6 @@
 
 #include <extensionHelpers.h>
 #include "typeWrappers.h"
-#include "procs.h"
 #include "testHarness.h"
 #include "mutable_command_basic.h"
 #include <vector>
@@ -166,8 +165,7 @@ struct MutableDispatchLocalSize : public InfoMutableCommandBufferTest
     cl_mutable_command_khr command = nullptr;
 };
 
-int test_mutable_dispatch_local_size(cl_device_id device, cl_context context,
-                                     cl_command_queue queue, int num_elements)
+REGISTER_TEST(mutable_dispatch_local_size)
 {
     return MakeAndRunTest<MutableDispatchLocalSize>(device, context, queue,
                                                     num_elements);

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_multiple_dispatches.cpp

@@ -207,10 +207,7 @@ struct MultipleCommandsDispatch : BasicMutableCommandBufferTest
 
 }
 
-int test_mutable_command_multiple_dispatches(cl_device_id device,
-                                             cl_context context,
-                                             cl_command_queue queue,
-                                             int num_elements)
+REGISTER_TEST(mutable_command_multiple_dispatches)
 {
     return MakeAndRunTest<MultipleCommandsDispatch>(device, context, queue,
                                                     num_elements);

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_overwrite_update.cpp

@@ -215,10 +215,7 @@ struct OverwriteUpdateDispatch : BasicMutableCommandBufferTest
 
 }
 
-int test_mutable_command_overwrite_update(cl_device_id device,
-                                          cl_context context,
-                                          cl_command_queue queue,
-                                          int num_elements)
+REGISTER_TEST(mutable_command_overwrite_update)
 {
     return MakeAndRunTest<OverwriteUpdateDispatch>(device, context, queue,
                                                    num_elements);

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_simultaneous.cpp

@@ -605,35 +605,25 @@ struct CrossQueueSimultaneousMutableDispatchTest
 
 } // anonymous namespace
 
-int test_mutable_dispatch_out_of_order(cl_device_id device, cl_context context,
-                                       cl_command_queue queue, int num_elements)
+REGISTER_TEST(mutable_dispatch_out_of_order)
 {
     return MakeAndRunTest<SimultaneousMutableDispatchTest<false, true>>(
         device, context, queue, num_elements);
 }
 
-int test_mutable_dispatch_simultaneous_out_of_order(cl_device_id device,
-                                                    cl_context context,
-                                                    cl_command_queue queue,
-                                                    int num_elements)
+REGISTER_TEST(mutable_dispatch_simultaneous_out_of_order)
 {
     return MakeAndRunTest<SimultaneousMutableDispatchTest<true, true>>(
         device, context, queue, num_elements);
 }
 
-int test_mutable_dispatch_simultaneous_in_order(cl_device_id device,
-                                                cl_context context,
-                                                cl_command_queue queue,
-                                                int num_elements)
+REGISTER_TEST(mutable_dispatch_simultaneous_in_order)
 {
     return MakeAndRunTest<SimultaneousMutableDispatchTest<true, false>>(
         device, context, queue, num_elements);
 }
 
-int test_mutable_dispatch_simultaneous_cross_queue(cl_device_id device,
-                                                   cl_context context,
-                                                   cl_command_queue queue,
-                                                   int num_elements)
+REGISTER_TEST(mutable_dispatch_simultaneous_cross_queue)
 {
     return MakeAndRunTest<CrossQueueSimultaneousMutableDispatchTest>(
         device, context, queue, num_elements);

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/mutable_command_work_groups.cpp

@@ -284,29 +284,21 @@ struct MutableDispatchWorkGroups : public BasicMutableCommandBufferTest
     const size_t sizeToAllocate = 64 * sizeof(cl_int);
 };
 
-int test_command_buffer_with_no_additional_work_groups(cl_device_id device,
-                                                       cl_context context,
-                                                       cl_command_queue queue,
-                                                       int num_elements)
+REGISTER_TEST(command_buffer_with_no_additional_work_groups)
 {
 
     return MakeAndRunTest<MutableDispatchWorkGroups<0>>(device, context, queue,
                                                         num_elements);
 }
 
-int test_ndrange_with_no_additional_work_groups(cl_device_id device,
-                                                cl_context context,
-                                                cl_command_queue queue,
-                                                int num_elements)
+REGISTER_TEST(ndrange_with_no_additional_work_groups)
 {
 
     return MakeAndRunTest<MutableDispatchWorkGroups<1>>(device, context, queue,
                                                         num_elements);
 }
 
-int test_ndrange_command_buffer_with_no_additional_work_groups(
-    cl_device_id device, cl_context context, cl_command_queue queue,
-    int num_elements)
+REGISTER_TEST(ndrange_command_buffer_with_no_additional_work_groups)
 {
 
     return MakeAndRunTest<MutableDispatchWorkGroups<2>>(device, context, queue,

test_conformance/extensions/cl_khr_command_buffer/cl_khr_command_buffer_mutable_dispatch/procs.h

@@ -1,143 +0,0 @@
-//
-// Copyright (c) 2023 The Khronos Group Inc.
-//
-// Licensed under the Apache License, Version 2.0 (the "License");
-// you may not use this file except in compliance with the License.
-// You may obtain a copy of the License at
-//
-//    http://www.apache.org/licenses/LICENSE-2.0
-//
-// Unless required by applicable law or agreed to in writing, software
-// distributed under the License is distributed on an "AS IS" BASIS,
-// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-// See the License for the specific language governing permissions and
-// limitations under the License.
-//
-#ifndef CL_KHR_COMMAND_BUFFER_MUTABLE_DISPATCH_PROCS_H
-#define CL_KHR_COMMAND_BUFFER_MUTABLE_DISPATCH_PROCS_H
-
-#include <CL/cl.h>
-
-
-// Basic mutable dispatch tests
-extern int test_mutable_command_info_device_query(cl_device_id device,
-                                                  cl_context context,
-                                                  cl_command_queue queue,
-                                                  int num_elements);
-extern int test_mutable_command_info_buffer(cl_device_id device,
-                                            cl_context context,
-                                            cl_command_queue queue,
-                                            int num_elements);
-extern int test_mutable_command_info_type(cl_device_id device,
-                                          cl_context context,
-                                          cl_command_queue queue,
-                                          int num_elements);
-extern int test_mutable_command_info_queue(cl_device_id device,
-                                           cl_context context,
-                                           cl_command_queue queue,
-                                           int num_elements);
-extern int test_mutable_command_properties_array(cl_device_id device,
-                                                 cl_context context,
-                                                 cl_command_queue queue,
-                                                 int num_elements);
-extern int test_mutable_command_kernel(cl_device_id device, cl_context context,
-                                       cl_command_queue queue,
-                                       int num_elements);
-extern int test_mutable_command_dimensions(cl_device_id device,
-                                           cl_context context,
-                                           cl_command_queue queue,
-                                           int num_elements);
-extern int test_mutable_command_info_global_work_offset(cl_device_id device,
-                                                        cl_context context,
-                                                        cl_command_queue queue,
-                                                        int num_elements);
-extern int test_mutable_command_info_local_work_size(cl_device_id device,
-                                                     cl_context context,
-                                                     cl_command_queue queue,
-                                                     int num_elements);
-extern int test_mutable_command_info_global_work_size(cl_device_id device,
-                                                      cl_context context,
-                                                      cl_command_queue queue,
-                                                      int num_elements);
-extern int test_mutable_dispatch_image_1d_arguments(cl_device_id device,
-                                                    cl_context context,
-                                                    cl_command_queue queue,
-                                                    int num_elements);
-extern int test_mutable_dispatch_image_2d_arguments(cl_device_id device,
-                                                    cl_context context,
-                                                    cl_command_queue queue,
-                                                    int num_elements);
-extern int test_mutable_dispatch_global_arguments(cl_device_id device,
-                                                  cl_context context,
-                                                  cl_command_queue queue,
-                                                  int num_elements);
-extern int test_mutable_dispatch_local_arguments(cl_device_id device,
-                                                 cl_context context,
-                                                 cl_command_queue queue,
-                                                 int num_elements);
-extern int test_mutable_dispatch_pod_arguments(cl_device_id device,
-                                               cl_context context,
-                                               cl_command_queue queue,
-                                               int num_elements);
-extern int test_mutable_dispatch_null_arguments(cl_device_id device,
-                                                cl_context context,
-                                                cl_command_queue queue,
-                                                int num_elements);
-extern int test_mutable_dispatch_svm_arguments(cl_device_id device,
-                                               cl_context context,
-                                               cl_command_queue queue,
-                                               int num_elements);
-extern int test_mutable_dispatch_out_of_order(cl_device_id device,
-                                              cl_context context,
-                                              cl_command_queue queue,
-                                              int num_elements);
-extern int test_mutable_dispatch_simultaneous_out_of_order(
-    cl_device_id device, cl_context context, cl_command_queue queue,
-    int num_elements);
-extern int test_mutable_dispatch_simultaneous_in_order(cl_device_id device,
-                                                       cl_context context,
-                                                       cl_command_queue queue,
-                                                       int num_elements);
-extern int test_mutable_dispatch_simultaneous_cross_queue(
-    cl_device_id device, cl_context context, cl_command_queue queue,
-    int num_elements);
-extern int test_mutable_dispatch_global_size(cl_device_id device,
-                                             cl_context context,
-                                             cl_command_queue queue,
-                                             int num_elements);
-extern int test_mutable_dispatch_local_size(cl_device_id device,
-                                            cl_context context,
-                                            cl_command_queue queue,
-                                            int num_elements);
-extern int test_mutable_dispatch_global_offset(cl_device_id device,
-                                               cl_context context,
-                                               cl_command_queue queue,
-                                               int num_elements);
-extern int test_mutable_command_full_dispatch(cl_device_id device,
-                                              cl_context context,
-                                              cl_command_queue queue,
-                                              int num_elements);
-extern int test_mutable_command_overwrite_update(cl_device_id device,
-                                                 cl_context context,
-                                                 cl_command_queue queue,
-                                                 int num_elements);
-extern int test_mutable_command_multiple_dispatches(cl_device_id device,
-                                                    cl_context context,
-                                                    cl_command_queue queue,
-                                                    int num_elements);
-extern int test_mutable_command_iterative_arg_update(cl_device_id device,
-                                                     cl_context context,
-                                                     cl_command_queue queue,
-                                                     int num_elements);
-extern int test_command_buffer_with_no_additional_work_groups(
-    cl_device_id device, cl_context context, cl_command_queue queue,
-    int num_elements);
-extern int test_ndrange_with_no_additional_work_groups(cl_device_id device,
-                                                       cl_context context,
-                                                       cl_command_queue queue,
-                                                       int num_elements);
-extern int test_ndrange_command_buffer_with_no_additional_work_groups(
-    cl_device_id device, cl_context context, cl_command_queue queue,
-    int num_elements);
-
-#endif /*_CL_KHR_COMMAND_BUFFER_MUTABLE_DISPATCH_PROCS_H*/

test_conformance/extensions/cl_khr_command_buffer/command_buffer_event_sync.cpp

@@ -15,7 +15,6 @@
 //
 
 #include "basic_command_buffer.h"
-#include "procs.h"
 
 #include <vector>
 #include <thread>
@@ -823,10 +822,7 @@ struct CommandBufferEventSync : public BasicCommandBufferTest
 
 //--------------------------------------------------------------------------
 // return-events test cases for regular queue
-int test_regular_wait_for_command_buffer(cl_device_id device,
-                                         cl_context context,
-                                         cl_command_queue queue,
-                                         int num_elements)
+REGISTER_TEST(regular_wait_for_command_buffer)
 {
     int status = TEST_PASS;
     // The approach here is that test scenario which involves out-of-order
@@ -848,10 +844,7 @@ int test_regular_wait_for_command_buffer(cl_device_id device,
     return status;
 }
 
-int test_command_buffer_wait_for_command_buffer(cl_device_id device,
-                                                cl_context context,
-                                                cl_command_queue queue,
-                                                int num_elements)
+REGISTER_TEST(command_buffer_wait_for_command_buffer)
 {
     int status = TEST_PASS;
     // out-of-order command queue test
@@ -871,10 +864,7 @@ int test_command_buffer_wait_for_command_buffer(cl_device_id device,
     return status;
 }
 
-int test_command_buffer_wait_for_sec_command_buffer(cl_device_id device,
-                                                    cl_context context,
-                                                    cl_command_queue queue,
-                                                    int num_elements)
+REGISTER_TEST(command_buffer_wait_for_sec_command_buffer)
 {
     int status = TEST_PASS;
     // out-of-order command queue test
@@ -894,8 +884,7 @@ int test_command_buffer_wait_for_sec_command_buffer(cl_device_id device,
     return status;
 }
 
-int test_return_event_callback(cl_device_id device, cl_context context,
-                               cl_command_queue queue, int num_elements)
+REGISTER_TEST(return_event_callback)
 {
     int status = TEST_PASS;
     // out-of-order command queue test
@@ -913,8 +902,7 @@ int test_return_event_callback(cl_device_id device, cl_context context,
     return status;
 }
 
-int test_clwaitforevents_single(cl_device_id device, cl_context context,
-                                cl_command_queue queue, int num_elements)
+REGISTER_TEST(clwaitforevents_single)
 {
     int status = TEST_PASS;
     // out-of-order command queue test
@@ -934,8 +922,7 @@ int test_clwaitforevents_single(cl_device_id device, cl_context context,
     return status;
 }
 
-int test_clwaitforevents(cl_device_id device, cl_context context,
-                         cl_command_queue queue, int num_elements)
+REGISTER_TEST(clwaitforevents)
 {
     int status = TEST_PASS;
     // out-of-order command queue test
@@ -953,10 +940,7 @@ int test_clwaitforevents(cl_device_id device, cl_context context,
     return status;
 }
 
-int test_command_buffer_wait_for_regular(cl_device_id device,
-                                         cl_context context,
-                                         cl_command_queue queue,
-                                         int num_elements)
+REGISTER_TEST(command_buffer_wait_for_regular)
 {
     int status = TEST_PASS;
     // out-of-order command queue test
@@ -976,8 +960,7 @@ int test_command_buffer_wait_for_regular(cl_device_id device,
     return status;
 }
 
-int test_wait_for_sec_queue_event(cl_device_id device, cl_context context,
-                                  cl_command_queue queue, int num_elements)
+REGISTER_TEST(wait_for_sec_queue_event)
 {
     int status = TEST_PASS;
     // out-of-order command queue test
@@ -1000,8 +983,7 @@ int test_wait_for_sec_queue_event(cl_device_id device, cl_context context,
 //--------------------------------------------------------------------------
 // user-events test cases
 
-int test_user_event_wait(cl_device_id device, cl_context context,
-                         cl_command_queue queue, int num_elements)
+REGISTER_TEST(user_event_wait)
 {
     int status = TEST_PASS;
     // out-of-order command queue test
@@ -1019,8 +1001,7 @@ int test_user_event_wait(cl_device_id device, cl_context context,
     return status;
 }
 
-int test_user_events_wait(cl_device_id device, cl_context context,
-                          cl_command_queue queue, int num_elements)
+REGISTER_TEST(user_events_wait)
 {
     int status = TEST_PASS;
     // out-of-order command queue test
@@ -1038,8 +1019,7 @@ int test_user_events_wait(cl_device_id device, cl_context context,
     return status;
 }
 
-int test_user_event_callback(cl_device_id device, cl_context context,
-                             cl_command_queue queue, int num_elements)
+REGISTER_TEST(user_event_callback)
 {
     int status = TEST_PASS;
     // out-of-order command queue test

test_conformance/extensions/cl_khr_command_buffer/command_buffer_finalize.cpp

@@ -14,7 +14,6 @@
 // limitations under the License.
 
 #include "basic_command_buffer.h"
-#include "procs.h"
 
 namespace {
 
@@ -71,15 +70,13 @@ struct FinalizeEmpty : public BasicCommandBufferTest
 };
 } // anonymous namespace
 
-int test_finalize_invalid(cl_device_id device, cl_context context,
-                          cl_command_queue queue, int num_elements)
+REGISTER_TEST(finalize_invalid)
 {
     return MakeAndRunTest<FinalizeInvalid>(device, context, queue,
                                            num_elements);
 }
 
-int test_finalize_empty(cl_device_id device, cl_context context,
-                        cl_command_queue queue, int num_elements)
+REGISTER_TEST(finalize_empty)
 {
     return MakeAndRunTest<FinalizeEmpty>(device, context, queue, num_elements);
 }

test_conformance/extensions/cl_khr_command_buffer/command_buffer_get_command_buffer_info.cpp

@@ -15,7 +15,6 @@
 //
 
 #include "basic_command_buffer.h"
-#include "procs.h"
 
 #include <vector>
 
@@ -374,40 +373,35 @@ struct CommandBufferGetCommandBufferInfo : public BasicCommandBufferTest
 } // anonymous namespace
 
 
-int test_info_queues(cl_device_id device, cl_context context,
-                     cl_command_queue queue, int num_elements)
+REGISTER_TEST(info_queues)
 {
     return MakeAndRunTest<
         CommandBufferGetCommandBufferInfo<CombufInfoTestMode::CITM_QUEUES>>(
         device, context, queue, num_elements);
 }
 
-int test_info_ref_count(cl_device_id device, cl_context context,
-                        cl_command_queue queue, int num_elements)
+REGISTER_TEST(info_ref_count)
 {
     return MakeAndRunTest<
         CommandBufferGetCommandBufferInfo<CombufInfoTestMode::CITM_REF_COUNT>>(
         device, context, queue, num_elements);
 }
 
-int test_info_state(cl_device_id device, cl_context context,
-                    cl_command_queue queue, int num_elements)
+REGISTER_TEST(info_state)
 {
     return MakeAndRunTest<
         CommandBufferGetCommandBufferInfo<CombufInfoTestMode::CITM_STATE>>(
         device, context, queue, num_elements);
 }
 
-int test_info_prop_array(cl_device_id device, cl_context context,
-                         cl_command_queue queue, int num_elements)
+REGISTER_TEST(info_prop_array)
 {
     return MakeAndRunTest<
         CommandBufferGetCommandBufferInfo<CombufInfoTestMode::CITM_PROP_ARRAY>>(
         device, context, queue, num_elements);
 }
 
-int test_info_context(cl_device_id device, cl_context context,
-                      cl_command_queue queue, int num_elements)
+REGISTER_TEST(info_context)
 {
     return MakeAndRunTest<
         CommandBufferGetCommandBufferInfo<CombufInfoTestMode::CITM_CONTEXT>>(

test_conformance/extensions/cl_khr_command_buffer/command_buffer_out_of_order.cpp

@@ -15,7 +15,6 @@
 //
 
 #include "basic_command_buffer.h"
-#include "procs.h"
 
 #include <vector>
 
@@ -337,15 +336,13 @@ struct OutOfOrderTest : public BasicCommandBufferTest
 
 } // anonymous namespace
 
-int test_out_of_order(cl_device_id device, cl_context context,
-                      cl_command_queue queue, int num_elements)
+REGISTER_TEST(out_of_order)
 {
     return MakeAndRunTest<OutOfOrderTest<false>>(device, context, queue,
                                                  num_elements);
 }
 
-int test_simultaneous_out_of_order(cl_device_id device, cl_context context,
-                                   cl_command_queue queue, int num_elements)
+REGISTER_TEST(simultaneous_out_of_order)
 {
     return MakeAndRunTest<OutOfOrderTest<true>>(device, context, queue,
                                                 num_elements);

test_conformance/extensions/cl_khr_command_buffer/command_buffer_printf.cpp

@@ -17,7 +17,6 @@
 #include <harness/os_helpers.h>
 
 #include "basic_command_buffer.h"
-#include "procs.h"
 
 #if !defined(_WIN32)
 #if defined(__APPLE__)
@@ -524,15 +523,13 @@ struct CommandBufferPrintfTest : public BasicCommandBufferTest
 
 } // anonymous namespace
 
-int test_basic_printf(cl_device_id device, cl_context context,
-                      cl_command_queue queue, int num_elements)
+REGISTER_TEST(basic_printf)
 {
     return MakeAndRunTest<CommandBufferPrintfTest<false>>(device, context,
                                                           queue, num_elements);
 }
 
-int test_simultaneous_printf(cl_device_id device, cl_context context,
-                             cl_command_queue queue, int num_elements)
+REGISTER_TEST(simultaneous_printf)
 {
     return MakeAndRunTest<CommandBufferPrintfTest<true>>(device, context, queue,
                                                          num_elements);

test_conformance/extensions/cl_khr_command_buffer/command_buffer_profiling.cpp

@@ -15,7 +15,6 @@
 //
 
 #include "basic_command_buffer.h"
-#include "procs.h"
 
 #include <vector>
 
@@ -357,22 +356,19 @@ struct CommandBufferSubstituteQueueProfiling : public BasicCommandBufferTest
 };
 } // anonymous namespace
 
-int test_basic_profiling(cl_device_id device, cl_context context,
-                         cl_command_queue queue, int num_elements)
+REGISTER_TEST(basic_profiling)
 {
     return MakeAndRunTest<CommandBufferProfiling<false>>(device, context, queue,
                                                          num_elements);
 }
 
-int test_simultaneous_profiling(cl_device_id device, cl_context context,
-                                cl_command_queue queue, int num_elements)
+REGISTER_TEST(simultaneous_profiling)
 {
     return MakeAndRunTest<CommandBufferProfiling<true>>(device, context, queue,
                                                         num_elements);
 }
 
-int test_substitute_queue_profiling(cl_device_id device, cl_context context,
-                                    cl_command_queue queue, int num_elements)
+REGISTER_TEST(substitute_queue_profiling)
 {
     return MakeAndRunTest<CommandBufferSubstituteQueueProfiling>(
         device, context, queue, num_elements);

test_conformance/extensions/cl_khr_command_buffer/command_buffer_queue_substitution.cpp

@@ -14,7 +14,6 @@
 // limitations under the License.
 //
 #include "basic_command_buffer.h"
-#include "procs.h"
 
 #include <vector>
 
@@ -396,38 +395,31 @@ struct QueueOrderTest : public BasicCommandBufferTest
 };
 } // anonymous namespace
 
-int test_queue_substitution(cl_device_id device, cl_context context,
-                            cl_command_queue queue, int num_elements)
+REGISTER_TEST(queue_substitution)
 {
     return MakeAndRunTest<SubstituteQueueTest<false, false>>(
         device, context, queue, num_elements);
 }
 
-int test_properties_queue_substitution(cl_device_id device, cl_context context,
-                                       cl_command_queue queue, int num_elements)
+REGISTER_TEST(properties_queue_substitution)
 {
     return MakeAndRunTest<SubstituteQueueTest<true, false>>(
         device, context, queue, num_elements);
 }
 
-int test_simultaneous_queue_substitution(cl_device_id device,
-                                         cl_context context,
-                                         cl_command_queue queue,
-                                         int num_elements)
+REGISTER_TEST(simultaneous_queue_substitution)
 {
     return MakeAndRunTest<SubstituteQueueTest<false, true>>(
         device, context, queue, num_elements);
 }
 
-int test_queue_substitute_in_order(cl_device_id device, cl_context context,
-                                   cl_command_queue queue, int num_elements)
+REGISTER_TEST(queue_substitute_in_order)
 {
     return MakeAndRunTest<QueueOrderTest<false>>(device, context, queue,
                                                  num_elements);
 }
 
-int test_queue_substitute_out_of_order(cl_device_id device, cl_context context,
-                                       cl_command_queue queue, int num_elements)
+REGISTER_TEST(queue_substitute_out_of_order)
 {
     return MakeAndRunTest<QueueOrderTest<true>>(device, context, queue,
                                                 num_elements);

test_conformance/extensions/cl_khr_command_buffer/command_buffer_set_kernel_arg.cpp

@@ -15,7 +15,6 @@
 //
 
 #include "basic_command_buffer.h"
-#include "procs.h"
 
 #include <vector>
 
@@ -306,15 +305,13 @@ struct CommandBufferSetKernelArg : public BasicCommandBufferTest
 
 } // anonymous namespace
 
-int test_basic_set_kernel_arg(cl_device_id device, cl_context context,
-                              cl_command_queue queue, int num_elements)
+REGISTER_TEST(basic_set_kernel_arg)
 {
     return MakeAndRunTest<CommandBufferSetKernelArg<false>>(
         device, context, queue, num_elements);
 }
 
-int test_pending_set_kernel_arg(cl_device_id device, cl_context context,
-                                cl_command_queue queue, int num_elements)
+REGISTER_TEST(pending_set_kernel_arg)
 {
     return MakeAndRunTest<CommandBufferSetKernelArg<true>>(device, context,
                                                            queue, num_elements);

test_conformance/extensions/cl_khr_command_buffer/command_buffer_test_barrier.cpp

@@ -14,7 +14,6 @@
 // limitations under the License.
 //
 #include "basic_command_buffer.h"
-#include "procs.h"
 
 #include <vector>
 
@@ -146,8 +145,7 @@ struct BarrierWithWaitListKHR : public BasicCommandBufferTest
 };
 
 
-int test_barrier_wait_list(cl_device_id device, cl_context context,
-                           cl_command_queue queue, int num_elements)
+REGISTER_TEST(barrier_wait_list)
 {
     return MakeAndRunTest<BarrierWithWaitListKHR>(device, context, queue,
                                                   num_elements);

test_conformance/extensions/cl_khr_command_buffer/command_buffer_test_copy.cpp

@@ -16,7 +16,6 @@
 #include "basic_command_buffer.h"
 #include "svm_command_basic.h"
 #include "harness/typeWrappers.h"
-#include "procs.h"
 
 #include <vector>
 
@@ -563,42 +562,36 @@ struct CopyBufferRectKHR : public BasicCommandBufferTest
 };
 };
 
-int test_copy_image(cl_device_id device, cl_context context,
-                    cl_command_queue queue, int num_elements)
+REGISTER_TEST(copy_image)
 {
     return MakeAndRunTest<CopyImageKHR>(device, context, queue, num_elements);
 }
 
-int test_copy_buffer(cl_device_id device, cl_context context,
-                     cl_command_queue queue, int num_elements)
+REGISTER_TEST(copy_buffer)
 {
     return MakeAndRunTest<CopyBufferKHR>(device, context, queue, num_elements);
 }
 
-int test_copy_svm_buffer(cl_device_id device, cl_context context,
-                         cl_command_queue queue, int num_elements)
+REGISTER_TEST(copy_svm_buffer)
 {
     return MakeAndRunTest<CopySVMBufferKHR>(device, context, queue,
                                             num_elements);
 }
 
 
-int test_copy_buffer_to_image(cl_device_id device, cl_context context,
-                              cl_command_queue queue, int num_elements)
+REGISTER_TEST(copy_buffer_to_image)
 {
     return MakeAndRunTest<CopyBufferToImageKHR>(device, context, queue,
                                                 num_elements);
 }
 
-int test_copy_image_to_buffer(cl_device_id device, cl_context context,
-                              cl_command_queue queue, int num_elements)
+REGISTER_TEST(copy_image_to_buffer)
 {
     return MakeAndRunTest<CopyImageToBufferKHR>(device, context, queue,
                                                 num_elements);
 }
 
-int test_copy_buffer_rect(cl_device_id device, cl_context context,
-                          cl_command_queue queue, int num_elements)
+REGISTER_TEST(copy_buffer_rect)
 {
     return MakeAndRunTest<CopyBufferRectKHR>(device, context, queue,
                                              num_elements);

test_conformance/extensions/cl_khr_command_buffer/command_buffer_test_event_info.cpp

@@ -14,7 +14,6 @@
 // limitations under the License.
 //
 #include "basic_command_buffer.h"
-#include "procs.h"
 #include <vector>
 
 namespace {
@@ -222,33 +221,28 @@ struct ReferenceCount : public BasicCommandBufferTest
 };
 };
 
-int test_event_info_command_type(cl_device_id device, cl_context context,
-                                 cl_command_queue queue, int num_elements)
+REGISTER_TEST(event_info_command_type)
 {
     return MakeAndRunTest<CommandType>(device, context, queue, num_elements);
 }
 
-int test_event_info_command_queue(cl_device_id device, cl_context context,
-                                  cl_command_queue queue, int num_elements)
+REGISTER_TEST(event_info_command_queue)
 {
     return MakeAndRunTest<CommandQueue>(device, context, queue, num_elements);
 }
 
-int test_event_info_context(cl_device_id device, cl_context context,
-                            cl_command_queue queue, int num_elements)
+REGISTER_TEST(event_info_context)
 {
     return MakeAndRunTest<Context>(device, context, queue, num_elements);
 }
 
-int test_event_info_execution_status(cl_device_id device, cl_context context,
-                                     cl_command_queue queue, int num_elements)
+REGISTER_TEST(event_info_execution_status)
 {
     return MakeAndRunTest<ExecutionStatus>(device, context, queue,
                                            num_elements);
 }
 
-int test_event_info_reference_count(cl_device_id device, cl_context context,
-                                    cl_command_queue queue, int num_elements)
+REGISTER_TEST(event_info_reference_count)
 {
     return MakeAndRunTest<ReferenceCount>(device, context, queue, num_elements);
 }

test_conformance/extensions/cl_khr_command_buffer/command_buffer_test_fill.cpp

@@ -16,7 +16,6 @@
 #include "basic_command_buffer.h"
 #include "svm_command_basic.h"
 #include "harness/typeWrappers.h"
-#include "procs.h"
 
 #include <vector>
 
@@ -230,22 +229,19 @@ struct FillSVMBufferKHR : public BasicSVMCommandBufferTest
 };
 };
 
-int test_fill_buffer(cl_device_id device, cl_context context,
-                     cl_command_queue queue, int num_elements)
+REGISTER_TEST(fill_buffer)
 {
     return MakeAndRunTest<FillBufferKHR>(device, context, queue, num_elements);
 }
 
-int test_fill_svm_buffer(cl_device_id device, cl_context context,
-                         cl_command_queue queue, int num_elements)
+REGISTER_TEST(fill_svm_buffer)
 {
     return MakeAndRunTest<FillSVMBufferKHR>(device, context, queue,
                                             num_elements);
 }
 
 
-int test_fill_image(cl_device_id device, cl_context context,
-                    cl_command_queue queue, int num_elements)
+REGISTER_TEST(fill_image)
 {
     return MakeAndRunTest<FillImageKHR>(device, context, queue, num_elements);
 }

test_conformance/extensions/cl_khr_command_buffer/command_buffer_with_immutable_memory.h

@@ -1,3 +1,4 @@
+//
 // Copyright (c) 2024 The Khronos Group Inc.
 //
 // Licensed under the Apache License, Version 2.0 (the "License");
@@ -11,17 +12,25 @@
 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 // See the License for the specific language governing permissions and
 // limitations under the License.
-//
-#ifndef CL_KHR_KERNEL_CLOCK_PROCS_H
-#define CL_KHR_KERNEL_CLOCK_PROCS_H
 
-#include <CL/cl.h>
+#pragma once
 
-int test_device_scope(cl_device_id device, cl_context context,
-                      cl_command_queue queue, int num_elements);
-int test_workgroup_scope(cl_device_id device, cl_context context,
-                         cl_command_queue queue, int num_elements);
-int test_subgroup_scope(cl_device_id device, cl_context context,
-                        cl_command_queue queue, int num_elements);
+#include "basic_command_buffer.h"
+#include <type_traits>
 
-#endif /*CL_KHR_KERNEL_CLOCK_PROCS_H*/
+template <class TBase>
+struct CommandBufferWithImmutableMemoryObjectsTest : public TBase
+{
+    using TBase::TBase;
+
+    static_assert(std::is_base_of<BasicCommandBufferTest, TBase>::value,
+                  "TBase must be BasicCommandBufferTest or a derived class");
+
+    bool Skip() override
+    {
+        bool is_immutable_memory_objects_supported = is_extension_available(
+            BasicCommandBufferTest::device, "cl_ext_immutable_memory_objects");
+
+        return !is_immutable_memory_objects_supported || TBase::Skip();
+    }
+};

test_conformance/extensions/cl_khr_command_buffer/main.cpp

@@ -12,157 +12,8 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
 //
-#include "procs.h"
 #include "harness/testHarness.h"
 
-test_definition test_list[] = {
-    ADD_TEST(multi_flag_creation),
-    ADD_TEST(single_ndrange),
-    ADD_TEST(interleaved_enqueue),
-    ADD_TEST(mixed_commands),
-    ADD_TEST(explicit_flush),
-    ADD_TEST(out_of_order),
-    ADD_TEST(simultaneous_out_of_order),
-    ADD_TEST(info_queues),
-    ADD_TEST(info_ref_count),
-    ADD_TEST(info_state),
-    ADD_TEST(info_prop_array),
-    ADD_TEST(info_context),
-    ADD_TEST(basic_profiling),
-    ADD_TEST(simultaneous_profiling),
-    ADD_TEST(substitute_queue_profiling),
-    ADD_TEST(regular_wait_for_command_buffer),
-    ADD_TEST(command_buffer_wait_for_command_buffer),
-    ADD_TEST(command_buffer_wait_for_sec_command_buffer),
-    ADD_TEST(return_event_callback),
-    ADD_TEST(clwaitforevents_single),
-    ADD_TEST(clwaitforevents),
-    ADD_TEST(command_buffer_wait_for_regular),
-    ADD_TEST(wait_for_sec_queue_event),
-    ADD_TEST(user_event_wait),
-    ADD_TEST(user_events_wait),
-    ADD_TEST(user_event_callback),
-    ADD_TEST(queue_substitution),
-    ADD_TEST(properties_queue_substitution),
-    ADD_TEST(simultaneous_queue_substitution),
-    ADD_TEST(queue_substitute_in_order),
-    ADD_TEST(queue_substitute_out_of_order),
-    ADD_TEST(fill_image),
-    ADD_TEST(fill_buffer),
-    ADD_TEST(fill_svm_buffer),
-    ADD_TEST(copy_image),
-    ADD_TEST(copy_buffer),
-    ADD_TEST(copy_svm_buffer),
-    ADD_TEST(copy_buffer_to_image),
-    ADD_TEST(copy_image_to_buffer),
-    ADD_TEST(copy_buffer_rect),
-    ADD_TEST(barrier_wait_list),
-    ADD_TEST(basic_printf),
-    ADD_TEST(simultaneous_printf),
-    ADD_TEST(basic_set_kernel_arg),
-    ADD_TEST(pending_set_kernel_arg),
-    ADD_TEST(event_info_command_type),
-    ADD_TEST(event_info_command_queue),
-    ADD_TEST(event_info_execution_status),
-    ADD_TEST(event_info_context),
-    ADD_TEST(event_info_reference_count),
-    ADD_TEST(finalize_invalid),
-    ADD_TEST(finalize_empty),
-    // Command-buffer negative tests
-    ADD_TEST(negative_retain_command_buffer_invalid_command_buffer),
-    ADD_TEST(negative_release_command_buffer_invalid_command_buffer),
-    ADD_TEST(negative_finalize_command_buffer_invalid_command_buffer),
-    ADD_TEST(negative_finalize_command_buffer_not_recording_state),
-    ADD_TEST(negative_command_buffer_command_fill_buffer_queue_not_null),
-    ADD_TEST(negative_command_buffer_command_fill_buffer_context_not_same),
-    ADD_TEST(
-        negative_command_buffer_command_fill_buffer_sync_points_null_or_num_zero),
-    ADD_TEST(
-        negative_command_buffer_command_fill_buffer_invalid_command_buffer),
-    ADD_TEST(
-        negative_command_buffer_command_fill_buffer_finalized_command_buffer),
-    ADD_TEST(
-        negative_command_buffer_command_fill_buffer_mutable_handle_not_null),
-    ADD_TEST(negative_command_buffer_command_fill_image_queue_not_null),
-    ADD_TEST(negative_command_buffer_command_fill_image_context_not_same),
-    ADD_TEST(
-        negative_command_buffer_command_fill_image_sync_points_null_or_num_zero),
-    ADD_TEST(negative_command_buffer_command_fill_image_invalid_command_buffer),
-    ADD_TEST(
-        negative_command_buffer_command_fill_image_finalized_command_buffer),
-    ADD_TEST(
-        negative_command_buffer_command_fill_image_mutable_handle_not_null),
-    ADD_TEST(negative_create_command_buffer_num_queues),
-    ADD_TEST(negative_create_command_buffer_null_queues),
-    ADD_TEST(negative_create_command_buffer_repeated_properties),
-    ADD_TEST(negative_create_command_buffer_not_supported_properties),
-    ADD_TEST(negative_command_ndrange_queue_not_null),
-    ADD_TEST(negative_command_ndrange_kernel_with_different_context),
-    ADD_TEST(negative_command_ndrange_kernel_sync_points_null_or_num_zero),
-    ADD_TEST(negative_command_ndrange_kernel_invalid_command_buffer),
-    ADD_TEST(negative_command_ndrange_kernel_invalid_properties),
-    ADD_TEST(negative_command_ndrange_kernel_command_buffer_finalized),
-    ADD_TEST(negative_command_ndrange_kernel_mutable_handle_not_null),
-    ADD_TEST(negative_command_ndrange_kernel_not_support_printf),
-    ADD_TEST(negative_command_ndrange_kernel_with_enqueue_call),
-    ADD_TEST(negative_command_buffer_command_copy_buffer_queue_not_null),
-    ADD_TEST(negative_command_buffer_command_copy_buffer_different_contexts),
-    ADD_TEST(
-        negative_command_buffer_command_copy_buffer_sync_points_null_or_num_zero),
-    ADD_TEST(
-        negative_command_buffer_command_copy_buffer_invalid_command_buffer),
-    ADD_TEST(
-        negative_command_buffer_command_copy_buffer_finalized_command_buffer),
-    ADD_TEST(
-        negative_command_buffer_command_copy_buffer_mutable_handle_not_null),
-    ADD_TEST(negative_command_buffer_command_copy_image_queue_not_null),
-    ADD_TEST(negative_command_buffer_command_copy_image_different_contexts),
-    ADD_TEST(
-        negative_command_buffer_command_copy_image_sync_points_null_or_num_zero),
-    ADD_TEST(negative_command_buffer_command_copy_image_invalid_command_buffer),
-    ADD_TEST(
-        negative_command_buffer_command_copy_image_finalized_command_buffer),
-    ADD_TEST(
-        negative_command_buffer_command_copy_image_mutable_handle_not_null),
-    ADD_TEST(negative_get_command_buffer_info_invalid_command_buffer),
-    ADD_TEST(negative_get_command_buffer_info_not_supported_param_name),
-    ADD_TEST(negative_get_command_buffer_info_queues),
-    ADD_TEST(negative_get_command_buffer_info_ref_count),
-    ADD_TEST(negative_get_command_buffer_info_state),
-    ADD_TEST(negative_get_command_buffer_info_prop_array),
-    ADD_TEST(negative_get_command_buffer_info_context),
-    ADD_TEST(negative_command_buffer_command_svm_queue_not_null),
-    ADD_TEST(negative_command_buffer_command_svm_sync_points_null_or_num_zero),
-    ADD_TEST(negative_command_buffer_command_svm_invalid_command_buffer),
-    ADD_TEST(negative_command_buffer_command_svm_finalized_command_buffer),
-    ADD_TEST(negative_command_buffer_command_svm_mutable_handle_not_null),
-    ADD_TEST(negative_command_buffer_copy_image_queue_not_null),
-    ADD_TEST(negative_command_buffer_copy_image_context_not_same),
-    ADD_TEST(negative_command_buffer_copy_image_sync_points_null_or_num_zero),
-    ADD_TEST(negative_command_buffer_copy_image_invalid_command_buffer),
-    ADD_TEST(negative_command_buffer_copy_image_finalized_command_buffer),
-    ADD_TEST(negative_command_buffer_copy_image_mutable_handle_not_null),
-    ADD_TEST(negative_command_buffer_barrier_not_null_queue),
-    ADD_TEST(negative_command_buffer_barrier_invalid_command_buffer),
-    ADD_TEST(negative_command_buffer_barrier_buffer_finalized),
-    ADD_TEST(negative_command_buffer_barrier_mutable_handle_not_null),
-    ADD_TEST(negative_command_buffer_barrier_sync_points_null_or_num_zero),
-    ADD_TEST(negative_enqueue_command_buffer_invalid_command_buffer),
-    ADD_TEST(negative_enqueue_command_buffer_not_finalized),
-    ADD_TEST(
-        negative_enqueue_command_buffer_without_simultaneous_no_pending_state),
-    ADD_TEST(negative_enqueue_command_buffer_null_queues_num_queues),
-    ADD_TEST(
-        negative_enqueue_command_buffer_num_queues_not_zero_different_while_buffer_creation),
-    ADD_TEST(negative_enqueue_command_buffer_not_valid_queue_in_queues),
-    ADD_TEST(negative_enqueue_queue_with_different_context),
-    ADD_TEST(negative_enqueue_command_buffer_different_context_than_event),
-    ADD_TEST(negative_enqueue_event_wait_list_null_or_events_null),
-    ADD_TEST(negative_enqueue_queue_without_reqd_properties),
-    ADD_TEST(negative_enqueue_with_unsupported_queue_property),
-    ADD_TEST(negative_enqueue_inconsistent_device),
-};
-
 int main(int argc, const char *argv[])
 {
     // A device may report the required properties of a queue that
@@ -171,6 +22,8 @@ int main(int argc, const char *argv[])
     // for this in the tests themselves, rather than here, where we have a
     // device to query.
     const cl_command_queue_properties queue_properties = 0;
-    return runTestHarnessWithCheck(argc, argv, ARRAY_SIZE(test_list), test_list,
+    return runTestHarnessWithCheck(argc, argv,
+                                   test_registry::getInstance().num_tests(),
+                                   test_registry::getInstance().definitions(),
                                    false, queue_properties, nullptr);
 }

test_conformance/extensions/cl_khr_command_buffer/negative_command_buffer_barrier.cpp

@@ -14,7 +14,6 @@
 // limitations under the License.
 //
 #include "basic_command_buffer.h"
-#include "procs.h"
 #include <vector>
 
 //--------------------------------------------------------------------------
@@ -159,42 +158,31 @@ struct CommandBufferBarrierSyncPointsNullOrNumZero
 };
 };
 
-int test_negative_command_buffer_barrier_not_null_queue(cl_device_id device,
-                                                        cl_context context,
-                                                        cl_command_queue queue,
-                                                        int num_elements)
+REGISTER_TEST(negative_command_buffer_barrier_not_null_queue)
 {
     return MakeAndRunTest<CommandBufferBarrierNotNullQueue>(
         device, context, queue, num_elements);
 }
 
-int test_negative_command_buffer_barrier_invalid_command_buffer(
-    cl_device_id device, cl_context context, cl_command_queue queue,
-    int num_elements)
+REGISTER_TEST(negative_command_buffer_barrier_invalid_command_buffer)
 {
     return MakeAndRunTest<CommandBufferBarrierInvalidCommandBuffer>(
         device, context, queue, num_elements);
 }
 
-int test_negative_command_buffer_barrier_buffer_finalized(
-    cl_device_id device, cl_context context, cl_command_queue queue,
-    int num_elements)
+REGISTER_TEST(negative_command_buffer_barrier_buffer_finalized)
 {
     return MakeAndRunTest<CommandBufferBarrierBufferFinalized>(
         device, context, queue, num_elements);
 }
 
-int test_negative_command_buffer_barrier_mutable_handle_not_null(
-    cl_device_id device, cl_context context, cl_command_queue queue,
-    int num_elements)
+REGISTER_TEST(negative_command_buffer_barrier_mutable_handle_not_null)
 {
     return MakeAndRunTest<CommandBufferBarrierMutableHandleNotNull>(
         device, context, queue, num_elements);
 }
 
-int test_negative_command_buffer_barrier_sync_points_null_or_num_zero(
-    cl_device_id device, cl_context context, cl_command_queue queue,
-    int num_elements)
+REGISTER_TEST(negative_command_buffer_barrier_sync_points_null_or_num_zero)
 {
     return MakeAndRunTest<CommandBufferBarrierSyncPointsNullOrNumZero>(
         device, context, queue, num_elements);