Merge pull request #25 from clEsperanto/use-new-backend-gpu

make it main

.github/workflows/build-and-run.yaml

@@ -0,0 +1,31 @@
+name: Build and Run
+
+on:
+  push:
+    branches: [ use-new-backend-gpu ]
+  pull_request:
+    branches: [ use-new-backend-gpu ]
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v2
+
+    - name: Set up JDK
+      uses: actions/setup-java@v2
+      with:
+        java-version: '11'
+        distribution: 'adopt'
+    - name: Find available GPUs
+      run: lspci | grep -i vga
+    - name: Install openCL
+      run: sudo apt install ocl-icd-opencl-dev
+    - name: Check installation
+      run: dpkg -L ocl-icd-opencl-dev
+    - name: Build with Maven
+      run: mvn clean install
+
+    - name: Run Java class
+      run: java -cp target/classes com.example.YourMainClass

native/clesperantoj/CMakeLists.txt

@@ -18,24 +18,20 @@ set(CMAKE_CXX_STANDARD_REQUIRED ON) # Require (at least) it
 set(CMAKE_CXX_EXTENSIONS OFF) # Don't use e.g. GNU extension (like -std=gnu++11) for portability
 set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON) # Export all symbols for shared library windows
 
-include(FetchContent)
-option(BUILD_TESTING OFF)
-option(BUILD_BENCHMARK OFF)
-FetchContent_Declare(
-  clic_lib
-  GIT_REPOSITORY https://github.com/clEsperanto/CLIc_prototype.git
-  GIT_TAG 0.6.4
-)
-FetchContent_MakeAvailable(clic_lib)
+## CLIc dependency
+option(BUILD_TESTING  OFF)
+option(BUILD_BENCHMARK  OFF)
+option(BUILD_SHARED_LIBS ON)
+set(CLIC_REPO_TAG 0.10.2) # branch name for dev
+set(CLIC_REPO_URL https://github.com/clEsperanto/CLIc_prototype.git)
+add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/clic)
 
 file(GLOB_RECURSE WRAPPER_SOURCES_FILES ${CMAKE_CURRENT_SOURCE_DIR}/src/*.cpp)
 file(GLOB_RECURSE WRAPPER_HEADERS_FILES ${CMAKE_CURRENT_SOURCE_DIR}/include/*.hpp)
 set(WRAPPER_DIR ${CMAKE_CURRENT_SOURCE_DIR}/include)
 
 add_library(${LIB_PACKAGE_NAME} SHARED ${WRAPPER_SOURCES_FILES})
-
 target_link_libraries(${LIB_PACKAGE_NAME} PUBLIC CLIc::CLIc)
-
 target_include_directories(${LIB_PACKAGE_NAME} PUBLIC "$<BUILD_INTERFACE:${WRAPPER_DIR}>")
 
 install(TARGETS ${LIB_PACKAGE_NAME} DESTINATION .)

native/clesperantoj/clic/CMakeLists.txt

@@ -0,0 +1,13 @@
+## fetch CLIc from repo release
+include(FetchContent)
+
+option(BUILD_SHARED_LIBS OFF)
+option(BUILD_TESTING OFF)
+option(BUILD_BENCHMARK OFF)
+
+FetchContent_Declare(
+  CLIc
+  GIT_REPOSITORY ${CLIC_REPO_URL}
+  GIT_TAG ${CLIC_REPO_TAG}
+)
+FetchContent_MakeAvailable(CLIc)

native/clesperantoj/cppbuild.sh

@@ -23,7 +23,7 @@ case $PLATFORM in
         ;;
     macosx-*)
         # the following line might not be necessary if make would be properly installed in the path
-        CMAKE=/Applications/CMake.app/Contents/bin/cmake
+        # CMAKE=/Applications/CMake.app/Contents/bin/cmake
 
         $CMAKE -DCMAKE_BUILD_TYPE=Release \
                -DCMAKE_INSTALL_PREFIX="../../../lib/macosx/" ..
@@ -36,8 +36,9 @@ case $PLATFORM in
     windows-x86_64)
         $CMAKE -G"NMake Makefiles" \
 		       -DCMAKE_BUILD_TYPE=Release \
-               -DCMAKE_INSTALL_PREFIX="../../../lib/win64/" ..
-            #    -DOPENCL_INCLUDE_DIR="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v11.2/include/" \
+               -DCMAKE_INSTALL_PREFIX="../../../lib/win64/"  ..
+            #   -DOpenCL_INCLUDE_DIR="C:/Program Files (x86)/OCL_SDK_Light/include" \
+            #   -DOpenCL_LIBRARY="C:/Program Files (x86)/OCL_SDK_Light/lib/x86_64/opencl.lib" ..
             #    -DCLIC_INCLUDE_DIR="../../../CLIc_prototype/clic/include/core" \
             #    -DOCLCL_INCLUDE_DIR="../../../CLIc_prototype/thirdparty/opencl/ocl-clhpp/include" \
 		    #    -DCLFFT_LIBRARY_DIR="C:/OpenCL/clFFT-2.12.2-Windows-x64/lib64/import/" .. 

native/clesperantoj/include/clesperantoj.hpp

@@ -5,68 +5,78 @@
 #include <string>
 #include <vector>
 
-#include "cleImage.hpp"
-#include "cleProcessor.hpp"
+#include "array.hpp"
+#include "backend.hpp"
+#include "device.hpp"
+#include "utils.hpp"
 
-class ProcessorJ
+class BackendJ
 {
-    friend class BufferJ;
-    friend class MemoryJ;
+public:
+    static void setBackend(const std::string &backendName);
+};
+
+class DeviceJ
+{
+    friend class BackendJ;
 
 private:
-    cle::Processor proc;
-    ProcessorJ(const cle::Processor &proc);
+    std::shared_ptr<cle::Device> device_;
 
 public:
-    ProcessorJ();
-    ProcessorJ(const std::string &name);
-    std::vector<std::string> getAvailableDevices();
-    void setDevice(const std::string &name);
-    std::string getDevice();
+    DeviceJ();
 
-    cle::Processor get() const;
-    std::shared_ptr<cle::Processor> getShared() const;
-};
+    static std::vector<std::string> getAvailableDevices(const std::string &deviceType = "all");
 
-class BufferJ
+    void setDevice(const std::string &deviceName = "", const std::string &deviceType = "all");
+    std::string getName();
+    std::string getInfo();
+
+    std::shared_ptr<cle::Device> get() const;
+};
+class ArrayJ
 {
-    friend class MemoryJ;
 
 private:
-    cle::Image buffer;
-    BufferJ(const cle::Image &buffer);
+    std::shared_ptr<cle::Array> array_;
+
+    friend class MemoryJ;
+
+protected:
+    static ArrayJ create(size_t width, size_t height, size_t depth, size_t dimension, const cle::dType &data_type, const cle::mType &memory_type, const DeviceJ &device);
+    void write(void *data) const;
+    void read(void *data) const;
 
 public:
-    BufferJ();
+    ArrayJ() = default;
+    ArrayJ(const std::shared_ptr<cle::Array> &array);
 
     size_t getWidth();
     size_t getHeight();
     size_t getDepth();
-    void getShape(size_t *shape);
     unsigned int getDimension();
 
     std::string getDataType();
     std::string getMemoryType();
     std::string getDevice();
 
     void fillMemory(float value);
-    void copyDataTo(BufferJ &dst);
+    void copyDataTo(ArrayJ &dst);
 
-    cle::Image get() const;
-    std::shared_ptr<cle::Image> getShared() const;
+    std::shared_ptr<cle::Array> get() const;
 };
 
 class MemoryJ
 {
 public:
-    static BufferJ makeFloatBuffer(const ProcessorJ &device, const size_t &width, const size_t &height, const size_t &depth, const std::string &memory_type);
-    static BufferJ makeIntBuffer(const ProcessorJ &device, const size_t &width, const size_t &height, const size_t &depth, const std::string &memory_type);
+    static ArrayJ makeFloatBuffer(const DeviceJ &device, const size_t &width, const size_t &height, const size_t &depth, const size_t &dimension, const std::string &memory_type);
+    static ArrayJ makeIntBuffer(const DeviceJ &device, const size_t &width, const size_t &height, const size_t &depth, const size_t &dimension, const std::string &memory_type);
 
-    static void writeFloatBuffer(const BufferJ &buffer, float *data, const size_t &size);
-    static void writeIntBuffer(const BufferJ &buffer, int *data, const size_t &size);
+    static void writeFloatBuffer(const ArrayJ &array, float *data, const size_t &size);
+    static void writeIntBuffer(const ArrayJ &array, int *data, const size_t &size);
 
-    static void readFloatBuffer(const BufferJ &buffer, float *data, const size_t &size);
-    static void readIntBuffer(const BufferJ &buffer, int *data, const size_t &size);
+    static void readFloatBuffer(const ArrayJ &array, float *data, const size_t &size);
+    static void readIntBuffer(const ArrayJ &array, int *data, const size_t &size);
 };
 
 #endif // __INCLUDE_CLESPERANTOJ_HPP

native/clesperantoj/include/kernelj.hpp

@@ -6,8 +6,10 @@
 class Tier1
 {
 public:
-    static void addImageAndScalar(const ProcessorJ &proc, const BufferJ &src, const BufferJ &dst, const float &scalar);
-    static void gaussianBlur(const ProcessorJ &proc, const BufferJ &src, const BufferJ &dst, const float &sigmaX, const float &sigmaY, const float &sigmaZ);
+    static ArrayJ absolute(const DeviceJ &dev, const ArrayJ &src, ArrayJ &dst);
+    static ArrayJ absolute(const DeviceJ &dev, const ArrayJ &src);
+    static void gaussianBlur(const DeviceJ &dev, const ArrayJ &src, const ArrayJ &dst, const float &sigmaX, const float &sigmaY, const float &sigmaZ);
+    static void addImageAndScalar(const DeviceJ &dev, const ArrayJ &src, const ArrayJ &dst, const float &scalar);
 };
 
 #endif // __INCLUDE_KERNELJ_HPP