Update alpaka dependency (#69)

* Remove alpaka submodule

* Delete alpaka extern folder

* Get alpaka with FetchContent

* Upgrade C++ standard to 20

* Update to latest alpaka API

* Delete unnecessary line

* Update benchmarking scripts

* Formatting

* Update alpaka API call

* Add docker image with alpaka preinstalled

* Update package version

.gitmodules

@@ -2,6 +2,3 @@
 	path = extern/pybind11
 	url = https://github.com/sbaldu/pybind11.git
 	branch = master
-[submodule "extern/alpaka"]
-	path = extern/alpaka
-	url = https://github.com/cms-patatrack/alpaka.git

CLUEstering/alpaka/AlpakaCore/CachingAllocator.h

@@ -334,8 +334,9 @@ namespace clue {
         return alpaka::allocBuf<std::byte, size_t>(device_, bytes);
       } else if constexpr (std::is_same_v<Device, alpaka::DevCpu>) {
         // allocate pinned host memory accessible by the queue's platform
-        return alpaka::allocMappedBuf<alpaka::Pltf<alpaka::Dev<Queue>>, std::byte, size_t>(
-            device_, bytes);
+        return alpaka::allocMappedBuf<alpaka::Platform<alpaka::Dev<Queue>>,
+                                      std::byte,
+                                      size_t>(device_, bytes);
       } else {
         // unsupported combination
         static_assert(std::is_same_v<Device, alpaka::Dev<Queue>> or

CLUEstering/alpaka/AlpakaCore/alpakaDevices.h

@@ -13,7 +13,7 @@ namespace clue {
 
   // alpaka host device
   inline const alpaka_common::DevHost host =
-      alpaka::getDevByIdx<alpaka_common::PltfHost>(0u);
+      alpaka::getDevByIdx(alpaka::PlatformCpu{}, 0u);
 
   // alpaka accelerator devices
   template <typename TPlatform>

CLUEstering/alpaka/AlpakaCore/getDeviceCachingAllocator.h

@@ -18,7 +18,7 @@ namespace clue {
     template <typename TDevice, typename TQueue>
     auto allocate_device_allocators() {
       using Allocator = CachingAllocator<TDevice, TQueue>;
-      auto const& devices = clue::enumerate<alpaka::Pltf<TDevice>>();
+      auto const& devices = clue::enumerate<alpaka::Platform<TDevice>>();
       auto const size = devices.size();
 
       // allocate the storage for the objects
@@ -57,9 +57,9 @@ namespace clue {
 
     size_t const index = getDeviceIndex(device);
 
-    std::vector<TDevice> devs = alpaka::getDevs<alpaka::Pltf<TDevice>>();
+    std::vector<TDevice> devs = alpaka::getDevs<alpaka::Platform<TDevice>>();
 
-    assert(index < clue::enumerate<alpaka::Pltf<TDevice>>().size());
+    assert(index < clue::enumerate<alpaka::Platform<TDevice>>().size());
 
     // the public interface is thread safe
     return allocators[index];

CLUEstering/alpaka/BindingModules/binding_cpu.cc

@@ -11,7 +11,7 @@
 namespace alpaka_serial_sync {
   void listDevices(const std::string& backend) {
     const char tab = '\t';
-    const std::vector<Device> devices = alpaka::getDevs<Platform>();
+    const std::vector<Device> devices = alpaka::getDevs(alpaka::Platform<Acc1D>());
     if (devices.empty()) {
       std::cout << "No devices found for the " << backend << " backend." << std::endl;
       return;
@@ -34,7 +34,7 @@ namespace alpaka_serial_sync {
                                         int Ndim,
                                         size_t block_size,
                                         size_t device_id) {
-    const auto dev_acc = alpaka::getDevByIdx<Acc1D>(device_id);
+    const auto dev_acc = alpaka::getDevByIdx(alpaka::Platform<Acc1D>{}, device_id);
 
     // Create the queue
     Queue queue_(dev_acc);

CLUEstering/alpaka/BindingModules/binding_cpu_tbb.cc

@@ -11,7 +11,7 @@
 namespace alpaka_tbb_async {
   void listDevices(const std::string& backend) {
     const char tab = '\t';
-    const std::vector<Device> devices = alpaka::getDevs<Platform>();
+    const std::vector<Device> devices = alpaka::getDevs(alpaka::Platform<Acc1D>());
     if (devices.empty()) {
       std::cout << "No devices found for the " << backend << " backend." << std::endl;
       return;
@@ -34,7 +34,7 @@ namespace alpaka_tbb_async {
                                         int Ndim,
                                         size_t block_size,
                                         size_t device_id) {
-    const auto dev_acc = alpaka::getDevByIdx<Acc1D>(device_id);
+    const auto dev_acc = alpaka::getDevByIdx(alpaka::Platform<Acc1D>{}, device_id);
 
     // Create the queue
     Queue queue_(dev_acc);

CLUEstering/alpaka/BindingModules/binding_gpu_cuda.cc

@@ -12,7 +12,7 @@ using cms::alpakatools::initialise;
 namespace alpaka_cuda_async {
   void listDevices(const std::string& backend) {
     const char tab = '\t';
-    const std::vector<Device> devices = alpaka::getDevs<Platform>();
+    const std::vector<Device> devices = alpaka::getDevs(alpaka::Platform<Acc1D>());
     if (devices.empty()) {
       std::cout << "No devices found for the " << backend << " backend." << std::endl;
       return;
@@ -35,7 +35,7 @@ namespace alpaka_cuda_async {
                                         int Ndim,
                                         size_t block_size,
                                         size_t device_id) {
-    const auto dev_acc = alpaka::getDevByIdx<Acc1D>(device_id);
+    const auto dev_acc = alpaka::getDevByIdx(alpaka::Platform<Acc1D>{}, device_id);
 
     // Create the queue
     Queue queue_(dev_acc);

CLUEstering/alpaka/BindingModules/binding_gpu_hip.cc

@@ -10,7 +10,7 @@
 namespace alpaka_rocm_async {
   void listDevices(const std::string& backend) {
     const char tab = '\t';
-    const std::vector<Device> devices = alpaka::getDevs<Platform>();
+    const std::vector<Device> devices = alpaka::getDevs(alpaka::Platform<Acc1D>());
     if (devices.empty()) {
       std::cout << "No devices found for the " << backend << " backend." << std::endl;
       return;
@@ -33,7 +33,7 @@ namespace alpaka_rocm_async {
                                         int Ndim,
                                         size_t block_size,
                                         size_t device_id) {
-    const auto dev_acc = alpaka::getDevByIdx<Acc1D>(device_id);
+    const auto dev_acc = alpaka::getDevByIdx(alpaka::Platform<Acc1D>{}, device_id);
 
     // Create the queue
     Queue queue_(dev_acc);

CMakeLists.txt

@@ -1,7 +1,7 @@
 cmake_minimum_required(VERSION 3.16.0)
 project(CLUEstering LANGUAGES CXX)
 
-set(CMAKE_CXX_STANDARD 17)
+set(CMAKE_CXX_STANDARD 20)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 
@@ -16,18 +16,33 @@ elseif(${CMAKE_BUILD_TYPE} STREQUAL "Release")
   set(CMAKE_CXX_FLAGS "-O2")
 endif()
 
-# include alpaka extern subfolder
-include_directories(extern/alpaka/include)
 # include pybind11 extern subfolder
 set(PYBIND11_FINDPYTHON ON)
 set(PYBIND11_PYTHON_VERSION ">=3.8")
 add_subdirectory(extern/pybind11)
 
-find_package(Boost 1.75.0)
+include(FetchContent)
+
+find_package(alpaka)
+if (NOT alpaka_FOUND)
+	FetchContent_Declare(
+		alpaka
+		GIT_REPOSITORY https://github.com/alpaka-group/alpaka.git
+		GIT_TAG develop
+	)
+
+	FetchContent_GetProperties(alpaka)
+	if(NOT alpaka_POPULATED)
+		FetchContent_Populate(alpaka)
+	endif()
+	set(alpaka_PATH ./build/_deps/alpaka-src/include)
+else()
+	set(alpaka_PATH ${alpaka_INCLUDE_DIRS})
+endif()
 
+find_package(Boost 1.75.0)
 # if boost is not found, it's fetched from the official boost repository
 if(NOT Boost_FOUND)
-  include(FetchContent)
   FetchContent_Declare(
     boost
     URL https://boostorg.jfrog.io/artifactory/main/release/1.76.0/source/boost_1_76_0.tar.gz
@@ -55,6 +70,7 @@ execute_process(COMMAND mkdir -p ./CLUEstering/lib)
 # Convolutional Kernels compile convolutional kernel module
 pybind11_add_module(CLUE_Convolutional_Kernels SHARED
                     ./CLUEstering/alpaka/BindingModules/binding_kernels.cc)
+target_include_directories(CLUE_Convolutional_Kernels PRIVATE ${alpaka_PATH})
 # link boost
 target_link_libraries(CLUE_Convolutional_Kernels PRIVATE ${Boost_LIBRARIES})
 target_include_directories(CLUE_Convolutional_Kernels PRIVATE ${Boost_PATH})
@@ -79,6 +95,7 @@ endforeach()
 # CPU Serial compile cpu serial module
 pybind11_add_module(CLUE_CPU_Serial SHARED
                     ./CLUEstering/alpaka/BindingModules/binding_cpu.cc)
+target_include_directories(CLUE_CPU_Serial PRIVATE ${alpaka_PATH})
 # link boost
 target_link_libraries(CLUE_CPU_Serial PRIVATE ${Boost_LIBRARIES})
 target_include_directories(CLUE_CPU_Serial PRIVATE ${Boost_PATH})
@@ -106,6 +123,7 @@ if(NOT ${SERIAL_ONLY})
     # compile cpu tbb module
     pybind11_add_module(CLUE_CPU_TBB SHARED
                         ./CLUEstering/alpaka/BindingModules/binding_cpu_tbb.cc)
+	target_include_directories(CLUE_CPU_TBB PRIVATE ${alpaka_PATH})
     target_link_libraries(CLUE_CPU_TBB PRIVATE ${Boost_LIBRARIES})
     target_include_directories(CLUE_CPU_TBB PRIVATE ${Boost_PATH})
     target_compile_options(
@@ -149,6 +167,7 @@ if((NOT ${CPU_ONLY}) AND (NOT ${SERIAL_ONLY}))
     # compile gpu cuda module
     pybind11_add_module(CLUE_GPU_CUDA SHARED
                         ./CLUEstering/alpaka/BindingModules/binding_gpu_cuda.cc)
+	target_include_directories(CLUE_GPU_CUDA PRIVATE ${alpaka_PATH})
     # link boost
     target_link_libraries(CLUE_GPU_CUDA PRIVATE ${Boost_LIBRARIES})
     target_include_directories(CLUE_GPU_CUDA PRIVATE ${Boost_PATH})
@@ -191,6 +210,7 @@ if((NOT ${CPU_ONLY}) AND (NOT ${SERIAL_ONLY}))
     # compile gpu hip module
     pybind11_add_module(CLUE_GPU_HIP SHARED
                         ./CLUEstering/alpaka/BindingModules/binding_gpu_hip.cc)
+	target_include_directories(CLUE_GPU_HIP PRIVATE ${alpaka_PATH})
     # link boost
     target_link_libraries(CLUE_GPU_HIP PRIVATE ${Boost_LIBRARIES})
     target_include_directories(CLUE_GPU_HIP PRIVATE ${Boost_PATH})

benchmark/dataset_size/CMakeLists.txt

@@ -16,15 +16,31 @@ elseif(${CMAKE_BUILD_TYPE} STREQUAL "Release")
 endif()
 
 # include alpaka extern subfolder
-include_directories(../../extern/alpaka/include)
 add_subdirectory(pybind11)
 
+include(FetchContent)
+
+find_package(alpaka)
+if (NOT alpaka_FOUND)
+	FetchContent_Declare(
+		alpaka
+		GIT_REPOSITORY https://github.com/alpaka-group/alpaka.git
+		GIT_TAG develop
+	)
+
+	FetchContent_GetProperties(alpaka)
+	if(NOT alpaka_POPULATED)
+		FetchContent_Populate(alpaka)
+	endif()
+	set(alpaka_PATH ./build/_deps/alpaka-src/include)
+else()
+	set(alpaka_PATH ${alpaka_INCLUDE_DIRS})
+endif()
+
 # look for boost
 find_package(Boost 1.75.0)
-
 # if boost is not found, it's fetched from the official boost repository
 if(NOT Boost_FOUND)
-  include(FetchContent)
   FetchContent_Declare(
     boost
     URL https://boostorg.jfrog.io/artifactory/main/release/1.76.0/source/boost_1_76_0.tar.gz
@@ -51,6 +67,7 @@ include_directories(../../CLUEstering/alpaka/)
 
 # CPU Serial
 add_executable(serial.out main.cc)
+target_include_directories(serial.out PRIVATE ${alpaka_PATH})
 target_link_libraries(serial.out PRIVATE pybind11::embed)
 # link boost
 target_link_libraries(serial.out PRIVATE ${Boost_LIBRARIES})
@@ -74,6 +91,7 @@ if(NOT ${SERIAL_ONLY})
   if(TBB_FOUND)
     # compile cpu tbb module
     add_executable(tbb.out main.cc)
+	target_include_directories(tbb.out PRIVATE ${alpaka_PATH})
 	target_link_libraries(tbb.out PRIVATE pybind11::embed)
     target_link_libraries(tbb.out PRIVATE ${Boost_LIBRARIES})
     target_include_directories(tbb.out PRIVATE ${Boost_PATH})
@@ -114,6 +132,7 @@ if((NOT ${CPU_ONLY}) AND (NOT ${SERIAL_ONLY}))
     set_source_files_properties(main.cc PROPERTIES LANGUAGE CUDA)
     # compile gpu cuda module
     add_executable(cuda.out main.cc)
+	target_include_directories(cuda.out PRIVATE ${alpaka_PATH})
 	target_link_libraries(cuda.out PRIVATE pybind11::embed)
     # link boost
     target_link_libraries(cuda.out PRIVATE ${Boost_LIBRARIES})

benchmark/dataset_size/main.cc

@@ -85,7 +85,7 @@ void to_csv(const TimeMeasures& measures, const std::string& filename) {
 
 namespace ALPAKA_ACCELERATOR_NAMESPACE {
   void run(const std::string& input_file) {
-    const auto dev_acc = alpaka::getDevByIdx<Acc1D>(0u);
+    const auto dev_acc = alpaka::getDevByIdx(alpaka::Platform<Acc1D>{}, 0u);
     Queue queue_(dev_acc);
     const auto data{read_csv<float, 2>(input_file)};
     Points<2> h_points(data.first, data.second);

benchmark/profiling/CMakeLists.txt

@@ -15,15 +15,30 @@ elseif(${CMAKE_BUILD_TYPE} STREQUAL "Release")
   set(CMAKE_CXX_FLAGS "-O2")
 endif()
 
-# include alpaka extern subfolder
-include_directories(../../extern/alpaka/include)
+include(FetchContent)
+
+find_package(alpaka)
+if (NOT alpaka_FOUND)
+	FetchContent_Declare(
+		alpaka
+		GIT_REPOSITORY https://github.com/alpaka-group/alpaka.git
+		GIT_TAG develop
+	)
+
+	FetchContent_GetProperties(alpaka)
+	if(NOT alpaka_POPULATED)
+		FetchContent_Populate(alpaka)
+	endif()
+	set(alpaka_PATH ./build/_deps/alpaka-src/include)
+else()
+	set(alpaka_PATH ${alpaka_INCLUDE_DIRS})
+endif()
+
 
 # look for boost
 find_package(Boost 1.75.0)
-
 # if boost is not found, it's fetched from the official boost repository
 if(NOT Boost_FOUND)
-  include(FetchContent)
   FetchContent_Declare(
     boost
     URL https://boostorg.jfrog.io/artifactory/main/release/1.76.0/source/boost_1_76_0.tar.gz
@@ -50,6 +65,7 @@ include_directories(../../CLUEstering/alpaka/)
 
 # CPU Serial
 add_executable(serial.out main.cc)
+target_include_directories(serial.out PRIVATE ${alpaka_PATH})
 # link boost
 target_link_libraries(serial.out PRIVATE ${Boost_LIBRARIES})
 target_include_directories(serial.out PRIVATE ${Boost_PATH})
@@ -72,6 +88,7 @@ if(NOT ${SERIAL_ONLY})
   if(TBB_FOUND)
     # compile cpu tbb module
     add_executable(tbb.out main.cc)
+	target_include_directories(tbb.out PRIVATE ${alpaka_PATH})
     target_link_libraries(tbb.out PRIVATE ${Boost_LIBRARIES})
     target_include_directories(tbb.out PRIVATE ${Boost_PATH})
     target_compile_options(
@@ -111,6 +128,7 @@ if((NOT ${CPU_ONLY}) AND (NOT ${SERIAL_ONLY}))
     set_source_files_properties(main.cc PROPERTIES LANGUAGE CUDA)
     # compile gpu cuda module
     add_executable(cuda.out main.cc)
+	target_include_directories(cuda.out PRIVATE ${alpaka_PATH})
     # link boost
     target_link_libraries(cuda.out PRIVATE ${Boost_LIBRARIES})
     target_include_directories(cuda.out PRIVATE ${Boost_PATH})

benchmark/profiling/main.cc

@@ -13,7 +13,7 @@
 
 namespace ALPAKA_ACCELERATOR_NAMESPACE {
   void run(const std::string& input_file) {
-    const auto dev_acc = alpaka::getDevByIdx<Acc1D>(0u);
+    const auto dev_acc = alpaka::getDevByIdx(alpaka::Platform<Acc1D>{}, 0u);
     Queue queue_(dev_acc);
     const auto data{read_csv<float, 2>(input_file)};
     Points<2> h_points(data.first, data.second);

docker/alpaka_preinstalled/Dockerfile

@@ -0,0 +1,18 @@
+
+FROM ubuntu:22.04
+
+RUN apt-get update && apt-get install -y \
+		python3-dev \
+		python3-pip \
+		cmake \
+		g++ \
+		libboost-all-dev \
+		git
+
+RUN git clone https://github.com/alpaka-group/alpaka.git \
+	&& cd alpaka \
+	&& cmake -B build -S . \
+	&& cmake --build build \
+	&& cmake --install build
+
+CMD ["/bin/bash"]

setup.py

@@ -4,7 +4,7 @@
 from setuptools import setup
 import subprocess
 
-__version__ = "2.2.11"
+__version__ = "2.2.12"
 
 this_directory = Path(__file__).parent
 long_description = (this_directory/'README.md').read_text()