Update CI to test with Vc, alpaka and CUDA

.github/workflows/ci.yaml

@@ -1,24 +1,15 @@
 name: CI
 on: [push, pull_request]
 
+env:
+  THREADS: 4
+  CONFIG: RelWithDebInfo
+
 jobs:
-  build-ubuntu:
+  amalgamation:
     runs-on: ubuntu-latest
     steps:
     - uses: actions/checkout@v2
-    - name: vcpkg install dependencies
-      run: |
-        vcpkg install catch2 fmt
-    - name: cmake
-      run: |
-        mkdir build
-        cd build
-        export BOOST_ROOT=$BOOST_ROOT_1_72_0
-        CXX=g++-10 cmake .. -DCMAKE_BUILD_TYPE=RelWithDebInfo -DASAN_FOR_TESTS=ON -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake
-    - name: build
-      run: cmake --build build
-    - name: tests
-      run: build/tests
     - name: clone amalgamate
       uses: actions/checkout@v2
       with:
@@ -34,6 +25,85 @@ jobs:
         name: llama.hpp
         path: build/llama-amalgamated/llama.hpp
 
+  build-ubuntu-gcc9-cuda:
+    env:
+      CUDA_INSTALLER: cuda_11.1.1_455.32.00_linux.run
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v2
+    - name: vcpkg install dependencies
+      run: |
+        vcpkg install catch2 fmt tbb
+    - name: cache CUDA 11
+      id: cache-cude
+      uses: actions/cache@v2
+      with:
+        path: ${{env.CUDA_INSTALLER}}
+        key: ${{env.CUDA_INSTALLER}}
+    - name: install CUDA 11
+      if: steps.cache-cuda.outputs.cache-hit != 'true'
+      run: |
+        wget --no-verbose https://developer.download.nvidia.com/compute/cuda/11.1.1/local_installers/$CUDA_INSTALLER
+        sudo sh $CUDA_INSTALLER --silent --toolkit
+    - name: install alpaka
+      run: |
+        git clone https://github.com/alpaka-group/alpaka.git
+        mkdir alpaka/build
+        cd alpaka/build
+        export BOOST_ROOT=$BOOST_ROOT_1_72_0
+        cmake .. -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake
+        sudo cmake --build . --target install
+    - name: install Vc
+      run: |
+        git clone https://github.com/VcDevel/Vc.git
+        mkdir Vc/build
+        cd Vc/build
+        cmake .. -DBUILD_TESTING=OFF -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake
+        sudo cmake --build . --target install
+    - name: cmake
+      run: |
+        mkdir build
+        cd build
+        export BOOST_ROOT=$BOOST_ROOT_1_72_0
+        CXX=g++-9 cmake .. -DCMAKE_BUILD_TYPE=$CONFIG -DASAN_FOR_TESTS=ON -DALPAKA_CXX_STANDARD=17 -DALPAKA_ACC_GPU_CUDA_ENABLE=ON -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake
+    - name: build tests + examples
+      run: cmake --build build -j $THREADS
+    - name: run tests
+      run: build/tests
+
+  build-ubuntu-gcc10:
+    runs-on: ubuntu-latest
+    steps:
+    - uses: actions/checkout@v2
+    - name: vcpkg install dependencies
+      run: |
+        vcpkg install catch2 fmt
+    - name: install alpaka
+      run: |
+        git clone https://github.com/alpaka-group/alpaka.git
+        mkdir alpaka/build
+        cd alpaka/build
+        export BOOST_ROOT=$BOOST_ROOT_1_72_0
+        cmake .. -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake
+        sudo cmake --build . --target install
+    - name: install Vc
+      run: |
+        git clone https://github.com/VcDevel/Vc.git
+        mkdir Vc/build
+        cd Vc/build
+        cmake .. -DBUILD_TESTING=OFF -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake
+        sudo cmake --build . --target install
+    - name: cmake
+      run: |
+        mkdir build
+        cd build
+        export BOOST_ROOT=$BOOST_ROOT_1_72_0
+        CXX=g++-10 cmake .. -DCMAKE_BUILD_TYPE=$CONFIG -DASAN_FOR_TESTS=ON -DALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE=ON -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake
+    - name: build tests + examples
+      run: cmake --build build -j $THREADS
+    - name: run tests
+      run: build/tests
+
   build-windows:
     runs-on: windows-latest
     env:
@@ -48,13 +118,30 @@ jobs:
     - name: vcpkg install dependencies
       run: |
         & $env:VCPKG_ROOT/vcpkg install catch2 fmt
+    - name: install alpaka
+      run: |
+        git clone https://github.com/alpaka-group/alpaka.git
+        mkdir alpaka/build
+        cd alpaka/build
+        $env:BOOST_ROOT = $env:BOOST_ROOT_1_72_0
+        cmake .. "-DCMAKE_TOOLCHAIN_FILE=$env:VCPKG_ROOT/scripts/buildsystems/vcpkg.cmake"
+        cmake --build . --target install --config $env:CONFIG
+# Vc does not currently compile with MSVC
+#    - name: install Vc
+#      run: |
+#        git clone https://github.com/VcDevel/Vc.git
+#        mkdir Vc/build
+#        cd Vc/build
+#        cmake .. -DBUILD_TESTING=OFF "-DCMAKE_TOOLCHAIN_FILE=$env:VCPKG_ROOT/scripts/buildsystems/vcpkg.cmake"
+#        cmake --build . --target install --config $env:CONFIG
     - name: cmake
       run: |
         mkdir build
         cd build
         $env:BOOST_ROOT = $env:BOOST_ROOT_1_72_0
-        cmake .. "-DCMAKE_TOOLCHAIN_FILE=$env:VCPKG_ROOT/scripts/buildsystems/vcpkg.cmake"
-    - name: build
-      run: cmake --build build --config RelWithDebInfo
-    - name: tests
-      run: build/RelWithDebInfo/tests
+        cmake .. -DALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE=ON "-DCMAKE_TOOLCHAIN_FILE=$env:VCPKG_ROOT/scripts/buildsystems/vcpkg.cmake"
+    - name: build tests + examples
+      run: cmake --build build -j $env:THREADS --config $env:CONFIG
+    - name: run tests
+      run: |
+        & build/$env:CONFIG/tests

examples/alpaka/asyncblur/asyncblur.cpp

@@ -84,13 +84,12 @@ struct BlurKernel
             {
                 // Using SoA for the shared memory
                 constexpr auto sharedChunkSize = ElemsPerBlock + 2 * KernelSize;
-                const auto sharedMapping = llama::mapping::tree::Mapping(
+                const auto sharedMapping = llama::mapping::SoA(
                     typename View::ArrayDomain{sharedChunkSize, sharedChunkSize},
-                    llama::Tuple{llama::mapping::tree::functor::LeafOnlyRT()},
                     typename View::DatumDomain{});
                 constexpr auto sharedMemSize = llama::sizeOf<PixelOnAcc> * sharedChunkSize * sharedChunkSize;
                 auto& sharedMem = alpaka::allocVar<std::byte[sharedMemSize], __COUNTER__>(acc);
-                return llama::View{sharedMapping, llama::Array{&sharedMem[0]}};
+                return llama::View(sharedMapping, llama::Array<std::byte*, 1>{&sharedMem[0]});
             }
             else
                 return int{}; // dummy

examples/alpaka/nbody/nbody.cpp

@@ -150,32 +150,32 @@ struct MoveKernel
     }
 };
 
+using Dim = alpaka::DimInt<1>;
+using Size = std::size_t;
+
+using Acc = alpaka::ExampleDefaultAcc<Dim, Size>;
+// using Acc = alpaka::AccGpuCudaRt<Dim, Size>;
+// using Acc = alpaka::AccCpuSerial<Dim, Size>;
+
+using DevHost = alpaka::DevCpu;
+using DevAcc = alpaka::Dev<Acc>;
+using PltfHost = alpaka::Pltf<DevHost>;
+using PltfAcc = alpaka::Pltf<DevAcc>;
+using Queue = alpaka::Queue<DevAcc, alpaka::Blocking>;
+
+constexpr std::size_t hardwareThreads = 2; // relevant for OpenMP2Threads
+using Distribution = common::ThreadsElemsDistribution<Acc, BLOCK_SIZE, hardwareThreads>;
+constexpr std::size_t elemCount = Distribution::elemCount;
+constexpr std::size_t threadCount = Distribution::threadCount;
+
 int main()
 {
-    using Dim = alpaka::DimInt<1>;
-    using Size = std::size_t;
-
-    using Acc = alpaka::ExampleDefaultAcc<Dim, Size>;
-    // using Acc = alpaka::AccGpuCudaRt<Dim, Size>;
-    // using Acc = alpaka::AccCpuSerial<Dim, Size>;
-
-    using DevHost = alpaka::DevCpu;
-    using DevAcc = alpaka::Dev<Acc>;
-    using PltfHost = alpaka::Pltf<DevHost>;
-    using PltfAcc = alpaka::Pltf<DevAcc>;
-    using Queue = alpaka::Queue<DevAcc, alpaka::Blocking>;
     const DevAcc devAcc(alpaka::getDevByIdx<PltfAcc>(0u));
     const DevHost devHost(alpaka::getDevByIdx<PltfHost>(0u));
     Queue queue(devAcc);
 
-    // NBODY
-    constexpr std::size_t hardwareThreads = 2; // relevant for OpenMP2Threads
-    using Distribution = common::ThreadsElemsDistribution<Acc, BLOCK_SIZE, hardwareThreads>;
-    constexpr std::size_t elemCount = Distribution::elemCount;
-    constexpr std::size_t threadCount = Distribution::threadCount;
     constexpr FP ts = 0.0001;
 
-    // LLAMA
     const auto arrayDomain = llama::ArrayDomain{PROBLEM_SIZE};
 
     const auto mapping = [&] {

include/llama/Array.hpp

@@ -66,7 +66,7 @@ namespace llama
 
         LLAMA_FN_HOST_ACC_INLINE constexpr friend auto operator+(const Array& a, const Array& b) -> Array
         {
-            Array temp;
+            Array temp{};
             for (std::size_t i = 0; i < N; ++i)
                 temp[i] = a[i] + b[i];
             return temp;