implement alpaka::concepts::Tag

.github/workflows/ci.yml

@@ -16,7 +16,7 @@ concurrency:
 # Therefore we aim to have each value been set in at lest one job.
 # ALPAKA_CI_CXX                                 : {g++, clang++}
 #   [g++] ALPAKA_CI_GCC_VER                     : {10, 11, 12, 13}
-#   [clang++] ALPAKA_CI_CLANG_VER               : {10, 11, 12, 13, 14}
+#   [clang++] ALPAKA_CI_CLANG_VER               : {13, 14, 15, 16, 17, 18, 19}
 #   [cl.exe] ALPAKA_CI_CL_VER                   : {2022}
 #   ALPAKA_CI_STDLIB                            : {libstdc++, [ALPAKA_CI_CXX==clang++]:libc++}
 # CMAKE_BUILD_TYPE                              : {Debug, Release}
@@ -173,18 +173,18 @@ jobs:
         #  - libc++
         #  - sanitizer jobs
         # clang++
-        - name: linux_clang-10_release
+        - name: linux_clang-13_release
           os: ubuntu-20.04
-          env: {ALPAKA_CI_CXX: clang++,  ALPAKA_CI_CLANG_VER: 10,     ALPAKA_CI_STDLIB: libc++,    CMAKE_BUILD_TYPE: Release, ALPAKA_BOOST_VERSION: 1.75.0, ALPAKA_CI_CMAKE_VER: 3.25.3, OMP_NUM_THREADS: 4, alpaka_ACC_CPU_B_TBB_T_SEQ_ENABLE: OFF, CMAKE_CXX_EXTENSIONS: OFF}
+          env: {ALPAKA_CI_CXX: clang++,  ALPAKA_CI_CLANG_VER: 13,     ALPAKA_CI_STDLIB: libc++,    CMAKE_BUILD_TYPE: Release, ALPAKA_BOOST_VERSION: 1.75.0, ALPAKA_CI_CMAKE_VER: 3.25.3, OMP_NUM_THREADS: 4, alpaka_ACC_CPU_B_TBB_T_SEQ_ENABLE: OFF, CMAKE_CXX_EXTENSIONS: OFF}
           container: ubuntu:20.04
         # clang-11 tested in GitLab CI
-        - name: linux_clang-12_release
+        - name: linux_clang-14_release
           os: ubuntu-20.04
-          env: {ALPAKA_CI_CXX: clang++,  ALPAKA_CI_CLANG_VER: 12,     ALPAKA_CI_STDLIB: libc++,    CMAKE_BUILD_TYPE: Release, ALPAKA_BOOST_VERSION: 1.77.0, ALPAKA_CI_CMAKE_VER: 3.25.3, OMP_NUM_THREADS: 4, alpaka_ACC_CPU_B_TBB_T_SEQ_ENABLE: OFF, CMAKE_CXX_EXTENSIONS: OFF}
+          env: {ALPAKA_CI_CXX: clang++,  ALPAKA_CI_CLANG_VER: 14,     ALPAKA_CI_STDLIB: libc++,    CMAKE_BUILD_TYPE: Release, ALPAKA_BOOST_VERSION: 1.77.0, ALPAKA_CI_CMAKE_VER: 3.25.3, OMP_NUM_THREADS: 4, alpaka_ACC_CPU_B_TBB_T_SEQ_ENABLE: OFF, CMAKE_CXX_EXTENSIONS: OFF}
           container: ubuntu:20.04
-        - name: linux_clang-13_debug
+        - name: linux_clang-15_debug
           os: ubuntu-22.04
-          env: {ALPAKA_CI_CXX: clang++,  ALPAKA_CI_CLANG_VER: 13,     ALPAKA_CI_STDLIB: libstdc++, CMAKE_BUILD_TYPE: Debug,   ALPAKA_BOOST_VERSION: 1.79.0, ALPAKA_CI_CMAKE_VER: 3.25.3, OMP_NUM_THREADS: 3, CMAKE_CXX_EXTENSIONS: OFF}
+          env: {ALPAKA_CI_CXX: clang++,  ALPAKA_CI_CLANG_VER: 15,     ALPAKA_CI_STDLIB: libstdc++, CMAKE_BUILD_TYPE: Debug,   ALPAKA_BOOST_VERSION: 1.79.0, ALPAKA_CI_CMAKE_VER: 3.25.3, OMP_NUM_THREADS: 3, CMAKE_CXX_EXTENSIONS: OFF}
           container: ubuntu:22.04
         - name: linux_clang-16_debug_ubsan
           os: ubuntu-22.04

example/bufferCopy/src/bufferCopy.cpp

@@ -59,7 +59,7 @@ struct FillBufferKernel
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Define the index domain

example/complex/src/complex.cpp

@@ -32,7 +32,7 @@ struct ComplexKernel
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     using Idx = std::size_t;

example/conv2DWithMdspan/src/conv2DWithMdspan.cpp

@@ -77,7 +77,7 @@ auto FuzzyEqual(float a, float b) -> bool
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Define the index domain

example/convolution1D/src/convolution1D.cpp

@@ -68,7 +68,7 @@ auto FuzzyEqual(float a, float b) -> bool
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Size of 1D arrays to be used in convolution integral

example/convolution2D/src/convolution2D.cpp

@@ -212,7 +212,7 @@ auto FuzzyEqual(float a, float b) -> bool
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Define the index domain

example/counterBasedRng/src/counterBasedRng.cpp

@@ -96,7 +96,7 @@ class CounterBasedRngKernel
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Define the index domain

example/heatEquation/src/heatEquation.cpp

@@ -67,7 +67,7 @@ auto exactSolution(double const x, double const t) -> double
 //! Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 //! selected accelerator only. If you use the example as the starting point for your project, you can rename the
 //! example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Parameters (a user is supposed to change numNodesX, numTimeSteps)

example/heatEquation2D/src/heatEquation2D.cpp

@@ -31,7 +31,7 @@
 //! Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 //! selected accelerator only. If you use the example as the starting point for your project, you can rename the
 //! example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Set Dim and Idx type

example/helloWorld/src/helloWorld.cpp

@@ -51,7 +51,7 @@ struct HelloWorldKernel
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Define the index domain

example/helloWorldLambda/src/helloWorldLambda.cpp

@@ -43,7 +43,7 @@ void ALPAKA_FN_ACC hiWorldFunction(TAcc const& acc, size_t const nExclamationMar
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
 // It requires support for extended lambdas when using nvcc as CUDA compiler.

example/kernelSpecialization/src/kernelSpecialization.cpp

@@ -57,7 +57,7 @@ struct Kernel
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Define the accelerator

example/matrixMulWithMdspan/src/matrixMulMdSpan.cpp

@@ -90,7 +90,7 @@ inline void initializeMatrx(TMdSpan& span)
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Set number of dimensions (i.e 2) as a type

example/monteCarloIntegration/src/monteCarloIntegration.cpp

@@ -77,7 +77,7 @@ struct Kernel
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Defines and setup.

example/parallelLoopPatterns/src/parallelLoopPatterns.cpp

@@ -404,7 +404,7 @@ void openMPSimdStyle(TDev& dev, TQueue& queue, TBufAcc& bufAcc)
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Define the index domain, this example is only for 1d

example/randomCells2D/src/randomCells2D.cpp

@@ -145,7 +145,7 @@ struct RunTimestepKernelVector
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     using Dim = alpaka::DimInt<2>;

example/randomStrategies/src/randomStrategies.cpp

@@ -25,7 +25,7 @@ using RandomEngine = alpaka::rand::Philox4x32x10;
 
 
 /// Parameters to set up the default accelerator, queue, and buffers
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 struct Box
 {
     // accelerator, queue, and work division typedefs
@@ -184,7 +184,7 @@ struct FillKernel
  *
  *  File is in TSV format. One line for each "point"; line length is the number of "rolls".
  */
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 void saveDataAndShowAverage(std::string filename, float const* buffer, Box<TAccTag> const& box)
 {
     std::ofstream output(filename);
@@ -207,7 +207,7 @@ struct Writer;
 template<>
 struct Writer<Strategy::seed>
 {
-    template<typename TAccTag>
+    template<alpaka::concepts::Tag TAccTag>
     static void save(float const* buffer, Box<TAccTag> const& box)
     {
         saveDataAndShowAverage("out_seed.csv", buffer, box);
@@ -217,7 +217,7 @@ struct Writer<Strategy::seed>
 template<>
 struct Writer<Strategy::subsequence>
 {
-    template<typename TAccTag>
+    template<alpaka::concepts::Tag TAccTag>
     static void save(float const* buffer, Box<TAccTag> const& box)
     {
         saveDataAndShowAverage("out_subsequence.csv", buffer, box);
@@ -227,14 +227,14 @@ struct Writer<Strategy::subsequence>
 template<>
 struct Writer<Strategy::offset>
 {
-    template<typename TAccTag>
+    template<alpaka::concepts::Tag TAccTag>
     static void save(float const* buffer, Box<TAccTag> const& box)
     {
         saveDataAndShowAverage("out_offset.csv", buffer, box);
     }
 };
 
-template<Strategy TStrategy, typename TAccTag>
+template<Strategy TStrategy, alpaka::concepts::Tag TAccTag>
 void runStrategy(Box<TAccTag>& box)
 {
     // Set up the pointer to the PRNG states buffer
@@ -326,7 +326,7 @@ void runStrategy(Box<TAccTag>& box)
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     Box<TAccTag> box; // Initialize the box

example/tagSpecialization/src/tagSpecialization.cpp

@@ -28,7 +28,7 @@ std::string host_function_ver1<alpaka::TagGpuCudaRt>()
 }
 
 //! Function specialization via overloading
-template<typename TTag>
+template<alpaka::concepts::Tag TTag>
 std::string host_function_ver2(TTag)
 {
     return "generic host function v2";
@@ -80,7 +80,7 @@ struct WrapperKernel
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Define the accelerator

example/vectorAdd/src/vectorAdd.cpp

@@ -48,7 +48,7 @@ class VectorAddKernel
 // Instead, a single accelerator is selected once from the active accelerators and the kernels are executed with the
 // selected accelerator only. If you use the example as the starting point for your project, you can rename the
 // example() function to main() and move the accelerator tag to the function body.
-template<typename TAccTag>
+template<alpaka::concepts::Tag TAccTag>
 auto example(TAccTag const&) -> int
 {
     // Define the index domain

include/alpaka/acc/AccGenericSycl.hpp

@@ -46,13 +46,13 @@
 
 namespace alpaka
 {
-    template<typename TTag, typename TAcc, typename TDim, typename TIdx, typename TKernelFnObj, typename... TArgs>
+    template<concepts::Tag TTag, typename TAcc, typename TDim, typename TIdx, typename TKernelFnObj, typename... TArgs>
     class TaskKernelGenericSycl;
 
     //! The SYCL accelerator.
     //!
     //! This accelerator allows parallel kernel execution on SYCL devices.
-    template<typename TTag, typename TDim, typename TIdx>
+    template<concepts::Tag TTag, typename TDim, typename TIdx>
     class AccGenericSycl
         : public WorkDivGenericSycl<TDim, TIdx>
         , public gb::IdxGbGenericSycl<TDim, TIdx>
@@ -103,26 +103,26 @@ namespace alpaka
 namespace alpaka::trait
 {
     //! The SYCL accelerator type trait specialization.
-    template<typename TTag, typename TDim, typename TIdx>
+    template<concepts::Tag TTag, typename TDim, typename TIdx>
     struct AccType<AccGenericSycl<TTag, TDim, TIdx>>
     {
         using type = AccGenericSycl<TTag, TDim, TIdx>;
     };
 
     //! The SYCL single thread accelerator type trait specialization.
-    template<typename TTag, typename TDim, typename TIdx>
+    template<concepts::Tag TTag, typename TDim, typename TIdx>
     struct IsSingleThreadAcc<AccGenericSycl<TTag, TDim, TIdx>> : std::false_type
     {
     };
 
     //! The SYCL multi thread accelerator type trait specialization.
-    template<typename TTag, typename TDim, typename TIdx>
+    template<concepts::Tag TTag, typename TDim, typename TIdx>
     struct IsMultiThreadAcc<AccGenericSycl<TTag, TDim, TIdx>> : std::true_type
     {
     };
 
     //! The SYCL accelerator device properties get trait specialization.
-    template<typename TTag, typename TDim, typename TIdx>
+    template<concepts::Tag TTag, typename TDim, typename TIdx>
     struct GetAccDevProps<AccGenericSycl<TTag, TDim, TIdx>>
     {
         static auto getAccDevProps(DevGenericSycl<TTag> const& dev) -> AccDevProps<TDim, TIdx>
@@ -159,7 +159,7 @@ namespace alpaka::trait
     };
 
     //! The SYCL accelerator name trait specialization.
-    template<typename TTag, typename TDim, typename TIdx>
+    template<concepts::Tag TTag, typename TDim, typename TIdx>
     struct GetAccName<AccGenericSycl<TTag, TDim, TIdx>>
     {
         static auto getAccName() -> std::string
@@ -170,21 +170,27 @@ namespace alpaka::trait
     };
 
     //! The SYCL accelerator device type trait specialization.
-    template<typename TTag, typename TDim, typename TIdx>
+    template<concepts::Tag TTag, typename TDim, typename TIdx>
     struct DevType<AccGenericSycl<TTag, TDim, TIdx>>
     {
         using type = DevGenericSycl<TTag>;
     };
 
     //! The SYCL accelerator dimension getter trait specialization.
-    template<typename TTag, typename TDim, typename TIdx>
+    template<concepts::Tag TTag, typename TDim, typename TIdx>
     struct DimType<AccGenericSycl<TTag, TDim, TIdx>>
     {
         using type = TDim;
     };
 
     //! The SYCL accelerator execution task type trait specialization.
-    template<typename TTag, typename TDim, typename TIdx, typename TWorkDiv, typename TKernelFnObj, typename... TArgs>
+    template<
+        concepts::Tag TTag,
+        typename TDim,
+        typename TIdx,
+        typename TWorkDiv,
+        typename TKernelFnObj,
+        typename... TArgs>
     struct CreateTaskKernel<AccGenericSycl<TTag, TDim, TIdx>, TWorkDiv, TKernelFnObj, TArgs...>
     {
         static auto createTaskKernel(TWorkDiv const& workDiv, TKernelFnObj const& kernelFnObj, TArgs&&... args)
@@ -197,14 +203,14 @@ namespace alpaka::trait
     };
 
     //! The SYCL execution task platform type trait specialization.
-    template<typename TTag, typename TDim, typename TIdx>
+    template<concepts::Tag TTag, typename TDim, typename TIdx>
     struct PlatformType<AccGenericSycl<TTag, TDim, TIdx>>
     {
         using type = PlatformGenericSycl<TTag>;
     };
 
     //! The SYCL accelerator idx type trait specialization.
-    template<typename TTag, typename TDim, typename TIdx>
+    template<concepts::Tag TTag, typename TDim, typename TIdx>
     struct IdxType<AccGenericSycl<TTag, TDim, TIdx>>
     {
         using type = TIdx;