[WIP] Add Counted Attributes

.github/workflows/build.yml

@@ -64,7 +64,7 @@ jobs:
           build-dir: build
           options:
             ENABLE_WARNINGS_AS_ERRORS=Off
-            BLT_CXX_STD=c++17
+            BLT_CXX_STD=c++20
             CMAKE_BUILD_TYPE=Release
             PERFSUITE_RUN_SHORT_TEST=On
             ${{ matrix.shared.args }}

CMakeLists.txt

@@ -17,17 +17,17 @@ set(COMPILERS_KNOWN_TO_CMAKE33 AppleClang Clang GNU MSVC)
 
 include(CheckCXXCompilerFlag)
 if(NOT DEFINED BLT_CXX_STD)
-  if("cxx_std_20" IN_LIST CMAKE_CXX_KNOWN_FEATURES)
-    set(BLT_CXX_STD c++20 CACHE STRING "Version of C++ standard")
+  if("cxx_std_23" IN_LIST CMAKE_CXX_KNOWN_FEATURES)
+    set(BLT_CXX_STD c++23 CACHE STRING "Version of C++ standard")
     message("Using C++ standard: ${BLT_CXX_STD}")
-  elseif("cxx_std_17" IN_LIST CMAKE_CXX_KNOWN_FEATURES)
-    set(BLT_CXX_STD c++17 CACHE STRING "Version of C++ standard")
+  elseif("cxx_std_20" IN_LIST CMAKE_CXX_KNOWN_FEATURES)
+    set(BLT_CXX_STD c++20 CACHE STRING "Version of C++ standard")
     message("Using C++ standard: ${BLT_CXX_STD}")
   elseif("${CMAKE_CXX_COMPILER_ID}" IN_LIST COMPILERS_KNOWN_TO_CMAKE33)
-    set(BLT_CXX_STD c++17 CACHE STRING "Version of C++ standard")
+    set(BLT_CXX_STD c++20 CACHE STRING "Version of C++ standard")
     message("Using C++ standard: ${BLT_CXX_STD}")
   else() #cmake has no idea what to do, do it ourselves...
-    set(flag_var "c++17")
+    set(flag_var "c++20")
     CHECK_CXX_COMPILER_FLAG("-std=${flag_var}" COMPILER_SUPPORTS_${flag_var})
     if(COMPILER_SUPPORTS_${flag_var})
       set(BLT_CXX_STD ${flag_var} CACHE STRING "Version of C++ standard")
@@ -38,8 +38,9 @@ if(NOT DEFINED BLT_CXX_STD)
 else() #check BLT_CXX_STD is high enough by disallowing the only invalid option
   if(("${BLT_CXX_STD}" STREQUAL "c++98") OR
      ("${BLT_CXX_STD}" STREQUAL "c++11") OR
-     ("${BLT_CXX_STD}" STREQUAL "c++14"))
-    message(FATAL_ERROR "RAJA requires minimum C++ standard of c++17")
+     ("${BLT_CXX_STD}" STREQUAL "c++14") OR
+     ("${BLT_CXX_STD}" STREQUAL "c++17"))
+    message(FATAL_ERROR "RAJA requires minimum C++ standard of c++20")
   endif()
 endif(NOT DEFINED BLT_CXX_STD)
 

scripts/alcf-builds/sycl.sh

@@ -26,7 +26,7 @@ cmake \
   -DENABLE_TARGET_OPENMP=Off \
   -DENABLE_ALL_WARNINGS=Off \
   -DENABLE_SYCL=On \
-  -DCMAKE_CXX_STANDARD=17 \
+  -DCMAKE_CXX_STANDARD=20 \
   -DCMAKE_LINKER=icpx \
   "$@" \
   ..

scripts/lc-builds/corona_sycl.sh

@@ -67,7 +67,7 @@ cmake \
   -DCMAKE_C_COMPILER=clang \
   -DCMAKE_CXX_COMPILER=clang++ \
   -DCMAKE_LINKER=clang++ \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -DENABLE_TESTS=On \
   -DENABLE_EXAMPLES=On \
   "$@" \

scripts/lc-builds/toss4_amdclang.sh

@@ -95,7 +95,7 @@ cmake \
   -DCMAKE_HIP_ARCHITECTURES="${COMP_ARCH}" \
   -DGPU_TARGETS="${COMP_ARCH}" \
   -DAMDGPU_TARGETS="${COMP_ARCH}" \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_HIP=ON \
   -DENABLE_OPENMP=ON \

scripts/lc-builds/toss4_amdclang_asan.sh

@@ -102,7 +102,7 @@ cmake \
   -DCMAKE_CXX_FLAGS="-fsanitize=address -fsanitize=undefined -shared-libsan" \
   -DCMAKE_HIP_FLAGS="-fsanitize=address -fsanitize=undefined -shared-libsan -fgpu-rdc --hip-version=${COMP_HIP_VER}" \
   -DCMAKE_EXE_LINKER_FLAGS="-L/opt/rocm-${COMP_HIP_VER}/lib/asan/ -L/opt/rocm-${COMP_HIP_VER}/llvm/lib/asan -Wl,-rpath,/opt/rocm-${COMP_HIP_VER}/lib/asan/:/opt/rocm-${COMP_HIP_VER}/llvm/lib/asan:/opt/rocm-${COMP_HIP_VER}/lib/llvm/lib/clang/${COMP_CLANG_MAJOR_VER}/lib/linux -fgpu-rdc --hip-version=${COMP_HIP_VER}" \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_HIP=ON \
   -DENABLE_OPENMP=ON \

scripts/lc-builds/toss4_cce_hip.sh

@@ -69,7 +69,7 @@ cmake \
   -DCMAKE_HIP_ARCHITECTURES=${HIP_ARCH} \
   -DGPU_TARGETS=${HIP_ARCH} \
   -DAMDGPU_TARGETS=${HIP_ARCH} \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_HIP=ON \
   -DENABLE_OPENMP=ON \

scripts/lc-builds/toss4_clang-mpi_caliper.sh

@@ -57,7 +57,7 @@ module load cmake/${CMAKE_VER}
 cmake \
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_CXX_COMPILER=/usr/tce/packages/clang/clang-${COMP_VER}/bin/clang++ \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_MPI=ON \
   -DENABLE_OPENMP=On \

scripts/lc-builds/toss4_clang.sh

@@ -51,7 +51,7 @@ module load cmake/${CMAKE_VER}
 cmake \
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_CXX_COMPILER=/usr/tce/packages/clang/clang-${COMP_VER}/bin/clang++ \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_OPENMP=On \
   -DCMAKE_INSTALL_PREFIX=../install_${BUILD_SUFFIX} \

scripts/lc-builds/toss4_clang_caliper.sh

@@ -57,7 +57,7 @@ module load cmake/${CMAKE_VER}
 cmake \
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_CXX_COMPILER=/usr/tce/packages/clang/clang-${COMP_VER}/bin/clang++ \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_OPENMP=On \
   -DCMAKE_INSTALL_PREFIX=../install_${BUILD_SUFFIX} \

scripts/lc-builds/toss4_cray-mpich_amdclang.sh

@@ -119,7 +119,7 @@ cmake \
   -DCMAKE_HIP_ARCHITECTURES="${COMP_ARCH}" \
   -DGPU_TARGETS="${COMP_ARCH}" \
   -DAMDGPU_TARGETS="${COMP_ARCH}" \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_MPI=ON \
   -DENABLE_HIP=ON \

scripts/lc-builds/toss4_cray-mpich_amdclang_asan.sh

@@ -126,7 +126,7 @@ cmake \
   -DCMAKE_CXX_FLAGS="-fsanitize=address -shared-libsan" \
   -DCMAKE_HIP_FLAGS="-fsanitize=address -shared-libsan -fgpu-rdc --hip-version=${COMP_HIP_VER}" \
   -DCMAKE_EXE_LINKER_FLAGS="-L/opt/rocm-${COMP_HIP_VER}/lib/asan/ -L/opt/rocm-${COMP_HIP_VER}/llvm/lib/asan -Wl,-rpath,/opt/rocm-${COMP_HIP_VER}/lib/asan/:/opt/rocm-${COMP_HIP_VER}/llvm/lib/asan:/opt/rocm-${COMP_HIP_VER}/lib/llvm/lib/clang/${COMP_CLANG_MAJOR_VER}/lib/linux -fgpu-rdc --hip-version=${COMP_HIP_VER}" \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_MPI=ON \
   -DENABLE_HIP=ON \

scripts/lc-builds/toss4_cray-mpich_amdclang_caliper.sh

@@ -123,7 +123,7 @@ cmake \
   -DCMAKE_HIP_ARCHITECTURES="${COMP_ARCH}" \
   -DGPU_TARGETS="${COMP_ARCH}" \
   -DAMDGPU_TARGETS="${COMP_ARCH}" \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_MPI=ON \
   -DENABLE_HIP=ON \

scripts/lc-builds/toss4_gcc-mpi_caliper.sh

@@ -58,7 +58,7 @@ cmake \
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_C_COMPILER=/usr/tce/packages/gcc/gcc-${COMP_VER}/bin/gcc \
   -DCMAKE_CXX_COMPILER=/usr/tce/packages/gcc/gcc-${COMP_VER}/bin/g++ \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_MPI=ON \
   -DENABLE_OPENMP=On \

scripts/lc-builds/toss4_gcc.sh

@@ -51,7 +51,7 @@ module load cmake/${CMAKE_VER}
 cmake \
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_CXX_COMPILER=/usr/tce/packages/gcc/gcc-${COMP_VER}/bin/g++ \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_OPENMP=On \
   -DCMAKE_INSTALL_PREFIX=../install_${BUILD_SUFFIX} \

scripts/lc-builds/toss4_gcc_caliper.sh

@@ -57,7 +57,7 @@ module load cmake/${CMAKE_VER}
 cmake \
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_CXX_COMPILER=/usr/tce/packages/gcc/gcc-${COMP_VER}/bin/g++ \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_OPENMP=On \
   -DCMAKE_INSTALL_PREFIX=../install_${BUILD_SUFFIX} \

scripts/lc-builds/toss4_hipcc.sh

@@ -85,7 +85,7 @@ cmake \
   -DCMAKE_HIP_ARCHITECTURES="${COMP_ARCH}" \
   -DGPU_TARGETS="${COMP_ARCH}" \
   -DAMDGPU_TARGETS="${COMP_ARCH}" \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_HIP=ON \
   -DENABLE_OPENMP=ON \

scripts/lc-builds/toss4_icpx.sh

@@ -64,7 +64,7 @@ cmake \
   -DCMAKE_BUILD_TYPE=Release \
   -DCMAKE_CXX_COMPILER=icpx \
   -DCMAKE_C_COMPILER=icx \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DRAJA_ENABLE_FORCEINLINE_RECURSIVE=Off \
   -DENABLE_OPENMP=On \

scripts/lc-builds/toss4_mvapich2_icpx.sh

@@ -17,7 +17,7 @@ if [[ $# -lt 2 ]]; then
   echo "You must pass 2 or more arguments to the script (in this order): "
   echo "   1) mvapich2 version number"
   echo "   2) icpx compiler version number"
-  echo "   3) optional cmake version
+  echo "   3) optional cmake version"
   echo
   echo "You must pass a compiler version number to script. For example,"
   echo "    toss4_mvapich2_gcc.sh 2.3.7 2022.1.0 [3.27.4]"
@@ -65,7 +65,7 @@ cmake \
   -DMPI_CXX_COMPILER="/usr/tce/packages/mvapich2/mvapich2-${MPI_VER}-intel-${COMP_VER}/bin/mpicxx" \
   -DCMAKE_CXX_COMPILER=icpx \
   -DCMAKE_C_COMPILER=icx \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_MPI=ON \
   -DRAJA_ENABLE_FORCEINLINE_RECURSIVE=Off \

scripts/ubuntu-builds/ubuntu_amdclang.sh

@@ -62,7 +62,7 @@ cmake \
   -DCMAKE_HIP_ARCHITECTURES="${COMP_ARCH}" \
   -DGPU_TARGETS="${COMP_ARCH}" \
   -DAMDGPU_TARGETS="${COMP_ARCH}" \
-  -DBLT_CXX_STD=c++17 \
+  -DBLT_CXX_STD=c++20 \
   -C ${RAJA_HOSTCONFIG} \
   -DENABLE_HIP=ON \
   -DENABLE_OPENMP=ON \

src/algorithm/ATOMIC.cpp

@@ -77,5 +77,53 @@ void ATOMIC::tearDown(VariantID RAJAPERF_UNUSED_ARG(vid), size_t RAJAPERF_UNUSED
 
 }
 
+
+// Only define setCountedAttributes functions past this point
+// BEWARE: data types (Index_type, Real_ptr, etc) become wrappers past this point
+#include "common/CountingMacros.hpp"
+
+void ATOMIC::setCountedAttributes()
+{
+  const size_t replication = getActualProblemSize();
+
+  VariantID vid = VariantID::Base_Seq;
+  size_t tune_idx = 0;
+
+  RAJAPERF_COUNTERS_INITIALIZE();
+
+  RAJAPERF_COUNTERS_CODE_WRAPPER(
+  setUp(vid, tune_idx);
+  );
+
+  {
+    RAJAPERF_COUNTERS_CODE_WRAPPER(
+    const Index_type ibegin = 0;
+    const Index_type iend = getActualProblemSize();
+
+    ATOMIC_DATA_SETUP(replication);
+    );
+
+    RAJAPERF_COUNTERS_REP_SCOPE()
+    {
+
+      RAJAPERF_COUNTERS_PAR_LOOP(for (Index_type i = ibegin; i < iend; ++i )) {
+        RAJAPERF_COUNTERS_LOOP_BODY(ATOMIC_BODY(RAJAPERF_ATOMIC_ADD_COUNTING, i, ATOMIC_VALUE));
+      }
+
+    }
+
+    RAJAPERF_COUNTERS_CODE_WRAPPER(
+    ATOMIC_DATA_TEARDOWN(replication);
+    );
+
+  }
+
+  RAJAPERF_COUNTERS_CODE_WRAPPER(
+  tearDown(vid, tune_idx);
+  );
+
+  RAJAPERF_COUNTERS_FINALIZE();
+}
+
 } // end namespace algorithm
 } // end namespace rajaperf

src/algorithm/ATOMIC.hpp

@@ -70,6 +70,7 @@ class ATOMIC : public KernelBase
   void setUp(VariantID vid, size_t tune_idx);
   void updateChecksum(VariantID vid, size_t tune_idx);
   void tearDown(VariantID vid, size_t tune_idx);
+  void setCountedAttributes();
 
   void defineSeqVariantTunings();
   void defineOpenMPVariantTunings();

src/algorithm/HISTOGRAM.cpp

@@ -145,5 +145,63 @@ void HISTOGRAM::tearDown(VariantID vid, size_t RAJAPERF_UNUSED_ARG(tune_idx))
   deallocData(DataSpace::Host, m_counts_final);
 }
 
+
+// // Only define setCountedAttributes functions past this point
+// // BEWARE: data types (Index_type, Real_ptr, etc) become wrappers past this point
+#include "common/CountingMacros.hpp"
+
+void HISTOGRAM::setCountedAttributes()
+{
+  VariantID vid = VariantID::Base_Seq;
+  size_t tune_idx = 0;
+
+  RAJAPERF_COUNTERS_INITIALIZE();
+
+  RAJAPERF_COUNTERS_CODE_WRAPPER(
+  setUp(vid, tune_idx);
+  );
+
+  {
+    RAJAPERF_COUNTERS_CODE_WRAPPER(
+    const Index_type ibegin = 0;
+    const Index_type iend = getActualProblemSize();
+
+    HISTOGRAM_DATA_SETUP;
+    );
+
+    RAJAPERF_COUNTERS_CODE_WRAPPER(
+    HISTOGRAM_SETUP_COUNTS;
+    );
+
+    RAJAPERF_COUNTERS_REP_SCOPE()
+    {
+
+      RAJAPERF_COUNTERS_CODE_WRAPPER(
+      HISTOGRAM_INIT_COUNTS;
+      );
+
+      RAJAPERF_COUNTERS_PAR_LOOP(for (Index_type i = ibegin; i < iend; ++i )) {
+        RAJAPERF_COUNTERS_LOOP_BODY(HISTOGRAM_BODY(RAJAPERF_ATOMIC_ADD_COUNTING));
+      }
+
+      RAJAPERF_COUNTERS_CODE_WRAPPER(
+      HISTOGRAM_FINALIZE_COUNTS;
+      );
+
+    }
+
+    RAJAPERF_COUNTERS_CODE_WRAPPER(
+    HISTOGRAM_TEARDOWN_COUNTS;
+    );
+
+  }
+
+  RAJAPERF_COUNTERS_CODE_WRAPPER(
+  tearDown(vid, 0);
+  );
+
+  RAJAPERF_COUNTERS_FINALIZE();
+}
+
 } // end namespace algorithm
 } // end namespace rajaperf

src/algorithm/HISTOGRAM.hpp

@@ -92,6 +92,7 @@ class HISTOGRAM : public KernelBase
   void setUp(VariantID vid, size_t tune_idx);
   void updateChecksum(VariantID vid, size_t tune_idx);
   void tearDown(VariantID vid, size_t tune_idx);
+  void setCountedAttributes();
 
   void defineSeqVariantTunings();
   void defineOpenMPVariantTunings();

src/algorithm/MEMCPY.cpp

@@ -77,5 +77,47 @@ void MEMCPY::tearDown(VariantID vid, size_t RAJAPERF_UNUSED_ARG(tune_idx))
   deallocData(m_y, vid);
 }
 
+
+// Only define setCountedAttributes functions past this point
+// BEWARE: data types (Index_type, Real_ptr, etc) become wrappers past this point
+#include "common/CountingMacros.hpp"
+
+void MEMCPY::setCountedAttributes()
+{
+  VariantID vid = VariantID::Base_Seq;
+  size_t tune_idx = 0;
+
+  RAJAPERF_COUNTERS_INITIALIZE();
+
+  RAJAPERF_COUNTERS_CODE_WRAPPER(
+  setUp(vid, tune_idx);
+  );
+
+  {
+    RAJAPERF_COUNTERS_CODE_WRAPPER(
+    const Index_type ibegin = 0;
+    const Index_type iend = getActualProblemSize();
+
+    MEMCPY_DATA_SETUP;
+    );
+
+    RAJAPERF_COUNTERS_REP_SCOPE()
+    {
+
+      RAJAPERF_COUNTERS_PAR_LOOP(for (Index_type i = ibegin; i < iend; ++i )) {
+        RAJAPERF_COUNTERS_LOOP_BODY(MEMCPY_BODY);
+      }
+
+    }
+
+  }
+
+  RAJAPERF_COUNTERS_CODE_WRAPPER(
+  tearDown(vid, tune_idx);
+  );
+
+  RAJAPERF_COUNTERS_FINALIZE();
+}
+
 } // end namespace algorithm
 } // end namespace rajaperf