Merge pull request #248 from LLNL/pr-from-fork/206

Pr for Basic Kokkos

Author: MrBurmark

.gitmodules

@@ -4,3 +4,6 @@
 [submodule "tpl/RAJA"]
 	path = tpl/RAJA
 	url = https://github.com/LLNL/RAJA.git
+[submodule "tpl/kokkos"]
+	path = tpl/kokkos
+	url = https://github.com/kokkos/kokkos

CMakeLists.txt

@@ -13,6 +13,7 @@ cmake_minimum_required(VERSION 3.14.5)
 
 option(ENABLE_RAJA_SEQUENTIAL "Run sequential variants of RAJA kernels. Disable
 this, and all other variants, to run _only_ raw C loops." On)
+option(ENABLE_KOKKOS "Include Kokkos implementations of the kernels in the RAJA Perfsuite" Off)
 
 #
 # Note: the BLT build system is inheritted by RAJA and is initialized by RAJA
@@ -22,8 +23,13 @@ if (PERFSUITE_ENABLE_WARNINGS)
   set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wextra -Werror")
 endif()
 
-set(CMAKE_CXX_STANDARD 14)
-set(BLT_CXX_STD c++14)
+if(ENABLE_KOKKOS)
+  set(CMAKE_CXX_STANDARD 17)
+  set(BLT_CXX_STD c++17)
+else()
+  set(CMAKE_CXX_STANDARD 14)
+  set(BLT_CXX_STD c++14)
+endif()
 
 include(blt/SetupBLT.cmake)
 
@@ -100,7 +106,12 @@ endif()
 if (ENABLE_CUDA)
   list(APPEND RAJA_PERFSUITE_DEPENDS cuda)
 endif()
-if (ENABLE_HIP)
+
+# Kokkos requires hipcc as the CMAKE_CXX_COMPILER for HIP AMD/VEGA GPU
+# platforms, whereas RAJAPerf Suite uses blt/CMake FindHIP to set HIP compiler. 
+# Separate RAJAPerf Suite and Kokkos handling of HIP compilers
+
+if ((ENABLE_HIP) AND (NOT ENABLE_KOKKOS))
   message(STATUS "HIP version: ${hip_VERSION}")
   if("${hip_VERSION}" VERSION_LESS "3.5")
     message(FATAL_ERROR "Trying to use HIP/ROCm version ${hip_VERSION}. RAJA Perf Suite requires HIP/ROCm version 3.5 or newer. ")
@@ -113,8 +124,13 @@ set(RAJAPERF_BUILD_SYSTYPE $ENV{SYS_TYPE})
 set(RAJAPERF_BUILD_HOST $ENV{HOSTNAME})
 
 if (ENABLE_CUDA)
-  set(CMAKE_CUDA_STANDARD 14)
-  set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -restrict -arch ${CUDA_ARCH} --expt-extended-lambda --expt-relaxed-constexpr")
+  if (ENABLE_KOKKOS)
+    set(CMAKE_CUDA_STANDARD 17)
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -restrict --extended-lambda --expt-relaxed-constexpr")
+  else()
+    set(CMAKE_CUDA_STANDARD 14)
+    set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -restrict -arch ${CUDA_ARCH} --expt-extended-lambda --expt-relaxed-constexpr")
+  endif()
 
   set(RAJAPERF_COMPILER "${CUDA_NVCC_EXECUTABLE}")
   list(APPEND RAJAPERF_COMPILER ${CMAKE_CXX_COMPILER})
@@ -135,13 +151,46 @@ configure_file(${CMAKE_SOURCE_DIR}/src/rajaperf_config.hpp.in
 
 include_directories($<BUILD_INTERFACE:${PROJECT_BINARY_DIR}/include>)
 
-# Make sure RAJA flag propagate (we need to do some house cleaning to
+# Make sure RAJA flags propagate (we need to do some tidying to
 # remove project-specific CMake variables that are no longer needed)
 set (CUDA_NVCC_FLAGS ${RAJA_NVCC_FLAGS})
 
 #
 # Each directory in the perf suite has its own CMakeLists.txt file.
-#
+
+# ENABLE_KOKKOS is A RAJAPerf Suite Option
+if(ENABLE_KOKKOS)
+  add_definitions(-DRUN_KOKKOS)
+  if(ENABLE_HIP)
+    set(Kokkos_ENABLE_HIP ON CACHE BOOL "Kokkos builds for AMD HIP set the
+Kokkos_ENABLE_HIP variable to ON")
+  endif()
+
+  if(ENABLE_TARGET_OPENMP)
+    set(Kokkos_ENABLE_OPENMPTARGET ON CACHE BOOL "Docstring")
+    if(NOT CMAKE_BUILD_TYPE MATCHES Debug)
+      if(NOT EXPERIMENTAL_BUILD)
+      message(FATAL_ERROR "Kokkos builds with OpenMPTarget require a Debug build to succeed at the moment. Rebuild with CMAKE_BUILD_TYPE=Debug. If you're a compiler developer, rebuild with -DEXPERIMENTAL_BUILD=ON")
+      endif()
+    endif()
+  endif()
+
+# ENABLE_CUDA IS A RAJA PERFSUITE OPTION
+  if(ENABLE_CUDA)
+    set(Kokkos_ENABLE_CUDA ON CACHE BOOL "Docstring")
+    set(Kokkos_ENABLE_CUDA_LAMBDA ON CACHE BOOL "Docstring")
+    enable_language(CUDA)
+  endif()
+  if(ENABLE_OPENMP)
+    set(Kokkos_ENABLE_OPENMP ON CACHE BOOL "Docstring")
+  endif()
+
+  add_subdirectory(tpl/kokkos)
+  get_property(KOKKOS_INCLUDE_DIRS DIRECTORY tpl/kokkos PROPERTY INCLUDE_DIRECTORIES)
+  include_directories(${KOKKOS_INCLUDE_DIRS})
+  list(APPEND RAJA_PERFSUITE_DEPENDS kokkos)
+endif()
+
 add_subdirectory(src)
 
 if (RAJA_PERFSUITE_ENABLE_TESTS)

src/CMakeLists.txt

@@ -9,8 +9,9 @@
 include_directories(.)
 
 add_subdirectory(common)
-add_subdirectory(apps)
 add_subdirectory(basic)
+add_subdirectory(basic-kokkos)
+add_subdirectory(apps)
 add_subdirectory(lcals)
 add_subdirectory(polybench)
 add_subdirectory(stream)
@@ -20,6 +21,7 @@ set(RAJA_PERFSUITE_EXECUTABLE_DEPENDS
     common
     apps
     basic
+    basic-kokkos
     lcals
     polybench
     stream

src/RAJAPerfSuiteDriver.cpp

@@ -6,6 +6,10 @@
 // SPDX-License-Identifier: (BSD-3-Clause)
 //~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~//
 
+#ifdef RUN_KOKKOS
+#include <Kokkos_Core.hpp>
+#endif
+
 #include "common/Executor.hpp"
 
 #include <iostream>
@@ -24,6 +28,9 @@ int main( int argc, char** argv )
   MPI_Comm_size(MPI_COMM_WORLD, &num_ranks);
   rajaperf::getCout() << "\n\nRunning with " << num_ranks << " MPI ranks..." << std::endl;
 #endif
+#ifdef RUN_KOKKOS
+  Kokkos::initialize(argc, argv);
+#endif
 
   // STEP 1: Create suite executor object
   rajaperf::Executor executor(argc, argv);
@@ -43,6 +50,9 @@ int main( int argc, char** argv )
 
   rajaperf::getCout() << "\n\nDONE!!!...." << std::endl;
 
+#ifdef RUN_KOKKOS
+  Kokkos::finalize();
+#endif
 #ifdef RAJA_PERFSUITE_ENABLE_MPI
   MPI_Finalize();
 #endif

src/basic-kokkos/CMakeLists.txt

@@ -0,0 +1,27 @@
+###############################################################################
+# Copyright (c) 2017-20, Lawrence Livermore National Security, LLC
+# and RAJA Performance Suite project contributors.
+# See the RAJAPerf/COPYRIGHT file for details.
+#
+# SPDX-License-Identifier: (BSD-3-Clause)
+###############################################################################
+
+#include_directories(SYSTEM ${CMAKE_CURRENT_SOURCE_DIR}/../basic)
+
+blt_add_library(
+  NAME basic-kokkos
+  SOURCES
+          PI_ATOMIC-Kokkos.cpp
+          DAXPY-Kokkos.cpp
+          IF_QUAD-Kokkos.cpp
+          INIT3-Kokkos.cpp
+          INIT_VIEW1D-Kokkos.cpp
+          INIT_VIEW1D_OFFSET-Kokkos.cpp
+          MULADDSUB-Kokkos.cpp
+          NESTED_INIT-Kokkos.cpp
+          REDUCE3_INT-Kokkos.cpp
+          TRAP_INT-Kokkos.cpp
+          DAXPY_ATOMIC-Kokkos.cpp
+  INCLUDES ${CMAKE_CURRENT_SOURCE_DIR}/../basic
+  DEPENDS_ON common ${RAJA_PERFSUITE_DEPENDS}
+  )

src/basic-kokkos/DAXPY-Kokkos.cpp

@@ -0,0 +1,69 @@
+//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~//
+// Copyright (c) 2017-20, Lawrence Livermore National Security, LLC
+// and RAJA Performance Suite project contributors.
+// See the RAJAPerf/COPYRIGHT file for details.
+//
+// SPDX-License-Identifier: (BSD-3-Clause)
+//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~//
+
+#include "DAXPY.hpp"
+#if defined(RUN_KOKKOS)
+#include "common/KokkosViewUtils.hpp"
+#include <iostream>
+
+namespace rajaperf {
+namespace basic {
+
+struct DaxpyFunctor {
+  Real_ptr x;
+  Real_ptr y;
+  Real_type a;
+  DaxpyFunctor(Real_ptr m_x, Real_ptr m_y, Real_type m_a)
+      : x(m_x), y(m_y), a(m_a) {}
+  void operator()(Index_type i) const { DAXPY_BODY; }
+};
+
+void DAXPY::runKokkosVariant(VariantID vid, size_t RAJAPERF_UNUSED_ARG(tune_idx)) {
+  const Index_type run_reps = getRunReps();
+  const Index_type ibegin = 0;
+  const Index_type iend = getActualProblemSize();
+
+  DAXPY_DATA_SETUP;
+
+  auto x_view = getViewFromPointer(x, iend);
+  auto y_view = getViewFromPointer(y, iend);
+
+  switch (vid) {
+
+  case Kokkos_Lambda: {
+
+    Kokkos::fence();
+    startTimer();
+
+    for (RepIndex_type irep = 0; irep < run_reps; ++irep) {
+      Kokkos::parallel_for(
+          "DAXPY-Kokkos Kokkos_Lambda",
+          Kokkos::RangePolicy<Kokkos::DefaultExecutionSpace>(ibegin, iend),
+          KOKKOS_LAMBDA(Index_type i) { y_view[i] += a * x_view[i]; });
+    }
+
+    Kokkos::fence();
+    stopTimer();
+
+    break;
+  }
+  default: {
+    std::cout << "\n  DAXPY : Unknown variant id = " << vid << std::endl;
+  }
+  }
+
+  // Move data (i.e., pointer, KokkosView-wrapped ponter) back to the host from
+  // the device
+
+  moveDataToHostFromKokkosView(x, x_view, iend);
+  moveDataToHostFromKokkosView(y, y_view, iend);
+}
+
+} // end namespace basic
+} // end namespace rajaperf
+#endif

src/basic-kokkos/DAXPY_ATOMIC-Kokkos.cpp

@@ -0,0 +1,70 @@
+//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~//
+// Copyright (c) 2017-21, Lawrence Livermore National Security, LLC
+// and RAJA Performance Suite project contributors.
+// See the RAJAPerf/LICENSE file for details.
+//
+// SPDX-License-Identifier: (BSD-3-Clause)
+//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~//
+
+#include "DAXPY_ATOMIC.hpp"
+#if defined(RUN_KOKKOS)
+#include "common/KokkosViewUtils.hpp"
+#include <iostream>
+
+// Delete me
+// For de-bugging:
+#include "RAJA/RAJA.hpp"
+
+namespace rajaperf {
+namespace basic {
+
+void DAXPY_ATOMIC::runKokkosVariant(VariantID vid, size_t RAJAPERF_UNUSED_ARG(tune_idx))
+{
+
+  const Index_type run_reps = getRunReps();
+  const Index_type ibegin = 0;
+  const Index_type iend = getActualProblemSize();
+
+  DAXPY_ATOMIC_DATA_SETUP;
+  //
+  // Kokkos Views to wrap pointers declared in DAXPY_ATOMIC.hpp
+  //
+
+  auto x_view = getViewFromPointer(x, iend);
+  auto y_view = getViewFromPointer(y, iend);
+
+  switch (vid) {
+
+  case Kokkos_Lambda: {
+
+    Kokkos::fence();
+    startTimer();
+
+    for (RepIndex_type irep = 0; irep < run_reps; ++irep) {
+
+      Kokkos::parallel_for(
+          "DAXPY_ATOMIC_Kokkos Kokkos_Lambda",
+          Kokkos::RangePolicy<Kokkos::DefaultExecutionSpace>(ibegin, iend),
+          KOKKOS_LAMBDA(Index_type i) {
+            Kokkos::atomic_add(&y_view[i], a * x_view[i]);
+          });
+    }
+
+    Kokkos::fence();
+    stopTimer();
+
+    break;
+  }
+
+  default: {
+    getCout() << "\n  DAXPY_ATOMIC : Unknown variant id = " << vid << std::endl;
+  }
+  }
+
+  moveDataToHostFromKokkosView(x, x_view, iend);
+  moveDataToHostFromKokkosView(y, y_view, iend);
+}
+
+} // end namespace basic
+} // end namespace rajaperf
+#endif