Add test for --kokkos-trace

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "sample/external/kokkos"]
+	path = sample/external/kokkos
+	url = https://github.com/kokkos/kokkos

CMakeLists.txt

@@ -194,6 +194,13 @@ message(STATUS "Pytest CPU threadcount: ${PYTEST_NUMPROCS}")
 # ---------------------------
 # profile mode tests
 # ---------------------------
+add_test(
+    NAME test_profile_kokkos_trace
+    COMMAND
+        ${Python3_EXECUTABLE} -m pytest -m kokkos
+        --junitxml=tests/test_profile_kokkos_trace.xml ${COV_OPTION}
+        ${PROJECT_SOURCE_DIR}/tests/test_profile_general.py
+    WORKING_DIRECTORY ${PROJECT_SOURCE_DIR})
 
 add_test(
     NAME test_profile_kernel_execution
@@ -245,6 +252,7 @@ add_test(
     WORKING_DIRECTORY ${PROJECT_SOURCE_DIR})
 
 set_tests_properties(
+    test_profile_kokkos_trace
     test_profile_kernel_execution
     test_profile_ipblocks
     test_profile_dispatch
@@ -531,4 +539,4 @@ set(CPACK_SOURCE_IGNORE_FILES
     /tests
     /build)
 
-include(CPack)
+include(CPack)

pyproject.toml

@@ -50,6 +50,7 @@ pythonpath = [
     ]
 
 markers = [
+	"kokkos",
 	"kernel_execution",
 	"block",
 	"misc",

sample/kokkos-no-uvm/Makefile

@@ -0,0 +1,44 @@
+#Set your Kokkos path to something appropriate
+KOKKOS_PATH = ${CURDIR}/../external/kokkos
+KOKKOS_DEVICES = "HIP"
+KOKKOS_ARCH = "AMD_GFX942"
+KOKKOS_CUDA_OPTIONS = enable_lambda
+SRC = $(wildcard *.cc)
+
+default: build
+	echo "Start Build"
+
+CXXFLAGS = -g -O3 -DUSE_OMP=1 -lineinfo -DRAJA_USE_GETTIME -I./includes -I./ -lineinfo
+
+ifeq ( $(USE_MPI), yes)
+CXX = hipcc
+CXXFLAGS += -DUSE_MPI=1
+else
+CXX = hipcc
+CXXFLAGS += -DUSE_MPI=0
+endif
+
+LINK = ${CXX}
+
+LINKFLAGS =
+EXE = lulesh.kk
+
+DEPFLAGS = -M
+
+OBJ = $(SRC:.cc=.o)
+LIB =
+
+include $(KOKKOS_PATH)/Makefile.kokkos
+
+build: $(EXE)
+
+$(EXE): $(OBJ) $(KOKKOS_LINK_DEPENDS)
+	$(LINK) $(KOKKOS_LDFLAGS) $(LINKFLAGS) $(EXTRA_PATH) $(OBJ) $(KOKKOS_LIBS) $(LIB) -o $(EXE)
+
+clean: kokkos-clean
+	rm -f *.o
+
+# Compilation rules
+
+%.o:%.cc $(KOKKOS_CPP_DEPENDS) lulesh.h lulesh_tuple.h
+	$(CXX) $(KOKKOS_CPPFLAGS) $(KOKKOS_CXXFLAGS) $(CXXFLAGS) $(EXTRA_INC) -c $<

sample/kokkos-no-uvm/Makefile.cuda

@@ -0,0 +1,47 @@
+#Set your Kokkos path to something appropriate
+KOKKOS_PATH = ${HOME}/Kokkos/kokkos
+KOKKOS_DEVICES = "Cuda,OpenMP"
+KOKKOS_ARCH = "Volta70"
+KOKKOS_CUDA_OPTIONS = enable_lambda
+SRC = $(wildcard *.cc)
+
+default: build
+	echo "Start Build"
+
+CXXFLAGS = -g -O3 -DUSE_OMP=1 -lineinfo -DRAJA_USE_GETTIME -I./includes -I./ -lineinfo
+
+ifeq ( $(USE_MPI), yes) 
+CXX = mpicxx
+CXXFLAGS += -DUSE_MPI=1
+else
+CXX = ${KOKKOS_PATH}/bin/nvcc_wrapper
+CXXFLAGS += -DUSE_MPI=0
+endif
+
+LINK = ${CXX}
+
+LINKFLAGS =  
+EXE = lulesh.cuda
+
+DEPFLAGS = -M
+
+OBJ = $(SRC:.cc=.o)
+LIB =
+
+include $(KOKKOS_PATH)/Makefile.kokkos
+
+build: $(EXE)
+
+$(EXE): $(OBJ) $(KOKKOS_LINK_DEPENDS)
+	$(LINK) $(KOKKOS_LDFLAGS) $(LINKFLAGS) $(EXTRA_PATH) $(OBJ) $(KOKKOS_LIBS) $(LIB) -o $(EXE)
+
+clean: kokkos-clean 
+	rm -f *.o
+
+# Compilation rules
+
+%.o:%.cc $(KOKKOS_CPP_DEPENDS) lulesh.h lulesh_tuple.h
+	$(CXX) $(KOKKOS_CPPFLAGS) $(KOKKOS_CXXFLAGS) $(CXXFLAGS) $(EXTRA_INC) -c $<
+
+
+

sample/kokkos-no-uvm/includes/Timer.hxx

@@ -0,0 +1,113 @@
+/*!
+ ******************************************************************************
+ *
+ * \file
+ *
+ * \brief   RAJA header file for simple class that can be used to 
+ *          time code sections.
+ *     
+ * \author  Rich Hornung, Center for Applied Scientific Computing, LLNL
+ * \author  Jeff Keasler, Applications, Simulations And Quality, LLNL
+ *
+ ******************************************************************************
+ */
+
+#ifndef RAJA_Timer_HXX
+#define RAJA_Timer_HXX
+
+
+#if defined(RAJA_USE_CYCLE)
+#include "./cycle.h"
+typedef ticks TimeType;
+
+#elif defined(RAJA_USE_CLOCK)
+#include <time.h>
+typedef clock_t TimeType;
+
+#elif defined(RAJA_USE_GETTIME)
+#include <time.h>
+typedef timespec TimeType;
+
+#else
+#error RAJA_TIMER_TYPE is undefined!
+
+#endif
+
+
+
+namespace RAJA {
+
+
+/*!
+ ******************************************************************************
+ *
+ * \brief  Simple timer class to time code sections.
+ *
+ ******************************************************************************
+ */
+class Timer
+{
+public:
+#if defined(RAJA_USE_CYCLE) || defined(RAJA_USE_CLOCK)
+   Timer() : telapsed(0) { ; }
+#endif
+#if defined(RAJA_USE_GETTIME)
+   Timer() : telapsed(0), stime_elapsed(0), nstime_elapsed(0) { ; }
+#endif
+
+#if defined(RAJA_USE_CYCLE)
+   void start() { tstart = getticks(); }
+   void stop()  { tstop = getticks();  set_elapsed(); }
+
+   long double elapsed()
+      { return static_cast<long double>(telapsed); }
+#endif
+
+#if defined(RAJA_USE_CLOCK)
+   void start() { tstart = clock(); }
+   void stop()  { tstop = clock();  set_elapsed(); }
+
+   long double elapsed()
+      { return static_cast<long double>(telapsed) / CLOCKS_PER_SEC; }
+#endif
+
+#if defined(RAJA_USE_GETTIME)
+
+#if 0
+   void start() { clock_gettime(CLOCK_REALTIME, &tstart); }
+   void stop()  { clock_gettime(CLOCK_REALTIME, &tstop); set_elapsed(); }
+#else
+   void start() { clock_gettime(CLOCK_MONOTONIC, &tstart); }
+   void stop()  { clock_gettime(CLOCK_MONOTONIC, &tstop); set_elapsed(); }
+#endif
+
+   long double elapsed()
+      { return (stime_elapsed + nstime_elapsed); }
+
+#endif
+
+private:
+   TimeType tstart;
+   TimeType tstop;
+   long double telapsed;
+
+#if defined(RAJA_USE_CYCLE) || defined(RAJA_USE_CLOCK)
+   void set_elapsed() { telapsed += (tstop - tstart); }
+
+#elif defined(RAJA_USE_GETTIME)
+   long double stime_elapsed;
+   long double nstime_elapsed;
+
+   void set_elapsed() { stime_elapsed += static_cast<long double>(
+                                         tstop.tv_sec - tstart.tv_sec); 
+                        nstime_elapsed += static_cast<long double>(
+                                          tstop.tv_nsec - tstart.tv_nsec ) /
+                                          1000000000.0; } 
+#endif
+
+};
+
+
+}  // closing brace for RAJA namespace
+
+#endif  // closing endif for header file include guard