Kokkos lib update

lib/kokkos/example/tutorial/Hierarchical_Parallelism/01_thread_teams/Makefile

@@ -5,7 +5,7 @@ default: build
 	echo "Start Build"
 
 ifneq (,$(findstring Cuda,$(KOKKOS_DEVICES)))
-CXX = nvcc_wrapper
+CXX = ../../../../config/nvcc_wrapper
 CXXFLAGS = -O3
 LINK = ${CXX}
 LINKFLAGS = 

lib/kokkos/example/tutorial/Hierarchical_Parallelism/01_thread_teams_lambda/Makefile

@@ -5,13 +5,14 @@ default: build
 	echo "Start Build"
 
 ifneq (,$(findstring Cuda,$(KOKKOS_DEVICES)))
-CXX = nvcc_wrapper
+CXX = ../../../../config/nvcc_wrapper
 CXXFLAGS = -O3
 LINK = ${CXX}
 LINKFLAGS = 
 EXE = $(SRC:.cpp=.cuda)
 KOKKOS_DEVICES = "Cuda,OpenMP"
 KOKKOS_ARCH = "SNB,Kepler35"
+KOKKOS_CUDA_OPTIONS = "enable_lambda"
 else
 CXX = g++
 CXXFLAGS = -O3

lib/kokkos/example/tutorial/Hierarchical_Parallelism/01_thread_teams_lambda/thread_teams_lambda.cpp

@@ -62,7 +62,8 @@ int main (int narg, char* args[]) {
 
   // Set up a policy that launches 12 teams, with the maximum number
   // of threads per team.
-  const team_policy policy (12, team_policy::team_size_max ( [=]{} ));
+
+  const team_policy policy (12, Kokkos::AUTO);
 
   // This is a reduction with a team policy.  The team policy changes
   // the first argument of the lambda.  Rather than an integer index

lib/kokkos/example/tutorial/Hierarchical_Parallelism/02_nested_parallel_for/Makefile

@@ -5,7 +5,7 @@ default: build
 	echo "Start Build"
 
 ifneq (,$(findstring Cuda,$(KOKKOS_DEVICES)))
-CXX = nvcc_wrapper
+CXX = ../../../../config/nvcc_wrapper
 CXXFLAGS = -O3
 LINK = ${CXX}
 LINKFLAGS = 

lib/kokkos/example/tutorial/Hierarchical_Parallelism/03_vectorization/Makefile

@@ -5,7 +5,7 @@ default: build
 	echo "Start Build"
 
 ifneq (,$(findstring Cuda,$(KOKKOS_DEVICES)))
-CXX = nvcc_wrapper
+CXX = ../../../../config/nvcc_wrapper
 CXXFLAGS = -O3
 LINK = ${CXX}
 LINKFLAGS = 

lib/kokkos/example/tutorial/Hierarchical_Parallelism/03_vectorization/vectorization.cpp

@@ -141,11 +141,11 @@ int main(int narg, char* args[]) {
 
   // Each team handles a slice of the data
   // Set up TeamPolicy with 512 teams with maximum number of threads per team and 16 vector lanes.
-  // The team_size_max function will determine the maximum number of threads taking into account
-  // shared memory requirements of the Functor.
+  // Kokkos::AUTO will determine the number of threads
   // The maximum vector length is hardware dependent but can always be smaller than the hardware allows.
   // The vector length must be a power of 2.
-  const Kokkos::TeamPolicy<> policy( 512 , Kokkos::TeamPolicy<>::team_size_max(SomeCorrelation(data,gsum)) , 16);
+
+  const Kokkos::TeamPolicy<> policy( 512 , Kokkos::AUTO , 16);
 
   Kokkos::parallel_for( policy , SomeCorrelation(data,gsum) );
 

lib/kokkos/example/tutorial/Hierarchical_Parallelism/04_team_scan/Makefile

@@ -5,7 +5,7 @@ default: build
 	echo "Start Build"
 
 ifneq (,$(findstring Cuda,$(KOKKOS_DEVICES)))
-CXX = nvcc_wrapper
+CXX = ../../../../config/nvcc_wrapper
 CXXFLAGS = -O3
 LINK = ${CXX}
 LINKFLAGS = 

lib/kokkos/example/tutorial/Hierarchical_Parallelism/04_team_scan/team_scan.cpp

@@ -117,7 +117,7 @@ int main(int narg, char* args[]) {
   Kokkos::DualView<int**> histogram("histogram",TEAM_SIZE,TEAM_SIZE);
 
 
-  Kokkos::Impl::Timer timer;
+  Kokkos::Timer timer;
   // threads/team is automatically limited to maximum supported by the device.
   Kokkos::parallel_for( team_policy( nchunks , TEAM_SIZE )
                       , find_2_tuples(chunk_size,data,histogram) );