Kokkos + KokkosKernels Promotion To Version 2.5.00

cmake/ProjectCompilerPostConfig.cmake

@@ -0,0 +1,61 @@
+IF (${Trilinos_ENABLE_Kokkos})
+
+  # This is where to generate the gen_kokkos.cmake and KokkosCore_config.h 
+  # that we will use in the configuration
+  set(Kokkos_GEN_DIR ${CMAKE_BINARY_DIR})
+
+  # Basic initialization (Used in KOKKOS_SETTINGS)
+  set(KOKKOS_SRC_PATH ${Kokkos_SOURCE_DIR})
+  set(KOKKOS_PATH ${KOKKOS_SRC_PATH})
+
+  #------------ COMPILER AND FEATURE CHECKS ------------------------------------
+  include(${KOKKOS_SRC_PATH}/cmake/kokkos_functions.cmake)
+  set_kokkos_cxx_compiler()
+  set_kokkos_cxx_standard()
+
+  #------------ GET OPTIONS ----------------------------------------------------
+  set(KOKKOS_CMAKE_VERBOSE True)
+  set(KOKKOS_HAS_TRILINOS True)
+  include(${KOKKOS_SRC_PATH}/cmake/kokkos_options.cmake)
+
+  #------------ COMPUTE KOKKOS_SETTINGS ----------------------------------------
+  include(${KOKKOS_SRC_PATH}/cmake/kokkos_settings.cmake)
+
+  #------------ GENERATE HEADER AND SOURCE FILES -------------------------------
+  execute_process(
+    COMMAND ${KOKKOS_SETTINGS} make -f ${KOKKOS_SRC_PATH}/cmake/Makefile.generate_cmake_settings CXX=${CMAKE_CXX_COMPILER} generate_build_settings
+    WORKING_DIRECTORY "${Kokkos_GEN_DIR}"
+    OUTPUT_FILE ${Kokkos_GEN_DIR}/core_src_make.out
+    RESULT_VARIABLE res
+  )
+  include(${Kokkos_GEN_DIR}/kokkos_generated_settings.cmake)
+
+  IF (NOT KOKKOS_ARCH STREQUAL "None")
+
+    # Convert CMakeList into string for CXX_FLAGS
+    set(CMAKE_CXX_FLAGSl "")
+    foreach(opt ${KOKKOS_CXX_FLAGS})
+      set(CMAKE_CXX_FLAGSl "${CMAKE_CXX_FLAGSl} ${opt}")
+    endforeach()
+
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${CMAKE_CXX_FLAGSl}")
+
+    # TODO -- need to remove the -lkokkos.  Check on LDFlags
+    #set(KOKKOS_LINK_DEPENDS libkokkos.a CACHE STRING "")
+    #set(KOKKOS_LIBS -lkokkos -ldl -lpthread CACHE STRING "")
+    #set(KOKKOS_LDFLAGS -L/scr_gabrielle/kruger/builds-ptsolve/trilinos/par2 --gcc-toolchain=/usr CACHE STRING "")
+
+    MESSAGE("-- " "Skip adding flags for C++11 because Kokkos flags does that ...")
+    SET(${PROJECT_NAME}_CXX11_FLAGS " ")
+
+    MESSAGE("-- " "Skip adding flags for OpenMP because Kokkos flags does that ...")
+    SET(OpenMP_CXX_FLAGS_OVERRIDE " ")
+
+  ENDIF()
+
+  # Above, It is important not to distrube the default configuraiton of
+  # Trilinos if KOKKOS_ARCH is not set.  But the implementation of the new
+  # Kokkos TriBITS CMake files requires kokkos_generated_settings.cmake be
+  # included.
+
+ENDIF()

cmake/tribits/core/package_arch/TribitsGlobalMacros.cmake

@@ -1951,10 +1951,10 @@ MACRO(TRIBITS_SETUP_ENV)
   IF(${PROJECT_NAME}_ENABLE_OpenMP)
     FIND_PACKAGE(OpenMP)
     IF(OPENMP_FOUND)
-      SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
-      SET(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} ${OpenMP_C_FLAGS}")
+      TRIBITS_SET_OPENMP_FLAGS(CXX)
+      TRIBITS_SET_OPENMP_FLAGS(C)
       IF(OpenMP_Fortran_FLAGS)
-        SET(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} ${OpenMP_Fortran_FLAGS}")
+        TRIBITS_SET_OPENMP_FLAGS(Fortran)
       ELSE()
       # Older versions of FindOpenMP.cmake don't find Fortran flags.  Mike H said this is safe.
         SET(CMAKE_Fortran_FLAGS "${CMAKE_Fortran_FLAGS} ${OpenMP_C_FLAGS}")
@@ -2012,6 +2012,16 @@ MACRO(TRIBITS_SETUP_ENV)
 
 ENDMACRO()
 
+
+MACRO(TRIBITS_SET_OPENMP_FLAGS  LANG)
+  IF (NOT "${OpenMP_${LANG}_FLAGS_OVERRIDE}" STREQUAL "")
+    SET(CMAKE_${LANG}_FLAGS "${CMAKE_${LANG}_FLAGS} ${OpenMP_${LANG}_FLAGS_OVERRIDE}")
+  ELSE()
+    SET(CMAKE_${LANG}_FLAGS "${CMAKE_${LANG}_FLAGS} ${OpenMP_${LANG}_FLAGS}")
+  ENDIF()
+ENDMACRO()
+
+
 #
 # Set mapping of labels to subprojects (i.e. TriBITS packages) for local CTest
 # only.

cmake/tribits/doc/build_ref/TribitsBuildReferenceBody.rst

@@ -802,6 +802,15 @@ cache variable::
 
   -D <Project>_CXX11_FLAGS="<compiler flags>"
 
+In order to enable C++11 but not have the default system set any flags for
+C++11, use::
+
+  -D <Project>_ENABLE_CXX11=ON
+  -D <Project>_CXX11_FLAGS=" "
+
+The empty space " " will result in the system assuming that no flags needs to
+be set.
+
 
 Enabling explicit template instantiation for C++
 ------------------------------------------------
@@ -1096,6 +1105,17 @@ To enable OpenMP support, one must set::
 Note that if you enable OpenMP directly through a compiler option (e.g.,
 ``-fopenmp``), you will NOT enable OpenMP inside <Project> source code.
 
+To skip adding flags for OpenMP for ``<LANG>`` = ``C``, ``CXX``, or
+``Fortran``, use::
+
+  -D OpenMP_<LANG>_FLAGS_OVERRIDE=" "
+
+The single space " " will result in no flags getting added.  This is needed
+since one can't set the flags ``OpenMP_<LANG>_FLAGS`` to an empty string or
+the ``FIND_PACKAGE(OpenMP)`` command will fail.  Setting the variable
+``-DOpenMP_<LANG>_FLAGS_OVERRIDE= " "`` is the only way to enable OpenMP but
+skip adding the OpenMP flags provided by ``FIND_PACKAGE(OpenMP)``.
+
 .. _BUILD_SHARED_LIBS:
 
 Building shared libraries

doc/build_ref/TrilinosBuildReferenceTemplate.rst

@@ -78,7 +78,7 @@ Enabling/disabling thread safety
 By default, many Trilinos classes are not thread-safe.  However, some of these
 classes can be made thread safe by configuring with::
 
-  -D <Project>_ENABLE_THREAD_SAFE:BOOL=ON
+  -D Trilinos_ENABLE_THREAD_SAFE:BOOL=ON
 
 This will set the default value ``Teuchos_ENABLE_THREAD_SAFE=ON`` which makes
 the Teuchos Memory Management classes (Teuchos::RCP, Teuchos::Ptr,
@@ -92,16 +92,73 @@ Enabling/disabling time monitors
 
 I order to enable instrumentation of select code to generate timing statistics, set::
 
- -D <Project>_ENABLE_TEUCHOS_TIME_MONITOR:BOOL=ON
+ -D Trilinos_ENABLE_TEUCHOS_TIME_MONITOR:BOOL=ON
 
 This will enable Teuchos time monitors by default in all Trilinos packages
 that support them.  To print the timers at the end of the program, call
 ``Teuchos::TimeMonitor::summarize()``.
 
-In order do co-development of TriBTS and Trilinos (see http://http://trac.trilinos.org/wiki/TriBITSTrilinosDev), set::
+Select different TriBITS implementation
+----------------------------------------
+
+In order do co-development of TriBTS and Trilinos (see
+http://http://trac.trilinos.org/wiki/TriBITSTrilinosDev), set, for example::
 
-   -D <Project>_TRIBITS_DIR:STRING=TriBITS \
-   -D <Project>_TRIBITS_PACKAGE_USE_TRIBITS_DIR=TRUE
+   -D Trilinos_TRIBITS_DIR:STRING=TriBITS/tribits
 
 (NOTE: You have to use the data-type ``STRING`` with ``Trilinos_TRIBITS_DIR``
 or CMake will automatically assume it is relative to the build dir!)
+
+
+Configuring with Kokkos and advanced back-ends
+----------------------------------------------
+
+Kokkos (https://github.com/kokkos/kokkos) is a C++ implementation of a
+cross-platform shared-memory parallel programming model. Many Trilinos packages,
+and other stand-alone applications, use it to implement parallel algorithms.
+
+If the Kokkos package is enabled (e.g. ``-DTrilinos_ENABLE_Kokkos=ON``), then
+the following CMake cache variables can be used to get the included Kokkos
+configuration system to select compiler and other build related flags for the
+target machine.  These build-related flags are selected to create correct and
+perforamnt code and for C++ software that uses Kokkos.
+
+============================    ======================================
+Functionality                   CMake Cache Varaible
+============================    ======================================
+Specify architecture            ``KOKKOS_HOST_ARCH``
+Debug builds                    ``KOKKOS_DEBUG``
+Device options:
+* Enable Cuda                   ``TPL_ENABLE_CUDA``
+* Enable OpenMP                 ``Trilinos_ENABLE_OpenMP``
+* Enable Pthread                ``TPL_ENABLE_PThread``
+* Specify Serial                ``TPL_ENABLE_MPI=FALSE``
+Advanced options:
+* Enable compiler warnings      ``KOKKOS_ENABLE_COMPILER_WARNINGS``
+* Aggressive Vectorization      ``KOKKOS_ENABLE_AGGRESSIVE_VECTORIZATION``
+* Profiling                     ``KOKKOS_ENABLE_PROFILING``
+* Enable profile load print     ``KOKKOS_ENABLE_PROFILE_LOAD_PRINT``
+* Enable dualview modify chk    ``KOKKOS_ENABLE_DUALVIEW_MODIFY_CHECK``
+Kokkos TPLs:                 
+* Use hwloc library             ``TPL_ENABLE_HWLOC``
+* Use memkind library           ``KOKKOS_ENABLE_MEMKIND``
+* Use librt                     ``KOKKOS_ENABLE_LIBRT``
+CUDA Options:                
+* Enable CUDA LDG               ``KOKKOS_ENABLE_CUDA_LDG_INTRINSIC`` (global mem load)
+* Enable CUDA UVM               ``KOKKOS_ENABLE_CUDA_UVM`` (unified virtual mem)
+* Enable CUDA RDC               ``KOKKOS_ENABLE_CUDA_RELOCATABLE_DEVICE_CODE``
+* Enable CUDA LAMBDA            ``KOKKOS_ENABLE_CUDA_LAMBDA``
+============================    ======================================
+
+If the cache var ``KOKKOS_HOST_ARCH`` is not set (or is set to ``None``) then
+the Kokkos settings are not used and the default Trilinos CMake configuration
+is used as described below.
+
+If ``KOKKOS_HOST_ARCH != None`` is set, then the correct compiler flags for
+C++11 and OpenMP are selected by the Kokkos system and the values of the cache
+vars ``Trilinos_CXX11_FLAGS`` and ``OpenMP_CXX_FLAGS`` set by the user will be
+ignored.
+
+To see more documentation for each of these options, run a configure with
+``-DTrilinos_ENABLE_Kokkos=ON`` and then look in the CMakeCache.txt file (as
+raw text or using the CMake QT GUI or ``ccmake``).

packages/ifpack2/src/Ifpack2_Relaxation_decl.hpp

@@ -609,7 +609,7 @@ class Relaxation :
   typedef typename local_matrix_type::StaticCrsGraphType::device_type PersistentWorkSpace;
   typedef typename local_matrix_type::StaticCrsGraphType::execution_space MyExecSpace;
   typedef typename KokkosKernels::Experimental::KokkosKernelsHandle
-      <lno_row_view_t,lno_nonzero_view_t, scalar_nonzero_view_t,
+      <typename lno_row_view_t::const_value_type, local_ordinal_type,typename scalar_nonzero_view_t::value_type,
       MyExecSpace, TemporaryWorkSpace,PersistentWorkSpace > mt_kernel_handle_type;
   Teuchos::RCP<mt_kernel_handle_type> mtKernelHandle_;
 #endif // HAVE_IFPACK2_EXPERIMENTAL_KOKKOSKERNELS_FEATURES

packages/kokkos-kernels/CHANGELOG.md

@@ -0,0 +1,63 @@
+# Change Log
+
+## [2.5.00](https://github.com/kokkos/kokkos-kernels/tree/2.5.00) (2017-12-15)
+[Full Changelog](https://github.com/kokkos/kokkos-kernels/compare/0.10.03...2.5.00)
+
+**Implemented enhancements:**
+
+- KokkosBlas:   Add GEMM interface  [\#105](https://github.com/kokkos/kokkos-kernels/issues/105)
+- KokkosBlas:   Add GEMM default Kernel [\#125](https://github.com/kokkos/kokkos-kernels/issues/125)
+- KokkosBlas:   Add GEMV that wraps BLAS \(and cuBLAS\) [\#16](https://github.com/kokkos/kokkos-kernels/issues/16)
+- KokkosSparse: Make SPMV test not print GBs of output if something goes wrong.  [\#111](https://github.com/kokkos/kokkos-kernels/issues/111)
+- KokkosSparse: ETI SpGEMM and Gauss Seidel and take it out of Experimental namespace [\#74](https://github.com/kokkos/kokkos-kernels/issues/74)
+- BuildSystem:  Fix Makesystem to correctly build library after aborted install [\#104](https://github.com/kokkos/kokkos-kernels/issues/104)
+- BuildSystem:  Add option ot generate\_makefile.bash to define memoryspaces for instantiation [\#89](https://github.com/kokkos/kokkos-kernels/issues/89)
+- BuildSystem:  generate makefile tpl option [\#66](https://github.com/kokkos/kokkos-kernels/issues/66)
+- BuildSystem:  Add a simpler compilation script, README update etc [\#96](https://github.com/kokkos/kokkos-kernels/issues/96)
+
+**Fixed bugs:**
+
+- Internal Compiler Error GCC in GEMM [\#129](https://github.com/kokkos/kokkos-kernels/issues/129)
+- Batched Team LU: bug for small team\_size [\#110](https://github.com/kokkos/kokkos-kernels/issues/110)
+- Compiler BUG in IBM XL pragma unrolling [\#92](https://github.com/kokkos/kokkos-kernels/issues/92)
+- Fix Blas TPL enables build [\#77](https://github.com/kokkos/kokkos-kernels/issues/77)
+- Batched Gemm Failure [\#73](https://github.com/kokkos/kokkos-kernels/issues/73)
+- CUDA 7.5 \(GCC 4.8.4\) build errors [\#72](https://github.com/kokkos/kokkos-kernels/issues/72)
+- Cuda BLAS tests fail with UVM if CUDA\_LAUNCH\_BLOCKING=1 is not defined on Kepler [\#51](https://github.com/kokkos/kokkos-kernels/issues/51)
+- CrsMatrix: sumIntoValues and replaceValues incorrectly count the number of valid column indices. [\#11](https://github.com/kokkos/kokkos-kernels/issues/11)
+- findRelOffset test assumes UVM [\#32](https://github.com/kokkos/kokkos-kernels/issues/32)
+
+## [0.10.03](https://github.com/kokkos/kokkos-kernels/tree/0.10.03) (2017-09-11)
+
+**Implemented enhancements:**
+
+- KokkosSparse: Fix unused variable warnings in spmv\_impl\_omp, spmv Test and graph color perf\_test [\#63](https://github.com/kokkos/kokkos-kernels/issues/63)
+- KokkosBlas:  dot: Add unit test [\#15](https://github.com/kokkos/kokkos-kernels/issues/15)
+- KokkosBlas:  dot: Add special case for multivector \* vector \(or vector \* multivector\) [\#13](https://github.com/kokkos/kokkos-kernels/issues/13)
+- BuildSystem: Make KokkosKernels build independently of Trilinos [\#1](https://github.com/kokkos/kokkos-kernels/issues/1)
+- BuildSystem: Fix ETI System not to depend on Tpetra ETI [\#5](https://github.com/kokkos/kokkos-kernels/issues/5)
+- BuildSystem: Change CMake to work with new ETI system [\#19](https://github.com/kokkos/kokkos-kernels/issues/19)
+- BuildSystem: Fix TpetraKernels names to KokkosKernels [\#4](https://github.com/kokkos/kokkos-kernels/issues/4)
+- BuildSystem: Trilinos/KokkosKernels reports no ETI in almost any circumstance [\#29](https://github.com/kokkos/kokkos-kernels/issues/29)
+- General:     Kokkos::ArithTraits\<double\>::nan\(\) is very slow [\#35](https://github.com/kokkos/kokkos-kernels/issues/35)
+- General:     Design and Define New UnitTest infrastructure [\#28](https://github.com/kokkos/kokkos-kernels/issues/28)
+- General:     Move Tpetra::Details::OrdinalTraits to KokkosKernels [\#22](https://github.com/kokkos/kokkos-kernels/issues/22)
+- General:     Rename files and NameSpace to KokkosKernels [\#12](https://github.com/kokkos/kokkos-kernels/issues/12)
+- General:      PrepareStandalone: Get rid of Teuchos usage [\#2](https://github.com/kokkos/kokkos-kernels/issues/2)
+- General:      Fix warning with char being either signed or unsigned in ArithTraits [\#60](https://github.com/kokkos/kokkos-kernels/issues/60)
+- Testing:      Make all tests run with -Werror [\#68](https://github.com/kokkos/kokkos-kernels/issues/68)
+
+**Fixed bugs:**
+
+- SPGEMM Test Fails for Cuda when compiled through Trilinos  [\#49](https://github.com/kokkos/kokkos-kernels/issues/49)
+- Fix ArithTraits min for floating points [\#47](https://github.com/kokkos/kokkos-kernels/issues/47)
+- Pthread ETI error [\#25](https://github.com/kokkos/kokkos-kernels/issues/25)
+- Fix CMake Based ETI for Threads backend [\#46](https://github.com/kokkos/kokkos-kernels/issues/46)
+- KokkosKernels\_ENABLE\_EXPERIMENTAL causes build error  [\#59](https://github.com/kokkos/kokkos-kernels/issues/59)
+- ArithTraits warnings in CUDA build [\#71](https://github.com/kokkos/kokkos-kernels/issues/71)
+- Graph coloring build warnings [\#3](https://github.com/kokkos/kokkos-kernels/issues/3)
+
+
+
+
+\* *This Change Log was automatically generated by [github_changelog_generator](https://github.com/skywinder/Github-Changelog-Generator)*

packages/kokkos-kernels/Makefile.kokkos-kernels

@@ -189,6 +189,7 @@ KOKKOSKERNELS_INTERNAL_INST_MEMSPACE_HOSTSPACE=$(strip $(shell echo $(KOKKOSKERN
 KOKKOSKERNELS_INTERNAL_INST_MEMSPACE_HBWSPACE=$(strip $(shell echo $(KOKKOSKERNELS_SPACES) | grep "HBWSpace" | wc -l))
 KOKKOSKERNELS_INTERNAL_INST_MEMSPACE_CUDASPACE=$(strip $(shell echo $(KOKKOSKERNELS_SPACES) | grep "CudaSpace" | wc -l))
 KOKKOSKERNELS_INTERNAL_INST_MEMSPACE_CUDAUVMSPACE=$(strip $(shell echo $(KOKKOSKERNELS_SPACES) | grep "CudaUVMSpace" | wc -l))
+KOKKOSKERNELS_INTERNAL_INST_MEMSPACE_CUDAHOSTPINNEDSPACE=$(strip $(shell echo $(KOKKOSKERNELS_SPACES) | grep "CudaHostPinnedSpace" | wc -l))
 
 tmp := $(shell echo "" >> KokkosKernels_config.tmp)
 tmp := $(shell echo "/* ---------------------------------------------" >> KokkosKernels_config.tmp)
@@ -207,6 +208,10 @@ endif
 ifeq ($(KOKKOSKERNELS_INTERNAL_INST_MEMSPACE_CUDAUVMSPACE), 1)
   tmp := $(shell echo "\#define KOKKOSKERNELS_INST_MEMSPACE_CUDAUVMSPACE" >> KokkosKernels_config.tmp )
 endif
+ifeq ($(KOKKOSKERNELS_INTERNAL_INST_MEMSPACE_CUDAHOSTPINNEDSPACE), 1)
+  tmp := $(shell echo "\#define KOKKOSKERNELS_INST_MEMSPACE_CUDAHOSTPINNEDSPACE" >> KokkosKernels_config.tmp )
+endif
+
 ifeq ($(KOKKOSKERNELS_INTERNAL_INST_EXECSPACE_CUDA), 1)
   ifeq ($(KOKKOSKERNELS_INTERNAL_INST_MEMSPACE_CUDASPACE), 0)
     ifeq ($(KOKKOSKERNELS_INTERNAL_INST_MEMSPACE_CUDAUVMSPACE), 0)
@@ -256,7 +261,7 @@ ifeq (${KOKKOSKERNELS_INTERNAL_ENABLE_CUBLAS}, 1)
   tmp := $(shell echo "\#define KOKKOSKERNELS_ENABLE_TPL_CUBLAS" >> KokkosKernels_config.tmp )
 endif
 
-  tmp := $(shell echo "\#if defined(KOKKOSKERNELS_ENABLE_TPL_MKL)" >> KokkosKernels_config.tmp )
+  tmp := $(shell echo "\#if defined(KOKKOSKERNELS_ENABLE_TPL_MKL) || defined(KOKKOSKERNELS_ENABLE_TPL_CUBLAS)" >> KokkosKernels_config.tmp )
   tmp := $(shell echo "\#if !defined(KOKKOSKERNELS_ENABLE_TPL_BLAS)" >> KokkosKernels_config.tmp )  
   tmp := $(shell echo "\#define KOKKOSKERNELS_ENABLE_TPL_BLAS" >> KokkosKernels_config.tmp )
   tmp := $(shell echo "\#endif" >> KokkosKernels_config.tmp )

packages/kokkos-kernels/README

@@ -1,3 +1,6 @@
+Kokkos C++ Performance Portability Programming EcoSystem: Math Kernels -
+Provides BLAS, Sparse BLAS and Graph Kernels 
+
 KokkosKernels implements local computational kernels for linear
 algebra and graph operations, using the Kokkos shared-memory parallel
 programming model.  "Local" means not using MPI, or running within a
@@ -19,7 +22,7 @@ Computational kernels in this subpackage include the following:
   - Gauss-Seidel with coloring (generalization of red-black)
   - Other linear algebra and graph operations
 
-Kokkos is licensed under standard 3-clause BSD terms of use.  For
+KokkosKernels is licensed under standard 3-clause BSD terms of use.  For
 specifics, please refer to the LICENSE file contained in the
 repository or distribution.
 
@@ -43,7 +46,9 @@ We organize this directory as follows:
   3. Correctness tests live in the unit_test/ subdirectory, and
      performance tests live in the perf_test/ subdirectory
 
-  4. Compact batched BLAS: Refer the README in src/stage/batched/.
+  4. Simple example scripts to build Kokkoskernels are in
+     example/buildlib/
+
 
 Do NOT use or rely on anything in the KokkosBlas::Impl namespace, or
 on anything in the impl/ subdirectory.
@@ -59,6 +64,12 @@ for certain template parameter combinations.  It also improves
 performance, since compilers have an easier time optimizing code in
 shorter .cpp files.
 
+Building Kokkoskernels
+----------------------
 
-
+  1. Modify example/buildlib/compileKokkosKernelsSimple.sh or
+     example/buildlib/compileKokkosKernels.sh for your environment
+     and run it to generate the required makefiles. 
+
+  2. Run "make build-test" to compile the tests.
 

packages/kokkos-kernels/example/buildlib/compileKokkosKernelsSimple.sh

@@ -0,0 +1,13 @@
+KOKKOS_PATH=${HOME}/proj/kokkos #path to kokkos source
+KOKKOSKERNELS_SCALARS=double #the scalar types to instantiate =double,float...
+KOKKOSKERNELS_LAYOUTS=LayoutLeft #the layout types to instantiate.
+KOKKOSKERNELS_ORDINALS=int #ordinal types to instantiate
+KOKKOSKERNELS_OFFSETS=int #offset types to instantiate
+KOKKOSKERNELS_PATH=../.. #path to kokkos-kernels top directory.
+CXX=icpc #${KOKKOS_PATH}/config/nvcc_wrapper #icpc #
+KOKKOSKERNELS_OPTIONS=eti-only #options for kokkoskernels  
+KOKKOS_DEVICES=OpenMP # other devices Cuda,Serial ..
+KOKKOS_ARCHS=KNL
+CXXFLAGS="-Wall -pedantic -Werror -O3 -g -Wshadow -Wsign-compare -Wtype-limits -Wuninitialized"
+
+../../scripts/generate_makefile.bash --kokkoskernels-path=${KOKKOSKERNELS_PATH} --with-scalars=${KOKKOSKERNELS_SCALARS} --with-ordinals=${KOKKOSKERNELS_ORDINALS} --with-offsets=${KOKKOSKERNELS_OFFSETS} --kokkos-path=${KOKKOS_PATH} --with-devices=${KOKKOS_DEVICES} --arch=${KOKKOS_ARCHS} --compiler=${CXX} --with-options=${KOKKOSKERNELS_OPTIONS}  --cxxflags="${CXXFLAGS}"

packages/kokkos-kernels/master_history.txt

@@ -1,2 +1,4 @@
 tag: 0.10.00    date: 08/22/1017  master: xxxxxxxx    develop: e323bcb0
 tag: 0.10.03    date: 09/11/2017  master: e323bcb0    develop: 4cb87390
+tag: 2.5.00     date: 12/15/2017  master: e4c645e9    develop: 04d58766
+