Merge pull request #725 from stanmoore1/kk_update

# Change Log

## [2.5.00](https://github.com/kokkos/kokkos/tree/2.5.00) (2017-12-15)

[Full Changelog](https://github.com/kokkos/kokkos/compare/2.04.11...2.5.00)

**Part of the Kokkos C++ Performance Portability Programming EcoSystem 2.5**      

**Implemented enhancements:**

- Provide Makefile.kokkos logic for CMake and TriBITS [\#878](https://github.com/kokkos/kokkos/issues/878)

- Add Scatter View [\#825](https://github.com/kokkos/kokkos/issues/825)

- Drop gcc 4.7 and intel 14 from supported compiler list [\#603](https://github.com/kokkos/kokkos/issues/603)

- Enable construction of unmanaged view using common\_view\_alloc\_prop [\#1170](https://github.com/kokkos/kokkos/issues/1170)

- Unused Function Warning with XL [\#1267](https://github.com/kokkos/kokkos/issues/1267)

- Add memory pool parameter check [\#1218](https://github.com/kokkos/kokkos/issues/1218)

- CUDA9: Fix warning for unsupported long double [\#1189](https://github.com/kokkos/kokkos/issues/1189)

- CUDA9: fix warning on defaulted function marking [\#1188](https://github.com/kokkos/kokkos/issues/1188)

- CUDA9: fix warnings for deprecated warp level functions [\#1187](https://github.com/kokkos/kokkos/issues/1187)

- Add CUDA 9.0 nightly testing [\#1174](https://github.com/kokkos/kokkos/issues/1174)

- {OMPI,MPICH}\_CXX hack breaks nvcc\_wrapper use case [\#1166](https://github.com/kokkos/kokkos/issues/1166)

- KOKKOS\_HAVE\_CUDA\_LAMBDA became KOKKOS\_CUDA\_USE\_LAMBDA [\#1274](https://github.com/kokkos/kokkos/issues/1274)

**Fixed bugs:**

- MinMax Reducer with tagged operator doesn't compile [\#1251](https://github.com/kokkos/kokkos/issues/1251)

- Reducers for Tagged operators give wrong answer [\#1250](https://github.com/kokkos/kokkos/issues/1250)

- Kokkos not Compatible with Big Endian Machines? [\#1235](https://github.com/kokkos/kokkos/issues/1235)

- Parallel Scan hangs forever on BG/Q [\#1234](https://github.com/kokkos/kokkos/issues/1234)

- Threads backend doesn't compile with Clang on OS X [\#1232](https://github.com/kokkos/kokkos/issues/1232)

- $\(shell date\) needs quote [\#1264](https://github.com/kokkos/kokkos/issues/1264)

- Unqualified parallel\_for call conflicts with user-defined parallel\_for [\#1219](https://github.com/kokkos/kokkos/issues/1219)

- KokkosAlgorithms: CMake issue in unit tests [\#1212](https://github.com/kokkos/kokkos/issues/1212)

- Intel 18 Error: "simd pragma has been deprecated" [\#1210](https://github.com/kokkos/kokkos/issues/1210)

- Memory leak in Kokkos::initialize [\#1194](https://github.com/kokkos/kokkos/issues/1194)

- CUDA9: compiler error with static assert template arguments [\#1190](https://github.com/kokkos/kokkos/issues/1190)

- Kokkos::Serial::is\_initialized returns always true [\#1184](https://github.com/kokkos/kokkos/issues/1184)

- Triple nested parallelism still fails on bowman [\#1093](https://github.com/kokkos/kokkos/issues/1093)

- OpenMP openmp.range on Develop Runs Forever on POWER7+ with RHEL7 and GCC4.8.5 [\#995](https://github.com/kokkos/kokkos/issues/995)

- Rendezvous performance at global scope [\#985](https://github.com/kokkos/kokkos/issues/985)

## [2.04.11](https://github.com/kokkos/kokkos/tree/2.04.11) (2017-10-28)

[Full Changelog](https://github.com/kokkos/kokkos/compare/2.04.04...2.04.11)

**Implemented enhancements:**

- Add Subview pattern. [\#648](https://github.com/kokkos/kokkos/issues/648)

- Add Kokkos "global" is\_initialized [\#1060](https://github.com/kokkos/kokkos/issues/1060)

- Add create\_mirror\_view\_and\_copy [\#1161](https://github.com/kokkos/kokkos/issues/1161)

- Add KokkosConcepts SpaceAccessibility function [\#1092](https://github.com/kokkos/kokkos/issues/1092)

- Option to Disable Initialize Warnings [\#1142](https://github.com/kokkos/kokkos/issues/1142)

- Mature task-DAG capability [\#320](https://github.com/kokkos/kokkos/issues/320)

- Promote Work DAG from experimental [\#1126](https://github.com/kokkos/kokkos/issues/1126)

- Implement new WorkGraph push/pop [\#1108](https://github.com/kokkos/kokkos/issues/1108)

- Kokkos\_ENABLE\_Cuda\_Lambda should default ON [\#1101](https://github.com/kokkos/kokkos/issues/1101)

- Add multidimensional parallel for example and improve unit test [\#1064](https://github.com/kokkos/kokkos/issues/1064)

- Fix ROCm:  Performance tests not building [\#1038](https://github.com/kokkos/kokkos/issues/1038)

- Make KOKKOS\_ALIGN\_SIZE a configure-time option [\#1004](https://github.com/kokkos/kokkos/issues/1004)

- Make alignment consistent [\#809](https://github.com/kokkos/kokkos/issues/809)

- Improve subview construction on Cuda backend [\#615](https://github.com/kokkos/kokkos/issues/615)

**Fixed bugs:**

- Kokkos::vector fixes for application [\#1134](https://github.com/kokkos/kokkos/issues/1134)

- DynamicView non-power of two value\_type [\#1177](https://github.com/kokkos/kokkos/issues/1177)

- Memory pool bug [\#1154](https://github.com/kokkos/kokkos/issues/1154)

- Cuda launch bounds performance regression bug [\#1140](https://github.com/kokkos/kokkos/issues/1140)

- Significant performance regression in LAMMPS after updating Kokkos [\#1139](https://github.com/kokkos/kokkos/issues/1139)

- CUDA compile error [\#1128](https://github.com/kokkos/kokkos/issues/1128)

- MDRangePolicy neg idx test failure in debug mode [\#1113](https://github.com/kokkos/kokkos/issues/1113)

- subview construction on Cuda backend [\#615](https://github.com/kokkos/kokkos/issues/615)

## [2.04.04](https://github.com/kokkos/kokkos/tree/2.04.04) (2017-09-11)

[Full Changelog](https://github.com/kokkos/kokkos/compare/2.04.00...2.04.04)

# Is this a build as part of Trilinos?

IF(COMMAND TRIBITS_PACKAGE_DECL)

  SET(KOKKOS_HAS_TRILINOS ON CACHE BOOL "")

ELSE()

IF(NOT KOKKOS_HAS_TRILINOS)

  cmake_minimum_required(VERSION 3.1 FATAL_ERROR)

  project(Kokkos CXX)

  INCLUDE(cmake/kokkos.cmake)

ELSE()

#------------------------------------------------------------------------------

#

# A) Forward delcare the package so that certain options are also defined for

# subpackages

#

TRIBITS_PACKAGE_DECL(Kokkos) # ENABLE_SHADOWING_WARNINGS)

#------------------------------------------------------------------------------

#

# B) Define the common options for Kokkos first so they can be used by

# subpackages as well.

#

  # Define Project Name if this is a standalone build

  IF(NOT DEFINED ${PROJECT_NAME})

    project(Kokkos CXX) 

  ENDIF()

# mfh 01 Aug 2016: See Issue #61:

#

# https://github.com/kokkos/kokkos/issues/61

#

# Don't use TRIBITS_ADD_DEBUG_OPTION() here, because that defines

# HAVE_KOKKOS_DEBUG.  We define KOKKOS_HAVE_DEBUG here instead,

# for compatibility with Kokkos' Makefile build system.

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_DEBUG

  KOKKOS_HAVE_DEBUG

  "Enable run-time debug checks.  These checks may be expensive, so they are disabled by default in a release build."

  ${${PROJECT_NAME}_ENABLE_DEBUG}

)

  # Basic initialization (Used in KOKKOS_SETTINGS)

  set(KOKKOS_SRC_PATH ${Kokkos_SOURCE_DIR})

  set(KOKKOS_PATH ${KOKKOS_SRC_PATH})

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_SIERRA_BUILD

  KOKKOS_FOR_SIERRA

  "Configure Kokkos for building within the Sierra build system."

  OFF

  )

  #------------ COMPILER AND FEATURE CHECKS ------------------------------------

  include(${KOKKOS_SRC_PATH}/cmake/kokkos_functions.cmake)

  set_kokkos_cxx_compiler()

  set_kokkos_cxx_standard()

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_Cuda

  KOKKOS_HAVE_CUDA

  "Enable CUDA support in Kokkos."

  "${TPL_ENABLE_CUDA}"

  )

  #------------ GET OPTIONS AND KOKKOS_SETTINGS --------------------------------

  # Add Kokkos' modules to CMake's module path.

  set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} "${Kokkos_SOURCE_DIR}/cmake/Modules/")

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_Cuda_UVM

  KOKKOS_USE_CUDA_UVM

  "Enable CUDA Unified Virtual Memory as the default in Kokkos."

  OFF

  )

  set(KOKKOS_CMAKE_VERBOSE True)

  include(${KOKKOS_SRC_PATH}/cmake/kokkos_options.cmake)

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_Cuda_RDC

  KOKKOS_HAVE_CUDA_RDC

  "Enable CUDA Relocatable Device Code support in Kokkos."

  OFF

  )

  include(${KOKKOS_SRC_PATH}/cmake/kokkos_settings.cmake)

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_Cuda_Lambda

  KOKKOS_HAVE_CUDA_LAMBDA

  "Enable CUDA LAMBDA support in Kokkos."

  OFF

  #------------ 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_BINARY_DIR}"

    OUTPUT_FILE ${Kokkos_BINARY_DIR}/core_src_make.out

    RESULT_VARIABLE res

  )

  include(${Kokkos_BINARY_DIR}/kokkos_generated_settings.cmake)

  set_kokkos_srcs(KOKKOS_SRC ${KOKKOS_SRC})

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_Pthread

  KOKKOS_HAVE_PTHREAD

  "Enable Pthread support in Kokkos."

  OFF

  )

  #------------ NOW BUILD ------------------------------------------------------

  include(${KOKKOS_SRC_PATH}/cmake/kokkos_build.cmake)

ASSERT_DEFINED(TPL_ENABLE_Pthread)

IF(Kokkos_ENABLE_Pthread AND NOT TPL_ENABLE_Pthread)

  MESSAGE(FATAL_ERROR "You set Kokkos_ENABLE_Pthread=ON, but Trilinos' support for Pthread(s) is not enabled (TPL_ENABLE_Pthread=OFF).  This is not allowed.  Please enable Pthreads in Trilinos before attempting to enable Kokkos' support for Pthreads.")

ENDIF()

IF(NOT TPL_ENABLE_Pthread)

  ADD_DEFINITIONS(-DGTEST_HAS_PTHREAD=0)

ENDIF()

  #------------ Add in Fake Tribits Handling to allow unit test builds- --------

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_OpenMP

  KOKKOS_HAVE_OPENMP

  "Enable OpenMP support in Kokkos."

  "${${PROJECT_NAME}_ENABLE_OpenMP}"

  )

  include(${KOKKOS_SRC_PATH}/cmake/tribits.cmake)

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_QTHREAD

  KOKKOS_HAVE_QTHREADS

  "Enable Qthreads support in Kokkos."

  "${TPL_ENABLE_QTHREAD}"

  )

  TRIBITS_PACKAGE_DECL(Kokkos)

# TODO: No longer an option in Kokkos.  Needs to be removed.

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_CXX11

  KOKKOS_HAVE_CXX11

  "Enable C++11 support in Kokkos."

  "${${PROJECT_NAME}_ENABLE_CXX11}"

  )

  ADD_SUBDIRECTORY(core)

  ADD_SUBDIRECTORY(containers)

  ADD_SUBDIRECTORY(algorithms)

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_HWLOC

  KOKKOS_HAVE_HWLOC

  "Enable HWLOC support in Kokkos."

  "${TPL_ENABLE_HWLOC}"

  )

# TODO: This is currently not used in Kokkos.  Should it be removed?

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_MPI

  KOKKOS_HAVE_MPI

  "Enable MPI support in Kokkos."

  "${TPL_ENABLE_MPI}"

  )

# Set default value of Kokkos_ENABLE_Debug_Bounds_Check option

#

# CMake is case sensitive.  The Kokkos_ENABLE_Debug_Bounds_Check

# option (defined below) is annoyingly not all caps, but we need to

# keep it that way for backwards compatibility.  If users forget and

# try using an all-caps variable, then make it count by using the

# all-caps version as the default value of the original, not-all-caps

# option.  Otherwise, the default value of this option comes from

# Kokkos_ENABLE_DEBUG (see Issue #367).

ASSERT_DEFINED(${PACKAGE_NAME}_ENABLE_DEBUG)

IF(DEFINED Kokkos_ENABLE_DEBUG_BOUNDS_CHECK)

  IF(Kokkos_ENABLE_DEBUG_BOUNDS_CHECK)

    SET(Kokkos_ENABLE_Debug_Bounds_Check_DEFAULT ON)

  ELSE()

    SET(Kokkos_ENABLE_Debug_Bounds_Check_DEFAULT "${${PACKAGE_NAME}_ENABLE_DEBUG}")

  ENDIF()

ELSE()

  SET(Kokkos_ENABLE_Debug_Bounds_Check_DEFAULT "${${PACKAGE_NAME}_ENABLE_DEBUG}")

ENDIF()

ASSERT_DEFINED(Kokkos_ENABLE_Debug_Bounds_Check_DEFAULT)

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_Debug_Bounds_Check

  KOKKOS_ENABLE_DEBUG_BOUNDS_CHECK

  "Enable Kokkos::View run-time bounds checking."

  "${Kokkos_ENABLE_Debug_Bounds_Check_DEFAULT}"

  )

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_Debug_DualView_Modify_Check

  KOKKOS_ENABLE_DEBUG_DUALVIEW_MODIFY_CHECK

  "Enable abort when Kokkos::DualView modified on host and device without sync."

  "${Kokkos_ENABLE_DEBUG}"

  )

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_Profiling

  KOKKOS_ENABLE_PROFILING

  "Enable KokkosP profiling support for kernel data collections."

  "${TPL_ENABLE_DLlib}"

  )

#------------------------------------------------------------------------------

#

# A) Forward declare the package so that certain options are also defined for

# subpackages

#

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_Profiling_Load_Print

  KOKKOS_ENABLE_PROFILING_LOAD_PRINT

  "Print to standard output which profiling library was loaded."

  OFF

  )

TRIBITS_PACKAGE_DECL(Kokkos) # ENABLE_SHADOWING_WARNINGS)

# placeholder for future device...

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_ENABLE_Winthread

  KOKKOS_HAVE_WINTHREAD

  "Enable Winthread support in Kokkos."

  "${TPL_ENABLE_Winthread}"

  )

# TODO: No longer an option in Kokkos.  Needs to be removed.

# use new/old View

TRIBITS_ADD_OPTION_AND_DEFINE(

  Kokkos_USING_DEPRECATED_VIEW

  KOKKOS_USING_DEPRECATED_VIEW

  "Choose whether to use the old, deprecated Kokkos::View"

  OFF

  )

#------------------------------------------------------------------------------

#

# B) Install Kokkos' build files

#

# If using the Makefile-generated files, then need to set things up.

# Here, assume that TriBITS has been run from ProjectCompilerPostConfig.cmake

# and already generated KokkosCore_config.h and kokkos_generated_settings.cmake

# in the previously define Kokkos_GEN_DIR

# We need to copy them over to the correct place and source the cmake file

if(NOT KOKKOS_LEGACY_TRIBITS)

  set(Kokkos_GEN_DIR ${CMAKE_BINARY_DIR})

  file(COPY "${Kokkos_GEN_DIR}/KokkosCore_config.h"

    DESTINATION "${CMAKE_CURRENT_BINARY_DIR}" USE_SOURCE_PERMISSIONS)

  install(FILES "${Kokkos_GEN_DIR}/KokkosCore_config.h"

    DESTINATION include)

  file(COPY "${Kokkos_GEN_DIR}/kokkos_generated_settings.cmake"

    DESTINATION "${CMAKE_CURRENT_BINARY_DIR}" USE_SOURCE_PERMISSIONS)

  include(${CMAKE_CURRENT_BINARY_DIR}/kokkos_generated_settings.cmake)

  # Sources come from makefile-generated kokkos_generated_settings.cmake file

  # Enable using the individual sources if needed

  set_kokkos_srcs(KOKKOS_SRC ${KOKKOS_SRC})

endif ()

#------------------------------------------------------------------------------

TRIBITS_EXCLUDE_AUTOTOOLS_FILES()

TRIBITS_EXCLUDE_FILES(

  classic/doc

  classic/LinAlg/doc/CrsRefactorNotesMay2012

  )

TRIBITS_PACKAGE_POSTPROCESS()

ENDIF()

	$(CXX) $(KOKKOS_CPPFLAGS) $(KOKKOS_CXXFLAGS) $(CXXFLAGS) -c $(KOKKOS_PATH)/core/src/impl/Kokkos_HostThreadTeam.cpp

Kokkos_Spinwait.o: $(KOKKOS_CPP_DEPENDS) $(KOKKOS_PATH)/core/src/impl/Kokkos_Spinwait.cpp

	$(CXX) $(KOKKOS_CPPFLAGS) $(KOKKOS_CXXFLAGS) $(CXXFLAGS) -c $(KOKKOS_PATH)/core/src/impl/Kokkos_Spinwait.cpp

Kokkos_Rendezvous.o: $(KOKKOS_CPP_DEPENDS) $(KOKKOS_PATH)/core/src/impl/Kokkos_Rendezvous.cpp

	$(CXX) $(KOKKOS_CPPFLAGS) $(KOKKOS_CXXFLAGS) $(CXXFLAGS) -c $(KOKKOS_PATH)/core/src/impl/Kokkos_Rendezvous.cpp

Kokkos_HostBarrier.o: $(KOKKOS_CPP_DEPENDS) $(KOKKOS_PATH)/core/src/impl/Kokkos_HostBarrier.cpp

	$(CXX) $(KOKKOS_CPPFLAGS) $(KOKKOS_CXXFLAGS) $(CXXFLAGS) -c $(KOKKOS_PATH)/core/src/impl/Kokkos_HostBarrier.cpp

Kokkos_Profiling_Interface.o: $(KOKKOS_CPP_DEPENDS) $(KOKKOS_PATH)/core/src/impl/Kokkos_Profiling_Interface.cpp

	$(CXX) $(KOKKOS_CPPFLAGS) $(KOKKOS_CXXFLAGS) $(CXXFLAGS) -c $(KOKKOS_PATH)/core/src/impl/Kokkos_Profiling_Interface.cpp

Kokkos_SharedAlloc.o: $(KOKKOS_CPP_DEPENDS) $(KOKKOS_PATH)/core/src/impl/Kokkos_SharedAlloc.cpp

============================================================================

Primary tested compilers on X86 are:

  GCC 4.7.2

  GCC 4.8.4

  GCC 4.9.2

  GCC 4.9.3

  GCC 5.1.0

  GCC 5.2.0

  Intel 14.0.4

  GCC 5.3.0

  GCC 6.1.0

  Intel 15.0.2

  Intel 16.0.1

  Intel 17.0.098

  Intel 17.1.132

  Intel 17.1.043

  Intel 17.4.196

  Intel 18.0.128

  Clang 3.5.2

  Clang 3.6.1

  Clang 3.7.1

  Clang 3.8.1

  Clang 3.9.0

  PGI 17.1

  Clang 4.0.0

  Clang 4.0.0 for CUDA (CUDA Toolkit 8.0.44)

  PGI 17.10

  NVCC 7.0 for CUDA (with gcc 4.8.4)

  NVCC 7.5 for CUDA (with gcc 4.8.4)

  NVCC 8.0.44 for CUDA (with gcc 5.3.0)

Primary tested compilers on Power 8 are:

  GCC 5.4.0 (OpenMP,Serial)

  IBM XL 13.1.3 (OpenMP, Serial) (There is a workaround in place to avoid a compiler bug)

  IBM XL 13.1.5 (OpenMP, Serial) (There is a workaround in place to avoid a compiler bug)

  NVCC 8.0.44 for CUDA (with gcc 5.4.0)

  NVCC 9.0.103 for CUDA (with gcc 6.3.0)

Primary tested compilers on Intel KNL are:

   GCC 6.2.0

   Intel 16.2.181 (with gcc 4.7.2)

   Intel 17.0.098 (with gcc 4.7.2)

   Intel 17.1.132 (with gcc 4.9.3)

   Intel 16.4.258 (with gcc 4.7.2)

   Intel 17.2.174 (with gcc 4.9.3)

   Intel 18.0.061 (beta) (with gcc 4.9.3)

Secondary tested compilers are:

  CUDA 7.0 (with gcc 4.8.4)

  CUDA 7.5 (with gcc 4.8.4)

  CUDA 8.0 (with gcc 5.3.0 on X86 and gcc 5.4.0 on Power8)

  CUDA/Clang 8.0 using Clang/Trunk compiler

   Intel 18.0.128 (with gcc 4.9.3)

Other compilers working:

  X86:

   Cygwin 2.1.0 64bit with gcc 4.9.3

Limited testing of the following compilers on POWER7+ systems:

  GCC 4.8.5 (on RHEL7.1 POWER7+)

Known non-working combinations:

  Power8:

   Pthreads backend

       -Wignored-qualifiers -Wempty-body -Wclobbered -Wuninitialized

Intel: -Wall -Wshadow -pedantic -Werror -Wsign-compare -Wtype-limits -Wuninitialized

Clang: -Wall -Wshadow -pedantic -Werror -Wsign-compare -Wtype-limits -Wuninitialized

NVCC: -Wall -Wshadow -pedantic -Werror -Wsign-compare -Wtype-limits -Wuninitialized

Secondary compilers are passing without -Werror.

Other compilers are tested occasionally, in particular when pushing from develop to 

master branch, without -Werror and only for a select set of backends.