summaryrefslogtreecommitdiff
path: root/var
diff options
context:
space:
mode:
authorChris White <white238@llnl.gov>2020-05-05 08:52:32 -0700
committerGitHub <noreply@github.com>2020-05-05 08:52:32 -0700
commit2cb4ae747d6dba8817f67ae98d78c39883cb2cff (patch)
tree7468062138384dd7ca2092962adb1c9e19fec416 /var
parent70c9dd37a7d63a5846ed373f6ff0c6df1ae7b2d5 (diff)
downloadspack-2cb4ae747d6dba8817f67ae98d78c39883cb2cff.tar.gz
spack-2cb4ae747d6dba8817f67ae98d78c39883cb2cff.tar.bz2
spack-2cb4ae747d6dba8817f67ae98d78c39883cb2cff.tar.xz
spack-2cb4ae747d6dba8817f67ae98d78c39883cb2cff.zip
Add Axom spack package (#15432)
* Add OpenMP variant to Umpire * Initial implementation of Axom package * Add Axom spack package and fix required dependencies * Fix overzealous tag replacement * Attempt to fix version error * Fix python version attempt #2 * Update raja and umpire * remove sys_type check * Address comments in axom package * Address Greg's comments * Fix flake8 * more flake8 * Simplify MPIEXEC and MPIEXEC_NUMPROC_FLAG * Fix typo * Revert back to slurm check, fix cuda_arch checks * Fix cuda_arch variant forwarding * Add cub variant * Add py-shroud * Address comments * Fix shroud path in axom * Fix merge conflict * Fix backwards if * Fix flake8 and add copyright * format for consistency
Diffstat (limited to 'var')
-rw-r--r--var/spack/repos/builtin/packages/axom/package.py515
-rw-r--r--var/spack/repos/builtin/packages/conduit/package.py58
-rw-r--r--var/spack/repos/builtin/packages/py-shroud/package.py23
-rw-r--r--var/spack/repos/builtin/packages/raja/package.py19
-rw-r--r--var/spack/repos/builtin/packages/umpire/package.py35
5 files changed, 613 insertions, 37 deletions
diff --git a/var/spack/repos/builtin/packages/axom/package.py b/var/spack/repos/builtin/packages/axom/package.py
new file mode 100644
index 0000000000..84eeac28b9
--- /dev/null
+++ b/var/spack/repos/builtin/packages/axom/package.py
@@ -0,0 +1,515 @@
+# Copyright 2013-2020 Lawrence Livermore National Security, LLC and other
+# Spack Project Developers. See the top-level COPYRIGHT file for details.
+#
+# SPDX-License-Identifier: (Apache-2.0 OR MIT)
+
+from spack import *
+
+import os
+import socket
+from os.path import join as pjoin
+
+import llnl.util.tty as tty
+
+
+def cmake_cache_entry(name, value, comment=""):
+ """Generate a string for a cmake cache variable"""
+ return 'set({0} "{1}" CACHE PATH "{2}")\n\n'.format(name, value, comment)
+
+
+def cmake_cache_option(name, boolean_value, comment=""):
+ """Generate a string for a cmake configuration option"""
+
+ value = "ON" if boolean_value else "OFF"
+ return 'set({0} {1} CACHE BOOL "{2}")\n\n'.format(name, value, comment)
+
+
+def get_spec_path(spec, package_name, path_replacements={}, use_bin=False):
+ """Extracts the prefix path for the given spack package
+ path_replacements is a dictionary with string replacements for the path.
+ """
+
+ if not use_bin:
+ path = spec[package_name].prefix
+ else:
+ path = spec[package_name].prefix.bin
+
+ path = os.path.realpath(path)
+
+ for key in path_replacements:
+ path = path.replace(key, path_replacements[key])
+
+ return path
+
+
+class Axom(CMakePackage, CudaPackage):
+ """Axom provides a robust, flexible software infrastructure for the development
+ of multi-physics applications and computational tools."""
+
+ maintainers = ['white238']
+
+ homepage = "https://github.com/LLNL/axom"
+ git = "https://github.com/LLNL/axom.git"
+
+ version('master', branch='master', submodules=True)
+ version('develop', branch='develop', submodules=True)
+ version('0.3.3', tag='v0.3.3', submodules="True")
+ version('0.3.2', tag='v0.3.2', submodules="True")
+ version('0.3.1', tag='v0.3.1', submodules="True")
+ version('0.3.0', tag='v0.3.0', submodules="True")
+ version('0.2.9', tag='v0.2.9', submodules="True")
+
+ phases = ["hostconfig", "cmake", "build", "install"]
+ root_cmakelists_dir = 'src'
+
+ # -----------------------------------------------------------------------
+ # Variants
+ # -----------------------------------------------------------------------
+ variant('debug', default=False,
+ description='Build debug instead of optimized version')
+
+ variant('fortran', default=True, description="Build with Fortran support")
+
+ variant("python", default=False, description="Build python support")
+
+ variant("mpi", default=True, description="Build MPI support")
+ variant('openmp', default=True, description='Turn on OpenMP support.')
+
+ variant("mfem", default=False, description="Build with mfem")
+ variant("hdf5", default=True, description="Build with hdf5")
+ variant("lua", default=False, description="Build with Lua")
+ variant("scr", default=False, description="Build with SCR")
+ variant("umpire", default=True, description="Build with umpire")
+
+ variant("raja", default=True, description="Build with raja")
+ variant("cub", default=True,
+ description="Build with RAJA's internal CUB support")
+
+ varmsg = "Build development tools (such as Sphinx, Uncrustify, etc...)"
+ variant("devtools", default=False, description=varmsg)
+
+ # -----------------------------------------------------------------------
+ # Dependencies
+ # -----------------------------------------------------------------------
+ # Basics
+ depends_on("cmake@3.8.2:", type='build')
+ depends_on("mpi", when="+mpi")
+
+ # Libraries
+ depends_on("conduit~shared+python", when="+python")
+ depends_on("conduit~shared~python", when="~python")
+ depends_on("conduit~shared+python+hdf5", when="+hdf5+python")
+ depends_on("conduit~shared+python~hdf5", when="~hdf5+python")
+ depends_on("conduit~shared~python+hdf5", when="+hdf5~python")
+ depends_on("conduit~shared~python~hdf5", when="~hdf5~python")
+
+ # HDF5 needs to be the same as Conduit's
+ depends_on("hdf5@1.8.19:1.8.999~mpi~cxx~shared~fortran", when="+hdf5")
+
+ depends_on("lua", when="+lua")
+
+ depends_on("scr", when="+scr")
+
+ depends_on("raja~openmp", when="+raja~openmp")
+ depends_on("raja+openmp", when="+raja+openmp")
+ depends_on("raja+cuda", when="+raja+cuda")
+
+ depends_on("umpire~openmp", when="+umpire~openmp")
+ depends_on("umpire+openmp", when="+umpire+openmp")
+ depends_on("umpire+cuda+deviceconst", when="+umpire+cuda")
+
+ for sm_ in CudaPackage.cuda_arch_values:
+ depends_on('raja cuda_arch={0}'.format(sm_),
+ when='+raja cuda_arch={0}'.format(sm_))
+ depends_on('umpire cuda_arch={0}'.format(sm_),
+ when='+umpire cuda_arch={0}'.format(sm_))
+
+ depends_on("mfem~mpi~hypre~metis~gzstream", when="+mfem")
+
+ depends_on("python", when="+python")
+
+ # Devtools
+ depends_on("cppcheck", when="+devtools")
+ depends_on("doxygen", when="+devtools")
+ depends_on("graphviz", when="+devtools")
+ depends_on("python", when="+devtools")
+ depends_on("py-sphinx", when="+devtools")
+ depends_on("py-shroud", when="+devtools")
+ depends_on("uncrustify@0.61", when="+devtools")
+
+ def _get_sys_type(self, spec):
+ sys_type = spec.architecture
+ # if on llnl systems, we can use the SYS_TYPE
+ if "SYS_TYPE" in env:
+ sys_type = env["SYS_TYPE"]
+ return sys_type
+
+ def _get_host_config_path(self, spec):
+ hostname = socket.gethostname()
+ if "SYS_TYPE" in env:
+ # Are we on a LLNL system then strip node number
+ hostname = hostname.rstrip('1234567890')
+ filename = "{0}-{1}-{2}.cmake".format(hostname,
+ self._get_sys_type(spec),
+ spec.compiler)
+ dest_dir = self.stage.source_path
+ fullpath = os.path.abspath(pjoin(dest_dir, filename))
+ return fullpath
+
+ def hostconfig(self, spec, prefix):
+ """
+ This method creates a 'host-config' file that specifies
+ all of the options used to configure and build Axom.
+ """
+
+ c_compiler = env["SPACK_CC"]
+ cpp_compiler = env["SPACK_CXX"]
+ f_compiler = None
+
+ # see if we should enable fortran support
+ if "SPACK_FC" in env.keys():
+ # even if this is set, it may not exist
+ # do one more sanity check
+ if os.path.isfile(env["SPACK_FC"]):
+ f_compiler = env["SPACK_FC"]
+
+ # cmake
+ if "+cmake" in spec:
+ cmake_exe = pjoin(spec['cmake'].prefix.bin, "cmake")
+ else:
+ cmake_exe = which("cmake")
+ if cmake_exe is None:
+ # error could not find cmake!
+ crash()
+ cmake_exe = cmake_exe.command
+ cmake_exe = os.path.realpath(cmake_exe)
+
+ host_config_path = self._get_host_config_path(spec)
+ cfg = open(host_config_path, "w")
+ cfg.write("#------------------{0}\n".format("-" * 60))
+ cfg.write("# !!!! This is a generated file, edit at own risk !!!!\n")
+ cfg.write("#------------------{0}\n".format("-" * 60))
+ cfg.write("# SYS_TYPE: {0}\n".format(self._get_sys_type(spec)))
+ cfg.write("# Compiler Spec: {0}\n".format(spec.compiler))
+ cfg.write("#------------------{0}\n".format("-" * 60))
+ # show path to cmake for reference and to be used by config-build.py
+ cfg.write("# CMake executable path: {0}\n".format(cmake_exe))
+ cfg.write("#------------------{0}\n\n".format("-" * 60))
+
+ # compiler settings
+ cfg.write("#------------------{0}\n".format("-" * 60))
+ cfg.write("# Compilers\n")
+ cfg.write("#------------------{0}\n\n".format("-" * 60))
+
+ cfg.write(cmake_cache_entry("CMAKE_C_COMPILER", c_compiler))
+ cfg.write(cmake_cache_entry("CMAKE_CXX_COMPILER", cpp_compiler))
+
+ if "+fortran" in spec or f_compiler is not None:
+ cfg.write(cmake_cache_option("ENABLE_FORTRAN", True))
+ cfg.write(cmake_cache_entry("CMAKE_Fortran_COMPILER", f_compiler))
+ else:
+ cfg.write(cmake_cache_option("ENABLE_FORTRAN", False))
+
+ # TPL locations
+ cfg.write("#------------------{0}\n".format("-" * 60))
+ cfg.write("# TPLs\n")
+ cfg.write("#------------------{0}\n\n".format("-" * 60))
+
+ # Try to find the common prefix of the TPL directory, including the
+ # compiler. If found, we will use this in the TPL paths
+ compiler_str = str(spec.compiler).replace('@', '-')
+ prefix_paths = prefix.split(compiler_str)
+ path_replacements = {}
+
+ if len(prefix_paths) == 2:
+ tpl_root = os.path.realpath(pjoin(prefix_paths[0], compiler_str))
+ path_replacements[tpl_root] = "${TPL_ROOT}"
+ cfg.write("# Root directory for generated TPLs\n")
+ cfg.write(cmake_cache_entry("TPL_ROOT", tpl_root))
+
+ conduit_dir = get_spec_path(spec, "conduit", path_replacements)
+ cfg.write(cmake_cache_entry("CONDUIT_DIR", conduit_dir))
+
+ # optional tpls
+
+ if "+mfem" in spec:
+ mfem_dir = get_spec_path(spec, "mfem", path_replacements)
+ cfg.write(cmake_cache_entry("MFEM_DIR", mfem_dir))
+ else:
+ cfg.write("# MFEM not built\n\n")
+
+ if "+hdf5" in spec:
+ hdf5_dir = get_spec_path(spec, "hdf5", path_replacements)
+ cfg.write(cmake_cache_entry("HDF5_DIR", hdf5_dir))
+ else:
+ cfg.write("# HDF5 not built\n\n")
+
+ if "+lua" in spec:
+ lua_dir = get_spec_path(spec, "lua", path_replacements)
+ cfg.write(cmake_cache_entry("LUA_DIR", lua_dir))
+ else:
+ cfg.write("# Lua not built\n\n")
+
+ if "+scr" in spec:
+ scr_dir = get_spec_path(spec, "scr", path_replacements)
+ cfg.write(cmake_cache_entry("SCR_DIR", scr_dir))
+ else:
+ cfg.write("# SCR not built\n\n")
+
+ if "+raja" in spec:
+ raja_dir = get_spec_path(spec, "raja", path_replacements)
+ cfg.write(cmake_cache_entry("RAJA_DIR", raja_dir))
+ else:
+ cfg.write("# RAJA not built\n\n")
+
+ if "+umpire" in spec:
+ umpire_dir = get_spec_path(spec, "umpire", path_replacements)
+ cfg.write(cmake_cache_entry("UMPIRE_DIR", umpire_dir))
+ else:
+ cfg.write("# Umpire not built\n\n")
+
+ cfg.write("#------------------{0}\n".format("-" * 60))
+ cfg.write("# MPI\n")
+ cfg.write("#------------------{0}\n\n".format("-" * 60))
+
+ if "+mpi" in spec:
+ cfg.write(cmake_cache_option("ENABLE_MPI", True))
+ cfg.write(cmake_cache_entry("MPI_C_COMPILER", spec['mpi'].mpicc))
+ cfg.write(cmake_cache_entry("MPI_CXX_COMPILER",
+ spec['mpi'].mpicxx))
+ if "+fortran" in spec or f_compiler is not None:
+ cfg.write(cmake_cache_entry("MPI_Fortran_COMPILER",
+ spec['mpi'].mpifc))
+
+ # Check for slurm
+ using_slurm = False
+ slurm_checks = ['+slurm',
+ 'schedulers=slurm',
+ 'process_managers=slurm']
+ if any(spec['mpi'].satisfies(variant) for variant in slurm_checks):
+ using_slurm = True
+
+ # Determine MPIEXEC
+ if using_slurm:
+ if spec['mpi'].external:
+ mpiexec = '/usr/bin/srun'
+ else:
+ mpiexec = os.path.join(spec['slurm'].prefix.bin, 'srun')
+ else:
+ mpiexec = os.path.join(spec['mpi'].prefix.bin, 'mpirun')
+ if not os.path.exists(mpiexec):
+ mpiexec = os.path.join(spec['mpi'].prefix.bin, 'mpiexec')
+
+ if not os.path.exists(mpiexec):
+ msg = "Unable to determine MPIEXEC, Axom tests may fail"
+ cfg.write("# {0}\n\n".format(msg))
+ tty.msg(msg)
+ else:
+ # starting with cmake 3.10, FindMPI expects MPIEXEC_EXECUTABLE
+ # vs the older versions which expect MPIEXEC
+ if self.spec["cmake"].satisfies('@3.10:'):
+ cfg.write(cmake_cache_entry("MPIEXEC_EXECUTABLE", mpiexec))
+ else:
+ cfg.write(cmake_cache_entry("MPIEXEC", mpiexec))
+
+ # Determine MPIEXEC_NUMPROC_FLAG
+ if using_slurm:
+ cfg.write(cmake_cache_entry("MPIEXEC_NUMPROC_FLAG", "-n"))
+ else:
+ cfg.write(cmake_cache_entry("MPIEXEC_NUMPROC_FLAG", "-np"))
+
+ if spec['mpi'].name == 'spectrum-mpi':
+ cfg.write(cmake_cache_entry("BLT_MPI_COMMAND_APPEND",
+ "mpibind"))
+ else:
+ cfg.write(cmake_cache_option("ENABLE_MPI", False))
+
+ ##################################
+ # Devtools
+ ##################################
+
+ cfg.write("#------------------{0}\n".format("-" * 60))
+ cfg.write("# Devtools\n")
+ cfg.write("#------------------{0}\n\n".format("-" * 60))
+
+ # Add common prefix to path replacement list
+ if "+devtools" in spec:
+ # Grab common devtools root and strip the trailing slash
+ path1 = os.path.realpath(spec["uncrustify"].prefix)
+ path2 = os.path.realpath(spec["doxygen"].prefix)
+ devtools_root = os.path.commonprefix([path1, path2])[:-1]
+ path_replacements[devtools_root] = "${DEVTOOLS_ROOT}"
+ cfg.write("# Root directory for generated developer tools\n")
+ cfg.write(cmake_cache_entry("DEVTOOLS_ROOT", devtools_root))
+
+ if "+python" in spec or "+devtools" in spec:
+ python_path = os.path.realpath(spec['python'].command.path)
+ for key in path_replacements:
+ python_path = python_path.replace(key, path_replacements[key])
+ cfg.write(cmake_cache_entry("PYTHON_EXECUTABLE", python_path))
+
+ if "doxygen" in spec or "py-sphinx" in spec:
+ cfg.write(cmake_cache_option("ENABLE_DOCS", True))
+
+ if "doxygen" in spec:
+ doxygen_bin_dir = get_spec_path(spec, "doxygen",
+ path_replacements,
+ use_bin=True)
+ cfg.write(cmake_cache_entry("DOXYGEN_EXECUTABLE",
+ pjoin(doxygen_bin_dir,
+ "doxygen")))
+
+ if "py-sphinx" in spec:
+ python_bin_dir = get_spec_path(spec, "python",
+ path_replacements,
+ use_bin=True)
+ cfg.write(cmake_cache_entry("SPHINX_EXECUTABLE",
+ pjoin(python_bin_dir,
+ "sphinx-build")))
+ else:
+ cfg.write(cmake_cache_option("ENABLE_DOCS", False))
+
+ if "py-shroud" in spec:
+ shroud_bin_dir = get_spec_path(spec, "py-shroud",
+ path_replacements, use_bin=True)
+ cfg.write(cmake_cache_entry("SHROUD_EXECUTABLE",
+ pjoin(shroud_bin_dir, "shroud")))
+
+ if "uncrustify" in spec:
+ uncrustify_bin_dir = get_spec_path(spec, "uncrustify",
+ path_replacements,
+ use_bin=True)
+ cfg.write(cmake_cache_entry("UNCRUSTIFY_EXECUTABLE",
+ pjoin(uncrustify_bin_dir,
+ "uncrustify")))
+
+ if "cppcheck" in spec:
+ cppcheck_bin_dir = get_spec_path(spec, "cppcheck",
+ path_replacements, use_bin=True)
+ cfg.write(cmake_cache_entry("CPPCHECK_EXECUTABLE",
+ pjoin(cppcheck_bin_dir, "cppcheck")))
+
+ ##################################
+ # Other machine specifics
+ ##################################
+
+ cfg.write("#------------------{0}\n".format("-" * 60))
+ cfg.write("# Other machine specifics\n")
+ cfg.write("#------------------{0}\n\n".format("-" * 60))
+
+ # OpenMP
+ if "+openmp" in spec:
+ cfg.write(cmake_cache_option("ENABLE_OPENMP", True))
+ else:
+ cfg.write(cmake_cache_option("ENABLE_OPENMP", False))
+
+ # Enable death tests
+ if spec.satisfies('target=ppc64le:') and "+cuda" in spec:
+ cfg.write(cmake_cache_option("ENABLE_GTEST_DEATH_TESTS", False))
+ else:
+ cfg.write(cmake_cache_option("ENABLE_GTEST_DEATH_TESTS", True))
+
+ # Override XL compiler family
+ familymsg = ("Override to proper compiler family for XL")
+ if "xlf" in f_compiler:
+ cfg.write(cmake_cache_entry("CMAKE_Fortran_COMPILER_ID", "XL",
+ familymsg))
+ if "xlc" in c_compiler:
+ cfg.write(cmake_cache_entry("CMAKE_C_COMPILER_ID", "XL",
+ familymsg))
+ if "xlC" in cpp_compiler:
+ cfg.write(cmake_cache_entry("CMAKE_CXX_COMPILER_ID", "XL",
+ familymsg))
+
+ if spec.satisfies('target=ppc64le:'):
+ if "xlf" in f_compiler:
+ description = ("Converts C-style comments to Fortran style "
+ "in preprocessed files")
+ cfg.write(cmake_cache_entry("BLT_FORTRAN_FLAGS",
+ "-WF,-C! -qxlf2003=polymorphic",
+ description))
+ # Grab lib directory for the current fortran compiler
+ libdir = os.path.join(os.path.dirname(
+ os.path.dirname(f_compiler)), "lib")
+ description = ("Adds a missing rpath for libraries "
+ "associated with the fortran compiler")
+ cfg.write(cmake_cache_entry("BLT_EXE_LINKER_FLAGS",
+ "-Wl,-rpath," + libdir,
+ description))
+
+ if "+cuda" in spec:
+ cfg.write("#------------------{0}\n".format("-" * 60))
+ cfg.write("# Cuda\n")
+ cfg.write("#------------------{0}\n\n".format("-" * 60))
+
+ cfg.write(cmake_cache_option("ENABLE_CUDA", True))
+
+ cudatoolkitdir = spec['cuda'].prefix
+ cfg.write(cmake_cache_entry("CUDA_TOOLKIT_ROOT_DIR",
+ cudatoolkitdir))
+ cudacompiler = "${CUDA_TOOLKIT_ROOT_DIR}/bin/nvcc"
+ cfg.write(cmake_cache_entry("CMAKE_CUDA_COMPILER",
+ cudacompiler))
+
+ cfg.write(cmake_cache_option("CUDA_SEPARABLE_COMPILATION",
+ True))
+
+ cfg.write(cmake_cache_option("AXOM_ENABLE_ANNOTATIONS", True))
+
+ if "+cub" in spec:
+ cfg.write(cmake_cache_option("AXOM_ENABLE_CUB", True))
+ else:
+ cfg.write(cmake_cache_option("AXOM_ENABLE_CUB", False))
+
+ # CUDA_FLAGS
+ cudaflags = "-restrict "
+
+ if not spec.satisfies('cuda_arch=none'):
+ cuda_arch = spec.variants['cuda_arch'].value
+ axom_arch = 'sm_{0}'.format(cuda_arch[0])
+ cfg.write(cmake_cache_entry("AXOM_CUDA_ARCH", axom_arch))
+ cudaflags += "-arch ${AXOM_CUDA_ARCH} "
+ else:
+ cfg.write("# cuda_arch could not be determined\n\n")
+
+ cudaflags += "-std=c++11 --expt-extended-lambda -G "
+ cfg.write(cmake_cache_entry("CMAKE_CUDA_FLAGS", cudaflags))
+
+ if "+mpi" in spec:
+ cfg.write(cmake_cache_entry("CMAKE_CUDA_HOST_COMPILER",
+ "${MPI_CXX_COMPILER}"))
+ else:
+ cfg.write(cmake_cache_entry("CMAKE_CUDA_HOST_COMPILER",
+ "${CMAKE_CXX_COMPILER}"))
+
+ cfg.write("# nvcc does not like gtest's 'pthreads' flag\n")
+ cfg.write(cmake_cache_option("gtest_disable_pthreads", True))
+
+ if ("gfortran" in f_compiler) and ("clang" in cpp_compiler):
+ clanglibdir = pjoin(os.path.dirname(
+ os.path.dirname(cpp_compiler)), "lib")
+ flags = "-Wl,-rpath,{0}".format(clanglibdir)
+ description = ("Adds a missing rpath for libraries "
+ "associated with the fortran compiler")
+ cfg.write(cmake_cache_entry("BLT_EXE_LINKER_FLAGS", flags,
+ description))
+
+ cfg.write("\n")
+ cfg.close()
+ tty.info("Spack generated Axom host-config file: " + host_config_path)
+
+ def cmake_args(self):
+ spec = self.spec
+ host_config_path = self._get_host_config_path(spec)
+
+ options = []
+ options.extend(['-C', host_config_path])
+ if self.run_tests is False:
+ options.append('-DENABLE_TESTS=OFF')
+ else:
+ options.append('-DENABLE_TESTS=ON')
+ return options
+
+ @run_after('install')
+ def install_cmake_cache(self):
+ install(self._get_host_config_path(spec), prefix)
diff --git a/var/spack/repos/builtin/packages/conduit/package.py b/var/spack/repos/builtin/packages/conduit/package.py
index ed1bd4f3d3..2e6ec09409 100644
--- a/var/spack/repos/builtin/packages/conduit/package.py
+++ b/var/spack/repos/builtin/packages/conduit/package.py
@@ -285,8 +285,10 @@ class Conduit(Package):
f_compiler = None
if self.compiler.fc:
- # even if this is set, it may not exist so do one more sanity check
- f_compiler = which(env["SPACK_FC"])
+ # even if this is set, it may not exist
+ # do one more sanity check
+ if os.path.isfile(env["SPACK_FC"]):
+ f_compiler = env["SPACK_FC"]
#######################################################################
# By directly fetching the names of the actual compilers we appear
@@ -299,6 +301,9 @@ class Conduit(Package):
if "SYS_TYPE" in env:
sys_type = env["SYS_TYPE"]
+ # are we on a specific machine
+ on_blueos = 'blueos' in sys_type
+
##############################################
# Find and record what CMake is used
##############################################
@@ -335,7 +340,7 @@ class Conduit(Package):
if "+fortran" in spec and f_compiler is not None:
cfg.write(cmake_cache_entry("ENABLE_FORTRAN", "ON"))
cfg.write(cmake_cache_entry("CMAKE_Fortran_COMPILER",
- f_compiler.path))
+ f_compiler))
else:
cfg.write("# no fortran compiler found\n\n")
cfg.write(cmake_cache_entry("ENABLE_FORTRAN", "OFF"))
@@ -345,16 +350,6 @@ class Conduit(Package):
else:
cfg.write(cmake_cache_entry("BUILD_SHARED_LIBS", "OFF"))
- # extra fun for blueos
- if 'blueos_3' in sys_type and "+fortran" in spec:
- if 'xl@coral' in os.getenv('SPACK_COMPILER_SPEC', ""):
- # Fix missing std linker flag in xlc compiler
- cfg.write(cmake_cache_entry("BLT_FORTRAN_FLAGS",
- "-WF,-C! -qxlf2003=polymorphic"))
- # Conduit can't link C++ into fortran for this spec, but works
- # fine in host code
- cfg.write(cmake_cache_entry("ENABLE_TESTS", "OFF"))
-
#######################
# Unit Tests
#######################
@@ -363,6 +358,32 @@ class Conduit(Package):
else:
cfg.write(cmake_cache_entry("ENABLE_TESTS", "OFF"))
+ # extra fun for blueos
+ if on_blueos:
+ # All of BlueOS compilers report clang due to nvcc,
+ # override to proper compiler family
+ if "xlc" in c_compiler:
+ cfg.write(cmake_cache_entry("CMAKE_C_COMPILER_ID", "XL"))
+ if "xlC" in cpp_compiler:
+ cfg.write(cmake_cache_entry("CMAKE_CXX_COMPILER_ID", "XL"))
+
+ if "+fortran" in spec:
+ if "xlf" in f_compiler:
+ cfg.write(cmake_cache_entry("CMAKE_Fortran_COMPILER_ID",
+ "XL"))
+
+ if 'xl@coral' in os.getenv('SPACK_COMPILER_SPEC', ""):
+ # Fix missing std linker flag in xlc compiler
+ flags = "-WF,-C! -qxlf2003=polymorphic"
+ cfg.write(cmake_cache_entry("BLT_FORTRAN_FLAGS",
+ flags))
+ # Grab lib directory for the current fortran compiler
+ libdir = os.path.join(os.path.dirname(
+ os.path.dirname(f_compiler)), "lib")
+ flags = "-lstdc++ -Wl,-rpath," + libdir
+ cfg.write(cmake_cache_entry("BLT_EXE_LINKER_FLAGS",
+ flags))
+
#######################
# Python
#######################
@@ -419,7 +440,10 @@ class Conduit(Package):
cfg.write(cmake_cache_entry("ENABLE_MPI", "ON"))
cfg.write(cmake_cache_entry("MPI_C_COMPILER", mpicc_path))
cfg.write(cmake_cache_entry("MPI_CXX_COMPILER", mpicxx_path))
- cfg.write(cmake_cache_entry("MPI_Fortran_COMPILER", mpifc_path))
+ if "+fortran" in spec:
+ cfg.write(cmake_cache_entry("MPI_Fortran_COMPILER",
+ mpifc_path))
+
mpiexe_bin = join_path(spec['mpi'].prefix.bin, 'mpiexec')
if os.path.isfile(mpiexe_bin):
# starting with cmake 3.10, FindMPI expects MPIEXEC_EXECUTABLE
@@ -456,12 +480,6 @@ class Conduit(Package):
if "+hdf5" in spec:
cfg.write(cmake_cache_entry("HDF5_DIR", spec['hdf5'].prefix))
- # extra fun for BG/Q
- if 'bgqos_0' in sys_type:
- cfg.write(cmake_cache_entry('HDF5_C_LIBRARY_m',
- '-lm', 'STRING'))
- cfg.write(cmake_cache_entry('HDF5_C_LIBRARY_dl',
- '-ldl', 'STRING'))
else:
cfg.write("# hdf5 not built by spack \n")
diff --git a/var/spack/repos/builtin/packages/py-shroud/package.py b/var/spack/repos/builtin/packages/py-shroud/package.py
new file mode 100644
index 0000000000..504c992f15
--- /dev/null
+++ b/var/spack/repos/builtin/packages/py-shroud/package.py
@@ -0,0 +1,23 @@
+# Copyright 2013-2020 Lawrence Livermore National Security, LLC and other
+# Spack Project Developers. See the top-level COPYRIGHT file for details.
+#
+# SPDX-License-Identifier: (Apache-2.0 OR MIT)
+
+from spack import *
+
+
+class PyShroud(PythonPackage):
+ """Create Fortran wrappers for a C++ library."""
+
+ homepage = "https://github.com/LLNL/shroud"
+ git = "https://github.com/LLNL/shroud.git"
+
+ version('develop', branch='develop')
+ version('master', branch='master')
+ version('0.11.0', tag='v0.11.0')
+ version('0.10.1', tag='v0.10.1')
+ version('0.9.0', tag='v0.9.0')
+ version('0.8.0', tag='v0.8.0')
+
+ depends_on("py-setuptools", type='build')
+ depends_on("py-pyyaml@4.2b1:", type=('build', 'run'))
diff --git a/var/spack/repos/builtin/packages/raja/package.py b/var/spack/repos/builtin/packages/raja/package.py
index 10a9265257..de9652dbc1 100644
--- a/var/spack/repos/builtin/packages/raja/package.py
+++ b/var/spack/repos/builtin/packages/raja/package.py
@@ -15,6 +15,7 @@ class Raja(CMakePackage, CudaPackage):
version('develop', branch='develop', submodules='True')
version('master', branch='master', submodules='True')
version('0.11.0', tag='v0.11.0', submodules="True")
+ version('0.10.1', tag='v0.10.1', submodules="True")
version('0.10.0', tag='v0.10.0', submodules="True")
version('0.9.0', tag='v0.9.0', submodules="True")
version('0.8.0', tag='v0.8.0', submodules="True")
@@ -27,11 +28,8 @@ class Raja(CMakePackage, CudaPackage):
version('0.4.1', tag='v0.4.1', submodules="True")
version('0.4.0', tag='v0.4.0', submodules="True")
- variant('cuda', default=False, description='Build with CUDA backend')
variant('openmp', default=True, description='Build OpenMP backend')
- depends_on('cuda', when='+cuda')
-
depends_on('cmake@3.8:', type='build')
depends_on('cmake@3.9:', when='+cuda', type='build')
@@ -47,9 +45,16 @@ class Raja(CMakePackage, CudaPackage):
'-DENABLE_CUDA=On',
'-DCUDA_TOOLKIT_ROOT_DIR=%s' % (spec['cuda'].prefix)])
- cuda_value = spec.variants['cuda_arch'].value
- cuda_arch = cuda_value[0]
- if cuda_arch is not None:
- options.append('-DCUDA_ARCH=sm_{0}'.format(cuda_arch))
+ if not spec.satisfies('cuda_arch=none'):
+ cuda_arch = spec.variants['cuda_arch'].value
+ options.append('-DCUDA_ARCH=sm_{0}'.format(cuda_arch[0]))
+
+ # Work around spack adding -march=ppc64le to SPACK_TARGET_ARGS which
+ # is used by the spack compiler wrapper. This can go away when BLT
+ # removes -Werror from GTest flags
+ if self.spec.satisfies('%clang target=ppc64le:') or not self.run_tests:
+ options.append('-DENABLE_TESTS=OFF')
+ else:
+ options.append('-DENABLE_TESTS=ON')
return options
diff --git a/var/spack/repos/builtin/packages/umpire/package.py b/var/spack/repos/builtin/packages/umpire/package.py
index 90ea23dde3..fb6b779d1b 100644
--- a/var/spack/repos/builtin/packages/umpire/package.py
+++ b/var/spack/repos/builtin/packages/umpire/package.py
@@ -7,7 +7,7 @@
from spack import *
-class Umpire(CMakePackage):
+class Umpire(CMakePackage, CudaPackage):
"""An application-focused API for memory management on NUMA & GPU
architectures"""
@@ -35,12 +35,13 @@ class Umpire(CMakePackage):
version('0.1.4', tag='v0.1.4', submodules='True')
version('0.1.3', tag='v0.1.3', submodules='True')
- variant('cuda', default=False, description='Build with CUDA support')
- variant('fortran', default=False, description='Build Fortran API')
+ variant('fortran', default=False, description='Build C/Fortran API')
variant('c', default=True, description='Build C API')
variant('numa', default=False, description='Enable NUMA support')
+ variant('openmp', default=False, description='Build with OpenMP support')
+ variant('deviceconst', default=False,
+ description='Enables support for constant device memory')
- depends_on('cuda', when='+cuda')
depends_on('cmake@3.8:', type='build')
depends_on('cmake@3.9:', when='+cuda', type='build')
@@ -56,16 +57,30 @@ class Umpire(CMakePackage):
options.extend([
'-DENABLE_CUDA=On',
'-DCUDA_TOOLKIT_ROOT_DIR=%s' % (spec['cuda'].prefix)])
+
+ if not spec.satisfies('cuda_arch=none'):
+ cuda_arch = spec.variants['cuda_arch'].value
+ flag = '-arch sm_{0}'.format(cuda_arch[0])
+ options.append('-DCMAKE_CUDA_FLAGS:STRING={0}'.format(flag))
+
+ if '+deviceconst' in spec:
+ options.append('-DENABLE_DEVICE_CONST=On')
else:
options.append('-DENABLE_CUDA=Off')
- if '+c' in spec:
- options.append('-DENABLE_C=On')
+ options.append('-DENABLE_C={0}'.format(
+ 'On' if '+c' in spec else 'Off'))
+
+ options.append('-DENABLE_FORTRAN={0}'.format(
+ 'On' if '+fortran' in spec else 'Off'))
+
+ options.append('-DENABLE_NUMA={0}'.format(
+ 'On' if '+numa' in spec else 'Off'))
- if '+fortran' in spec:
- options.append('-DENABLE_FORTRAN=On')
+ options.append('-DENABLE_OPENMP={0}'.format(
+ 'On' if '+openmp' in spec else 'Off'))
- if '+numa' in spec:
- options.append('-DENABLE_NUMA=On')
+ options.append('-DENABLE_TESTS={0}'.format(
+ 'On' if self.run_tests else 'Off'))
return options