[WIP] CEED 2.0 (#10903)

* Initial commit for v2.0 of the CEED software suite.

* Update Nek packages and gslib

* Help spack concretize the hypre version for ceed-2.0.

* Fix nekcem install error

* Add support for gfortran v8 in nek5000 and nekcem.

* Split Nek5000 into Nek5000 and Nektools

* Get Nektools to build fine in Theta

* Fix travis failure: remove unused 'import numbers' from nek5000.

* Check for gfortran if it is wrapped

* Tweak the detection of gfortran in nek5000.

* Fix Nek packages to add -std=legacy when FC=gcc

* spack install ceed~petsc works fine on Theta

* Fix flake8 errors

* Fix more flake8 tests

* Fix an import issue

* Tweak the suite-sparse package to avoid interaction with existing system
installations of suite-sparse.

* petsc: update superlu-dist dependency

* Updates in the packages: occa, libceed, and ceed.

* In the libceed package, explicitly tell nvcc which host compiler to use.

* Fix python formatting.

* Simplify the test for gfortran in nek* packages.

* ceed: 2.0 uses petsc@3.11.0

* hpgmg-0.4; use from ceed@2.0.0

* Update the hypre dependency for ceed 2.0.

* Disable the superlu-dist dependency (through hypre) when using a
+quickbuild of ceed 2.0.

* petsc-3.11.0: add xlf fix

* nekcem: has a build dependency on Python 2.7+

* hpgmg: better setting of compiler options and use python for configure

* libceed: use v0.4 tag

* libceed: fix 0.4 release oops (pkgconfig version)

* Add a patch for magma-2.5.0 that brings it up the current 'master'.

* In the mfem package, install the examples, miniapps, and data under
$prefix/share/mfem.

* In the magma package, apply a patch to v2.5.0 that disables
magma_sparse - for testing purposes.

* In the magma package, link the 'magma' library with the
'nvToolsExt' library.

* In the magma package, update the 'magma-2.5.0.patch' with the latest
commits from the magma source repository. Also, remove the library
'nvToolsExt' from the 'magma-2.5.0-cmake.patch' - now it is not
needed.

* In the magma package, disable OpenMP when using v2.5.0 with the
IBM XL compiler.

 Please enter the commit message for your changes. Lines starting

* In the mfem package, add version for the 'laghos-v2.0' tag; also,
prefix the versions `laghos-v*` with their respective development
version numbers -- this way they are properly ordered within spack
relative to the official numbered versions.

* petsc: add version 3.11.1 (#11179)


(cherry picked from commit 1eab6e3c865836d209db4621d0cd44bc8663656b)

* ceed-2.0: use petsc-3.11.1

* this-is-so-dumb.f -> empty.f

18 files changed, 926 insertions, 144 deletions

diff --git a/var/spack/repos/builtin/packages/ceed/package.py b/var/spack/repos/builtin/packages/ceed/package.py
index 4ed5925da1..b066da4446 100644
--- a/var/spack/repos/builtin/packages/ceed/package.py
+++ b/var/spack/repos/builtin/packages/ceed/package.py
@@ -20,6 +20,7 @@ class Ceed(Package):
     url  = 'file://' + os.path.dirname(__file__) + '/README.md'
     sha1 = 'b2eefd95c09ba573f663a761b84811a2d7e39788'
 
+    version('2.0.0', sha1, expand=False)
     version('1.0.0', sha1, expand=False)
 
     variant('cuda', default=False,
@@ -38,22 +39,53 @@ class Ceed(Package):
     # TODO: Add 'int64' variant?
 
     # LibCEED
+    # ceed-2.0
+    depends_on('libceed@0.4~cuda', when='@2.0.0~cuda')
+    depends_on('libceed@0.4+cuda', when='@2.0.0+cuda')
+    depends_on('libceed@0.4+occa', when='@2.0.0+occa')
+    depends_on('libceed@0.4~occa', when='@2.0.0~occa')
+    # ceed-1.0
     depends_on('libceed@0.2~cuda', when='@1.0.0~cuda')
     depends_on('libceed@0.2+cuda', when='@1.0.0+cuda')
     depends_on('libceed@0.2+occa', when='@1.0.0+occa')
     depends_on('libceed@0.2~occa', when='@1.0.0~occa')
 
     # OCCA
-    depends_on('occa@v1.0.0-alpha.5~cuda', when='@1.0.0+occa~cuda')
-    depends_on('occa@v1.0.0-alpha.5+cuda', when='@1.0.0+occa+cuda')
+    # ceed-2.0
+    depends_on('occa@1.0.8~cuda', when='@2.0.0+occa~cuda')
+    depends_on('occa@1.0.8+cuda', when='@2.0.0+occa+cuda')
+    # ceed-1.0
+    depends_on('occa@1.0.0-alpha.5~cuda', when='@1.0.0+occa~cuda')
+    depends_on('occa@1.0.0-alpha.5+cuda', when='@1.0.0+occa+cuda')
 
     # Nek5000, GSLIB, Nekbone, and NekCEM
+    # ceed-2.0
+    depends_on('nek5000@17.0', when='@2.0.0+nek')
+    depends_on('nektools@17.0%gcc', when='@2.0.0+nek')
+    depends_on('gslib@1.0.2', when='@2.0.0+nek')
+    depends_on('nekbone@17.0', when='@2.0.0+nek')
+    depends_on('nekcem@7332619', when='@2.0.0+nek')
+    # ceed-1.0
     depends_on('nek5000@17.0', when='@1.0.0+nek')
+    depends_on('nektools@17.0%gcc', when='@1.0.0+nek')
     depends_on('gslib@1.0.2', when='@1.0.0+nek')
     depends_on('nekbone@17.0', when='@1.0.0+nek')
     depends_on('nekcem@0b8bedd', when='@1.0.0+nek')
 
     # PETSc, HPGMG
+    # ceed-2.0
+    # For a +quickbuild we disable hdf5, and superlu-dist in PETSc.
+    # Ideally, these can be turned into recommendations to Spack for
+    # concretizing the PETSc spec, if Spack ever supports recommendations.
+    depends_on('petsc@3.11.1~hdf5~superlu-dist',
+               when='@2.0.0+petsc+quickbuild')
+    depends_on('petsc@3.11.1+mpi+double~int64', when='@2.0.0+petsc~mfem')
+    # The mfem petsc examples need the petsc variants +hypre, +suite-sparse,
+    # and +mumps:
+    depends_on('petsc@3.11.1+mpi+hypre+suite-sparse+mumps+double~int64',
+               when='@2.0.0+petsc+mfem')
+    depends_on('hpgmg@0.4+fe', when='@2.0.0+petsc')
+    # ceed-1.0
     # For a +quickbuild we disable hdf5, and superlu-dist in PETSc.
     # Ideally, these can be turned into recommendations to Spack for
     # concretizing the PETSc spec, if Spack ever supports recommendations.
@@ -67,12 +99,33 @@ class Ceed(Package):
     depends_on('hpgmg@a0a5510df23b+fe', when='@1.0.0+petsc')
 
     # MAGMA
+    # ceed-2.0
+    depends_on('magma@2.5.0', when='@2.0.0+cuda')
+    # ceed-1.0
     depends_on('magma@2.3.0', when='@1.0.0+cuda')
 
     # PUMI
+    # ceed-2.0
+    depends_on('pumi@2.2.0', when='@2.0.0+pumi')
+    # ceed-1.0
     depends_on('pumi@2.1.0', when='@1.0.0+pumi')
 
     # MFEM, Laghos
+    # ceed-2.0
+    depends_on('mfem@3.4.0+mpi+examples+miniapps', when='@2.0.0+mfem~petsc')
+    depends_on('mfem@3.4.0+mpi+petsc+examples+miniapps',
+               when='@2.0.0+mfem+petsc')
+    depends_on('mfem@3.4.0+pumi', when='@2.0.0+mfem+pumi')
+    depends_on('laghos@2.0', when='@2.0.0+mfem')
+    # Help the spack concretizer find a suitable version of hypre:
+    depends_on('hypre~internal-superlu', when='@2.0.0+mfem')
+    depends_on('hypre~internal-superlu~superlu-dist',
+               when='@2.0.0+mfem+quickbuild')
+
+    # If using gcc version <= 4.8 build suite-sparse version <= 5.1.0
+    depends_on('suite-sparse@:5.1.0', when='@2.0.0%gcc@:4.8+mfem+petsc')
+
+    # ceed-1.0
     depends_on('mfem@3.3.2+mpi+examples+miniapps', when='@1.0.0+mfem~petsc')
     depends_on('mfem@3.3.2+mpi+petsc+examples+miniapps',
                when='@1.0.0+mfem+petsc')
diff --git a/var/spack/repos/builtin/packages/gslib/package.py b/var/spack/repos/builtin/packages/gslib/package.py
index 611f745895..59a077b628 100644
--- a/var/spack/repos/builtin/packages/gslib/package.py
+++ b/var/spack/repos/builtin/packages/gslib/package.py
@@ -12,6 +12,7 @@ class Gslib(Package):
     homepage = "https://github.com/gslib/gslib"
     git      = "https://github.com/gslib/gslib.git"
 
+    version('1.0.3', tag='v1.0.3')
     version('1.0.2', tag='v1.0.2')
     version('1.0.1', tag='v1.0.1')
     version('1.0.0', tag='v1.0.0')
diff --git a/var/spack/repos/builtin/packages/hpgmg/package.py b/var/spack/repos/builtin/packages/hpgmg/package.py
index 5a649e2289..b9c33c118a 100644
--- a/var/spack/repos/builtin/packages/hpgmg/package.py
+++ b/var/spack/repos/builtin/packages/hpgmg/package.py
@@ -23,6 +23,7 @@ class Hpgmg(Package):
     tags = ['proxy-app']
 
     version('develop', branch='master')
+    version('0.4',          'abdabfe09453487299500b5bd8da4e6dc3d88477199bcfa38ac41d0b3c780f6f')
     version('a0a5510df23b', 'b9c50f25e541428d4735fb07344d1d0ed9fc821bdde918d8e0defa78c0d9b4f9')
     version('0.3',          '12a65da216fec91daea78594ae4b5a069c8f1a700f1ba21eed9f45a79a68c793')
 
@@ -64,7 +65,14 @@ class Hpgmg(Package):
         if '+debug' in self.spec:
             cflags.append('-g')
         elif any(map(self.spec.satisfies, ['%gcc', '%clang', '%intel'])):
-            cflags += ['-O3', '-march=native']
+            cflags.append('-O3')
+            if self.compiler.target in ['x86_64']:
+                cflags.append('-march=native')
+            else:
+                cflags.append('-mcpu=native')
+                cflags.append('-mtune=native')
+        else:
+            cflags.append('-O3')
 
         args.append('--CFLAGS=' + ' '.join(cflags))
 
@@ -75,7 +83,7 @@ class Hpgmg(Package):
         return args
 
     def configure(self, spec, prefix):
-        configure(*self.configure_args())
+        python('configure', *self.configure_args())
 
     def build(self, spec, prefix):
         make('-C', 'build')
diff --git a/var/spack/repos/builtin/packages/laghos/package.py b/var/spack/repos/builtin/packages/laghos/package.py
index 6e2a0f60f0..953886c0da 100644
--- a/var/spack/repos/builtin/packages/laghos/package.py
+++ b/var/spack/repos/builtin/packages/laghos/package.py
@@ -28,11 +28,13 @@ class Laghos(MakefilePackage):
     depends_on('mfem@develop+mpi+metis', when='@develop+metis')
     depends_on('mfem@develop+mpi~metis', when='@develop~metis')
 
-    depends_on('mfem@laghos-v2.0,3.4.0:+mpi+metis', when='@2.0:+metis')
-    depends_on('mfem@laghos-v2.0,3.4.0:+mpi~metis', when='@2.0:~metis')
+    # Recommended mfem version for laghos v2.0 is: ^mfem@3.4.1-laghos-v2.0
+    depends_on('mfem@3.4.0:+mpi+metis', when='@2.0+metis')
+    depends_on('mfem@3.4.0:+mpi~metis', when='@2.0~metis')
 
-    depends_on('mfem@laghos-v1.0,3.3.2:+mpi+metis', when='@1.0:+metis')
-    depends_on('mfem@laghos-v1.0,3.3.2:+mpi~metis', when='@1.0:~metis')
+    # Recommended mfem version for laghos v1.x is: ^mfem@3.3.1-laghos-v1.0
+    depends_on('mfem@3.3.1-laghos-v1.0:+mpi+metis', when='@1.0,1.1+metis')
+    depends_on('mfem@3.3.1-laghos-v1.0:+mpi~metis', when='@1.0,1.1~metis')
 
     @property
     def build_targets(self):
diff --git a/var/spack/repos/builtin/packages/libceed/package.py b/var/spack/repos/builtin/packages/libceed/package.py
index 578da42181..ecce991e8f 100644
--- a/var/spack/repos/builtin/packages/libceed/package.py
+++ b/var/spack/repos/builtin/packages/libceed/package.py
@@ -13,18 +13,29 @@ class Libceed(Package):
     git      = "https://github.com/CEED/libCEED.git"
 
     version('develop', branch='master')
+    version('0.4', tag='v0.4')
     version('0.2', tag='v0.2')
     version('0.1', tag='v0.1')
 
     variant('occa', default=True, description='Enable OCCA backends')
     variant('cuda', default=False, description='Enable CUDA support')
     variant('debug', default=False, description='Enable debug build')
+    variant('libxsmm', default=False, description='Enable LIBXSMM backend')
+
+    conflicts('+libxsmm', when='@:0.2')
+
+    depends_on('cuda', when='+cuda')
 
-    depends_on('occa@v1.0.0-alpha.5,develop', when='+occa')
     depends_on('occa@develop', when='@develop+occa')
+    depends_on('occa@1.0.8:', when='@0.4+occa')
+    depends_on('occa@1.0.0-alpha.5,develop', when='@:0.2+occa')
     depends_on('occa+cuda', when='+occa+cuda')
     depends_on('occa~cuda', when='+occa~cuda')
 
+    depends_on('libxsmm', when='+libxsmm')
+
+    patch('pkgconfig-version-0.4.diff', when='@0.4')
+
     # occa: do not occaFree kernels
     # Repeated creation and freeing of kernels appears to expose a caching
     # bug in Occa.
@@ -32,18 +43,71 @@ class Libceed(Package):
 
     phases = ['build', 'install']
 
-    def build(self, spec, prefix):
+    @property
+    def common_make_opts(self):
+        spec = self.spec
+        compiler = self.compiler
         # Note: The occa package exports OCCA_DIR in the environment
 
+        # Use verbose building output
         makeopts = ['V=1']
-        makeopts += ['NDEBUG=%s' % ('' if '+debug' in spec else '1')]
+
+        if '@:0.2' in spec:
+            makeopts += ['NDEBUG=%s' % ('' if '+debug' in spec else '1')]
+
+        elif '@0.4:' in spec:
+            # Determine options based on the compiler:
+            if '+debug' in spec:
+                opt = '-g'
+            elif compiler.name == 'gcc':
+                opt = '-O -g -ffp-contract=fast'
+                if compiler.target in ['x86_64']:
+                    opt += ' -march=native'
+                elif compiler.target in ['ppc64le']:
+                    opt += ' -mcpu=native -mtune=native'
+                if compiler.version >= ver(4.9):
+                    opt += ' -fopenmp-simd'
+            elif compiler.name == 'clang':
+                opt = '-O -g -march=native -ffp-contract=fast'
+                if compiler.version.string.endswith('-apple'):
+                    if compiler.version >= ver(10):
+                        opt += ' -fopenmp-simd'
+                else:  # not apple clang
+                    if compiler.version >= ver(6):
+                        opt += ' -fopenmp-simd'
+            elif compiler.name in ['xl', 'xl_r']:
+                opt = '-O -g -qsimd=auto'
+            else:
+                opt = '-O -g'
+            makeopts += ['OPT=%s' % opt]
+
+            if '+cuda' in spec:
+                makeopts += ['CUDA_DIR=%s' % spec['cuda'].prefix]
+                nvccflags = ['-ccbin %s -Xcompiler "%s" -Xcompiler %s' %
+                             (compiler.cxx, opt, compiler.pic_flag)]
+                nvccflags = ' '.join(nvccflags)
+                makeopts += ['NVCCFLAGS=%s' % nvccflags]
+            else:
+                # Disable CUDA auto-detection:
+                makeopts += ['CUDA_DIR=/disable-cuda']
+
+            if '+libxsmm' in spec:
+                makeopts += ['XSMM_DIR=%s' % spec['libxsmm'].prefix]
+
+        return makeopts
+
+    def build(self, spec, prefix):
+        makeopts = self.common_make_opts
+        make('info', *makeopts)
         make(*makeopts)
 
         if self.run_tests:
             make('prove', *makeopts, parallel=False)
 
     def install(self, spec, prefix):
-        make('install', 'prefix=%s' % prefix, parallel=False)
+        installopts = ['prefix=%s' % prefix]
+        installopts += self.common_make_opts
+        make('install', *installopts, parallel=False)
 
     @when('@0.1')
     def install(self, spec, prefix):
diff --git a/var/spack/repos/builtin/packages/libceed/pkgconfig-version-0.4.diff b/var/spack/repos/builtin/packages/libceed/pkgconfig-version-0.4.diff
new file mode 100644
index 0000000000..5f08d9d454
--- /dev/null
+++ b/var/spack/repos/builtin/packages/libceed/pkgconfig-version-0.4.diff
@@ -0,0 +1,12 @@
+diff --git c/ceed.pc.template w/ceed.pc.template
+index 3216f08..5ada754 100644
+--- c/ceed.pc.template
++++ w/ceed.pc.template
+@@ -4,6 +4,6 @@ libdir=${prefix}/lib
+ 
+ Name: CEED
+ Description: Code for Efficient Extensible Discretization
+-Version: 0.2.1
++Version: 0.4
+ Cflags: -I${includedir}
+ Libs: -L${libdir} -lceed
diff --git a/var/spack/repos/builtin/packages/magma/magma-2.5.0-cmake.patch b/var/spack/repos/builtin/packages/magma/magma-2.5.0-cmake.patch
new file mode 100644
index 0000000000..56b58d85e7
--- /dev/null
+++ b/var/spack/repos/builtin/packages/magma/magma-2.5.0-cmake.patch
@@ -0,0 +1,77 @@
+diff -ru magma-2.5.0-orig/CMakeLists.txt magma-2.5.0/CMakeLists.txt
+--- magma-2.5.0-orig/CMakeLists.txt	2019-01-02 11:18:39.000000000 -0800
++++ magma-2.5.0/CMakeLists.txt	2019-04-03 15:58:01.871234891 -0700
+@@ -440,18 +440,20 @@
+ # compile MAGMA sparse library
+ 
+ # sparse doesn't have Fortran at the moment, so no need for above shenanigans
+-include_directories( sparse/include )
+-include_directories( sparse/control )
+-include_directories( testing )
+-cuda_add_library( magma_sparse ${libsparse_all} )
+-target_link_libraries( magma_sparse
+-	magma
++if (MAGMA_SPARSE)
++  include_directories( sparse/include )
++  include_directories( sparse/control )
++  include_directories( testing )
++  cuda_add_library( magma_sparse ${libsparse_all} )
++  target_link_libraries( magma_sparse
++	  magma
+ 	${LAPACK_LIBRARIES}
+ 	${CUDA_CUDART_LIBRARY}
+ 	${CUDA_CUBLAS_LIBRARIES}
+ 	${CUDA_cusparse_LIBRARY}
+-)
+-set( LIBS_SPARSE ${LIBS} magma_sparse )
++  )
++  set( LIBS_SPARSE ${LIBS} magma_sparse )
++endif()
+ 
+ 
+ # ----------------------------------------
+@@ -480,23 +482,31 @@
+ 
+ # ----------------------------------------
+ # compile each sparse tester
+-set( CMAKE_RUNTIME_OUTPUT_DIRECTORY sparse/testing )
+-foreach( TEST ${sparse_testing_all} )
++if (MAGMA_SPARSE)
++  set( CMAKE_RUNTIME_OUTPUT_DIRECTORY sparse/testing )
++  foreach( TEST ${sparse_testing_all} )
+ 	string( REGEX REPLACE "\\.(cpp|f90|F90)"     "" EXE ${TEST} )
+ 	string( REGEX REPLACE "sparse/testing/" "" EXE ${EXE} )
+ 	#message( "${TEST} --> ${EXE}" )
+ 	add_executable( ${EXE} ${TEST} )
+ 	target_link_libraries( ${EXE} ${LIBS_SPARSE} )
+-endforeach()
++  endforeach()
++endif()
+ 
+ 
+ # ----------------------------------------
+ # what to install
+-install( TARGETS magma magma_sparse ${blas_fix}
++set(MAGMA_TARGETS magma)
++set(MAGMA_HEADERS_PATTERNS include/*.h)
++if (MAGMA_SPARSE)
++  set(MAGMA_TARGETS ${MAGMA_TARGETS} magma_sparse)
++  set(MAGMA_HEADERS_PATTERNS ${MAGMA_HEADERS_PATTERNS} sparse/include/*.h)
++endif()
++install( TARGETS ${MAGMA_TARGETS} ${blas_fix}
+          RUNTIME DESTINATION bin
+          LIBRARY DESTINATION lib
+          ARCHIVE DESTINATION lib )
+-file( GLOB headers include/*.h sparse/include/*.h )
++file( GLOB headers ${MAGMA_HEADERS_PATTERNS} )
+ install( FILES ${headers}
+          DESTINATION include )
+ 
+@@ -509,4 +519,6 @@
+ message( STATUS "    NFLAGS       ${CUDA_NVCC_FLAGS}" )
+ message( STATUS "    FFLAGS       ${CMAKE_Fortran_FLAGS}" )
+ message( STATUS "    LIBS         ${LIBS}" )
+-message( STATUS "    LIBS_SPARSE  ${LIBS_SPARSE}" )
++if (MAGMA_SPARSE)
++  message( STATUS "    LIBS_SPARSE  ${LIBS_SPARSE}" )
++endif()
diff --git a/var/spack/repos/builtin/packages/magma/magma-2.5.0.patch b/var/spack/repos/builtin/packages/magma/magma-2.5.0.patch
new file mode 100644
index 0000000000..1ac800c5f8
--- /dev/null
+++ b/var/spack/repos/builtin/packages/magma/magma-2.5.0.patch
@@ -0,0 +1,428 @@
+diff -r 89706c0efbdb .hgtags
+--- a/.hgtags	Wed Jan 02 14:17:26 2019 -0500
++++ b/.hgtags	Wed Apr 03 15:50:54 2019 -0700
+@@ -1,3 +1,4 @@
+ 9c7e7cffa7d0e2decd23cde36a4830dfb55bea13 v2.2.0
+ b2b2e21c22a59a79eefbf1e5cff8e7d539a52c0c v2.3.0
+ 04d08aaa27dc8a551513d268c68fc299e81b6780 v2.4.0
++89706c0efbdbfd48bf8a2c20cc0d73e53c3f387e v2.5.0
+diff -r 89706c0efbdb include/magma_types.h
+--- a/include/magma_types.h	Wed Jan 02 14:17:26 2019 -0500
++++ b/include/magma_types.h	Wed Apr 03 15:50:54 2019 -0700
+@@ -77,7 +77,7 @@
+     typedef magma_int_t    magma_device_t;
+ 
+     // Half precision in CUDA 
+-    #if defined(__cplusplus) && CUDA_VERSION > 7500
++    #if defined(__cplusplus) && CUDA_VERSION >= 7500
+     #include <cuda_fp16.h>
+     typedef __half           magmaHalf;
+     #else
+diff -r 89706c0efbdb sparse/blas/magma_zsampleselect.cu
+--- a/sparse/blas/magma_zsampleselect.cu	Wed Jan 02 14:17:26 2019 -0500
++++ b/sparse/blas/magma_zsampleselect.cu	Wed Apr 03 15:50:54 2019 -0700
+@@ -15,9 +15,12 @@
+ 
+ #define PRECISION_z
+ 
++
+ namespace magma_sampleselect {
+ 
+-__global__ void compute_abs(const magmaDoubleComplex* __restrict__ in, double* __restrict__ out, int32_t size) {
++__global__ void compute_abs(const magmaDoubleComplex* __restrict__ in, double* __restrict__ out, int32_t size) 
++{
++#if (__CUDA_ARCH__ >= 350)
+     auto idx = threadIdx.x + blockDim.x * blockIdx.x;
+     if (idx >= size) {
+         return;
+@@ -25,6 +28,7 @@
+ 
+     auto v = in[idx];
+     out[idx] = real(v) * real(v) + imag(v) * imag(v);
++#endif
+ }
+ 
+ } // namespace magma_sampleselect
+@@ -164,36 +168,43 @@
+     magma_queue_t queue )
+ {
+     magma_int_t info = 0;
++    magma_int_t arch = magma_getdevice_arch();
+ 
+-    auto num_blocks = magma_ceildiv(total_size, block_size);
+-    auto local_work = (total_size + num_threads - 1) / num_threads;
+-    auto required_size = sizeof(double) * (total_size + searchtree_size)
++    if( arch >= 350 ) {
++        auto num_blocks = magma_ceildiv(total_size, block_size);
++        auto local_work = (total_size + num_threads - 1) / num_threads;
++        auto required_size = sizeof(double) * (total_size + searchtree_size)
+                          + sizeof(int32_t) * (searchtree_width * (num_grouped_blocks + 1) + 1);
+-    auto realloc_result = realloc_if_necessary(tmp_ptr, tmp_size, required_size);
++        auto realloc_result = realloc_if_necessary(tmp_ptr, tmp_size, required_size);
+ 
+-    double* gputmp = (double*)*tmp_ptr;
+-    double* gputree = gputmp + total_size;
+-    uint32_t* gpubucketidx = (uint32_t*)(gputree + searchtree_size);
+-    int32_t* gpurankout = (int32_t*)(gpubucketidx + 1);
+-    int32_t* gpucounts = gpurankout + 1;
+-    int32_t* gpulocalcounts = gpucounts + searchtree_width;
+-    uint32_t bucketidx{};
++        double* gputmp = (double*)*tmp_ptr;
++        double* gputree = gputmp + total_size;
++        uint32_t* gpubucketidx = (uint32_t*)(gputree + searchtree_size);
++        int32_t* gpurankout = (int32_t*)(gpubucketidx + 1);
++        int32_t* gpucounts = gpurankout + 1;
++        int32_t* gpulocalcounts = gpucounts + searchtree_width;
++        uint32_t bucketidx{};
+ 
+-    CHECK(realloc_result);
++        CHECK(realloc_result);
+ 
+-    compute_abs<<<num_blocks, block_size, 0, queue->cuda_stream()>>>
+-        (val, gputmp, total_size);
+-    build_searchtree<<<1, sample_size, 0, queue->cuda_stream()>>>
+-        (gputmp, gputree, total_size);
+-    count_buckets<<<num_grouped_blocks, block_size, 0, queue->cuda_stream()>>>
+-        (gputmp, gputree, gpulocalcounts, total_size, local_work);
+-    reduce_counts<<<searchtree_width, num_grouped_blocks, 0, queue->cuda_stream()>>>
+-        (gpulocalcounts, gpucounts, num_grouped_blocks);
+-    sampleselect_findbucket<<<1, searchtree_width / 2, 0, queue->cuda_stream()>>>
+-        (gpucounts, subset_size, gpubucketidx, gpurankout);
+-    magma_getvector(1, sizeof(uint32_t), gpubucketidx, 1, &bucketidx, 1, queue);
+-    magma_dgetvector(1, gputree + searchtree_width - 1 + bucketidx, 1, thrs, 1, queue);
+-    *thrs = std::sqrt(*thrs);
++        compute_abs<<<num_blocks, block_size, 0, queue->cuda_stream()>>>
++            (val, gputmp, total_size);
++        build_searchtree<<<1, sample_size, 0, queue->cuda_stream()>>>
++            (gputmp, gputree, total_size);
++        count_buckets<<<num_grouped_blocks, block_size, 0, queue->cuda_stream()>>>
++            (gputmp, gputree, gpulocalcounts, total_size, local_work);
++        reduce_counts<<<searchtree_width, num_grouped_blocks, 0, queue->cuda_stream()>>>
++            (gpulocalcounts, gpucounts, num_grouped_blocks);
++        sampleselect_findbucket<<<1, searchtree_width / 2, 0, queue->cuda_stream()>>>
++            (gpucounts, subset_size, gpubucketidx, gpurankout);
++        magma_getvector(1, sizeof(uint32_t), gpubucketidx, 1, &bucketidx, 1, queue);
++        magma_dgetvector(1, gputree + searchtree_width - 1 + bucketidx, 1, thrs, 1, queue);
++        *thrs = std::sqrt(*thrs);
++    }
++    else {
++        printf("error: this functionality needs CUDA architecture >= 3.5\n");
++        info = MAGMA_ERR_NOT_SUPPORTED;
++    }
+ 
+ cleanup:
+     return info;
+diff -r 89706c0efbdb src/xhsgetrf_gpu.cpp
+--- a/src/xhsgetrf_gpu.cpp	Wed Jan 02 14:17:26 2019 -0500
++++ b/src/xhsgetrf_gpu.cpp	Wed Apr 03 15:50:54 2019 -0700
+@@ -16,6 +16,131 @@
+ #include <cuda_fp16.h>
+ #endif
+ 
++#if CUDA_VERSION < 9020
++// conversion float to half are not defined for host in CUDA version <9.2
++// thus uses the conversion below when CUDA VERSION is < 9.2.
++#include <string.h>
++//
++// Copyright (c) 1993-2016, NVIDIA CORPORATION. All rights reserved.
++//
++// Redistribution and use in source and binary forms, with or without
++// modification, are permitted provided that the following conditions
++// are met:
++//  * Redistributions of source code must retain the above copyright
++//    notice, this list of conditions and the following disclaimer.
++//  * Redistributions in binary form must reproduce the above copyright
++//    notice, this list of conditions and the following disclaimer in the
++//    documentation and/or other materials provided with the distribution.
++//  * Neither the name of NVIDIA CORPORATION nor the names of its
++//    contributors may be used to endorse or promote products derived
++//    from this software without specific prior written permission.
++//
++// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
++// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
++// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
++// PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
++// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
++// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
++// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
++// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
++// OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
++// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
++// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
++
++// This code modified from the public domain code here: 
++// https://gist.github.com/rygorous/2156668
++// The URL above includes more robust conversion routines
++// that handle Inf and NaN correctly. 
++// 
++// It is recommended to use the more robust versions in production code.
++
++typedef unsigned uint;
++
++union FP32
++{
++    uint u;
++    float f;
++    struct
++    {
++        uint Mantissa : 23;
++        uint Exponent : 8;
++        uint Sign : 1;
++    };
++};
++
++union FP16
++{
++    unsigned short u;
++    struct
++    {
++        uint Mantissa : 10;
++        uint Exponent : 5;
++        uint Sign : 1;
++    };
++};
++
++// Approximate solution. This is faster but converts some sNaNs to
++// infinity and doesn't round correctly. Handle with care.
++// Approximate solution. This is faster but converts some sNaNs to
++// infinity and doesn't round correctly. Handle with care.
++static half approx_float_to_half(float fl)
++{
++    FP32 f32infty = { 255 << 23 };
++    FP32 f16max = { (127 + 16) << 23 };
++    FP32 magic = { 15 << 23 };
++    FP32 expinf = { (255 ^ 31) << 23 };
++    uint sign_mask = 0x80000000u;
++    FP16 o = { 0 };
++
++    FP32 f = *((FP32*)&fl);
++
++    uint sign = f.u & sign_mask;
++    f.u ^= sign;
++
++    if (!(f.f < f32infty.u)) // Inf or NaN
++        o.u = f.u ^ expinf.u;
++    else
++    {
++        if (f.f > f16max.f) f.f = f16max.f;
++        f.f *= magic.f;
++    }
++
++    o.u = f.u >> 13; // Take the mantissa bits
++    o.u |= sign >> 16;
++    half tmp;
++    memcpy(&tmp, &o, sizeof(half));
++    //return *((half*)&o);
++    return tmp;
++}
++
++// from half->float code - just for verification.
++static float half_to_float(half hf)
++{
++    FP16 h;
++    memcpy(&h, &hf, sizeof(half));
++
++    static const FP32 magic = { 113 << 23 };
++    static const uint shifted_exp = 0x7c00 << 13; // exponent mask after shift
++    FP32 o;
++
++    o.u = (h.u & 0x7fff) << 13;     // exponent/mantissa bits
++    uint exp = shifted_exp & o.u;   // just the exponent
++    o.u += (127 - 15) << 23;        // exponent adjust
++
++    // handle exponent special cases
++    if (exp == shifted_exp) // Inf/NaN?
++        o.u += (128 - 16) << 23;    // extra exp adjust
++    else if (exp == 0) // Zero/Denormal?
++    {
++        o.u += 1 << 23;             // extra exp adjust
++        o.f -= magic.f;             // renormalize
++    }
++
++    o.u |= (h.u & 0x8000) << 16;    // sign bit
++    return o.f;
++}
++#endif
++
+ #include "magma_internal.h"
+ //#include "nvToolsExt.h"
+ 
+@@ -106,10 +231,13 @@
+     float c_one     = MAGMA_S_ONE;
+     float c_neg_one = MAGMA_S_NEG_ONE;
+     #if 1
++    #if CUDA_VERSION >= 9020
+     const magmaHalf h_one     = (magmaHalf) 1.0;
+     const magmaHalf h_neg_one = (magmaHalf)-1.0;
+-    //const magmaHalf h_one = approx_float_to_half(1.0);
+-    //const magmaHalf h_neg_one = approx_float_to_half(-1.0);
++    #else
++    const magmaHalf h_one = approx_float_to_half(1.0);
++    const magmaHalf h_neg_one = approx_float_to_half(-1.0);
++    #endif
+     #else
+     FP32 float_one    = *((FP32*)&c_one);
+     FP16 half_one     = float_to_half_full(float_one);
+diff -r 89706c0efbdb src/xshgetrf_gpu.cpp
+--- a/src/xshgetrf_gpu.cpp	Wed Jan 02 14:17:26 2019 -0500
++++ b/src/xshgetrf_gpu.cpp	Wed Apr 03 15:50:54 2019 -0700
+@@ -92,7 +92,7 @@
+     magma_mp_type_t enable_tc,
+     magma_mp_type_t mp_algo_type )
+ {
+-#if CUDA_VERSION >= 7500
++#if CUDA_VERSION >= 9000
+     #ifdef HAVE_clBLAS
+     #define  dA(i_, j_) dA,  (dA_offset  + (i_)       + (j_)*ldda)
+     #define dAT(i_, j_) dAT, (dAT_offset + (i_)*lddat + (j_))
+diff -r 89706c0efbdb testing/testing_hgemm.cpp
+--- a/testing/testing_hgemm.cpp	Wed Jan 02 14:17:26 2019 -0500
++++ b/testing/testing_hgemm.cpp	Wed Apr 03 15:50:54 2019 -0700
+@@ -22,6 +22,131 @@
+ #include "magma_operators.h"
+ #include "testings.h"
+ 
++#if CUDA_VERSION < 9020
++// conversion float to half are not defined for host in CUDA version <9.2
++// thus uses the conversion below when CUDA VERSION is < 9.2.
++#include <string.h>
++//
++// Copyright (c) 1993-2016, NVIDIA CORPORATION. All rights reserved.
++//
++// Redistribution and use in source and binary forms, with or without
++// modification, are permitted provided that the following conditions
++// are met:
++//  * Redistributions of source code must retain the above copyright
++//    notice, this list of conditions and the following disclaimer.
++//  * Redistributions in binary form must reproduce the above copyright
++//    notice, this list of conditions and the following disclaimer in the
++//    documentation and/or other materials provided with the distribution.
++//  * Neither the name of NVIDIA CORPORATION nor the names of its
++//    contributors may be used to endorse or promote products derived
++//    from this software without specific prior written permission.
++//
++// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
++// EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
++// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
++// PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
++// CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
++// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
++// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
++// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
++// OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
++// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
++// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
++
++// This code modified from the public domain code here: 
++// https://gist.github.com/rygorous/2156668
++// The URL above includes more robust conversion routines
++// that handle Inf and NaN correctly. 
++// 
++// It is recommended to use the more robust versions in production code.
++
++typedef unsigned uint;
++
++union FP32
++{
++    uint u;
++    float f;
++    struct
++    {
++        uint Mantissa : 23;
++        uint Exponent : 8;
++        uint Sign : 1;
++    };
++};
++
++union FP16
++{
++    unsigned short u;
++    struct
++    {
++        uint Mantissa : 10;
++        uint Exponent : 5;
++        uint Sign : 1;
++    };
++};
++
++// Approximate solution. This is faster but converts some sNaNs to
++// infinity and doesn't round correctly. Handle with care.
++// Approximate solution. This is faster but converts some sNaNs to
++// infinity and doesn't round correctly. Handle with care.
++static half approx_float_to_half(float fl)
++{
++    FP32 f32infty = { 255 << 23 };
++    FP32 f16max = { (127 + 16) << 23 };
++    FP32 magic = { 15 << 23 };
++    FP32 expinf = { (255 ^ 31) << 23 };
++    uint sign_mask = 0x80000000u;
++    FP16 o = { 0 };
++
++    FP32 f = *((FP32*)&fl);
++
++    uint sign = f.u & sign_mask;
++    f.u ^= sign;
++
++    if (!(f.f < f32infty.u)) // Inf or NaN
++        o.u = f.u ^ expinf.u;
++    else
++    {
++        if (f.f > f16max.f) f.f = f16max.f;
++        f.f *= magic.f;
++    }
++
++    o.u = f.u >> 13; // Take the mantissa bits
++    o.u |= sign >> 16;
++    half tmp;
++    memcpy(&tmp, &o, sizeof(half));
++    //return *((half*)&o);
++    return tmp;
++}
++
++// from half->float code - just for verification.
++static float half_to_float(half hf)
++{
++    FP16 h;
++    memcpy(&h, &hf, sizeof(half));
++
++    static const FP32 magic = { 113 << 23 };
++    static const uint shifted_exp = 0x7c00 << 13; // exponent mask after shift
++    FP32 o;
++
++    o.u = (h.u & 0x7fff) << 13;     // exponent/mantissa bits
++    uint exp = shifted_exp & o.u;   // just the exponent
++    o.u += (127 - 15) << 23;        // exponent adjust
++
++    // handle exponent special cases
++    if (exp == shifted_exp) // Inf/NaN?
++        o.u += (128 - 16) << 23;    // extra exp adjust
++    else if (exp == 0) // Zero/Denormal?
++    {
++        o.u += 1 << 23;             // extra exp adjust
++        o.f -= magic.f;             // renormalize
++    }
++
++    o.u |= (h.u & 0x8000) << 16;    // sign bit
++    return o.f;
++}
++#endif
++
+ /* ////////////////////////////////////////////////////////////////////////////
+    -- Testing sgemm
+ */
+@@ -47,8 +172,13 @@
+     float c_neg_one = MAGMA_S_NEG_ONE;
+     float alpha = MAGMA_S_MAKE(  0.29, -0.86 );
+     float beta  = MAGMA_S_MAKE( -0.48,  0.38 );
+-    magmaHalf h_alpha = (magmaHalf)alpha;
+-    magmaHalf h_beta  = (magmaHalf)beta;
++    #if CUDA_VERSION >= 9020
++    const magmaHalf h_alpha = (magmaHalf) alpha;
++    const magmaHalf h_beta  = (magmaHalf) beta;
++    #else
++    const magmaHalf h_alpha = approx_float_to_half(alpha);
++    const magmaHalf h_beta  = approx_float_to_half(beta);
++    #endif
+     magma_opts opts;
+     opts.parse_opts( argc, argv );
+     
diff --git a/var/spack/repos/builtin/packages/magma/package.py b/var/spack/repos/builtin/packages/magma/package.py
index a4ca3e0e40..192bab856a 100644
--- a/var/spack/repos/builtin/packages/magma/package.py
+++ b/var/spack/repos/builtin/packages/magma/package.py
@@ -36,6 +36,8 @@ class Magma(CMakePackage):
     patch('ibm-xl.patch', when='@2.2:%xl')
     patch('ibm-xl.patch', when='@2.2:%xl_r')
     patch('magma-2.3.0-gcc-4.8.patch', when='@2.3.0%gcc@:4.8')
+    patch('magma-2.5.0.patch', when='@2.5.0')
+    patch('magma-2.5.0-cmake.patch', when='@2.5.0')
 
     def cmake_args(self):
         spec = self.spec
@@ -70,6 +72,11 @@ class Magma(CMakePackage):
             else:
                 options.extend(['-DGPU_TARGET=sm_30'])
 
+        if '@2.5.0' in spec:
+            options.extend(['-DMAGMA_SPARSE=OFF'])
+            if spec.compiler.name in ['xl', 'xl_r']:
+                options.extend(['-DCMAKE_DISABLE_FIND_PACKAGE_OpenMP=TRUE'])
+
         return options
 
     @run_after('install')
diff --git a/var/spack/repos/builtin/packages/mfem/package.py b/var/spack/repos/builtin/packages/mfem/package.py
index f471b37de0..16a977d48c 100644
--- a/var/spack/repos/builtin/packages/mfem/package.py
+++ b/var/spack/repos/builtin/packages/mfem/package.py
@@ -45,6 +45,9 @@ class Mfem(Package):
     # other version.
     version('develop', branch='master')
 
+    # Tagged development version used by the laghos package:
+    version('3.4.1-laghos-v2.0', tag='laghos-v2.0')
+
     version('3.4.0',
             '4e73e4fe0482636de3c5dc983cd395839a83cb16f6f509bd88b053e8b3858e05',
             url='https://bit.ly/mfem-3-4', extension='.tar.gz',
@@ -54,7 +57,8 @@ class Mfem(Package):
             'b70fa3c5080b9ec514fc05f4a04ff74322b99ac4ecd6d99c229f0ed5188fc0ce',
             url='https://goo.gl/Kd7Jk8', extension='.tar.gz')
 
-    version('laghos-v1.0', tag='laghos-v1.0')
+    # Tagged development version used by the laghos package:
+    version('3.3.1-laghos-v1.0', tag='laghos-v1.0')
 
     version('3.3',
             'b17bd452593aada93dc0fee748fcfbbf4f04ce3e7d77fdd0341cc9103bcacd0b',
@@ -424,16 +428,18 @@ class Mfem(Package):
                 copy(str(self.config_mk), 'config.mk')
                 shutil.copystat('config.mk.orig', 'config.mk')
 
+        prefix_share = join_path(prefix, 'share', 'mfem')
+
         if '+examples' in spec:
             make('examples')
-            install_tree('examples', join_path(prefix, 'examples'))
+            install_tree('examples', join_path(prefix_share, 'examples'))
 
         if '+miniapps' in spec:
             make('miniapps')
-            install_tree('miniapps', join_path(prefix, 'miniapps'))
+            install_tree('miniapps', join_path(prefix_share, 'miniapps'))
 
         if install_em:
-            install_tree('data', join_path(prefix, 'data'))
+            install_tree('data', join_path(prefix_share, 'data'))
 
     @property
     def suitesparse_components(self):
diff --git a/var/spack/repos/builtin/packages/nek5000/package.py b/var/spack/repos/builtin/packages/nek5000/package.py
index a522af8cf0..9e2af1d5fe 100644
--- a/var/spack/repos/builtin/packages/nek5000/package.py
+++ b/var/spack/repos/builtin/packages/nek5000/package.py
@@ -5,19 +5,9 @@
 
 from spack import *
 
-import numbers
 import os
 
 
-def is_integral(x):
-    """Any integer value"""
-    try:
-        return isinstance(int(x), numbers.Integral) and \
-            not isinstance(x, bool) and int(x) > 0
-    except ValueError:
-        return False
-
-
 class Nek5000(Package):
     """A fast and scalable high-order solver for computational fluid
        dynamics"""
@@ -40,33 +30,8 @@ class Nek5000(Package):
     # TODO: add a variant 'blas' or 'external-blas' to enable the usage of
     #       Spack installed/configured blas.
 
-    # Variant for MAXNEL, we need to read this from user
-    variant(
-        'MAXNEL',
-        default=150000,
-        description='Maximum number of elements for Nek5000 tools.',
-        values=is_integral
-    )
-
-    # Variants for Nek tools
-    variant('genbox',   default=True, description='Build genbox tool.')
-    variant('int_tp',   default=True, description='Build int_tp tool.')
-    variant('n2to3',    default=True, description='Build n2to3 tool.')
-    variant('postnek',  default=True, description='Build postnek tool.')
-    variant('reatore2', default=True, description='Build reatore2 tool.')
-    variant('genmap',   default=True, description='Build genmap tool.')
-    variant('nekmerge', default=True, description='Build nekmerge tool.')
-    variant('prenek',   default=True, description='Build prenek tool.')
-
     # Dependencies
     depends_on('mpi', when="+mpi")
-    depends_on('libx11', when="+prenek")
-    depends_on('libx11', when="+postnek")
-    # libxt is needed for X11/Intrinsic.h but not for linking
-    depends_on('libxt', when="+prenek")
-    depends_on('xproto', when="+prenek")
-    depends_on('libxt', when="+postnek")
-    depends_on('visit', when="+visit")
 
     @run_before('install')
     def fortran_check(self):
@@ -83,7 +48,6 @@ class Nek5000(Package):
                 raise RuntimeError(msg)
 
     def install(self, spec, prefix):
-        tools_dir   = 'tools'
         bin_dir     = 'bin'
 
         # Do not use the Spack compiler wrappers.
@@ -93,101 +57,22 @@ class Nek5000(Package):
 
         fflags = spec.compiler_flags['fflags']
         cflags = spec.compiler_flags['cflags']
-        if ('+prenek' in spec) or ('+postnek' in spec):
-            libx11_h = find_headers('Xlib', spec['libx11'].prefix.include,
-                                    recursive=True)
-            if not libx11_h:
-                raise RuntimeError('Xlib.h not found in %s' %
-                                   spec['libx11'].prefix.include)
-            cflags += ['-I%s' % os.path.dirname(libx11_h.directories[0])]
-
-            xproto_h = find_headers('X', spec['xproto'].prefix.include,
-                                    recursive=True)
-            if not xproto_h:
-                raise RuntimeError('X.h not found in %s' %
-                                   spec['xproto'].prefix.include)
-            cflags += ['-I%s' % os.path.dirname(xproto_h.directories[0])]
-
-            libxt_h = find_headers('Intrinsic', spec['libxt'].prefix.include,
-                                   recursive=True)
-            if not libxt_h:
-                raise RuntimeError('X11/Intrinsic.h not found in %s' %
-                                   spec['libxt'].prefix.include)
-            cflags += ['-I%s' % os.path.dirname(libxt_h.directories[0])]
+
         if self.compiler.name in ['xl', 'xl_r']:
             # Use '-qextname' to add underscores.
             # Use '-WF,-qnotrigraph' to fix an error about a string: '... ??'
             fflags += ['-qextname', '-WF,-qnotrigraph']
-        fflags = ' '.join(fflags)
-        cflags = ' '.join(cflags)
 
-        # Build the tools, maketools copy them to Nek5000/bin by default.
-        # We will then install Nek5000/bin under prefix after that.
-        with working_dir(tools_dir):
-            # Update the maketools script to use correct compilers
-            filter_file(r'^#FC\s*=.*', 'FC="{0}"'.format(fc), 'maketools')
-            filter_file(r'^#CC\s*=.*', 'CC="{0}"'.format(cc), 'maketools')
-            if fflags:
-                filter_file(r'^#FFLAGS=.*', 'FFLAGS="{0}"'.format(fflags),
-                            'maketools')
-            if cflags:
-                filter_file(r'^#CFLAGS=.*', 'CFLAGS="{0}"'.format(cflags),
-                            'maketools')
+        error = Executable(fc)('empty.f', output=str, error=str,
+                               fail_on_error=False)
 
-            if self.compiler.name in ['xl', 'xl_r']:
-                # Patch 'maketools' to use '-qextname' when checking for
-                # underscore becasue 'xl'/'xl_r' use this option to enable the
-                # addition of the underscore.
-                filter_file(r'^\$FC -c ', '$FC -qextname -c ', 'maketools')
-
-            libx11_lib = find_libraries('libX11', spec['libx11'].prefix.lib,
-                                        shared=True, recursive=True)
-            if not libx11_lib:
-                libx11_lib = \
-                    find_libraries('libX11', spec['libx11'].prefix.lib64,
-                                   shared=True, recursive=True)
-            if not libx11_lib:
-                raise RuntimeError('libX11 not found in %s/{lib,lib64}' %
-                                   spec['libx11'].prefix)
-            # There is no other way to set the X11 library path except brute
-            # force:
-            filter_file(r'-L\$\(X\)', libx11_lib.search_flags,
-                        join_path('prenek', 'makefile'))
-            filter_file(r'-L\$\(X\)', libx11_lib.search_flags,
-                        join_path('postnek', 'makefile'))
+        if 'gfortran' in error or 'GNU' in error or 'gfortran' in fc:
+            # Use '-std=legacy' to suppress an error that used to be a
+            # warning in previous versions of gfortran.
+            fflags += ['-std=legacy']
 
-            if self.compiler.name in ['xl', 'xl_r']:
-                # Use '-qextname' when compiling mxm.f
-                filter_file(r'\$\(OLAGS\)', '-qextname $(OLAGS)',
-                            join_path('postnek', 'makefile'))
-            # Define 'rename_' function that calls 'rename'
-            with open(join_path('postnek', 'xdriver.c'), 'a') as xdriver:
-                xdriver.write('\nvoid rename_(char *from, char *to)\n{\n'
-                              '   rename(from, to);\n}\n')
-
-            maxnel = self.spec.variants['MAXNEL'].value
-            filter_file(r'^#MAXNEL\s*=.*', 'MAXNEL=' + maxnel, 'maketools')
-
-            maketools = Executable('./maketools')
-
-            # Build the tools
-            if '+genbox' in spec:
-                maketools('genbox')
-            # "ERROR: int_tp does not exist!"
-            # if '+int_tp' in spec:
-            #     maketools('int_tp')
-            if '+n2to3' in spec:
-                maketools('n2to3')
-            if '+postnek' in spec:
-                maketools('postnek')
-            if '+reatore2' in spec:
-                maketools('reatore2')
-            if '+genmap' in spec:
-                maketools('genmap')
-            if '+nekmerge' in spec:
-                maketools('nekmerge')
-            if '+prenek' in spec:
-                maketools('prenek')
+        fflags = ' '.join(fflags)
+        cflags = ' '.join(cflags)
 
         with working_dir(bin_dir):
             if '+mpi' in spec:
diff --git a/var/spack/repos/builtin/packages/nekbone/package.py b/var/spack/repos/builtin/packages/nekbone/package.py
index 2666e77bf3..9d46ee8623 100644
--- a/var/spack/repos/builtin/packages/nekbone/package.py
+++ b/var/spack/repos/builtin/packages/nekbone/package.py
@@ -52,11 +52,21 @@ class Nekbone(Package):
         install(makenek, prefix.bin)
         install(nekpmpi, prefix.bin)
 
+        error = Executable(fc)('empty.f', output=str, error=str,
+                               fail_on_error=False)
+
+        fflags = ''
+        if 'gfortran' in error or 'GNU' in error or 'gfortran' in fc:
+            # Use '-std=legacy' to suppress an error that used to be a
+            # warning in previous versions of gfortran.
+            fflags = ' -std=legacy'
+
         with working_dir(prefix.bin):
             filter_file(r'^SOURCE_ROOT\s*=.*', 'SOURCE_ROOT=\"' +
                         prefix.bin.Nekbone + '/src\"', 'makenek')
             filter_file(r'^CC\s*=.*', 'CC=\"' + cc + '\"', 'makenek')
-            filter_file(r'^F77\s*=.*', 'F77=\"' + fc + '\"', 'makenek')
+            filter_file(r'^F77\s*=.*', 'F77=\"' + fc + fflags + '\"',
+                        'makenek')
 
             if '+mpi' not in spec:
                 filter_file(r'^#IFMPI=\"false\"', 'IFMPI=\"false\"', 'makenek')
diff --git a/var/spack/repos/builtin/packages/nekcem/package.py b/var/spack/repos/builtin/packages/nekcem/package.py
index 4dea04fb08..710c465b07 100644
--- a/var/spack/repos/builtin/packages/nekcem/package.py
+++ b/var/spack/repos/builtin/packages/nekcem/package.py
@@ -22,11 +22,13 @@ class Nekcem(Package):
     # We only have a development version
     version('develop')
     version('0b8bedd', commit='0b8beddfdcca646bfcc866dfda1c5f893338399b')
+    version('7332619', commit='7332619b73d03868a256614b61794dce2d95b360')
 
     # dependencies
     depends_on('mpi', when='+mpi')
     depends_on('blas')
     depends_on('lapack')
+    depends_on('python@2.7:', type='build')
 
     @run_before('install')
     def fortran_check(self):
@@ -81,6 +83,14 @@ class Nekcem(Package):
                 fflags += ['-r8']
                 cflags += ['-DUNDERSCORE']
 
+            error = Executable(fc)('empty.f', output=str, error=str,
+                                   fail_on_error=False)
+
+            if 'gfortran' in error or 'GNU' in error or 'gfortran' in fc:
+                # Use '-std=legacy' to suppress an error that used to be a
+                # warning in previous versions of gfortran.
+                fflags += ['-std=legacy']
+
             if '+mpi' in spec:
                 fflags += ['-DMPI', '-DMPIIO']
                 cflags += ['-DMPI', '-DMPIIO']
diff --git a/var/spack/repos/builtin/packages/nektools/package.py b/var/spack/repos/builtin/packages/nektools/package.py
new file mode 100644
index 0000000000..43dd014b0a
--- /dev/null
+++ b/var/spack/repos/builtin/packages/nektools/package.py
@@ -0,0 +1,177 @@
+# Copyright 2013-2019 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 numbers
+import os
+
+
+def is_integral(x):
+    """Any integer value"""
+    try:
+        return isinstance(int(x), numbers.Integral) and \
+            not isinstance(x, bool) and int(x) > 0
+    except ValueError:
+        return False
+
+
+class Nektools(Package):
+    """Tools reuqired by Nek5000"""
+
+    homepage = "https://nek5000.mcs.anl.gov/"
+    url      = "https://github.com/Nek5000/Nek5000/releases/download/v17.0/Nek5000-v17.0.tar.gz"
+    git      = "https://github.com/Nek5000/Nek5000.git"
+
+    tags = ['cfd', 'flow', 'hpc', 'solver', 'navier-stokes',
+            'spectral-elements', 'fluid', 'ecp', 'ecp-apps']
+
+    version('develop', branch='master')
+    version('17.0', '6a13bfad2ce023897010dd88f54a0a87')
+
+    # Variant for MAXNEL, we need to read this from user
+    variant(
+        'MAXNEL',
+        default=150000,
+        description='Maximum number of elements for Nek5000 tools.',
+        values=is_integral
+    )
+
+    # Variants for Nek tools
+    variant('genbox',   default=True, description='Build genbox tool.')
+    variant('n2to3',    default=True, description='Build n2to3 tool.')
+    variant('postnek',  default=True, description='Build postnek tool.')
+    variant('reatore2', default=True, description='Build reatore2 tool.')
+    variant('genmap',   default=True, description='Build genmap tool.')
+    variant('nekmerge', default=True, description='Build nekmerge tool.')
+    variant('prenek',   default=True, description='Build prenek tool.')
+
+    depends_on('libx11', when="+prenek")
+    depends_on('libx11', when="+postnek")
+    # libxt is needed for X11/Intrinsic.h but not for linking
+    depends_on('libxt', when="+prenek")
+    depends_on('xproto', when="+prenek")
+    depends_on('libxt', when="+postnek")
+    depends_on('visit', when="+visit")
+
+    @run_before('install')
+    def fortran_check(self):
+        if not self.compiler.f77:
+            msg = 'Cannot build Nek5000 without a Fortran 77 compiler.'
+            raise RuntimeError(msg)
+
+    def install(self, spec, prefix):
+        tools_dir   = 'tools'
+        bin_dir     = 'bin'
+
+        fc = env['FC']
+        cc = env['CC']
+
+        fflags = spec.compiler_flags['fflags']
+        cflags = spec.compiler_flags['cflags']
+        if ('+prenek' in spec) or ('+postnek' in spec):
+            libx11_h = find_headers('Xlib', spec['libx11'].prefix.include,
+                                    recursive=True)
+            if not libx11_h:
+                raise RuntimeError('Xlib.h not found in %s' %
+                                   spec['libx11'].prefix.include)
+            cflags += ['-I%s' % os.path.dirname(libx11_h.directories[0])]
+
+            xproto_h = find_headers('X', spec['xproto'].prefix.include,
+                                    recursive=True)
+            if not xproto_h:
+                raise RuntimeError('X.h not found in %s' %
+                                   spec['xproto'].prefix.include)
+            cflags += ['-I%s' % os.path.dirname(xproto_h.directories[0])]
+
+            libxt_h = find_headers('Intrinsic', spec['libxt'].prefix.include,
+                                   recursive=True)
+            if not libxt_h:
+                raise RuntimeError('X11/Intrinsic.h not found in %s' %
+                                   spec['libxt'].prefix.include)
+            cflags += ['-I%s' % os.path.dirname(libxt_h.directories[0])]
+        if self.compiler.name in ['xl', 'xl_r']:
+            # Use '-qextname' to add underscores.
+            # Use '-WF,-qnotrigraph' to fix an error about a string: '... ??'
+            fflags += ['-qextname', '-WF,-qnotrigraph']
+
+        error = Executable(fc)('empty.f', output=str, error=str,
+                               fail_on_error=False)
+
+        if 'gfortran' in error or 'GNU' in error or 'gfortran' in fc:
+            # Use '-std=legacy' to suppress an error that used to be a
+            # warning in previous versions of gfortran.
+            fflags += ['-std=legacy']
+
+        fflags = ' '.join(fflags)
+        cflags = ' '.join(cflags)
+
+        # Build the tools, maketools copy them to Nek5000/bin by default.
+        # We will then install Nek5000/bin under prefix after that.
+        with working_dir(tools_dir):
+            # Update the maketools script to use correct compilers
+            filter_file(r'^#FC\s*=.*', 'FC="{0}"'.format(fc), 'maketools')
+            filter_file(r'^#CC\s*=.*', 'CC="{0}"'.format(cc), 'maketools')
+            if fflags:
+                filter_file(r'^#FFLAGS=.*', 'FFLAGS="{0}"'.format(fflags),
+                            'maketools')
+            if cflags:
+                filter_file(r'^#CFLAGS=.*', 'CFLAGS="{0}"'.format(cflags),
+                            'maketools')
+
+            if self.compiler.name in ['xl', 'xl_r']:
+                # Patch 'maketools' to use '-qextname' when checking for
+                # underscore becasue 'xl'/'xl_r' use this option to enable the
+                # addition of the underscore.
+                filter_file(r'^\$FC -c ', '$FC -qextname -c ', 'maketools')
+
+            libx11_lib = find_libraries('libX11', spec['libx11'].prefix.lib,
+                                        shared=True, recursive=True)
+            if not libx11_lib:
+                libx11_lib = \
+                    find_libraries('libX11', spec['libx11'].prefix.lib64,
+                                   shared=True, recursive=True)
+            if not libx11_lib:
+                raise RuntimeError('libX11 not found in %s/{lib,lib64}' %
+                                   spec['libx11'].prefix)
+            # There is no other way to set the X11 library path except brute
+            # force:
+            filter_file(r'-L\$\(X\)', libx11_lib.search_flags,
+                        join_path('prenek', 'makefile'))
+            filter_file(r'-L\$\(X\)', libx11_lib.search_flags,
+                        join_path('postnek', 'makefile'))
+
+            if self.compiler.name in ['xl', 'xl_r']:
+                # Use '-qextname' when compiling mxm.f
+                filter_file(r'\$\(OLAGS\)', '-qextname $(OLAGS)',
+                            join_path('postnek', 'makefile'))
+            # Define 'rename_' function that calls 'rename'
+            with open(join_path('postnek', 'xdriver.c'), 'a') as xdriver:
+                xdriver.write('\nvoid rename_(char *from, char *to)\n{\n'
+                              '   rename(from, to);\n}\n')
+
+            maxnel = self.spec.variants['MAXNEL'].value
+            filter_file(r'^#MAXNEL\s*=.*', 'MAXNEL=' + maxnel, 'maketools')
+
+            maketools = Executable('./maketools')
+
+            # Build the tools
+            if '+genbox' in spec:
+                maketools('genbox')
+            if '+n2to3' in spec:
+                maketools('n2to3')
+            if '+postnek' in spec:
+                maketools('postnek')
+            if '+reatore2' in spec:
+                maketools('reatore2')
+            if '+genmap' in spec:
+                maketools('genmap')
+            if '+nekmerge' in spec:
+                maketools('nekmerge')
+            if '+prenek' in spec:
+                maketools('prenek')
+
+        # Install Nek5000/bin in prefix/bin
+        install_tree(bin_dir, prefix.bin)
diff --git a/var/spack/repos/builtin/packages/occa/package.py b/var/spack/repos/builtin/packages/occa/package.py
index cb60d80d29..c6528f2960 100644
--- a/var/spack/repos/builtin/packages/occa/package.py
+++ b/var/spack/repos/builtin/packages/occa/package.py
@@ -20,9 +20,10 @@ class Occa(Package):
     git      = 'https://github.com/libocca/occa.git'
 
     version('develop')
-    version('v1.0.0-alpha.5', tag='v1.0.0-alpha.5')
-    version('v0.2.0', tag='v0.2.0')
-    version('v0.1.0', tag='v0.1.0')
+    version('1.0.8', tag='v1.0.8')
+    version('1.0.0-alpha.5', tag='v1.0.0-alpha.5')
+    version('0.2.0', tag='v0.2.0')
+    version('0.1.0', tag='v0.1.0')
 
     variant('cuda',
             default=True,
diff --git a/var/spack/repos/builtin/packages/petsc/package.py b/var/spack/repos/builtin/packages/petsc/package.py
index 3d9686eb21..8cc961d361 100644
--- a/var/spack/repos/builtin/packages/petsc/package.py
+++ b/var/spack/repos/builtin/packages/petsc/package.py
@@ -108,6 +108,8 @@ class Petsc(Package):
 
     patch('xcode_stub_out_of_sync.patch', when='@:3.10.4')
 
+    patch('xlf_fix-dup-petscfecreate.patch', when='@3.11.0')
+
     # Virtual dependencies
     # Git repository needs sowing to build Fortran interface
     depends_on('sowing', when='@develop')
diff --git a/var/spack/repos/builtin/packages/petsc/xlf_fix-dup-petscfecreate.patch b/var/spack/repos/builtin/packages/petsc/xlf_fix-dup-petscfecreate.patch
new file mode 100644
index 0000000000..cce573c97e
--- /dev/null
+++ b/var/spack/repos/builtin/packages/petsc/xlf_fix-dup-petscfecreate.patch
@@ -0,0 +1,31 @@
+commit 5849cfbb8c629127894f722457da7b2cbb7f6a0b
+Author: Satish Balay <balay@mcs.anl.gov>
+Date:   Sun Mar 31 21:43:12 2019 -0500
+
+    dm: remove duplicate PetscFECreate() definition from f90 modules. This fixes IBM xlf build.
+    
+    "/tmp/dobrev1/spack-stage/spack-stage-Vp2QRG/petsc-3.11.0/src/tao/f90-mod/petsctaomod.F", line 38.13: 1514-264 (S) Procedure petscfecreate has more than one interface accessible by use association. The interfaces are assumed to be the same.
+    1501-511  Compilation failed for file petsctaomod.F.
+    
+    Reported-by: "Dobrev, Veselin A." <dobrev1@llnl.gov>
+
+diff --git a/src/dm/f90-mod/petscdt.h90 b/src/dm/f90-mod/petscdt.h90
+index 339f7fd41e..766f0e37dc 100644
+--- a/src/dm/f90-mod/petscdt.h90
++++ b/src/dm/f90-mod/petscdt.h90
+@@ -49,15 +49,6 @@
+         End Subroutine
+       End Interface
+ 
+-      Interface
+-        Subroutine PetscFECreate(c,f,ierr)
+-          use petscdmdef
+-          MPI_Comm, intent(in) :: c
+-          PetscFE, intent(out) :: f
+-          PetscErrorCode, intent(out) ::  ierr
+-        End Subroutine
+-      End Interface
+-
+       Interface
+         Subroutine PetscFECreateDefault(c,d,n,i,str,o,f,ierr)
+           use petscdmdef
diff --git a/var/spack/repos/builtin/packages/suite-sparse/package.py b/var/spack/repos/builtin/packages/suite-sparse/package.py
index ad3b984cc2..fc53051825 100644
--- a/var/spack/repos/builtin/packages/suite-sparse/package.py
+++ b/var/spack/repos/builtin/packages/suite-sparse/package.py
@@ -114,6 +114,14 @@ class SuiteSparse(Package):
                 'TBB=-L%s -ltbb' % spec['tbb'].prefix.lib,
             ]
 
+        if '@5.3:' in spec:
+            # Without CMAKE_LIBRARY_PATH defined, the CMake file in the
+            # Mongoose directory finds libsuitesparseconfig.so in system
+            # directories like /usr/lib.
+            make_args += [
+                'CMAKE_OPTIONS=-DCMAKE_INSTALL_PREFIX=%s' % prefix +
+                ' -DCMAKE_LIBRARY_PATH=%s' % prefix.lib]
+
         make('install', *make_args)
 
     @property