[NEW] Added amdfftw, amdlibflame and amdscalapack recipes (#19457)

* [NEW] Added amdfftw, amdlibflame and amdscalapack recipes

Updated base fftw, libflame and netlib-scalapack recipes
to accommodate the above listed AMD Optimizing CPU Libraries
which are a set of numerical routines optimized for AMD platforms.

Updated amdblis spack recipe

amdblis:
	1. updated with amdblis 2.2 release
amdfftw:
	1. "--enable-single" now work as synonym for "--enable-float"
amdlibflame:
	1. Added enable_or_disable_threads() to set value for "--enable-multithreading" flag
Libflame:
	1. Added enable_or_disable_threads() to set value for "--enable-multithreading" flag
	2. Corrected invocation of "enable_or_disable('threads')"

Change-Id: I9da0a2c2c4e2075b7fa2776e7cfe6548a2e0b32f

* Added amd-toolchain-support as maintainers

Added team github account amd-toolchain-support
as maintainers for all the recipes owned by
AMD Optimizing CPU Libraries (AOCL) team

Change-Id: I9a7969bd48fc42cfbb88dd7bd93e0802c6138582

* Incorporated review comments

Updated packages.yaml with aocl components
Handled Flake8 test failures

Change-Id: I0a03f02d8c9f326b2434ec907958c3de3a8e18eb

* Readded accidental removal of stream recipe

amdfftw:
	1. Updated the aocc clang selection as per spack standards
fftw:
	1. Currently apple-clang section is redundant,
	already it is handled in the conflict checks.

Change-Id: Idef4a3f61717eb81f321e0cd16e7ba9619eac846

* Fix for style and docs/validate (pull_request) test

unnumbered format placeholders from {} to {0}

Change-Id: If67a3374177ec067573e5504462d257712fafc05

* changed compiler references to Spack's compiler wrapper:spack_cc, spack_cxx, spack_fc

Change-Id: I7ae29c978fff16e37773913f14c84df232499763

* Removed 'single' variant from amdfftw recipe

Instead of conflict for apple-clang + openmp, handled this senario
via below available feature:
depends_on('llvm-openmp', when='%apple-clang +openmp')

Change-Id: I701b23d83e822a500ca3aaf2b60cc9ace09e13dc

* Added relevant info for users who prefers to use single precision

Change-Id: I3506e21da428ddef5fb7895b5aaed32c2a061ef6

* Minor changes on fftw, amdfftw and libflame

amdfftw:
	1. Removed escape symbol to the single quotes
	2. Rewording the conflict line from Recommended
	to Required

fftw:
	1. Reorded to following recommended sections:
	versions, variants, dependencies, providers,
	patches

libflame:
	1. Added provides entry for 5.1.0 version

Change-Id: I21ebff99b6dfde031763154693ecb3f1fa47b476

* Removed single quote from amdfftw docstring to fix style failures

Change-Id: Ife939a5a2f5ccbc8879b730c7bebfe2fcfef9332

8 files changed, 361 insertions, 70 deletions

diff --git a/var/spack/repos/builtin/packages/amdblis/package.py b/var/spack/repos/builtin/packages/amdblis/package.py
index 82d1f18938..e4657aeb8c 100644
--- a/var/spack/repos/builtin/packages/amdblis/package.py
+++ b/var/spack/repos/builtin/packages/amdblis/package.py
@@ -17,8 +17,9 @@ class Amdblis(BlisBase):
 
     _name = 'amdblis'
     homepage = "https://developer.amd.com/amd-aocl/blas-library/"
-    url = "https://github.com/amd/blis/archive/2.1.tar.gz"
+    url = "https://github.com/amd/blis/archive/2.2.tar.gz"
     git = "https://github.com/amd/blis.git"
 
+    maintainers = ['amd-toolchain-support']
+
     version('2.2', sha256='e1feb60ac919cf6d233c43c424f6a8a11eab2c62c2c6e3f2652c15ee9063c0c9')
-    version('2.1', sha256='3b1d611d46f0f13b3c0917e27012e0f789b23dbefdddcf877b20327552d72fb3')
diff --git a/var/spack/repos/builtin/packages/amdfftw/package.py b/var/spack/repos/builtin/packages/amdfftw/package.py
new file mode 100644
index 0000000000..9e4887ae98
--- /dev/null
+++ b/var/spack/repos/builtin/packages/amdfftw/package.py
@@ -0,0 +1,130 @@
+# 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)
+
+import os
+from spack import *
+from spack.pkg.builtin.fftw import FftwBase
+
+
+class Amdfftw(FftwBase):
+    """FFTW (AMD Optimized version) is a comprehensive collection of
+    fast C routines for computing the Discrete Fourier Transform (DFT)
+    and various special cases thereof.
+
+    It is an open-source implementation of the Fast Fourier transform
+    algorithm. It can compute transforms of real and complex-values
+    arrays of arbitrary size and dimension.
+    AMD Optimized FFTW is the optimized FFTW implementation targeted
+    for AMD CPUs.
+
+    For single precision build, please use precision value as float.
+    Example : spack install amdfftw precision=float
+    """
+
+    _name = 'amdfftw'
+    homepage = "https://developer.amd.com/amd-aocl/fftw/"
+    url = "https://github.com/amd/amd-fftw/archive/2.2.tar.gz"
+    git = "https://github.com/amd/amd-fftw.git"
+
+    maintainers = ['amd-toolchain-support']
+
+    version('2.2', sha256='de9d777236fb290c335860b458131678f75aa0799c641490c644c843f0e246f8')
+
+    variant('shared', default=True, description='Builds a shared version of the library')
+    variant('openmp', default=True, description="Enable OpenMP support")
+    variant('debug', default=False, description='Builds a debug version of the library')
+
+    depends_on('texinfo')
+
+    provides('fftw-api@3', when='@2:')
+
+    conflicts('precision=quad', when='%aocc', msg="AOCC clang doesn't support quad precision")
+    conflicts('+debug', when='%aocc', msg="AOCC clang doesn't support debug")
+    conflicts('%gcc@:7.2', when="@2.2:", msg="Required GCC version above 7.2 for AMDFFTW")
+
+    def configure(self, spec, prefix):
+        """Configure function"""
+        # Base options
+        options = [
+            '--prefix={0}'.format(prefix),
+            '--enable-amd-opt',
+            '--enable-threads'
+        ]
+
+        # Check if compiler is AOCC
+        if spec.satisfies('%aocc'):
+            options.append("CC={0}".format(os.path.basename(spack_cc)))
+            options.append("CXX={0}".format(os.path.basename(spack_cxx)))
+            options.append("FC={0}".format(os.path.basename(spack_fc)))
+
+        if '+shared' in spec:
+            options.append('--enable-shared')
+        else:
+            options.append('--disable-shared')
+
+        if '+openmp' in spec:
+            options.append('--enable-openmp')
+        else:
+            options.append('--disable-openmp')
+
+        if '+mpi' in spec:
+            options.append('--enable-mpi')
+            options.append('--enable-amd-mpifft')
+        else:
+            options.append('--disable-mpi')
+            options.append('--disable-amd-mpifft')
+
+        if not self.compiler.f77 or not self.compiler.fc:
+            options.append("--disable-fortran")
+
+        # Specific SIMD support.
+        # float and double precisions are supported
+        simd_features = ['sse2', 'avx', 'avx2', 'avx512', 'avx-128-fma',
+                         'kcvi', 'vsx', 'neon']
+        simd_options = []
+        for feature in simd_features:
+            msg = '--enable-{0}' if feature in spec.target else '--disable-{0}'
+            simd_options.append(msg.format(feature))
+
+        simd_options += [
+            '--enable-fma' if 'fma' in spec.target else '--disable-fma'
+        ]
+
+        float_simd_features = ['altivec', 'sse']
+
+        # When enabling configure option "--enable-amd-opt", do not use the
+        # configure option "--enable-generic-simd128" or
+        # "--enable-generic-simd256"
+
+        # Double is the default precision, for all the others we need
+        # to enable the corresponding option.
+        enable_precision = {
+            'float': ['--enable-float'],
+            'double': None,
+            'long_double': ['--enable-long-double'],
+            'quad': ['--enable-quad-precision']
+        }
+
+        # Different precisions must be configured and compiled one at a time
+        configure = Executable('../configure')
+        for precision in self.selected_precisions:
+
+            opts = (enable_precision[precision] or []) + options[:]
+
+            # SIMD optimizations are available only for float and double
+            if precision in ('float', 'double'):
+                opts += simd_options
+
+            # float-only acceleration
+            if precision == 'float':
+                for feature in float_simd_features:
+                    if feature in spec.target:
+                        msg = '--enable-{0}'
+                    else:
+                        msg = '--disable-{0}'
+                    opts.append(msg.format(feature))
+
+            with working_dir(precision, create=True):
+                configure(*opts)
diff --git a/var/spack/repos/builtin/packages/amdlibflame/aocc-2.2.0.patch b/var/spack/repos/builtin/packages/amdlibflame/aocc-2.2.0.patch
new file mode 100644
index 0000000000..95eb06dbc7
--- /dev/null
+++ b/var/spack/repos/builtin/packages/amdlibflame/aocc-2.2.0.patch
@@ -0,0 +1,12 @@
+--- a/configure	2020-09-09 18:44:01.241188553 +0530
++++ b/configure	2020-09-09 18:46:51.672191301 +0530
+@@ -6513,6 +6513,9 @@
+ 				gcc)
+ 			fla_c_openmp_flags='-fopenmp'
+ 		;;
++                 clang)
++            fla_c_openmp_flags='-fopenmp'
++        ;;
+ 				pathcc)
+ 			fla_c_openmp_flags='-mp'
+ 		;;
diff --git a/var/spack/repos/builtin/packages/amdlibflame/package.py b/var/spack/repos/builtin/packages/amdlibflame/package.py
new file mode 100644
index 0000000000..2a19f6df89
--- /dev/null
+++ b/var/spack/repos/builtin/packages/amdlibflame/package.py
@@ -0,0 +1,62 @@
+# 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.pkg.builtin.libflame import LibflameBase
+
+
+class Amdlibflame(LibflameBase):
+    """libFLAME (AMD Optimized version) is a portable library for
+    dense matrix computations, providing much of the functionality
+    present in Linear Algebra Package (LAPACK). It includes a
+    compatibility layer, FLAPACK, which includes complete LAPACK
+    implementation.
+
+    The library provides scientific and numerical computing communities
+    with a modern, high-performance dense linear algebra library that is
+    extensible, easy to use, and available under an open source
+    license. libFLAME is a C-only implementation and does not
+    depend on any external FORTRAN libraries including LAPACK.
+    There is an optional backward compatibility layer, lapack2flame
+    that maps LAPACK routine invocations to their corresponding
+    native C implementations in libFLAME. This allows legacy
+    applications to start taking advantage of libFLAME with
+    virtually no changes to their source code.
+
+    In combination with BLIS library which includes optimizations
+    for the AMD EPYC processor family, libFLAME enables running
+    high performing LAPACK functionalities on AMD platform.
+    """
+
+    _name = 'amdlibflame'
+    homepage = "http://developer.amd.com/amd-cpu-libraries/blas-library/#libflame"
+    url = "https://github.com/amd/libflame/archive/2.2.tar.gz"
+    git = "https://github.com/amd/libflame.git"
+
+    maintainers = ['amd-toolchain-support']
+
+    version('2.2', sha256='12b9c1f92d2c2fa637305aaa15cf706652406f210eaa5cbc17aaea9fcfa576dc')
+
+    patch('aocc-2.2.0.patch', when="@:2.999", level=1)
+
+    provides('flame@5.2', when='@2:')
+
+    @property
+    def lapack_libs(self):
+        """find lapack_libs function"""
+        shared = True if '+shared' in self.spec else False
+        return find_libraries(
+            'libflame', root=self.prefix, shared=shared, recursive=True
+        )
+
+    def configure_args(self):
+        """configure_args function"""
+        args = super(Amdlibflame, self).configure_args()
+        args.append("--enable-external-lapack-interfaces")
+        return args
+
+    def install(self, spec, prefix):
+        """make install function"""
+        # make install in parallel fails with message 'File already exists'
+        make("install", parallel=False)
diff --git a/var/spack/repos/builtin/packages/amdscalapack/package.py b/var/spack/repos/builtin/packages/amdscalapack/package.py
new file mode 100644
index 0000000000..270641c020
--- /dev/null
+++ b/var/spack/repos/builtin/packages/amdscalapack/package.py
@@ -0,0 +1,54 @@
+# 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 *
+from spack.pkg.builtin.netlib_scalapack import ScalapackBase
+
+
+class Amdscalapack(ScalapackBase):
+    """
+    ScaLAPACK is a library of high-performance linear algebra routines
+    for parallel distributed memory machines. It depends on external
+    libraries including BLAS and LAPACK for Linear Algebra computations.
+
+    AMD's optimized version of ScaLAPACK enables using BLIS and
+    LibFLAME library that have optimized dense matrix functions and
+    solvers for AMD EPYC processor family CPUs.
+    """
+
+    _name = 'amdscalapack'
+    homepage = "https://developer.amd.com/amd-aocl/scalapack/"
+    url = "https://github.com/amd/scalapack/archive/2.2.tar.gz"
+    git = "https://github.com/amd/scalapack.git"
+
+    maintainers = ['amd-toolchain-support']
+
+    version('2.2', sha256='2d64926864fc6d12157b86e3f88eb1a5205e7fc157bf67e7577d0f18b9a7484c')
+
+    variant(
+        'build_type',
+        default='Release',
+        description='CMake build type',
+        values=('Release', 'RelWithDebInfo'))
+
+    def cmake_args(self):
+        """ cmake_args function"""
+        args = super(Amdscalapack, self).cmake_args()
+        spec = self.spec
+
+        args.extend([
+            "-DUSE_DOTC_WRAPPER:BOOL=%s" % (
+                'ON' if '%aocc ^amdblis' in spec else 'OFF'
+            )
+        ])
+
+        args.extend([
+            '-DUSE_F2C=ON',
+            '-DLAPACK_FOUND=true',
+            '-DCMAKE_C_COMPILER=%s' % spec['mpi'].mpicc,
+            '-DCMAKE_Fortran_COMPILER=%s' % spec['mpi'].mpifc,
+        ])
+
+        return args
diff --git a/var/spack/repos/builtin/packages/fftw/package.py b/var/spack/repos/builtin/packages/fftw/package.py
index 05df011d74..f6a5e83908 100644
--- a/var/spack/repos/builtin/packages/fftw/package.py
+++ b/var/spack/repos/builtin/packages/fftw/package.py
@@ -5,35 +5,14 @@
 
 import os
 import os.path
-
 import llnl.util.lang
-
 from spack import *
 
 
-class Fftw(AutotoolsPackage):
-    """FFTW is a C subroutine library for computing the discrete Fourier
-       transform (DFT) in one or more dimensions, of arbitrary input
-       size, and of both real and complex data (as well as of even/odd
-       data, i.e. the discrete cosine/sine transforms or DCT/DST). We
-       believe that FFTW, which is free software, should become the FFT
-       library of choice for most applications."""
-
-    homepage = "http://www.fftw.org"
-    url = "http://www.fftw.org/fftw-3.3.4.tar.gz"
-    list_url = "http://www.fftw.org/download.html"
-
-    version('3.3.8', sha256='6113262f6e92c5bd474f2875fa1b01054c4ad5040f6b0da7c03c98821d9ae303')
-    version('3.3.7', sha256='3b609b7feba5230e8f6dd8d245ddbefac324c5a6ae4186947670d9ac2cd25573')
-    version('3.3.6-pl2', sha256='a5de35c5c824a78a058ca54278c706cdf3d4abba1c56b63531c2cb05f5d57da2')
-    version('3.3.5', sha256='8ecfe1b04732ec3f5b7d279fdb8efcad536d555f9d1e8fabd027037d45ea8bcf')
-    version('3.3.4', sha256='8f0cde90929bc05587c3368d2f15cd0530a60b8a9912a8e2979a72dbe5af0982')
-    version('2.1.5', sha256='f8057fae1c7df8b99116783ef3e94a6a44518d49c72e2e630c24b689c6022630')
-
-    patch('pfft-3.3.5.patch', when="@3.3.5:+pfft_patches", level=0)
-    patch('pfft-3.3.4.patch', when="@3.3.4+pfft_patches", level=0)
-    patch('pgi-3.3.6-pl2.patch', when="@3.3.6-pl2%pgi", level=0)
-    patch('intel-configure.patch', when="@3:3.3.8%intel", level=0)
+class FftwBase(AutotoolsPackage):
+    """Base class for building Fftw, shared with the AMD optimized version
+    of the library in the 'amdfftw' package.
+    """
 
     variant(
         'precision', values=any_combination_of(
@@ -43,14 +22,9 @@ class Fftw(AutotoolsPackage):
     )
     variant('openmp', default=False, description="Enable OpenMP support.")
     variant('mpi', default=True, description='Activate MPI support')
-    variant(
-        'pfft_patches', default=False,
-        description='Add extra transpose functions for PFFT compatibility')
 
     depends_on('mpi', when='+mpi')
-    depends_on('automake', type='build', when='+pfft_patches')
-    depends_on('autoconf', type='build', when='+pfft_patches')
-    depends_on('libtool', type='build', when='+pfft_patches')
+    depends_on('llvm-openmp', when='%apple-clang +openmp')
 
     # https://github.com/FFTW/fftw3/commit/902d0982522cdf6f0acd60f01f59203824e8e6f3
     conflicts('%gcc@8:8.9999', when="@3.3.7")
@@ -58,10 +32,6 @@ class Fftw(AutotoolsPackage):
               msg='Long double precision is not supported in FFTW 2')
     conflicts('precision=quad', when='@2.1.5',
               msg='Quad precision is not supported in FFTW 2')
-    conflicts('+openmp', when='%apple-clang', msg="Apple's clang does not support OpenMP")
-
-    provides('fftw-api@2', when='@2.1.5')
-    provides('fftw-api@3', when='@3:')
 
     @property
     def libs(self):
@@ -218,3 +188,39 @@ class Fftw(AutotoolsPackage):
 
     def install(self, spec, prefix):
         self.for_each_precision_make('install')
+
+
+class Fftw(FftwBase):
+    """FFTW is a C subroutine library for computing the discrete Fourier
+       transform (DFT) in one or more dimensions, of arbitrary input
+       size, and of both real and complex data (as well as of even/odd
+       data, i.e. the discrete cosine/sine transforms or DCT/DST). We
+       believe that FFTW, which is free software, should become the FFT
+       library of choice for most applications."""
+
+    homepage = "http://www.fftw.org"
+    url = "http://www.fftw.org/fftw-3.3.4.tar.gz"
+    list_url = "http://www.fftw.org/download.html"
+
+    version('3.3.8', sha256='6113262f6e92c5bd474f2875fa1b01054c4ad5040f6b0da7c03c98821d9ae303')
+    version('3.3.7', sha256='3b609b7feba5230e8f6dd8d245ddbefac324c5a6ae4186947670d9ac2cd25573')
+    version('3.3.6-pl2', sha256='a5de35c5c824a78a058ca54278c706cdf3d4abba1c56b63531c2cb05f5d57da2')
+    version('3.3.5', sha256='8ecfe1b04732ec3f5b7d279fdb8efcad536d555f9d1e8fabd027037d45ea8bcf')
+    version('3.3.4', sha256='8f0cde90929bc05587c3368d2f15cd0530a60b8a9912a8e2979a72dbe5af0982')
+    version('2.1.5', sha256='f8057fae1c7df8b99116783ef3e94a6a44518d49c72e2e630c24b689c6022630')
+
+    variant(
+        'pfft_patches', default=False,
+        description='Add extra transpose functions for PFFT compatibility')
+
+    depends_on('automake', type='build', when='+pfft_patches')
+    depends_on('autoconf', type='build', when='+pfft_patches')
+    depends_on('libtool', type='build', when='+pfft_patches')
+
+    provides('fftw-api@2', when='@2.1.5')
+    provides('fftw-api@3', when='@3:')
+
+    patch('pfft-3.3.5.patch', when="@3.3.5:+pfft_patches", level=0)
+    patch('pfft-3.3.4.patch', when="@3.3.4+pfft_patches", level=0)
+    patch('pgi-3.3.6-pl2.patch', when="@3.3.6-pl2%pgi", level=0)
+    patch('intel-configure.patch', when="@3:3.3.8%intel", level=0)
diff --git a/var/spack/repos/builtin/packages/libflame/package.py b/var/spack/repos/builtin/packages/libflame/package.py
index 7d7a9837d3..c205e54dcf 100644
--- a/var/spack/repos/builtin/packages/libflame/package.py
+++ b/var/spack/repos/builtin/packages/libflame/package.py
@@ -6,22 +6,9 @@
 from spack import *
 
 
-class Libflame(AutotoolsPackage):
-    """libflame is a portable library for dense matrix computations,
-    providing much of the functionality present in LAPACK, developed
-    by current and former members of the Science of High-Performance
-    Computing (SHPC) group in the Institute for Computational
-    Engineering and Sciences at The University of Texas at Austin.
-    libflame includes a compatibility layer, lapack2flame, which
-    includes a complete LAPACK implementation."""
-
-    homepage = "https://www.cs.utexas.edu/~flame/web/libFLAME.html"
-    url      = "https://github.com/flame/libflame/archive/5.1.0.tar.gz"
-    git      = "https://github.com/flame/libflame.git"
-
-    version('master', branch='master')
-    version('5.2.0', sha256='997c860f351a5c7aaed8deec00f502167599288fd0559c92d5bfd77d0b4d475c')
-    version('5.1.0', sha256='e7189b750890bd781fe773f366b374518dd1d89a6513d3d6261bf549826384d1')
+class LibflameBase(AutotoolsPackage):
+    """Base class for building Libflame, shared with the AMD
+    optimized version of the library in the 'libflame' package"""
 
     provides('lapack', when='+lapack2flame')
 
@@ -70,9 +57,9 @@ class Libflame(AutotoolsPackage):
             flags.append('-std=gnu99')
         return (flags, None, None)
 
-    def enable_or_disable_threads(self, variant, options):
+    def enable_or_disable_threads(self):
         opt_val = self.spec.variants['threads'].value
-        if variant_val == 'none':
+        if opt_val == 'none':
             opt_val = 'no'
         return ['--enable-multithreading={0}'.format(opt_val)]
 
@@ -102,9 +89,9 @@ class Libflame(AutotoolsPackage):
         else:
             config_args.append("--disable-debug")
 
-        config_args.extend(self.enable_or_disable('threads'))
+        config_args.extend(self.enable_or_disable_threads())
 
-        if 'none' != self.spec.variants['threads'].value:
+        if self.spec.variants['threads'].value != 'none':
             config_args.append("--enable-supermatrix")
         else:
             config_args.append("--disable-supermatrix")
@@ -119,3 +106,24 @@ class Libflame(AutotoolsPackage):
         # The shared library is not installed correctly on Darwin; fix this
         if self.spec.satisfies('platform=darwin'):
             fix_darwin_install_name(self.prefix.lib)
+
+
+class Libflame(LibflameBase):
+    """libflame is a portable library for dense matrix computations,
+    providing much of the functionality present in LAPACK, developed
+    by current and former members of the Science of High-Performance
+    Computing (SHPC) group in the Institute for Computational
+    Engineering and Sciences at The University of Texas at Austin.
+    libflame includes a compatibility layer, lapack2flame, which
+    includes a complete LAPACK implementation."""
+
+    homepage = "https://www.cs.utexas.edu/~flame/web/libFLAME.html"
+    url = "https://github.com/flame/libflame/archive/5.1.0.tar.gz"
+    git = "https://github.com/flame/libflame.git"
+
+    version('master', branch='master')
+    version('5.2.0', sha256='997c860f351a5c7aaed8deec00f502167599288fd0559c92d5bfd77d0b4d475c')
+    version('5.1.0', sha256='e7189b750890bd781fe773f366b374518dd1d89a6513d3d6261bf549826384d1')
+
+    provides('flame@5.2', when='@5.2.0')
+    provides('flame@5.1', when='@5.1.0')
diff --git a/var/spack/repos/builtin/packages/netlib-scalapack/package.py b/var/spack/repos/builtin/packages/netlib-scalapack/package.py
index 64381c5b28..11d9b9a486 100644
--- a/var/spack/repos/builtin/packages/netlib-scalapack/package.py
+++ b/var/spack/repos/builtin/packages/netlib-scalapack/package.py
@@ -7,20 +7,15 @@ from spack import *
 import sys
 
 
-class NetlibScalapack(CMakePackage):
-    """ScaLAPACK is a library of high-performance linear algebra routines for
-    parallel distributed memory machines
+class ScalapackBase(CMakePackage):
+    """Base class for building ScaLAPACK, shared with the AMD optimized version
+    of the library in the 'amdscalapack' package.
     """
-
-    homepage = "http://www.netlib.org/scalapack/"
-    url = "http://www.netlib.org/scalapack/scalapack-2.0.2.tgz"
-
-    version('2.1.0', sha256='61d9216cf81d246944720cfce96255878a3f85dec13b9351f1fa0fd6768220a6')
-    version('2.0.2', sha256='0c74aeae690fe5ee4db7926f49c5d0bb69ce09eea75beb915e00bba07530395c')
-    version('2.0.1', sha256='a9b34278d4e10b40cbe084c6d87d09af8845e874250719bfbbc497b2a88bfde1')
-    version('2.0.0', sha256='e51fbd9c3ef3a0dbd81385b868e2355900148eea689bf915c5383d72daf73114')
-    # versions before 2.0.0 are not using cmake and requires blacs as
-    # a separated package
+    variant(
+        'build_type',
+        default='Release',
+        description='CMake build type',
+        values=('Debug', 'Release', 'RelWithDebInfo', 'MinSizeRel'))
 
     variant(
         'shared',
@@ -91,3 +86,26 @@ class NetlibScalapack(CMakePackage):
         # The shared libraries are not installed correctly on Darwin:
         if (sys.platform == 'darwin') and ('+shared' in self.spec):
             fix_darwin_install_name(self.spec.prefix.lib)
+
+
+class NetlibScalapack(ScalapackBase):
+    """ScaLAPACK is a library of high-performance linear algebra routines for
+    parallel distributed memory machines
+    """
+
+    homepage = "http://www.netlib.org/scalapack/"
+    url = "http://www.netlib.org/scalapack/scalapack-2.0.2.tgz"
+
+    version('2.1.0', sha256='61d9216cf81d246944720cfce96255878a3f85dec13b9351f1fa0fd6768220a6')
+    version('2.0.2', sha256='0c74aeae690fe5ee4db7926f49c5d0bb69ce09eea75beb915e00bba07530395c')
+    version('2.0.1', sha256='a9b34278d4e10b40cbe084c6d87d09af8845e874250719bfbbc497b2a88bfde1')
+    version('2.0.0', sha256='e51fbd9c3ef3a0dbd81385b868e2355900148eea689bf915c5383d72daf73114')
+    # versions before 2.0.0 are not using cmake and requires blacs as
+    # a separated package
+
+    # See: https://github.com/Reference-ScaLAPACK/scalapack/issues/9
+    patch("cmake_fortran_mangle.patch", when='@2.0.2:2.0.99')
+    # See: https://github.com/Reference-ScaLAPACK/scalapack/pull/10
+    patch("mpi2-compatibility.patch", when='@2.0.2:2.0.99')
+    # See: https://github.com/Reference-ScaLAPACK/scalapack/pull/16
+    patch("int_overflow.patch", when='@2.0.0:2.1.0')