spack/var/spack/repos/builtin/packages/quantum-espresso/package.py
AMD Toolchain Support 1144666c09
AOCC support for QE 6.7 (#22664)
Adding qe_6.7_aocc_support
2021-03-31 09:34:21 -06:00

477 lines
21 KiB
Python

# Copyright 2013-2021 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)
class QuantumEspresso(Package):
"""Quantum ESPRESSO is an integrated suite of Open-Source computer codes
for electronic-structure calculations and materials modeling at the
nanoscale. It is based on density-functional theory, plane waves, and
pseudopotentials.
"""
homepage = 'http://quantum-espresso.org'
url = 'https://gitlab.com/QEF/q-e/-/archive/qe-6.6/q-e-qe-6.6.tar.gz'
git = 'https://gitlab.com/QEF/q-e.git'
maintainers = ['naromero77']
version('develop', branch='develop')
version('6.7', sha256='fe0ce74ff736b10d2a20c9d59025c01f88f86b00d229c123b1791f1edd7b4315',
url='https://gitlab.com/QEF/q-e/-/archive/qe-6.7MaX-Release/q-e-qe-6.7MaX-Release.tar.gz'
)
version('6.6', sha256='924656cb083f52e5d2fe71ade05881389dac64b45316f1bdd6dee1c6170a672c')
version('6.5', sha256='258b2a8a6280e86dad779e5c56356d8b35dc96d12ff33dabeee914bc03d6d602')
version('6.4.1', sha256='b0d7e9f617b848753ad923d8c6ca5490d5d82495f82b032b71a0ff2f2e9cfa08')
version('6.4', sha256='781366d03da75516fdcf9100a1caadb26ccdd1dedd942a6f8595ff0edca74bfe')
version('6.3', sha256='4067c8fffa957aabbd5cf2439e2fcb6cf3752325393c67a17d99fd09edf8689c')
version('6.2.1', sha256='11fe24b4a9d85834f8b6d429baebed8b360a685ecfae222887ed451e118a9156')
version('6.2.0', sha256='e204df367c8ea1a50c7534b44481841d835a542a23ae71c3e33ad712fc636c8b')
version('6.1.0', sha256='fd2c2eb346b3ca8f08138df5ef3f69b466c256d2119db40eea1b578b0a42c66e')
version('6.0.0', sha256='bc77d9553bf5a9253ae74058dffb1d6e5fb61093188e78d3b8d8564755136f19')
version('5.4', sha256='e3993fccae9cea04a5c6492e8b961a053a63727051cb5c4eb6008f62cda8f335')
version('5.3', sha256='3b26038efb9e3f8ac7a2b950c31d8c29169a3556c0b68c299eb88a4be8dc9048')
resource(name='environ',
git='https://github.com/environ-developers/Environ.git',
tag='v1.1',
when='@6.3:6.4.99 +environ',
destination='.'
)
resource(name='environ',
git='https://github.com/environ-developers/Environ.git',
tag='v1.0',
when='@6.2.1:6.2.99 +environ',
destination='.'
)
variant('mpi', default=True, description='Builds with mpi support')
variant('openmp', default=False, description='Enables openMP support')
variant('scalapack', default=True, description='Enables scalapack support')
variant('elpa', default=False, description='Uses elpa as an eigenvalue solver')
# Support for HDF5 has been added starting in version 6.1.0 and is
# still experimental, therefore we default to False for the variant
variant(
'hdf5', default='none', description='Builds with HDF5 support',
values=('parallel', 'serial', 'none'), multi=False
)
# Enables building Electron-phonon Wannier 'epw.x' executable
# http://epw.org.uk/Main/About
variant('epw', default=False,
description='Builds Electron-phonon Wannier executable')
# Apply upstream patches by default. Variant useful for 3rd party
# patches which are incompatible with upstream patches
desc = 'Apply recommended upstream patches. May need to be set '
desc = desc + 'to False for third party patches or plugins'
variant('patch', default=True, description=desc)
# QMCPACK converter patch
# https://github.com/QMCPACK/qmcpack/tree/develop/external_codes/quantum_espresso
variant('qmcpack', default=False,
description='Build QE-to-QMCPACK wave function converter')
variant('environ', default=False,
description='Enables support for introducing environment effects '
'into atomistic first-principles simulations.'
'See http://quantum-environ.org/about.html')
# Dependencies
depends_on('blas')
depends_on('lapack')
depends_on('fftw-api@3')
depends_on('mpi', when='+mpi')
depends_on('scalapack', when='+scalapack+mpi')
depends_on('elpa+openmp', when='+elpa+openmp')
depends_on('elpa~openmp', when='+elpa~openmp')
# Versions of HDF5 prior to 1.8.16 lead to QE runtime errors
depends_on('hdf5@1.8.16:+fortran+hl+mpi', when='hdf5=parallel')
depends_on('hdf5@1.8.16:+fortran+hl~mpi', when='hdf5=serial')
depends_on('hdf5', when='+qmcpack')
# TODO: enable building EPW when ~mpi
depends_on('mpi', when='+epw')
# CONFLICTS SECTION
# Omitted for now due to concretizer bug
# MKL with 64-bit integers not supported.
# conflicts(
# '^mkl+ilp64',
# msg='Quantum ESPRESSO does not support MKL 64-bit integer variant'
# )
# We can't ask for scalapack or elpa if we don't want MPI
conflicts(
'+scalapack',
when='~mpi',
msg='scalapack is a parallel library and needs MPI support'
)
conflicts(
'+elpa',
when='~mpi',
msg='elpa is a parallel library and needs MPI support'
)
# HDF5 support introduced in 6.1.0, but the configure had some limitations.
# In recent tests (Oct 2019), GCC and Intel work with the HDF5 Spack
# package for the default variant. This is only for hdf5=parallel variant.
# Support, for hdf5=serial was introduced in 6.4.1 but required a patch
# for the serial (no MPI) case. This patch was to work around an issue
# that only manifested itself inside the Spack environment.
conflicts(
'hdf5=parallel',
when='@:6.0',
msg='parallel HDF5 support only in QE 6.1.0 and later'
)
conflicts(
'hdf5=serial',
when='@:6.4.0',
msg='serial HDF5 support only in QE 6.4.1 and later'
)
conflicts(
'hdf5=parallel',
when='~mpi',
msg='parallel HDF5 requires MPI support'
)
# Elpa is formally supported by @:5.4.0, but QE configure searches
# for it in the wrong folders (or tries to download it within
# the build directory). Instead of patching Elpa to provide the
# folder QE expects as a link, we issue a conflict here.
conflicts('+elpa', when='@:5.4.0')
# Some QMCPACK converters are incompatible with upstream patches.
# HDF5 is a hard requirement. Need to do two HDF5 cases explicitly
# since Spack lacks support for expressing NOT operation.
conflicts(
'@6.4+patch',
when='+qmcpack',
msg='QE-to-QMCPACK wave function converter requires '
'deactivatation of upstream patches'
)
conflicts(
'@6.3:6.4.0 hdf5=serial',
when='+qmcpack',
msg='QE-to-QMCPACK wave function converter only '
'supported with parallel HDF5'
)
conflicts(
'hdf5=none',
when='+qmcpack',
msg='QE-to-QMCPACK wave function converter requires HDF5'
)
# The first version of Q-E to feature integrated EPW is 6.0.0,
# as per http://epw.org.uk/Main/DownloadAndInstall .
# Complain if trying to install a version older than this.
conflicts('+epw', when='@:5',
msg='EPW only available from version 6.0.0 and on')
# Below goes some constraints as shown in the link above.
# Constraints may be relaxed as successful reports
# of different compiler+mpi combinations arrive
# TODO: enable building EPW when ~mpi
conflicts('+epw', when='~mpi', msg='EPW needs MPI')
# EPW doesn't gets along well with OpenMPI 2.x.x
conflicts('+epw', when='^openmpi@2.0.0:2.999.999',
msg='OpenMPI version incompatible with EPW')
# EPW also doesn't gets along well with PGI 17.x + OpenMPI 1.10.7
conflicts('+epw', when='^openmpi@1.10.7%pgi@17.0:17.12',
msg='PGI+OpenMPI version combo incompatible with EPW')
# PATCHES SECTION
# THIRD-PARTY PATCHES
# NOTE: *SOME* third-party patches will require deactivation of
# upstream patches using `~patch` variant
# QMCPACK converter patches for QE 6.4.1, 6.4, and 6.3
conflicts('@:6.2,6.5:', when='+qmcpack',
msg='QMCPACK converter NOT available for this version of QE')
# Internal compiler error gcc8 and a64fx, I check only 6.5 and 6.6
conflicts('@5.3:', when='%gcc@8 target=a64fx',
msg='Internal compiler error with gcc8 and a64fx')
conflicts('@6.5:', when='+environ',
msg='6.4.x is the latest QE series supported by Environ')
# 6.4.1
patch_url = 'https://raw.githubusercontent.com/QMCPACK/qmcpack/develop/external_codes/quantum_espresso/add_pw2qmcpack_to_qe-6.4.1.diff'
patch_checksum = '57cb1b06ee2653a87c3acc0dd4f09032fcf6ce6b8cbb9677ae9ceeb6a78f85e2'
patch(patch_url, sha256=patch_checksum, when='@6.4.1+qmcpack')
# 6.4
patch_url = 'https://raw.githubusercontent.com/QMCPACK/qmcpack/develop/external_codes/quantum_espresso/add_pw2qmcpack_to_qe-6.4.diff'
patch_checksum = 'ef08f5089951be902f0854a4dbddaa7b01f08924cdb27decfade6bef0e2b8994'
patch(patch_url, sha256=patch_checksum, when='@6.4:6.4.0+qmcpack')
# 6.3
patch_url = 'https://raw.githubusercontent.com/QMCPACK/qmcpack/develop/external_codes/quantum_espresso/add_pw2qmcpack_to_qe-6.3.diff'
patch_checksum = '2ee346e24926479f5e96f8dc47812173a8847a58354bbc32cf2114af7a521c13'
patch(patch_url, sha256=patch_checksum, when='@6.3+qmcpack')
# ELPA
patch('dspev_drv_elpa.patch', when='@6.1.0:+elpa ^elpa@2016.05.004')
patch('dspev_drv_elpa.patch', when='@6.1.0:+elpa ^elpa@2016.05.003')
# QE UPSTREAM PATCHES
# QE 6.6 fix conpile error when FFT_LIBS is specified.
patch('https://gitlab.com/QEF/q-e/-/commit/cf1fedefc20d39f5cd7551ded700ea4c77ad6e8f.diff',
sha256='8f179663a8d031aff9b1820a32449942281195b6e7b1ceaab1f729651b43fa58',
when='+patch@6.6')
# QE 6.5 INTENT(OUT) without settig value in tetra_weights_only(..., ef):
# For Fujitsu compiler
patch('https://gitlab.com/QEF/q-e/-/commit/8f096b53e75026701c681c508e2c24a9378c0950.diff',
sha256='f4f1cce4182b57ac797c8f6ec8460fe375ee96385fcd8f6a61e1460bc957eb67',
when='+patch@6.5')
# QE 6.5 Fix INTENT
# For Fujitsu compiler
patch('https://gitlab.com/QEF/q-e/-/commit/c2a86201ed72693ffa50cc99b22f5d3365ae2c2b.diff',
sha256='b2dadc0bc008a3ad4b74ae85cc380dd2b63f2ae43a634e6f9d8db8077efcea6c',
when='+patch@6.5')
# QE 6.3 requires multiple patches to fix MKL detection
# There may still be problems on Mac with MKL detection
patch('https://gitlab.com/QEF/q-e/commit/0796e1b7c55c9361ecb6515a0979280e78865e36.diff',
sha256='bc8c5b8523156cee002d97dab42a5976dffae20605da485a427b902a236d7e6b',
when='+patch@6.3:6.3.0')
# QE 6.3 `make install` broken and a patch must be applied
patch('https://gitlab.com/QEF/q-e/commit/88e6558646dbbcfcafa5f3fa758217f6062ab91c.diff',
sha256='b776890d008e16cca28c31299c62f47de0ba606b900b17cbc27c041f45e564ca',
when='+patch@6.3:6.3.0')
# QE 6.4.1 patch to work around configure issues that only appear in the
# Spack environment. We now are able to support:
# `spack install qe~mpi~scalapack hdf5=serial`
patch('https://gitlab.com/QEF/q-e/commit/5fb1195b0844e1052b7601f18ab5c700f9cbe648.diff',
sha256='b1aa3179ee1c069964fb9c21f3b832aebeae54947ce8d3cc1a74e7b154c3c10f',
when='+patch@6.4.1:6.5.0')
# QE 6.4.1 Fix intent for Fujitsu compiler
patch('fj-intent.6.4.1.patch', when='+patch@6.4.1')
# QE 6.4.1 Fix intent
patch('https://gitlab.com/QEF/q-e/-/commit/c2a86201ed72693ffa50cc99b22f5d3365ae2c2b.diff',
sha256='b2dadc0bc008a3ad4b74ae85cc380dd2b63f2ae43a634e6f9d8db8077efcea6c',
when='+patch@6.4.1')
# QE 6.4.1 Small fixes for XLF compilation
patch('https://gitlab.com/QEF/q-e/-/commit/cf088926d68792cbaea48960c222e336a3965df6.diff',
sha256='bbceba1fb08d01d548d4393bbcaeae966def13f75884268a0f84448457b8eaa3',
when='+patch@6.4.1:6.5.0')
# Configure updated to work with AOCC compilers
patch('configure_aocc.patch', when='@6.7 %aocc')
# Configure updated to work with NVIDIA compilers
patch('nvhpc.patch', when='@6.5 %nvhpc')
# Configure updated to work with Fujitsu compilers
patch('fj.6.5.patch', when='@6.5+patch %fj')
patch('fj.6.6.patch', when='@6.6:6.7+patch %fj')
# extlibs_makefile updated to work with fujitsu compilers
patch('fj-fox.patch', when='+patch %fj')
# Spurious problems running in parallel the Makefile
# generated by the configure
parallel = False
def install(self, spec, prefix):
prefix_path = prefix.bin if '@:5.4.0' in spec else prefix
options = ['-prefix={0}'.format(prefix_path)]
# This additional flag is needed anytime the target architecture
# does not match the host architecture, which results in a binary that
# configure cannot execute on the login node. This is how we detect
# cross compilation: If the platform is NOT either Linux or Darwin
# and the target=backend, that we are in the cross-compile scenario
# scenario. This should cover Cray, BG/Q, and other custom platforms.
# The other option is to list out all the platform where you would be
# cross compiling explicitly.
if not (spec.satisfies('platform=linux') or
spec.satisfies('platform=darwin')):
if spec.satisfies('target=backend'):
options.append('--host')
# QE autoconf compiler variables has some limitations:
# 1. There is no explicit MPICC variable so we must re-purpose
# CC for the case of MPI.
# 2. F90 variable is set to be consistent with MPIF90 wrapper
# 3. If an absolute path for F90 is set, the build system breaks.
#
# Thus, due to 2. and 3. the F90 variable is not explictly set
# because it would be mostly pointless and could lead to erroneous
# behaviour.
if '+mpi' in spec:
mpi = spec['mpi']
options.append('--enable-parallel=yes')
options.append('MPIF90={0}'.format(mpi.mpifc))
options.append('CC={0}'.format(mpi.mpicc))
else:
options.append('--enable-parallel=no')
options.append('CC={0}'.format(env['SPACK_CC']))
options.append('F77={0}'.format(env['SPACK_F77']))
options.append('F90={0}'.format(env['SPACK_FC']))
if '+openmp' in spec:
options.append('--enable-openmp')
# QE external BLAS, FFT, SCALAPACK detection is a bit tricky.
# More predictable to pass in the correct link line to QE.
# If external detection of BLAS, LAPACK and FFT fails, QE
# is supposed to revert to internal versions of these libraries
# instead -- but more likely it will pickup versions of these
# libraries found in its the system path, e.g. Red Hat or
# Ubuntu's FFTW3 package.
# FFT
# FFT detection gets derailed if you pass into the CPPFLAGS, instead
# you need to pass it in the FFTW_INCLUDE and FFT_LIBS directory.
# QE supports an internal FFTW2, but only an external FFTW3 interface.
if '^mkl' in spec:
# A seperate FFT library is not needed when linking against MKL
options.append(
'FFTW_INCLUDE={0}'.format(join_path(env['MKLROOT'],
'include/fftw')))
if '^fftw@3:' in spec:
fftw_prefix = spec['fftw'].prefix
options.append('FFTW_INCLUDE={0}'.format(fftw_prefix.include))
fftw_ld_flags = spec['fftw'].libs.ld_flags
options.append('FFT_LIBS={0}'.format(fftw_ld_flags))
if '^amdfftw' in spec:
fftw_prefix = spec['amdfftw'].prefix
options.append('FFTW_INCLUDE={0}'.format(fftw_prefix.include))
if '+openmp' in spec:
fftw_ld_flags = spec['amdfftw:openmp'].libs.ld_flags
else:
fftw_ld_flags = spec['amdfftw'].libs.ld_flags
options.append('FFT_LIBS={0}'.format(fftw_ld_flags))
# External BLAS and LAPACK requires the correct link line into
# BLAS_LIBS, do no use LAPACK_LIBS as the autoconf scripts indicate
# that this variable is largely ignored/obsolete.
# For many Spack packages, lapack.libs = blas.libs, hence it will
# appear twice in in link line but this is harmless
lapack_blas = spec['lapack'].libs + spec['blas'].libs
# qe-6.5 fails to detect MKL for FFT if BLAS_LIBS is set due to
# an unfortunate upsteam change in their autoconf/configure:
# - qe-6.5/install/m4/x_ac_qe_blas.m4 only sets 'have_blas'
# but no 'have_mkl' if BLAS_LIBS is set (which seems to be o.k.)
# - however, qe-6.5/install/m4/x_ac_qe_fft.m4 in 6.5 unfortunately
# relies on x_ac_qe_blas.m4 to detect MKL and set 'have_mkl'
# - qe-5.4 up to 6.4.1 had a different logic and worked fine with
# BLAS_LIBS being set
# However, MKL is correctly picked up by qe-6.5 for BLAS and FFT if
# MKLROOT is set (which SPACK does automatically for ^mkl)
if not ('quantum-espresso@6.5' in spec and '^mkl' in spec):
options.append('BLAS_LIBS={0}'.format(lapack_blas.ld_flags))
if '+scalapack' in spec:
if '^mkl' in spec:
if '^openmpi' in spec:
scalapack_option = 'yes'
else: # mpich, intel-mpi
scalapack_option = 'intel'
else:
scalapack_option = 'yes'
options.append('--with-scalapack={0}'.format(scalapack_option))
scalapack_lib = spec['scalapack'].libs
options.append('SCALAPACK_LIBS={0}'.format(scalapack_lib.ld_flags))
if '+elpa' in spec:
# Spec for elpa
elpa = spec['elpa']
# Compute the include directory from there: versions
# of espresso prior to 6.1 requires -I in front of the directory
elpa_include = '' if '@6.1:' in spec else '-I'
elpa_include += join_path(
elpa.headers.directories[0],
'modules'
)
options.extend([
'--with-elpa-include={0}'.format(elpa_include),
'--with-elpa-version={0}'.format(elpa.version.version[0]),
])
elpa_suffix = '_openmp' if '+openmp' in elpa else ''
# Currently AOCC support only static libraries of ELPA
if '%aocc' in spec:
options.extend([
'--with-elpa-lib={0}'.format(
join_path(elpa.prefix.lib,
'libelpa{elpa_suffix}.a'
.format(elpa_suffix=elpa_suffix)))
])
else:
options.extend([
'--with-elpa-lib={0}'.format(elpa.libs[0])
])
if spec.variants['hdf5'].value != 'none':
options.append('--with-hdf5={0}'.format(spec['hdf5'].prefix))
if spec.satisfies('@6.4.1,6.5'):
options.extend([
'--with-hdf5-include={0}'.format(
spec['hdf5'].headers.directories[0]
),
'--with-hdf5-libs={0}'.format(
spec['hdf5:hl,fortran'].libs.ld_flags
)
])
configure(*options)
# Filter file must be applied after configure executes
# QE 6.1.0 to QE 6.4 have `-L` missing in front of zlib library
# This issue is backported through an internal patch in 6.4.1, but
# can't be applied to the '+qmcpack' variant
if spec.variants['hdf5'].value != 'none':
if (spec.satisfies('@6.1.0:6.4.0') or
(spec.satisfies('@6.4.1') and '+qmcpack' in spec)):
make_inc = join_path(self.stage.source_path, 'make.inc')
zlib_libs = spec['zlib'].prefix.lib + ' -lz'
filter_file(
zlib_libs, format(spec['zlib'].libs.ld_flags), make_inc
)
if '+epw' in spec:
make('all', 'epw')
else:
make('all')
if '+environ' in spec:
addsonpatch = Executable('./install/addsonpatch.sh')
environpatch = Executable('./Environ/patches/environpatch.sh')
makedeps = Executable('./install/makedeps.sh')
addsonpatch('Environ', 'Environ/src', 'Modules', '-patch')
environpatch('-patch')
makedeps()
make('pw')
if 'platform=darwin' in spec:
mkdirp(prefix.bin)
install('bin/*.x', prefix.bin)
else:
make('install')