Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

enhance FFTW easyblock to support SVE CPU feature and building with Fujitsu compiler #2425

Merged
merged 2 commits into from
May 26, 2021
Merged

enhance FFTW easyblock to support SVE CPU feature and building with Fujitsu compiler #2425

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0c355c2
add sve and fujitsu support to FFTW easyblock
migueldiascosta May 14, 2021
63de44d
add fujitsu specific flags
migueldiascosta May 22, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 19 additions & 5 deletions easybuild/easyblocks/f/fftw.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,6 +33,7 @@
from easybuild.easyblocks.generic.configuremake import ConfigureMake
from easybuild.framework.easyconfig import CUSTOM
from easybuild.toolchains.compiler.gcc import TC_CONSTANT_GCC
from easybuild.toolchains.compiler.fujitsu import TC_CONSTANT_FUJITSU
from easybuild.tools.build_log import EasyBuildError
from easybuild.tools.config import build_option
from easybuild.tools.modules import get_software_version
Expand All @@ -46,7 +47,7 @@
FFTW_CPU_FEATURE_FLAGS_SINGLE_DOUBLE = ['avx', 'avx2', 'avx512', 'fma4', 'sse2', 'vsx']
# Altivec (POWER), SSE (x86), NEON (ARM), FMA (x86_64)
# asimd is CPU feature for extended NEON on AARCH64
FFTW_CPU_FEATURE_FLAGS = FFTW_CPU_FEATURE_FLAGS_SINGLE_DOUBLE + ['altivec', 'asimd', 'neon', 'sse']
FFTW_CPU_FEATURE_FLAGS = FFTW_CPU_FEATURE_FLAGS_SINGLE_DOUBLE + ['altivec', 'asimd', 'neon', 'sse', 'sve']
FFTW_PRECISION_FLAGS = ['single', 'double', 'long-double', 'quad-precision']


Expand Down Expand Up @@ -187,9 +188,14 @@ def run_all_steps(self, *args, **kwargs):
if prec == 'single' and getattr(self, flag):
prec_configopts.append('--enable-%s' % flag)

# NEON (ARM) only for single precision and double precision (on AARCH64)
if (prec == 'single' and (self.asimd or self.neon)) or (prec == 'double' and self.asimd):
prec_configopts.append('--enable-neon')
if self.sve:
# SVE (ARM) only for single precision and double precision (on AARCH64 if sve feature is present)
if prec == 'single' or prec == 'double':
prec_configopts.append('--enable-fma --enable-sve --enable-armv8-cntvct-el0')
elif self.asimd or self.neon:
# NEON (ARM) only for single precision and double precision (on AARCH64)
if prec == 'single' or (prec == 'double' and self.asimd):
prec_configopts.append('--enable-neon')

# For POWER with GCC 5/6/7 and FFTW/3.3.6 we need to disable some settings for tests to pass
# (we do it last so as not to affect previous logic)
Expand All @@ -207,6 +213,14 @@ def run_all_steps(self, *args, **kwargs):
self.log.info("Disabling vsx for double precision on POWER with GCC for FFTW/%s" % self.version)
prec_configopts.append('--disable-vsx')

# Fujitsu specific flags (from build instructions at https://github.com/fujitsu/fftw3)
if self.toolchain.comp_family() == TC_CONSTANT_FUJITSU:
prec_configopts.append('CFLAGS="-Ofast"')
prec_configopts.append('FFLAGS="-Kfast"')
prec_configopts.append('ac_cv_prog_f77_v="-###"')
if self.cfg['with_openmp']:
prec_configopts.append('OPENMP_CFLAGS="-Kopenmp"')

# append additional configure options (may be empty string, but that's OK)
self.cfg.update('configopts', [' '.join(prec_configopts) + ' ' + common_config_opts])

Expand All @@ -217,7 +231,7 @@ def run_all_steps(self, *args, **kwargs):
def test_step(self):
"""Custom implementation of test step for FFTW."""

if self.toolchain.mpi_family() == toolchain.OPENMPI:
if self.toolchain.mpi_family() == toolchain.OPENMPI and not self.toolchain.comp_family() == TC_CONSTANT_FUJITSU:

# allow oversubscription of number of processes over number of available cores with OpenMPI 3.0 & newer,
# to avoid that some tests fail if only a handful of cores are available
Expand Down