# Copyright (C) 2015 Joshua Willis
#
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 3 of the License, or (at your
# option) any later version.
#
# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
# Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
"""
This module defines optimization flags and determines hardware features that some
other modules and packages may use in addition to some optimized utilities.
"""
import os, sys
import logging
from collections import OrderedDict
import pycbc
# Work around different Python versions to get runtime
# info on hardware cache sizes
_USE_SUBPROCESS = False
HAVE_GETCONF = False
if os.environ.get("LEVEL2_CACHE_SIZE", None) or os.environ.get("NO_GETCONF", None):
HAVE_GETCONF = False
elif sys.platform == 'darwin':
# Mac has getconf, but we can do nothing useful with it
HAVE_GETCONF = False
elif sys.version_info >= (2, 7):
import subprocess
_USE_SUBPROCESS = True
HAVE_GETCONF = True
else:
try:
import commands # Only available on Unix?
HAVE_GETCONF = True
except:
pass
if pycbc.HAVE_OMP:
omp_support = """
#include <omp.h>
"""
omp_libs = ['gomp']
omp_flags = ['-fopenmp']
else:
omp_support = ""
omp_libs = []
omp_flags = []
# The following are intended to also be used by weave
# functions. It would be good to figure out a way to
# test for the presence of <x86intrin.h>, and make
# this conditional, or else define something to
# flag to other codes that it cannot be used.
#
# The constant ALGN is how much to align in bytes, and
# ALGN_FLT and ALGN_DBL are that value as a number of
# floats and doubles, respectively.
simd_intel_intrin_support = """
#include <x86intrin.h>
#ifdef __AVX2__
#define _HAVE_AVX2 1
#else
#define _HAVE_AVX2 0
#endif
#ifdef __AVX__
#define _HAVE_AVX 1
#else
#define _HAVE_AVX 0
#endif
#ifdef __SSE4_1__
#define _HAVE_SSE4_1 1
#else
#define _HAVE_SSE4_1 0
#endif
#ifdef __SSE3__
#define _HAVE_SSE3 1
#else
#define _HAVE_SSE3 0
#endif
#if _HAVE_AVX
#define ALGN 32
#define ALGN_FLT 8
#define ALGN_DBL 4
#else
#define ALGN 16
#define ALGN_FLT 4
#define ALGN_DBL 2
#endif
"""
if os.environ.get("LEVEL2_CACHE_SIZE", None):
LEVEL2_CACHE_SIZE = int(os.environ["LEVEL2_CACHE_SIZE"])
logging.info("opt: using LEVEL2_CACHE_SIZE %d from environment" % LEVEL2_CACHE_SIZE)
elif HAVE_GETCONF:
if _USE_SUBPROCESS:
def getconf(confvar):
return int(subprocess.check_output(['getconf', confvar]))
else:
[docs] def getconf(confvar):
retlist = commands.getstatusoutput('getconf ' + confvar)
return int(retlist[1])
LEVEL1_DCACHE_SIZE = getconf('LEVEL1_DCACHE_SIZE')
LEVEL1_DCACHE_ASSOC = getconf('LEVEL1_DCACHE_ASSOC')
LEVEL1_DCACHE_LINESIZE = getconf('LEVEL1_DCACHE_LINESIZE')
LEVEL2_CACHE_SIZE = getconf('LEVEL2_CACHE_SIZE')
LEVEL2_CACHE_ASSOC = getconf('LEVEL2_CACHE_ASSOC')
LEVEL2_CACHE_LINESIZE = getconf('LEVEL2_CACHE_LINESIZE')
LEVEL3_CACHE_SIZE = getconf('LEVEL3_CACHE_SIZE')
LEVEL3_CACHE_ASSOC = getconf('LEVEL3_CACHE_ASSOC')
LEVEL3_CACHE_LINESIZE = getconf('LEVEL3_CACHE_LINESIZE')
[docs]def insert_optimization_option_group(parser):
"""
Adds the options used to specify optimization-specific options.
Parameters
----------
parser : object
OptionParser instance
"""
optimization_group = parser.add_argument_group("Options for selecting "
"optimization-specific settings")
optimization_group.add_argument("--cpu-affinity", help="""
A set of CPUs on which to run, specified in a format suitable
to pass to taskset.""")
optimization_group.add_argument("--cpu-affinity-from-env", help="""
The name of an enivornment variable containing a set
of CPUs on which to run, specified in a format suitable
to pass to taskset.""")
[docs]def verify_optimization_options(opt, parser):
"""Parses the CLI options, verifies that they are consistent and
reasonable, and acts on them if they are
Parameters
----------
opt : object
Result of parsing the CLI with OptionParser, or any object with the
required attributes
parser : object
OptionParser instance.
"""
# Pin to specified CPUs if requested
requested_cpus = None
if opt.cpu_affinity_from_env is not None:
if opt.cpu_affinity is not None:
logging.error("Both --cpu_affinity_from_env and --cpu_affinity specified")
sys.exit(1)
requested_cpus = os.environ.get(opt.cpu_affinity_from_env)
if requested_cpus is None:
logging.error("CPU affinity requested from environment variable %s "
"but this variable is not defined" % opt.cpu_affinity_from_env)
sys.exit(1)
if requested_cpus == '':
logging.error("CPU affinity requested from environment variable %s "
"but this variable is empty" % opt.cpu_affinity_from_env)
sys.exit(1)
if requested_cpus is None:
requested_cpus = opt.cpu_affinity
if requested_cpus is not None:
command = 'taskset -pc %s %d' % (requested_cpus, os.getpid())
retcode = os.system(command)
if retcode != 0:
logging.error('taskset command <%s> failed with return code %d' % \
(command, retcode))
sys.exit(1)
logging.info("Pinned to CPUs %s " % requested_cpus)
[docs]class LimitedSizeDict(OrderedDict):
""" Fixed sized dict for FIFO caching"""
def __init__(self, *args, **kwds):
self.size_limit = kwds.pop("size_limit", None)
OrderedDict.__init__(self, *args, **kwds)
self._check_size_limit()
def __setitem__(self, key, value):
OrderedDict.__setitem__(self, key, value)
self._check_size_limit()
def _check_size_limit(self):
if self.size_limit is not None:
while len(self) > self.size_limit:
self.popitem(last=False)