Mercurial > hg
view mercurial/policy.py @ 38732:be4984261611
merge: mark file gets as not thread safe (issue5933)
In default installs, this has the effect of disabling the thread-based
worker on Windows when manifesting files in the working directory. My
measurements have shown that with revlog-based repositories, Mercurial
spends a lot of CPU time in revlog code resolving file data. This ends
up incurring a lot of context switching across threads and slows down
`hg update` operations when going from an empty working directory to
the tip of the repo.
On mozilla-unified (246,351 files) on an i7-6700K (4+4 CPUs):
before: 487s wall
after: 360s wall (equivalent to worker.enabled=false)
cpus=2: 379s wall
Even with only 2 threads, the thread pool is still slower.
The introduction of the thread-based worker (02b36e860e0b) states that
it resulted in a "~50%" speedup for `hg sparse --enable-profile` and
`hg sparse --disable-profile`. This disagrees with my measurement
above. I theorize a few reasons for this:
1) Removal of files from the working directory is I/O - not CPU - bound
and should benefit from a thread pool (unless I/O is insanely fast
and the GIL release is near instantaneous). So tests like `hg sparse
--enable-profile` may exercise deletion throughput and aren't good
benchmarks for worker tasks that are CPU heavy.
2) The patch was authored by someone at Facebook. The results were
likely measured against a repository using remotefilelog. And I
believe that revision retrieval during working directory updates with
remotefilelog will often use a remote store, thus being I/O and not
CPU bound. This probably resulted in an overstated performance gain.
Since there appears to be a need to enable the thread-based worker with
some stores, I've made the flagging of file gets as thread safe
configurable. I've made it experimental because I don't want to formalize
a boolean flag for this option and because this attribute is best
captured against the store implementation. But we don't have a proper
store API for this yet. I'd rather cross this bridge later.
It is possible there are revlog-based repositories that do benefit from
a thread-based worker. I didn't do very comprehensive testing. If there
are, we may want to devise a more proper algorithm for whether to use
the thread-based worker, including possibly config options to limit the
number of threads to use. But until I see evidence that justifies
complexity, simplicity wins.
Differential Revision: https://phab.mercurial-scm.org/D3963
author | Gregory Szorc <gregory.szorc@gmail.com> |
---|---|
date | Wed, 18 Jul 2018 09:49:34 -0700 |
parents | 0304f22497fa |
children | 6104b203bec8 adacefb0b7ea |
line wrap: on
line source
# policy.py - module policy logic for Mercurial. # # Copyright 2015 Gregory Szorc <gregory.szorc@gmail.com> # # This software may be used and distributed according to the terms of the # GNU General Public License version 2 or any later version. from __future__ import absolute_import import os import sys # Rules for how modules can be loaded. Values are: # # c - require C extensions # allow - allow pure Python implementation when C loading fails # cffi - required cffi versions (implemented within pure module) # cffi-allow - allow pure Python implementation if cffi version is missing # py - only load pure Python modules # # By default, fall back to the pure modules so the in-place build can # run without recompiling the C extensions. This will be overridden by # __modulepolicy__ generated by setup.py. policy = b'allow' _packageprefs = { # policy: (versioned package, pure package) b'c': (r'cext', None), b'allow': (r'cext', r'pure'), b'cffi': (r'cffi', None), b'cffi-allow': (r'cffi', r'pure'), b'py': (None, r'pure'), } try: from . import __modulepolicy__ policy = __modulepolicy__.modulepolicy except ImportError: pass # PyPy doesn't load C extensions. # # The canonical way to do this is to test platform.python_implementation(). # But we don't import platform and don't bloat for it here. if r'__pypy__' in sys.builtin_module_names: policy = b'cffi' # Environment variable can always force settings. if sys.version_info[0] >= 3: if r'HGMODULEPOLICY' in os.environ: policy = os.environ[r'HGMODULEPOLICY'].encode(r'utf-8') else: policy = os.environ.get(r'HGMODULEPOLICY', policy) def _importfrom(pkgname, modname): # from .<pkgname> import <modname> (where . is looked through this module) fakelocals = {} pkg = __import__(pkgname, globals(), fakelocals, [modname], level=1) try: fakelocals[modname] = mod = getattr(pkg, modname) except AttributeError: raise ImportError(r'cannot import name %s' % modname) # force import; fakelocals[modname] may be replaced with the real module getattr(mod, r'__doc__', None) return fakelocals[modname] # keep in sync with "version" in C modules _cextversions = { (r'cext', r'base85'): 1, (r'cext', r'bdiff'): 3, (r'cext', r'mpatch'): 1, (r'cext', r'osutil'): 4, (r'cext', r'parsers'): 5, } # map import request to other package or module _modredirects = { (r'cext', r'charencode'): (r'cext', r'parsers'), (r'cffi', r'base85'): (r'pure', r'base85'), (r'cffi', r'charencode'): (r'pure', r'charencode'), (r'cffi', r'parsers'): (r'pure', r'parsers'), } def _checkmod(pkgname, modname, mod): expected = _cextversions.get((pkgname, modname)) actual = getattr(mod, r'version', None) if actual != expected: raise ImportError(r'cannot import module %s.%s ' r'(expected version: %d, actual: %r)' % (pkgname, modname, expected, actual)) def importmod(modname): """Import module according to policy and check API version""" try: verpkg, purepkg = _packageprefs[policy] except KeyError: raise ImportError(r'invalid HGMODULEPOLICY %r' % policy) assert verpkg or purepkg if verpkg: pn, mn = _modredirects.get((verpkg, modname), (verpkg, modname)) try: mod = _importfrom(pn, mn) if pn == verpkg: _checkmod(pn, mn, mod) return mod except ImportError: if not purepkg: raise pn, mn = _modredirects.get((purepkg, modname), (purepkg, modname)) return _importfrom(pn, mn)