Mercurial > hg
view mercurial/statichttprepo.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 | 0664be4f0c1f |
children | 089fc0db0954 |
line wrap: on
line source
# statichttprepo.py - simple http repository class for mercurial # # This provides read-only repo access to repositories exported via static http # # Copyright 2005-2007 Matt Mackall <mpm@selenic.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 errno from .i18n import _ from . import ( changelog, error, localrepo, manifest, namespaces, pathutil, scmutil, store, url, util, vfs as vfsmod, ) urlerr = util.urlerr urlreq = util.urlreq class httprangereader(object): def __init__(self, url, opener): # we assume opener has HTTPRangeHandler self.url = url self.pos = 0 self.opener = opener self.name = url def __enter__(self): return self def __exit__(self, exc_type, exc_value, traceback): self.close() def seek(self, pos): self.pos = pos def read(self, bytes=None): req = urlreq.request(self.url) end = '' if bytes: end = self.pos + bytes - 1 if self.pos or end: req.add_header('Range', 'bytes=%d-%s' % (self.pos, end)) try: f = self.opener.open(req) data = f.read() code = f.code except urlerr.httperror as inst: num = inst.code == 404 and errno.ENOENT or None raise IOError(num, inst) except urlerr.urlerror as inst: raise IOError(None, inst.reason[1]) if code == 200: # HTTPRangeHandler does nothing if remote does not support # Range headers and returns the full entity. Let's slice it. if bytes: data = data[self.pos:self.pos + bytes] else: data = data[self.pos:] elif bytes: data = data[:bytes] self.pos += len(data) return data def readlines(self): return self.read().splitlines(True) def __iter__(self): return iter(self.readlines()) def close(self): pass # _RangeError and _HTTPRangeHandler were originally in byterange.py, # which was itself extracted from urlgrabber. See the last version of # byterange.py from history if you need more information. class _RangeError(IOError): """Error raised when an unsatisfiable range is requested.""" class _HTTPRangeHandler(urlreq.basehandler): """Handler that enables HTTP Range headers. This was extremely simple. The Range header is a HTTP feature to begin with so all this class does is tell urllib2 that the "206 Partial Content" response from the HTTP server is what we expected. """ def http_error_206(self, req, fp, code, msg, hdrs): # 206 Partial Content Response r = urlreq.addinfourl(fp, hdrs, req.get_full_url()) r.code = code r.msg = msg return r def http_error_416(self, req, fp, code, msg, hdrs): # HTTP's Range Not Satisfiable error raise _RangeError('Requested Range Not Satisfiable') def build_opener(ui, authinfo): # urllib cannot handle URLs with embedded user or passwd urlopener = url.opener(ui, authinfo) urlopener.add_handler(_HTTPRangeHandler()) class statichttpvfs(vfsmod.abstractvfs): def __init__(self, base): self.base = base def __call__(self, path, mode='r', *args, **kw): if mode not in ('r', 'rb'): raise IOError('Permission denied') f = "/".join((self.base, urlreq.quote(path))) return httprangereader(f, urlopener) def join(self, path): if path: return pathutil.join(self.base, path) else: return self.base return statichttpvfs class statichttppeer(localrepo.localpeer): def local(self): return None def canpush(self): return False class statichttprepository(localrepo.localrepository): supported = localrepo.localrepository._basesupported def __init__(self, ui, path): self._url = path self.ui = ui self.root = path u = util.url(path.rstrip('/') + "/.hg") self.path, authinfo = u.authinfo() vfsclass = build_opener(ui, authinfo) self.vfs = vfsclass(self.path) self.cachevfs = vfsclass(self.vfs.join('cache')) self._phasedefaults = [] self.names = namespaces.namespaces() self.filtername = None try: requirements = scmutil.readrequires(self.vfs, self.supported) except IOError as inst: if inst.errno != errno.ENOENT: raise requirements = set() # check if it is a non-empty old-style repository try: fp = self.vfs("00changelog.i") fp.read(1) fp.close() except IOError as inst: if inst.errno != errno.ENOENT: raise # we do not care about empty old-style repositories here msg = _("'%s' does not appear to be an hg repository") % path raise error.RepoError(msg) # setup store self.store = store.store(requirements, self.path, vfsclass) self.spath = self.store.path self.svfs = self.store.opener self.sjoin = self.store.join self._filecache = {} self.requirements = requirements self.manifestlog = manifest.manifestlog(self.svfs, self) self.changelog = changelog.changelog(self.svfs) self._tags = None self.nodetagscache = None self._branchcaches = {} self._revbranchcache = None self.encodepats = None self.decodepats = None self._transref = None def _restrictcapabilities(self, caps): caps = super(statichttprepository, self)._restrictcapabilities(caps) return caps.difference(["pushkey"]) def url(self): return self._url def local(self): return False def peer(self): return statichttppeer(self) def wlock(self, wait=True): raise error.LockUnavailable(0, _('lock not available'), 'lock', _('cannot lock static-http repository')) def lock(self, wait=True): raise error.Abort(_('cannot lock static-http repository')) def _writecaches(self): pass # statichttprepository are read only def instance(ui, path, create, intents=None): if create: raise error.Abort(_('cannot create new static-http repository')) return statichttprepository(ui, path[7:])