Mercurial > hg
view tests/test-context.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 | a75896bf5ccb |
children | 0bd56c291359 |
line wrap: on
line source
from __future__ import absolute_import, print_function import os import stat import sys from mercurial.node import hex from mercurial import ( context, diffutil, encoding, hg, scmutil, ui as uimod, ) print_ = print def print(*args, **kwargs): """print() wrapper that flushes stdout buffers to avoid py3 buffer issues We could also just write directly to sys.stdout.buffer the way the ui object will, but this was easier for porting the test. """ print_(*args, **kwargs) sys.stdout.flush() def printb(data, end=b'\n'): out = getattr(sys.stdout, 'buffer', sys.stdout) out.write(data + end) out.flush() u = uimod.ui.load() repo = hg.repository(u, b'test1', create=1) os.chdir('test1') # create 'foo' with fixed time stamp f = open('foo', 'wb') f.write(b'foo\n') f.close() os.utime('foo', (1000, 1000)) # add+commit 'foo' repo[None].add([b'foo']) repo.commit(text=b'commit1', date=b"0 0") d = repo[None][b'foo'].date() if os.name == 'nt': d = d[:2] print("workingfilectx.date = (%d, %d)" % d) # test memctx with non-ASCII commit message def filectxfn(repo, memctx, path): return context.memfilectx(repo, memctx, b"foo", b"") ctx = context.memctx(repo, [b'tip', None], encoding.tolocal(b"Gr\xc3\xbcezi!"), [b"foo"], filectxfn) ctx.commit() for enc in "ASCII", "Latin-1", "UTF-8": encoding.encoding = enc printb(b"%-8s: %s" % (enc.encode('ascii'), repo[b"tip"].description())) # test performing a status def getfilectx(repo, memctx, f): fctx = memctx.parents()[0][f] data, flags = fctx.data(), fctx.flags() if f == b'foo': data += b'bar\n' return context.memfilectx( repo, memctx, f, data, b'l' in flags, b'x' in flags) ctxa = repo[0] ctxb = context.memctx(repo, [ctxa.node(), None], b"test diff", [b"foo"], getfilectx, ctxa.user(), ctxa.date()) print(ctxb.status(ctxa)) # test performing a diff on a memctx diffopts = diffutil.diffallopts(repo.ui, {b'git': True}) for d in ctxb.diff(ctxa, opts=diffopts): printb(d, end=b'') # test safeness and correctness of "ctx.status()" print('= checking context.status():') # ancestor "wcctx ~ 2" actx2 = repo[b'.'] repo.wwrite(b'bar-m', b'bar-m\n', b'') repo.wwrite(b'bar-r', b'bar-r\n', b'') repo[None].add([b'bar-m', b'bar-r']) repo.commit(text=b'add bar-m, bar-r', date=b"0 0") # ancestor "wcctx ~ 1" actx1 = repo[b'.'] repo.wwrite(b'bar-m', b'bar-m bar-m\n', b'') repo.wwrite(b'bar-a', b'bar-a\n', b'') repo[None].add([b'bar-a']) repo[None].forget([b'bar-r']) # status at this point: # M bar-m # A bar-a # R bar-r # C foo from mercurial import scmutil print('== checking workingctx.status:') wctx = repo[None] print('wctx._status=%s' % (str(wctx._status))) print('=== with "pattern match":') print(actx1.status(other=wctx, match=scmutil.matchfiles(repo, [b'bar-m', b'foo']))) print('wctx._status=%s' % (str(wctx._status))) print(actx2.status(other=wctx, match=scmutil.matchfiles(repo, [b'bar-m', b'foo']))) print('wctx._status=%s' % (str(wctx._status))) print('=== with "always match" and "listclean=True":') print(actx1.status(other=wctx, listclean=True)) print('wctx._status=%s' % (str(wctx._status))) print(actx2.status(other=wctx, listclean=True)) print('wctx._status=%s' % (str(wctx._status))) print("== checking workingcommitctx.status:") wcctx = context.workingcommitctx(repo, scmutil.status([b'bar-m'], [b'bar-a'], [], [], [], [], []), text=b'', date=b'0 0') print('wcctx._status=%s' % (str(wcctx._status))) print('=== with "always match":') print(actx1.status(other=wcctx)) print('wcctx._status=%s' % (str(wcctx._status))) print(actx2.status(other=wcctx)) print('wcctx._status=%s' % (str(wcctx._status))) print('=== with "always match" and "listclean=True":') print(actx1.status(other=wcctx, listclean=True)) print('wcctx._status=%s' % (str(wcctx._status))) print(actx2.status(other=wcctx, listclean=True)) print('wcctx._status=%s' % (str(wcctx._status))) print('=== with "pattern match":') print(actx1.status(other=wcctx, match=scmutil.matchfiles(repo, [b'bar-m', b'foo']))) print('wcctx._status=%s' % (str(wcctx._status))) print(actx2.status(other=wcctx, match=scmutil.matchfiles(repo, [b'bar-m', b'foo']))) print('wcctx._status=%s' % (str(wcctx._status))) print('=== with "pattern match" and "listclean=True":') print(actx1.status(other=wcctx, match=scmutil.matchfiles(repo, [b'bar-r', b'foo']), listclean=True)) print('wcctx._status=%s' % (str(wcctx._status))) print(actx2.status(other=wcctx, match=scmutil.matchfiles(repo, [b'bar-r', b'foo']), listclean=True)) print('wcctx._status=%s' % (str(wcctx._status))) os.chdir('..') # test manifestlog being changed print('== commit with manifestlog invalidated') repo = hg.repository(u, b'test2', create=1) os.chdir('test2') # make some commits for i in [b'1', b'2', b'3']: with open(i, 'wb') as f: f.write(i) status = scmutil.status([], [i], [], [], [], [], []) ctx = context.workingcommitctx(repo, status, text=i, user=b'test@test.com', date=(0, 0)) ctx.p1().manifest() # side effect: cache manifestctx n = repo.commitctx(ctx) printb(b'commit %s: %s' % (i, hex(n))) # touch 00manifest.i mtime so storecache could expire. # repo.__dict__['manifestlog'] is deleted by transaction releasefn. st = repo.svfs.stat(b'00manifest.i') repo.svfs.utime(b'00manifest.i', (st[stat.ST_MTIME] + 1, st[stat.ST_MTIME] + 1)) # read the file just committed try: if repo[n][i].data() != i: print('data mismatch') except Exception as ex: print('cannot read data: %r' % ex) with repo.wlock(), repo.lock(), repo.transaction(b'test'): with open(b'4', 'wb') as f: f.write(b'4') repo.dirstate.normal(b'4') repo.commit(b'4') revsbefore = len(repo.changelog) repo.invalidate(clearfilecache=True) revsafter = len(repo.changelog) if revsbefore != revsafter: print('changeset lost by repo.invalidate()')