Mercurial > hg
view tests/test-worker.t @ 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 | 4f0439981a8a |
children | bad59bbd9bec |
line wrap: on
line source
Test UI worker interaction $ cat > t.py <<EOF > from __future__ import absolute_import, print_function > import time > from mercurial import ( > error, > registrar, > ui as uimod, > worker, > ) > def abort(ui, args): > if args[0] == 0: > # by first worker for test stability > raise error.Abort(b'known exception') > return runme(ui, []) > def exc(ui, args): > if args[0] == 0: > # by first worker for test stability > raise Exception('unknown exception') > return runme(ui, []) > def runme(ui, args): > for arg in args: > ui.status(b'run\n') > yield 1, arg > time.sleep(0.1) # easier to trigger killworkers code path > functable = { > b'abort': abort, > b'exc': exc, > b'runme': runme, > } > cmdtable = {} > command = registrar.command(cmdtable) > @command(b'test', [], b'hg test [COST] [FUNC]') > def t(ui, repo, cost=1.0, func=b'runme'): > cost = float(cost) > func = functable[func] > ui.status(b'start\n') > runs = worker.worker(ui, cost, func, (ui,), range(8)) > for n, i in runs: > pass > ui.status(b'done\n') > EOF $ abspath=`pwd`/t.py $ hg init Run tests with worker enable by forcing a heigh cost $ hg --config "extensions.t=$abspath" test 100000.0 start run run run run run run run run done Run tests without worker by forcing a low cost $ hg --config "extensions.t=$abspath" test 0.0000001 start run run run run run run run run done #if no-windows Known exception should be caught, but printed if --traceback is enabled $ hg --config "extensions.t=$abspath" --config 'worker.numcpus=8' \ > test 100000.0 abort 2>&1 start abort: known exception [255] $ hg --config "extensions.t=$abspath" --config 'worker.numcpus=8' \ > test 100000.0 abort --traceback 2>&1 | egrep '^(SystemExit|Abort)' Abort: known exception SystemExit: 255 Traceback must be printed for unknown exceptions $ hg --config "extensions.t=$abspath" --config 'worker.numcpus=8' \ > test 100000.0 exc 2>&1 | grep '^Exception' Exception: unknown exception Workers should not do cleanups in all cases $ cat > $TESTTMP/detectcleanup.py <<EOF > from __future__ import absolute_import > import atexit > import os > import time > oldfork = os.fork > count = 0 > parentpid = os.getpid() > def delayedfork(): > global count > count += 1 > pid = oldfork() > # make it easier to test SIGTERM hitting other workers when they have > # not set up error handling yet. > if count > 1 and pid == 0: > time.sleep(0.1) > return pid > os.fork = delayedfork > def cleanup(): > if os.getpid() != parentpid: > os.write(1, 'should never happen\n') > atexit.register(cleanup) > EOF $ hg --config "extensions.t=$abspath" --config worker.numcpus=8 --config \ > "extensions.d=$TESTTMP/detectcleanup.py" test 100000 abort start abort: known exception [255] #endif