Mercurial > hg
view tests/generate-working-copy-states.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 | ed46d48453e8 |
children | 2372284d9457 |
line wrap: on
line source
# Helper script used for generating history and working copy files and content. # The file's name corresponds to its history. The number of changesets can # be specified on the command line. With 2 changesets, files with names like # content1_content2_content1-untracked are generated. The first two filename # segments describe the contents in the two changesets. The third segment # ("content1-untracked") describes the state in the working copy, i.e. # the file has content "content1" and is untracked (since it was previously # tracked, it has been forgotten). # # This script generates the filenames and their content, but it's up to the # caller to tell hg about the state. # # There are two subcommands: # filelist <numchangesets> # state <numchangesets> (<changeset>|wc) # # Typical usage: # # $ python $TESTDIR/generate-working-copy-states.py state 2 1 # $ hg addremove --similarity 0 # $ hg commit -m 'first' # # $ python $TESTDIR/generate-working-copy-states.py state 2 1 # $ hg addremove --similarity 0 # $ hg commit -m 'second' # # $ python $TESTDIR/generate-working-copy-states.py state 2 wc # $ hg addremove --similarity 0 # $ hg forget *_*_*-untracked # $ rm *_*_missing-* from __future__ import absolute_import, print_function import os import sys # Generates pairs of (filename, contents), where 'contents' is a list # describing the file's content at each revision (or in the working copy). # At each revision, it is either None or the file's actual content. When not # None, it may be either new content or the same content as an earlier # revisions, so all of (modified,clean,added,removed) can be tested. def generatestates(maxchangesets, parentcontents): depth = len(parentcontents) if depth == maxchangesets + 1: for tracked in (b'untracked', b'tracked'): filename = b"_".join([(content is None and b'missing' or content) for content in parentcontents]) + b"-" + tracked yield (filename, parentcontents) else: for content in ({None, b'content' + (b"%d" % (depth + 1))} | set(parentcontents)): for combination in generatestates(maxchangesets, parentcontents + [content]): yield combination # retrieve the command line arguments target = sys.argv[1] maxchangesets = int(sys.argv[2]) if target == 'state': depth = sys.argv[3] # sort to make sure we have stable output combinations = sorted(generatestates(maxchangesets, [])) # compute file content content = [] for filename, states in combinations: if target == 'filelist': print(filename.decode('ascii')) elif target == 'state': if depth == 'wc': # Make sure there is content so the file gets written and can be # tracked. It will be deleted outside of this script. content.append((filename, states[maxchangesets] or b'TOBEDELETED')) else: content.append((filename, states[int(depth) - 1])) else: print("unknown target:", target, file=sys.stderr) sys.exit(1) # write actual content for filename, data in content: if data is not None: f = open(filename, 'wb') f.write(data + b'\n') f.close() elif os.path.exists(filename): os.remove(filename)