Mercurial > hg
view tests/test-import-eol.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 | 230eb9594150 |
children | 5abc47d4ca6b |
line wrap: on
line source
$ cat > makepatch.py <<EOF > import sys > f = open(sys.argv[2], 'wb') > w = f.write > w(b'test message\n') > w(b'diff --git a/a b/a\n') > w(b'--- a/a\n') > w(b'+++ b/a\n') > w(b'@@ -1,5 +1,5 @@\n') > w(b' a\n') > w(b'-bbb\r\n') > w(b'+yyyy\r\n') > w(b' cc\r\n') > w({'empty:lf': b' \n', > 'empty:crlf': b' \r\n', > 'empty:stripped-lf': b'\n', > 'empty:stripped-crlf': b'\r\n'}[sys.argv[1]]) > w(b' d\n') > w(b'-e\n') > w(b'\ No newline at end of file\n') > w(b'+z\r\n') > w(b'\ No newline at end of file\r\n') > EOF $ hg init repo $ cd repo $ echo '\.diff' > .hgignore Test different --eol values $ $PYTHON -c 'open("a", "wb").write(b"a\nbbb\ncc\n\nd\ne")' $ hg ci -Am adda adding .hgignore adding a $ $PYTHON ../makepatch.py empty:lf eol.diff $ $PYTHON ../makepatch.py empty:crlf eol-empty-crlf.diff $ $PYTHON ../makepatch.py empty:stripped-lf eol-empty-stripped-lf.diff $ $PYTHON ../makepatch.py empty:stripped-crlf eol-empty-stripped-crlf.diff invalid eol $ hg --config patch.eol='LFCR' import eol.diff applying eol.diff abort: unsupported line endings type: LFCR [255] $ hg revert -a force LF $ hg --traceback --config patch.eol='LF' import eol.diff applying eol.diff $ hg id 9e4ef7b3d4af tip $ cat a a yyyy cc d e (no-eol) $ hg st (test empty-line variants: all of them should generate the same revision) $ hg up -qC 0 $ hg --config patch.eol='LF' import eol-empty-crlf.diff applying eol-empty-crlf.diff $ hg id 9e4ef7b3d4af tip $ hg up -qC 0 $ hg --config patch.eol='LF' import eol-empty-stripped-lf.diff applying eol-empty-stripped-lf.diff $ hg id 9e4ef7b3d4af tip $ hg up -qC 0 $ hg --config patch.eol='LF' import eol-empty-stripped-crlf.diff applying eol-empty-stripped-crlf.diff $ hg id 9e4ef7b3d4af tip force CRLF $ hg up -C 0 1 files updated, 0 files merged, 0 files removed, 0 files unresolved $ hg --traceback --config patch.eol='CRLF' import eol.diff applying eol.diff $ cat a a\r (esc) yyyy\r (esc) cc\r (esc) \r (esc) d\r (esc) e (no-eol) $ hg st auto EOL on LF file $ hg up -C 0 1 files updated, 0 files merged, 0 files removed, 0 files unresolved $ hg --traceback --config patch.eol='auto' import eol.diff applying eol.diff $ cat a a yyyy cc d e (no-eol) $ hg st auto EOL on CRLF file $ $PYTHON -c 'open("a", "wb").write(b"a\r\nbbb\r\ncc\r\n\r\nd\r\ne")' $ hg commit -m 'switch EOLs in a' $ hg --traceback --config patch.eol='auto' import eol.diff applying eol.diff $ cat a a\r (esc) yyyy\r (esc) cc\r (esc) \r (esc) d\r (esc) e (no-eol) $ hg st auto EOL on new file or source without any EOL $ $PYTHON -c 'open("noeol", "wb").write(b"noeol")' $ hg add noeol $ hg commit -m 'add noeol' $ $PYTHON -c 'open("noeol", "wb").write(b"noeol\r\nnoeol\n")' $ $PYTHON -c 'open("neweol", "wb").write(b"neweol\nneweol\r\n")' $ hg add neweol $ hg diff --git > noeol.diff $ hg revert --no-backup noeol neweol $ rm neweol $ hg --traceback --config patch.eol='auto' import -m noeol noeol.diff applying noeol.diff $ cat noeol noeol\r (esc) noeol $ cat neweol neweol neweol\r (esc) $ hg st Test --eol and binary patches $ $PYTHON -c 'open("b", "wb").write(b"a\x00\nb\r\nd")' $ hg ci -Am addb adding b $ $PYTHON -c 'open("b", "wb").write(b"a\x00\nc\r\nd")' $ hg diff --git > bin.diff $ hg revert --no-backup b binary patch with --eol $ hg import --config patch.eol='CRLF' -m changeb bin.diff applying bin.diff $ cat b a\x00 (esc) c\r (esc) d (no-eol) $ hg st $ cd ..