bundle2: increase payload part chunk size to 32kb
Bundle2 payload parts are framed chunks. Esentially, we obtain
data in equal size chunks of size `preferedchunksize` and emit those
to a generator. That generator is fed into a compressor (which can
be the no-op compressor, which just re-emits the generator). And
the output from the compressor likely goes to a file descriptor
or socket.
What this means is that small chunk sizes create more Python objects
and Python function calls than larger chunk sizes. And as we know,
Python object and function call overhead in performance sensitive
code matters (at least with CPython).
This commit increases the bundle2 part payload chunk size from 4k
to 32k. Practically speaking, this means that the chunks we feed
into a compressor (implemented in C code) or feed directly into a
file handle or socket write() are larger. It's possible the chunks
might be larger than what the receiver can handle in one logical
operation. But at that point, we're in C code, which is much more
efficient at dealing with splitting up the chunk and making multiple
function calls than Python is.
A downside to larger chunks is that the receiver has to wait for that
much data to arrive (either raw or from a decompressor) before it
can process the chunk. But 32kb still feels like a small buffer to
have to wait for. And in many cases, the client will convert from
8 read(4096) to 1 read(32768). That's happening in Python land. So
we cut down on the number of Python objects and function calls,
making the client faster as well. I don't think there are any
significant concerns to increasing the payload chunk size to 32kb.
The impact of this change on performance significant. Using `curl`
to obtain a stream clone bundle2 payload from a server on localhost
serving the mozilla-unified repository:
before: 20.78 user; 7.71 system; 80.5 MB/s
after: 13.90 user; 3.51 system; 132 MB/s
legacy: 9.72 user; 8.16 system; 132 MB/s
bundle2 stream clone generation is still more resource intensive than
legacy stream clone (that's likely because of the use of a
util.chunkbuffer). But the throughput is the same. We might
be in territory we're this is effectively a benchmark of the
networking stack or Python's syscall throughput.
From the client perspective, `hg clone -U --stream`:
before: 33.50 user; 7.95 system; 53.3 MB/s
after: 22.82 user; 7.33 system; 72.7 MB/s
legacy: 29.96 user; 7.94 system; 58.0 MB/s
And for `hg clone --stream` with a working directory update of
~230k files:
after: 119.55 user; 26.47 system; 0:57.08 wall
legacy: 126.98 user; 26.94 system; 1:05.56 wall
So, it appears that bundle2's stream clone is now definitively faster
than legacy stream clone!
Differential Revision: https://phab.mercurial-scm.org/D1932
$ cat <<EOF >> $HGRCPATH
> [extensions]
> mq =
> [mq]
> git = keep
> [diff]
> nodates = 1
> EOF
init:
$ hg init repo
$ cd repo
$ echo a > a
$ hg ci -Am adda
adding a
$ echo a >> a
$ hg qnew -f p1
$ echo b >> a
$ hg qnew -f p2
$ echo c >> a
$ hg qnew -f p3
Fold in the middle of the queue:
(this tests also that editor is not invoked if '--edit' is not
specified)
$ hg qpop p1
popping p3
popping p2
now at: p1
$ hg qdiff
diff -r 07f494440405 a
--- a/a
+++ b/a
@@ -1,1 +1,2 @@
a
+a
$ HGEDITOR=cat hg qfold p2
$ grep git .hg/patches/p1 && echo 'git patch found!'
[1]
$ hg qser
p1
p3
$ hg qdiff
diff -r 07f494440405 a
--- a/a
+++ b/a
@@ -1,1 +1,3 @@
a
+a
+b
Fold with local changes:
$ echo d >> a
$ hg qfold p3
abort: local changes found, qrefresh first
[255]
$ hg diff -c .
diff -r 07f494440405 -r ???????????? a (glob)
--- a/a
+++ b/a
@@ -1,1 +1,3 @@
a
+a
+b
$ hg revert -a --no-backup
reverting a
Fold git patch into a regular patch, expect git patch:
$ echo a >> a
$ hg qnew -f regular
$ hg cp a aa
$ hg qnew --git -f git
$ hg qpop
popping git
now at: regular
$ hg qfold git
$ cat .hg/patches/regular
# HG changeset patch
# Parent ???????????????????????????????????????? (glob)
diff --git a/a b/a
--- a/a
+++ b/a
@@ -1,3 +1,4 @@
a
a
b
+a
diff --git a/a b/aa
copy from a
copy to aa
--- a/a
+++ b/aa
@@ -1,3 +1,4 @@
a
a
b
+a
$ hg qpop
popping regular
now at: p1
$ hg qdel regular
Fold regular patch into a git patch, expect git patch:
$ hg cp a aa
$ hg qnew --git -f git
$ echo b >> aa
$ hg qnew -f regular
$ hg qpop
popping regular
now at: git
$ hg qfold regular
$ cat .hg/patches/git
# HG changeset patch
# Parent ???????????????????????????????????????? (glob)
diff --git a/a b/aa
copy from a
copy to aa
--- a/a
+++ b/aa
@@ -1,3 +1,4 @@
a
a
b
+b
Test saving last-message.txt:
$ hg qrefresh -m "original message"
$ cat > $TESTTMP/commitfailure.py <<EOF
> from mercurial import error
> def reposetup(ui, repo):
> class commitfailure(repo.__class__):
> def commit(self, *args, **kwargs):
> raise error.Abort('emulating unexpected abort')
> repo.__class__ = commitfailure
> EOF
$ cat >> .hg/hgrc <<EOF
> [extensions]
> # this failure occurs before editor invocation
> commitfailure = $TESTTMP/commitfailure.py
> EOF
$ cat > $TESTTMP/editor.sh << EOF
> echo "==== before editing"
> cat \$1
> echo "===="
> (echo; echo "test saving last-message.txt") >> \$1
> EOF
$ hg qapplied
p1
git
$ hg tip --template "{files}\n"
aa
(test that editor is not invoked before transaction starting,
and that combination of '--edit' and '--message' doesn't abort execution)
$ rm -f .hg/last-message.txt
$ HGEDITOR="sh $TESTTMP/editor.sh" hg qfold -e -m MESSAGE p3
qrefresh interrupted while patch was popped! (revert --all, qpush to recover)
abort: emulating unexpected abort
[255]
$ test -f .hg/last-message.txt
[1]
(reset applied patches and directory status)
$ cat >> .hg/hgrc <<EOF
> [extensions]
> # this failure occurs after editor invocation
> commitfailure = !
> EOF
$ hg qapplied
p1
$ hg status -A aa
? aa
$ rm aa
$ hg status -m
M a
$ hg revert --no-backup -q a
$ hg qpush -q git
now at: git
(test that editor is invoked and commit message is saved into
"last-message.txt")
$ cat >> .hg/hgrc <<EOF
> [hooks]
> # this failure occurs after editor invocation
> pretxncommit.unexpectedabort = false
> EOF
$ rm -f .hg/last-message.txt
$ HGEDITOR="sh $TESTTMP/editor.sh" hg qfold -e p3
==== before editing
original message
HG: Enter commit message. Lines beginning with 'HG:' are removed.
HG: Leave message empty to use default message.
HG: --
HG: user: test
HG: branch 'default'
HG: added aa
HG: changed a
====
note: commit message saved in .hg/last-message.txt
transaction abort!
rollback completed
qrefresh interrupted while patch was popped! (revert --all, qpush to recover)
abort: pretxncommit.unexpectedabort hook exited with status 1
[255]
$ cat .hg/last-message.txt
original message
test saving last-message.txt
(confirm whether files listed up in the commit message editing are correct)
$ cat >> .hg/hgrc <<EOF
> [hooks]
> pretxncommit.unexpectedabort =
> EOF
$ hg status -u | while read f; do rm ${f}; done
$ hg revert --no-backup -q --all
$ hg qpush -q git
now at: git
$ hg qpush -q --move p3
now at: p3
$ hg status --rev "git^1" --rev . -arm
M a
A aa
$ cd ..