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
worker: ability to disable thread unsafe tasks
The worker on Windows is implemented using a thread pool. If worker
tasks are not thread safe, badness can occur. In addition, if tasks
are executing CPU bound code and holding onto the GIL, there will be
non-substantial overhead in Python context switching between active
threads. This can result in significant slowdowns of tasks.
This commit teaches the code for determining whether to use a worker
to take thread safety into account. Effectively, thread unsafe tasks
don't use the thread-based worker on Windows.
Differential Revision: https://phab.mercurial-scm.org/D3962
worker: rename variable to reflect constant
Differential Revision: https://phab.mercurial-scm.org/D3961
worker: use one pipe per posix worker and select() in parent process
This allows us to pass results larger than PIPE_BUF through the pipes without
interleaving them. This is necessary now because "hg fix" sends file contents
as the result from workers.
Differential Revision: https://phab.mercurial-scm.org/D3960
tests: refactor common bundle2 capabilities
Differential Revision: https://phab.mercurial-scm.org/D1945
debug: move extensions debug behind a dedicated flag
Since
b86664c81833, we process the `--debug` flag earlier. This is overall
good and useful, but has at least one negative side effect.
Previously the debug message we report when trying to import extensions were
issued before we processed the `--debug` flag. Now they happen after.
Before:
$ ./hg id --debug
21f507b8de2f9c1606e9aeb5ec7d2a6dedb7a4a7 tip
After:
$ ./hg id --debug ☿ (revset-bench)
could not import hgext.evolve (No module named evolve): trying hgext3rd.evolve
could not import hgext.mercurial_keyring (No module named mercurial_keyring): trying hgext3rd.mercurial_keyring
could not import hgext3rd.mercurial_keyring (No module named mercurial_keyring): trying mercurial_keyring
could not import hgext.hggit (No module named hggit): trying hgext3rd.hggit
could not import hgext3rd.hggit (No module named hggit): trying hggit
21f507b8de2f9c1606e9aeb5ec7d2a6dedb7a4a7 tip
(This get worse if --traceback is used).
To work around this, we move this extensions related debug message behind a
new flag 'devel.debug.extensions' and restore the previous output.
I'm not fully happy about using the 'devel' section for a flag that can be
used by legitimate users to debug extensions issues. However, it fits well
next to other `devel.devel.*` options and is mostly used by extensions author
anyway.
We might move it to another, more appropriate section in the future (using
alias).
curses: use "text" interface if TERM=dumb
Differential Revision: https://phab.mercurial-scm.org/D3948