view mercurial/lsprof.py @ 42043:1fac9b931d46

compression: introduce a `storage.revlog.zlib.level` configuration

This option control the zlib compression level used when compression revlog
chunk.

This is also a good excuse to pave the way for a similar configuration option
for the zstd compression engine. Having a dedicated option for each compression
algorithm is useful because they don't support the same range of values.

Using a higher zlib compression impact CPU consumption at compression time, but
does not directly affected decompression time. However dealing with small
compressed chunk can directly help decompression and indirectly help other
revlog logic.

I ran some basic test on repositories using different level. I am using the
mercurial, pypy, netbeans and mozilla-central clone from our benchmark suite.
All tested repository use sparse-revlog and got all their delta recomputed.

The different compression level has a small effect on the repository size
(about 10% variation in the total range). My quick analysis is that revlog
mostly store small delta, that are not affected by the compression level much.
So the variation probably mostly comes from better compression of the snapshots
revisions, and snapshot revision only represent a small portion of the
repository content.

I also made some basic timings measurements. The "read" timings are gathered using
simple run of `hg perfrevlogrevisions`, the "write" timings using `hg
perfrevlogwrite` (restricted to the last 5000 revisions for netbeans and
mozilla central). The timings are gathered on a generic machine, (not one  of
our performance locked machine), so small variation might not be meaningful.
However large trend remains relevant.

Keep in mind that these numbers are not pure compression/decompression time.
They also involve the full revlog logic. In particular the difference in chunk
size has an impact on the delta chain structure, affecting performance when
writing or reading them.

On read/write performance, the compression level has a bigger impact.
Counter-intuitively, the higher compression levels improve "write" performance
for the large repositories in our tested setting. Maybe because the last 5000
delta chain end up having a very different shape in this specific spot? Or maybe
because of a more general trend of better delta chains thanks to the smaller
chunk and snapshot.

This series does not intend to change the default compression level. However,
these result call for a deeper analysis of this performance difference in the
future.

Full data
=========

repo   level  .hg/store size  00manifest.d read       write
----------------------------------------------------------------
mercurial  1      49,402,813     5,963,475   0.170159  53.250304
mercurial  6      47,197,397     5,875,730   0.182820  56.264320
mercurial  9      47,121,596     5,849,781   0.189219  56.293612

pypy       1     370,830,572    28,462,425   2.679217 460.721984
pypy       6     340,112,317    27,648,747   2.768691 467.537158
pypy       9     338,360,736    27,639,003   2.763495 476.589918

netbeans   1   1,281,847,810   165,495,457 122.477027 520.560316
netbeans   6   1,205,284,353   159,161,207 139.876147 715.930400
netbeans   9   1,197,135,671   155,034,586 141.620281 678.297064

mozilla    1   2,775,497,186   298,527,987 147.867662 751.263721
mozilla    6   2,596,856,420   286,597,671 170.572118 987.056093
mozilla    9   2,587,542,494   287,018,264 163.622338 739.803002

from __future__ import absolute_import, print_function

import _lsprof
import sys

Profiler = _lsprof.Profiler

# PyPy doesn't expose profiler_entry from the module.
profiler_entry = getattr(_lsprof, 'profiler_entry', None)

__all__ = ['profile', 'Stats']

def profile(f, *args, **kwds):
    """XXX docstring"""
    p = Profiler()
    p.enable(subcalls=True, builtins=True)
    try:
        f(*args, **kwds)
    finally:
        p.disable()
    return Stats(p.getstats())


class Stats(object):
    """XXX docstring"""

    def __init__(self, data):
        self.data = data

    def sort(self, crit=r"inlinetime"):
        """XXX docstring"""
        # profiler_entries isn't defined when running under PyPy.
        if profiler_entry:
            if crit not in profiler_entry.__dict__:
                raise ValueError("Can't sort by %s" % crit)
        elif self.data and not getattr(self.data[0], crit, None):
            raise ValueError("Can't sort by %s" % crit)

        self.data.sort(key=lambda x: getattr(x, crit), reverse=True)
        for e in self.data:
            if e.calls:
                e.calls.sort(key=lambda x: getattr(x, crit), reverse=True)

    def pprint(self, top=None, file=None, limit=None, climit=None):
        """XXX docstring"""
        if file is None:
            file = sys.stdout
        d = self.data
        if top is not None:
            d = d[:top]
        cols = "% 12d %12d %11.4f %11.4f   %s\n"
        hcols = "% 12s %12s %12s %12s %s\n"
        file.write(hcols % ("CallCount", "Recursive", "Total(s)",
                            "Inline(s)", "module:lineno(function)"))
        count = 0
        for e in d:
            file.write(cols % (e.callcount, e.reccallcount, e.totaltime,
                               e.inlinetime, label(e.code)))
            count += 1
            if limit is not None and count == limit:
                return
            ccount = 0
            if climit and e.calls:
                for se in e.calls:
                    file.write(cols % (se.callcount, se.reccallcount,
                                       se.totaltime, se.inlinetime,
                                       "    %s" % label(se.code)))
                    count += 1
                    ccount += 1
                    if limit is not None and count == limit:
                        return
                    if climit is not None and ccount == climit:
                        break

    def freeze(self):
        """Replace all references to code objects with string
        descriptions; this makes it possible to pickle the instance."""

        # this code is probably rather ickier than it needs to be!
        for i in range(len(self.data)):
            e = self.data[i]
            if not isinstance(e.code, str):
                self.data[i] = type(e)((label(e.code),) + e[1:])
            if e.calls:
                for j in range(len(e.calls)):
                    se = e.calls[j]
                    if not isinstance(se.code, str):
                        e.calls[j] = type(se)((label(se.code),) + se[1:])

_fn2mod = {}

def label(code):
    if isinstance(code, str):
        if sys.version_info.major >= 3:
            code = code.encode('latin-1')
        return code
    try:
        mname = _fn2mod[code.co_filename]
    except KeyError:
        for k, v in list(sys.modules.iteritems()):
            if v is None:
                continue
            if not isinstance(getattr(v, '__file__', None), str):
                continue
            if v.__file__.startswith(code.co_filename):
                mname = _fn2mod[code.co_filename] = k
                break
        else:
            mname = _fn2mod[code.co_filename] = r'<%s>' % code.co_filename

    res = r'%s:%d(%s)' % (mname, code.co_firstlineno, code.co_name)

    if sys.version_info.major >= 3:
        res = res.encode('latin-1')

    return res

if __name__ == '__main__':
    import os
    sys.argv = sys.argv[1:]
    if not sys.argv:
        print("usage: lsprof.py <script> <arguments...>", file=sys.stderr)
        sys.exit(2)
    sys.path.insert(0, os.path.abspath(os.path.dirname(sys.argv[0])))
    stats = profile(execfile, sys.argv[0], globals(), locals())
    stats.sort()
    stats.pprint()