Patchwork [1,of,2,zstd-revlogs,V2] revlog: use compression engine APIs for decompression

login
register
mail settings
Submitter Gregory Szorc
Date Jan. 14, 2017, 4:45 a.m.
Message ID <081a7a0c0d5665056476.1484369156@ubuntu-vm-main>
Download mbox | patch
Permalink /patch/18215/
State Accepted
Headers show

Comments

Gregory Szorc - Jan. 14, 2017, 4:45 a.m.
# HG changeset patch
# User Gregory Szorc <gregory.szorc@gmail.com>
# Date 1484366280 28800
#      Fri Jan 13 19:58:00 2017 -0800
# Node ID 081a7a0c0d5665056476ed35875d663ea5eaac73
# Parent  8614546154cbffe8df6a56d53e404491fa092636
revlog: use compression engine APIs for decompression

Now that compression engines declare their header in revlog chunks
and can decompress revlog chunks, we refactor revlog.decompress()
to use them.

Making full use of the property that revlog compressor objects are
reusable, revlog instances now maintain a dict mapping an engine's
revlog header to a compressor object. This is not only a performance
optimization for engines where compressor object reuse can result in
better performance, but it also serves as a cache of header values
so we don't need to perform redundant lookups against the compression
engine manager. (Yes, I measured and the overhead of a function call
versus a dict lookup was observed.)

Replacing the previous inline lookup table with a dict lookup was
measured to make chunk reading ~2.5% slower on changelogs and ~4.5%
slower on manifests. So, the inline lookup table has been mostly
preserved so we don't lose performance. This is unfortunate. But
many decompression operations complete in microseconds, so Python
attribute lookup, dict lookup, and function calls do matter.

The impact of this change on mozilla-unified is as follows:

$ hg perfrevlogchunks -c
! chunk
! wall 1.953663 comb 1.950000 user 1.920000 sys 0.030000 (best of 6)
! wall 1.946000 comb 1.940000 user 1.910000 sys 0.030000 (best of 6)
! chunk batch
! wall 1.791075 comb 1.800000 user 1.760000 sys 0.040000 (best of 6)
! wall 1.785690 comb 1.770000 user 1.750000 sys 0.020000 (best of 6)

$ hg perfrevlogchunks -m
! chunk
! wall 2.587262 comb 2.580000 user 2.550000 sys 0.030000 (best of 4)
! wall 2.616330 comb 2.610000 user 2.560000 sys 0.050000 (best of 4)
! chunk batch
! wall 2.427092 comb 2.420000 user 2.400000 sys 0.020000 (best of 5)
! wall 2.462061 comb 2.460000 user 2.400000 sys 0.060000 (best of 4)

Changelog chunk reading is slightly faster but manifest reading is
slower. What gives?

On this repo, 99.85% of changelog entries are zlib compressed (the 'x'
header). On the manifest, 67.5% are zlib and 32.4% are '\0'. This patch
swapped the test order of 'x' and '\0' so now 'x' is tested first. This
makes changelogs faster since they almost always hit the first branch.
This makes a significant percentage of manifest '\0' chunks slower
because that code path now performs an extra test. Yes, I too can't
believe we're able to measure the impact of an if..elif with simple
string compares. I reckon this code would benefit from being written
in C...

Patch

diff --git a/mercurial/revlog.py b/mercurial/revlog.py
--- a/mercurial/revlog.py
+++ b/mercurial/revlog.py
@@ -39,7 +39,8 @@  from . import (
 
 _pack = struct.pack
 _unpack = struct.unpack
-_decompress = zlib.decompress
+# Aliased for performance.
+_zlibdecompress = zlib.decompress
 
 # revlog header flags
 REVLOGV0 = 0
@@ -339,6 +340,8 @@  class revlog(object):
             self._chunkclear()
         # revnum -> (chain-length, sum-delta-length)
         self._chaininfocache = {}
+        # revlog header -> revlog compressor
+        self._decompressors = {}
 
     @util.propertycache
     def _compressor(self):
@@ -1491,17 +1494,52 @@  class revlog(object):
         """
         if not data:
             return data
+
+        # Revlogs are read much more frequently than they are written and many
+        # chunks only take microseconds to decompress, so performance is
+        # important here.
+        #
+        # We can make a few assumptions about revlogs:
+        #
+        # 1) the majority of chunks will be compressed (as opposed to inline
+        #    raw data).
+        # 2) decompressing *any* data will likely by at least 10x slower than
+        #    returning raw inline data.
+        # 3) we want to prioritize common and officially supported compression
+        #    engines
+        #
+        # It follows that we want to optimize for "decompress compressed data
+        # when encoded with common and officially supported compression engines"
+        # case over "raw data" and "data encoded by less common or non-official
+        # compression engines." That is why we have the inline lookup first
+        # followed by the compengines lookup.
+        #
+        # According to `hg perfrevlogchunks`, this is ~0.5% faster for zlib
+        # compressed chunks. And this matters for changelog and manifest reads.
         t = data[0]
-        if t == '\0':
-            return data
+
         if t == 'x':
             try:
-                return _decompress(data)
+                return _zlibdecompress(data)
             except zlib.error as e:
                 raise RevlogError(_('revlog decompress error: %s') % str(e))
-        if t == 'u':
+        # '\0' is more common than 'u' so it goes first.
+        elif t == '\0':
+            return data
+        elif t == 'u':
             return util.buffer(data, 1)
-        raise RevlogError(_('unknown compression type %r') % t)
+
+        try:
+            compressor = self._decompressors[t]
+        except KeyError:
+            try:
+                engine = util.compengines.forrevlogheader(t)
+                compressor = engine.revlogcompressor()
+                self._decompressors[t] = compressor
+            except KeyError:
+                raise RevlogError(_('unknown compression type %r') % t)
+
+        return compressor.decompress(data)
 
     def _isgooddelta(self, d, textlen):
         """Returns True if the given delta is good. Good means that it is within