sparse-revlog: put the native implementation of slicechunktodensity to use
When possible, the C implementation of delta chain slicing will be used.
providing a large boost in performance for this operation.
To take a practical example of restoring manifest revision '59547c40bc4c' for
a reference NetBeans repository (using sparse-revlog). The media time of the
step `slice-sparse-chain` of `perfrevlogrevision` improve from 0.660 ms to
0.098 ms;
The full series move delta chain slicing from 1.120 ms to 0.098 ms;
Implementing _slicechunktosize into C would yield further improvements.
However, the performance seems good enough for now.
--- a/mercurial/revlogutils/deltas.py Thu Nov 15 11:10:52 2018 +0100
+++ b/mercurial/revlogutils/deltas.py Thu Nov 15 11:11:38 2018 +0100
@@ -44,6 +44,7 @@
self._srdensitythreshold = density
self._srmingapsize = mingap
self._snapshot = set(snapshot)
+ self.index = None
def start(self, rev):
if rev == 0:
@@ -120,9 +121,12 @@
targetsize = max(targetsize, revlog._srmingapsize)
# targetsize should not be specified when evaluating delta candidates:
# * targetsize is used to ensure we stay within specification when reading,
- for chunk in _slicechunktodensity(revlog, revs,
- revlog._srdensitythreshold,
- revlog._srmingapsize):
+ densityslicing = getattr(revlog.index, 'slicechunktodensity', None)
+ if densityslicing is None:
+ densityslicing = lambda x, y, z: _slicechunktodensity(revlog, x, y, z)
+ for chunk in densityslicing(revs,
+ revlog._srdensitythreshold,
+ revlog._srmingapsize):
for subchunk in _slicechunktosize(revlog, chunk, targetsize):
yield subchunk