[PATCH 0 of 6 A2 series] lowerencode and dotauxencode in C

Adrian Buehlmann adrian at cadifra.com
Sat Oct 6 04:52:21 CDT 2012


This is the combined pile of patches for hashed paths with the current encoding,
using a C version of those functions only (keeping the complicated "hashmangle"
step in Python).

Not really meant to be pushed. Just for consideration.

This series as a whole would bring the Netbeans repo speed from

  $ hg perffncacheencode
  ! wall 0.853000 comb 0.858006 user 0.858006 sys 0.000000 (best of 12)

down to

  $ hg perffncacheencode
  ! wall 0.488000 comb 0.483603 user 0.483603 sys 0.000000 (best of 21)

which is 42% faster (Windows 7 x64).

This series is basically a part of Bryan's patch

  store: implement fncache hash encoding in C
  http://selenic.com/pipermail/mercurial-devel/2012-September/044135.html

from 2012-09-10.



More information about the Mercurial-devel mailing list