[darcs-users] report on using darcs-22.214.171.124 (performance of "darcs query contents")
zooko at zooko.com
Sat Jul 11 20:22:54 UTC 2009
I tried out darcs-126.96.36.199 (upgrading from darcs-2.2.0). Everything
that I tried works fine, but the performance of "darcs query
contents" (http://bugs.darcs.net/issue1477 ) is not better. In fact,
it seems to be slightly worse.
For darcs-188.8.131.52 compiled with ghc-6.10.3, the best run out of nine
darcs query contents --quiet --match "hash
was 3.13s and the worst was 3.48s (all measurements are user CPU time
reported by "time").
For darcs-2.2.0 compiled with ghc-6.10.1 the best was 2.98s and the
worst was 3.19s.
I started to wonder if part of it was a performance regression from
ghc-6.10.1 to ghc-6.10.3 so I recompiled darcs-2.2.0 with ghc-6.10.3
and got: best 3.21s, worst: 4.11s. Huh. So, unless my benchmarks
were perturbed by shifting load on this server, it almost looks like
darcs-184.108.40.206 improves performance over darcs-2.2.0, but ghc-6.10.3
reduces performance over ghc-6.10.1. In order to experiment further,
I compiled darcs-220.127.116.11 with ghc-6.10.1:
best: 3.05s, worst: 3.29s
Oh well, I guess it's just random variation in my benchmarks. Just
to double-check, I ran nine more measurements with darcs-2.2.0
compiled with ghc-6.10.1: best: 3.01s, worst: 3.25s.
Of course, this hardly matters anyway to my tracdarcs use case; what
I really need for tracdarcs is for "darcs query contents" to return
the same answers in about 1/100 of the time it currently takes (i.e.
about 30 milliseconds would be an improvement), or perhaps to allow
queries on multiple files in a single call so that the tracdarcs
plugin doesn't need to invoke it dozens of times in order to render a
directory full of dozens of files. See http://bugs.darcs.net/
issue1477 for details.
More information about the darcs-users