Hi again,
to back up my claims that this fixes repair performance, I have done some
benchmarking. As a sidenote, while doing that, I have derived a small
darcs-benchmarking haskell module from bench.hs from
http://code.haskell.org/darcs/darcs-benchmark -- I'll be pushing that in a
while. It's currently pretty stupid, but it does let you express your benchmark
cases in haskell, providing you with a convenient way to run the right darcs
binary, say:
import Bench
-- ... more imports
check darcs =
darcs [ "check", "--no-test" ]
main = do
t <- run check "darcs check"
putStrLn $ TA.render id $ tabulate "big-zoo/darcs-repo" t
where s = setup [ "big-zoo/darcs-repo" ]
[ "../issue971/darcs", "../mainline/darcs", "darcs" ]
run = test s
Of course, eventually I'll provide a better interface for the benchmark runner,
so you'd just give it the setup and a list of tests and it'll do everything for
you.
Anyway, that was quite a detour, so back to check/repair ...
I run three iterations in a row for every case and take the best one, just to
minimise cache effects and such. The table looks like this:
|| ../issue971/darcs | ../mainline/darcs | darcs
=============++===================+===================+======
darcs check || 12.5 | 38.4 | 14.2
darcs repair || 12.3 | 39.7 | 150.3
where:
../issue971/darcs = the current unstable with the two repair bundles on top
../mainline/darcs = the current unstable on which the bundles are based
darcs = 2.1.0 from debian package
All the tests have been run on big-zoo/darcs-repo. Unfortunately, I don't have
memory usage covered yet, that'll come in later, hopefully...
Yours,
Petr.
--
Peter Rockai | me()mornfall!net | prockai()redhat!com
http://blog.mornfall.net | http://web.mornfall.net
"In My Egotistical Opinion, most people's C programs should be
indented six feet downward and covered with dirt."
-- Blair P. Houghton on the subject of C program indentation
_______________________________________________
darcs-users mailing list
[email protected]
http://lists.osuosl.org/mailman/listinfo/darcs-users
