On Mon, Nov 24, 2008 at 10:41 AM, Nathan Gray <[EMAIL PROTECTED]>wrote:
> I mentioned a week or two ago that we are unable to upgrade to darcs2 > because it is slower than darcs1. > > I did some benchmarking, and find that darcs1 is much faster than > darcs2. I also found that non-debian versions of darcs2 are quite a > bit faster than debian versions of darcs2. > > Attached to this e-mail message are the perl script I used to run my > benchmarking and the shell scripts I used to invoke gnuplot. Thanks! That makes you this weeks MVC, or Most Valuable Contributor :) We've been building a benchmark suite and we're happy to receive additions, especially cases that real users really care about. > > > perl darcs_benchmark.pl $source_repo $target_dir $darcsexecutable > sh plot_darcs_time.sh $datafile1 $datafile2 ... > sh plot_darcs_cpu.sh $datafile1 $datafile2 ... > > The benchmarking is very simple, and just gives a rough picture of the > speed of darcs doing common things on a repo. I used four versions of > darcs on the same type-1 repo, and collected stats on all of them. I > also made a type-2 repo from the type-1 repo and used three versions of > darcs2 to collect stats on those. > > darcs109 - non-debian darcs 1.0.9 (release) > darcs202 - debian darcs 2.0.2 (release) > darcs210 - non-debian darcs 2.1.0 (release) > darcs211 - debian darcs 2.1.1rc2 (+ 281 patches) Hmm...you may be missing some important optimizations from after the darcs sprint. We did some good low level (meaning not algorithmic) speed-ups during the sprint. Thanks! Jason
_______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
