> (Removing David Roundy from the Cc: line.) Yeah. I need to stop spamming him :-( > So, if I understand correctly "darcs whatsnew" on a GHC repo with no > changes takes 7.60 seconds with old-fashioned-inventory format and > 8.08 seconds with hashed-format.
Actually, this was with some changes introduced by replacing all instances of the string "id" with "ego" in the Haskell sources (id is the identity function, for what it's worth). Naturally, It would have been sensible to do the same with no changes. > Is it this added 0.48 seconds which is what deterred the GHC > developers from switching to hashed format? I seriously doubt that! > Or did they experience significantly worse slowdowns on their systems > (Windows) and their workflows (perhaps lots of unrecorded changes?). I suspect the slowdows are due to not-Linux. Anecdotally speaking, I had the impression that things were much slower on my home Mac than on my work Linux box, though the specs seemed comparable. > Also, could anyone state why the GHC developers didn't switch to > darcs-2-format? Too slow (hence the bug above) -- Eric Kow <http://www.nltg.brighton.ac.uk/home/Eric.Kow> PGP Key ID: 08AC04F9
pgpWRKByAvjyl.pgp
Description: PGP signature
_______________________________________________ darcs-users mailing list [email protected] http://lists.osuosl.org/mailman/listinfo/darcs-users
