On Jan 16, 2006, at 8:09 AM, Juliusz Chroboczek wrote:
Sat Jan 14 04:22:52 PST 2006 Jason Dagit <[EMAIL PROTECTED]>
* Do not reread freshly written patch when recording.
I've looked at the code again -- and I'm puzzled. I s'pose it's
linesPS biting us again.
Maybe, I had at least one profiling report (retainers) point to
linesPS but I don't buy it.
The experimental results are definitely promising, but there's
obviously something going on here that we don't understand. I'm going
to wait for Ian to work out what's wrong with the current code before
committing this thing. Ian, please let me know if you decide you're
happy with this patch.
Ian isn't the only one looking into this :) I want to understand it
as well. Time permitting I'm going to send an email to Haskell-Cafe
asking for advice with this as I'm still relatively a newbie with
Haskell.
Sorry for the delay, Jason. More experimental results would be
welcome.
No problem, as I noted in my other email I understand now why we
don't want to immediately apply this patch. I should do an
experiment with committing the linux kernel in one commit. That
would give me a test similar to Ian's original benchmark in this area.
Thanks,
Jason
_______________________________________________
darcs-devel mailing list
[email protected]
http://www.abridgegame.org/cgi-bin/mailman/listinfo/darcs-devel
