[darcs-users] darcs patch: refactor Slurpy to common up name compon... (and 9 more)

[Ganesh Sittampalam]

On Sun, 2 Nov 2008, David Roundy wrote:

> On Thu, Oct 30, 2008 at 09:31:31PM +0000, Ganesh Sittampalam wrote:
>> On Wed, 29 Oct 2008, Ganesh Sittampalam wrote:
>>
>>> OK, I've now benchmarked this work on the GHC repo in kowey's
>>> zoo, and the overall picture suggests significant speedups (20% or
>>> so) on most of the tests (get, pull, unpull, record, unrecord,
>>> revert), with possible slowdowns in whatsnew and get-lazy of
>>> about 10%.
>>
>> My apologies for lazy benchmarking. This is all complete rubbish, as I
>> should have realised if I stopped to think about what was getting faster
>> and whether it was plausible.
>>
>> After some more careful runs thanks to kowey's scripts, the basic
>> conclusion is that there isn't any obvious (>5%) difference in performance
>> either way, although in general the new code is probably a bit faster than
>> the old code.
>
> Do you still think this code is worth putting in? I lean against it, if it
> has no measurable performance impact.  On the other hand, using a map does
> seem like a good idea, so if you think this should go in, I'll go ahead and
> review it.

Sorry, I was unclear. It still has a huge impact on 'whatnew -sl' on the 
articifial repo with 1000 new files that I've been testing with (quadratic 
-> something a lot better, probably n log n). It's just the testing on 
"normal" repos like ghc etc that shows no noticeable difference; which is 
at it should be really, I was benchmarking to ensure there was no slowdown 
there, and got carried away by results showing a speedup across the board 
without really thinking about them.

So I do think this should go in.

Ganesh