[PATCH] manifest.add speedups w/delta generation

[Matt Mackall]

On Sat, Jun 25, 2005 at 07:12:01PM -0400, Chris Mason wrote:
> On Saturday 25 June 2005 17:22, Matt Mackall wrote:
> > - removing the sanity check (sortdiff seems to be sane now)
> 
> Yes, it didn't make any difference for my short runs.  I didn't try for the 
> full 2700 patch run.  I've found pretty good linear relation between things 
> that help the small runs and things that help the big ones.

Well that's good, I suppose.
> 
> > - adding a sortdiff() to bdiff.c
> 
> Yes, no difference.  I copied the sortdiff algorithm exactly and just put it 
> into C (I've still got the code if you want).  This was after installing 
> pysco, I should do a compare run with pyso off.

I suspect psyco could do pretty well with sortdiff. Both mpatch and
sortdiff are nicely O(N + M). So I suspect they're still getting
swamped by O(NM) behavior elsewhere.

-- 
Mathematics is the supreme nostalgia of our time.