On Aug 28, 2012, at 7:23 PM, Richard Trieu <[email protected]> wrote:

> If I read these numbers correctly, the hash table algorithm (with O(n) 
> performance) takes about 1.6-2% percent more than the control for runs 1-3, 
> and hardly anything noticeable for the clang code base.  Were runs 1-3 used 
> in your earlier measurements,
> Yes, these are the same runs I have been using.  I earlier did some more runs 
> with smaller files, but the improvements to this warning made the differences 
> too small to detect, so they were dropped.
>  
> where the sorting-based approach took about ~4% longer (or is that not the 
> correct number)?
> That is correct.  The fastest sorting-based reached 4% difference.

Ok, sounds great.  The patch looks fantastic to me now, and the performance is 
good.  I'd be happy to see this go in.
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to