Review Request: little faster sycoca

[David Faure]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
http://git.reviewboard.kde.org/r/102723/#review6887
-----------------------------------------------------------

Ship it!


You rock :)

I'm surprised that the compiler doesn't cache constEnd() though... Did you measure performance in a debug build or in a release build?

In any case I'm quite happy to see a performance improvement in that method, which I had seen as majorly guilty in profilings of kbuildsycoca.

- David Faure


On Sept. 27, 2011, 8:42 p.m., Jaime Torres Amate wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> http://git.reviewboard.kde.org/r/102723/
> -----------------------------------------------------------
> 
> (Updated Sept. 27, 2011, 8:42 p.m.)
> 
> 
> Review request for kdelibs.
> 
> 
> Description
> -------
> 
> Using this way to build a for loop: for (const_iterator it=list.constBegin, end=list.constEnd ; it != end; ++it)
> and replacing the count of bits with the QBitArray method for doing that.
> callgrind says:
> before: 2984 calls to constBegin, 0,00%. 2960531 calls to constEnd, 2.33%
> after: 2921 calls to constBegin, 0,00%. 2921 calls to constEnd, 0.00%
> 
> before: calcDiversity, 55.83%  (debug enabled)
> after: calcDiversity, 14,46%  (debug enabled)
> 
> buildsycoca is still not faster than light, but it is only a four lines patch.
> 
> 
> Diffs
> -----
> 
>   kdecore/sycoca/ksycocadict.cpp 17ed9ac 
> 
> Diff: http://git.reviewboard.kde.org/r/102723/diff/diff
> 
> 
> Testing
> -------
> 
> 
> Thanks,
> 
> Jaime Torres Amate
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-core-devel/attachments/20110928/7df9d2e4/attachment.htm>