Vc branch ready for testing

[Sven Langkamp]

On Thu, Sep 6, 2012 at 10:27 PM, Sven Langkamp <sven.langkamp at gmail.com>wrote:

>
>
> On Thu, Sep 6, 2012 at 7:36 PM, JL VT <pentalis at gmail.com> wrote:
>
>> On Thu, Sep 6, 2012 at 12:16 AM, Sven Langkamp <sven.langkamp at gmail.com>wrote:
>>
>>>  Hi,
>>>
>>>
>>> I have pushed the krita-vc-langkamp branch. It does implement the brush
>>> mask computation with vc.
>>> To build the branch you need a cpu with SSE2 and Vc 0.61 or git:
>>> http://code <http://code.compeng.uni-frankfurt.de/projects/vc>
>>> *
>>> *
>>> *I need to look up *.compeng.uni-frankfurt.de/projects/vc<http://code.compeng.uni-frankfurt.de/projects/vc>
>>>  AVX should work too, but requires a change in the Krita cmake file.
>>>
>>> With the branch there is around 15x speedup in the mask benchmark and
>>> about 2-3x speedup in the stroke benchmark (for brushes without random and
>>> density).
>>>
>>
>> You are a hero.
>>
>> And I juuust got an i7 Core 2600K Sandy Bridge... When I have this
>> installed, it'll be time to compile your branch.
>> And play with Krita. This will be good.
>>
>
> Don't get too excited. Animtim has reported an actual speedup of 2,
> although that was still without AVX so it might get better. I also have a
> few more idea how to optimize it further.
>

I pushed some changes, so now AVX should be used if CPU is Sandy Bridge. I
have also made another improvement to the apply alpha mask, not vectorized
but stroke benchmark is about 15-20% faster. Callgrind show that the mask
calculation is now longer the major performance bottleneck, instead the
composite op is now in the top spot.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kimageshop/attachments/20120907/c0fbaebf/attachment.html>