[prev in list] [next in list] [prev in thread] [next in thread]
List: kde-kimageshop
Subject: Re: Speed of Krita 2.4
From: Sven Langkamp <sven.langkamp () gmail ! com>
Date: 2011-08-06 17:46:57
Message-ID: CAAmsBfkAmA1OC_Ed4H2njomeqVb-a4mYCf-T9C-fimZVPBM7ig () mail ! gmail ! com
[Download RAW message or body]
[Attachment #2 (multipart/alternative)]
On Wed, Aug 3, 2011 at 12:03 AM, Sven Langkamp <sven.langkamp@gmail.com>wrote:
> On Tue, Aug 2, 2011 at 5:09 PM, Sven Langkamp <sven.langkamp@gmail.com>wrote:
>
>> On Mon, Aug 1, 2011 at 11:42 PM, Sven Langkamp <sven.langkamp@gmail.com>wrote:
>>
>>> On Sat, Jul 30, 2011 at 11:41 AM, Silvio Heinrich <plassy@web.de> wrote:
>>>
>>>> Am 29.07.2011 23:30, schrieb JL VT:
>>>> > I don't know how fast Krita is in the other branches (I haven't
>>>> tested)
>>>> > however I'd like to ask if there's a good date to start feeling really
>>>> > worried about speed?, I don't want 2.4 to be a single bit slower than
>>>> > 2.3, so I'd like to start hacking to try to find where our current
>>>> > bottlenecks are; but I don't know which parts deep in Krita to touch
>>>> > without stepping on other developer's toes, moreover, my fears may be
>>>> > unfounded, maybe we're just a couple weeks shy of a branch merge
>>>> solving
>>>> > all these problems, but, again, I haven't kept up enough with our IRC
>>>> > backlogs and Krita branches to be sure.
>>>> >
>>>> > Should I be worried?.
>>>> >
>>>> > Is there an upcoming branch I should be testing to help with the speed
>>>> > bottlenecks instead?.
>>>> >
>>>>
>>>> As far as I know we have two big bottle necks.
>>>> The first is the creation and transformation (rotation, scaling) of the
>>>> brush masks. The last time I profiled krita it spent 20-30% (I can't
>>>> remember exactly anymore) of the whole processing time while painting
>>>> with this operations.
>>>>
>>>
>>>
>>> Depends on which brush is used. Autobrush shows about 30% of the time in
>>> processing the mask, but it's not using 100% cpu. I think there is another
>>> bottleneck that callgrind doesn't show. Painting with predefined brushes
>>> shows a big performance bottleneck in scaling the brush (callgrind file:
>>> http://depot.tu-dortmund.de/get/syvg6 ) in the stroke benchmark. There
>>> is a single method that takes most of the time, so that should be the first
>>> target when trying to speed up things.
>>>
>>
>> I have attached a patch that turns off interpolation. It should give about
>> double the performance when painting with predefined brushes. Can the
>> artists here check how big the impact on quality is?
>>
>
> As first tested showed that it my patch also reduces some artifacts too, I
> have commited it. Please check out if it's ok. If not I will revert it.
>
So far I got the confirmation that there are no visual regression from
Timothee and Ramon, so I think we can continue to use it.
[Attachment #5 (text/html)]
<div class="gmail_quote">On Wed, Aug 3, 2011 at 12:03 AM, Sven Langkamp <span \
dir="ltr"><<a href="mailto:sven.langkamp@gmail.com">sven.langkamp@gmail.com</a>></span> \
wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px \
#ccc solid;padding-left:1ex;"> <div><div></div><div class="h5"><div \
class="gmail_quote">On Tue, Aug 2, 2011 at 5:09 PM, Sven Langkamp <span \
dir="ltr"><<a href="mailto:sven.langkamp@gmail.com" \
target="_blank">sven.langkamp@gmail.com</a>></span> wrote:<br> <blockquote \
class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc \
solid;padding-left:1ex"> <div><div class="gmail_quote">On Mon, Aug 1, 2011 at 11:42 \
PM, Sven Langkamp <span dir="ltr"><<a href="mailto:sven.langkamp@gmail.com" \
target="_blank">sven.langkamp@gmail.com</a>></span> wrote:<br><blockquote \
class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc \
solid;padding-left:1ex">
<div class="gmail_quote"><div>On Sat, Jul 30, 2011 at 11:41 AM, Silvio Heinrich <span \
dir="ltr"><<a href="mailto:plassy@web.de" \
target="_blank">plassy@web.de</a>></span> wrote:<br></div><div><blockquote \
class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc \
solid;padding-left:1ex">
Am 29.07.2011 23:30, schrieb JL VT:<br>
<div>> I don't know how fast Krita is in the other branches (I haven't \
tested)<br> > however I'd like to ask if there's a good date to start \
feeling really<br> > worried about speed?, I don't want 2.4 to be a single bit \
slower than<br> > 2.3, so I'd like to start hacking to try to find where our \
current<br> > bottlenecks are; but I don't know which parts deep in Krita to \
touch<br> > without stepping on other developer's toes, moreover, my fears may \
be<br> > unfounded, maybe we're just a couple weeks shy of a branch merge \
solving<br> > all these problems, but, again, I haven't kept up enough with \
our IRC<br> > backlogs and Krita branches to be sure.<br>
><br>
> Should I be worried?.<br>
><br>
> Is there an upcoming branch I should be testing to help with the speed<br>
> bottlenecks instead?.<br>
><br>
<br>
</div>As far as I know we have two big bottle necks.<br>
The first is the creation and transformation (rotation, scaling) of the<br>
brush masks. The last time I profiled krita it spent 20-30% (I can't<br>
remember exactly anymore) of the whole processing time while painting<br>
with this operations.<br></blockquote></div><div><br><br>Depends on which brush is \
used. Autobrush shows about 30% of the time in processing the mask, but it's not \
using 100% cpu. I think there is another bottleneck that callgrind doesn't show. \
Painting with predefined brushes shows a big performance bottleneck in scaling the
brush (callgrind file: <br>
<a href="http://depot.tu-dortmund.de/get/syvg6" \
target="_blank">http://depot.tu-dortmund.de/get/syvg6</a> ) in the stroke benchmark. \
There is a single method that takes most of the time, so that should be the first \
target when trying to speed up things.<br></div></div></blockquote></div><br></div>I \
have attached a patch that turns off interpolation. It should give about double the \
performance when painting with predefined brushes. Can the artists here check how big \
the impact on quality is?<br>
</blockquote></div><br></div></div>As first tested showed that it my patch also \
reduces some artifacts too, I have commited it. Please check out if it's ok. If \
not I will revert it.<br> </blockquote></div><br>So far I got the confirmation that \
there are no visual regression from Timothee and Ramon, so I think we can continue to \
use it.<br>
_______________________________________________
kimageshop mailing list
kimageshop@kde.org
https://mail.kde.org/mailman/listinfo/kimageshop
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic