Bug 344738 - Brushperformance 8bit/32bit color deep
Summary: Brushperformance 8bit/32bit color deep
Status: RESOLVED UPSTREAM
Alias: None
Product: krita
Classification: Applications
Component: Brush engines (show other bugs)
Version: 2.9
Platform: unspecified All
: NOR minor
Target Milestone: ---
Assignee: Krita Bugs
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2015-03-02 11:32 UTC by reikenberg
Modified: 2016-03-19 09:57 UTC (History)
2 users (show)

See Also:
Latest Commit:
Version Fixed In:
Sentry Crash Report:


Attachments
Performance analyse records (633.20 KB, application/zip)
2015-03-07 01:30 UTC, reikenberg
Details

Note You need to log in before you can comment on or make changes to this bug.
Description reikenberg 2015-03-02 11:32:36 UTC
The brush performance varies not as expected if compare color deeps
This result from the same hardware with different os system with and without opengl on:

Nvidia GTX660
Win7 64bit
color deep: 8bit slower than 32bit (unexpected)
Linux Mint 14
color deep: 32bit slower than 8bit (expected)



Reproducible: Always




There is a bug or restriction that drops performance for nvidia cards with 8bit color deep or maybe nvidia is much more faster with 32bit or their drivers are buggy.

This info is perhaps useful for other users or a dev get an idea what is going on there.
Comment 1 reikenberg 2015-03-07 01:30:24 UTC
Created attachment 91462 [details]
Performance analyse records

Here is a compare for a played macro in 8bit and 32bit, maybe it helps.
Recorded with AMDs CodeXL.
Comment 2 wolthera 2015-03-08 20:17:19 UTC
This is expected. Nvidea card are optimised for floating point values, and you will find they're even faster on 16bit float images!

Issue though with these sizes is that they do eat ram.

I'll leave this open for now, for the others devs to take a look.
Comment 3 reikenberg 2015-03-09 07:12:14 UTC
Well, thanks for the info.

But i wonder why the different bit modes seems to use different dll calls and counts.
Comment 4 Halla Rempt 2016-03-19 09:57:37 UTC
I'm sorry, I know it's observed behavior, and it would be better if it wouldn't happen, but we cannot influence the performance characteristics of display drivers....