490700 - Let qcms directly transform to Cairo pixel format to save later conversions

Status: RESOLVED INACTIVE

(Core :: Graphics: Color Management, defect)

Description

[Alfred Kayser]

Image decoders that apply cms transforms using qcms can be improved by
letting the cmstranform function directly write to Cairo Packed pixel format.
This saves another loop through the pixel data after the transform.
Next step could to extend the SSE code in the common transform paths to do also the Cairo packed pixel conversion in SSE code as well.
See http://wontfix-en.blogspot.com/2009/01/need-sse2-optimization-for.html
for some ideas on that.

Comment 1

[Alfred Kayser]

Attachment: Proof of concept patch: compiles and runs, even passes Mochitests

Comment 2

[Alfred Kayser]

Attachment: Whitespace ignore version, as lots of lines are only indenting issues

Comment 3

[Jeff Muizelaar [:jrmuizel]]

Here are some quick comments:
- I'd like to avoid using the NSPR times to allow others to reuse qcms easily
- the GFX macros would probably need a different prefix, though I'd prefer static inline functions
- I'm not sure if I like getting rid of the packed and non-premultiplied output formats, these are the formats used by libpng and libjpeg as well as lcms which makes integration easier.
- Have you done any sort of measurement to see what kind of difference this patch makes?

Comment 4

[BugBot [:suhaib / :marco/ :calixte]]

The bug assignee didn't login in Bugzilla in the last 7 months.
:aosmond, could you have a look please?
For more information, please visit auto_nag documentation.