343655 - optimize win32 image rendering

Status: RESOLVED FIXED

(Core :: Graphics, defect)

Description

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

There's a lot of things that can be done to speed up our image rendering under Win32; right now we just use 24/32-bpp DIBs for everything, and we can do better than that.

Comment 1

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Attachment: use DDBs when we can (checkpoint)

This patch adds support for DDB surfaces to cairo and allows nsThebesImage to create them.  There's no usage of MaskBlt or PatBlt yet.  It also optimizes 1x1 image rendering to just do a solid fill instead of going through the image rendering routines.

Comment 2

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

Attachment: better patch

Note that this has StretchBlt disabled (the flag is never set on the surfaces, it's commented out); there are still some issues with that, and I'd rather work on those seperately in a followup.

Comment 3

[Stuart Parmenter]

Comment on attachment 232948 [details] [diff] [review]
better patch

in cairo-win32.c you have a lot of mixed tab/space stuff where one line is tabs and the next is spaces.  please make it consistent with the rest of the file (tab to 8/space the rest -- whatever cairo's annoying default is?)
make gfxPlatform::DoesARGBImageDataHaveAlpha() protected.  Don't think we want this as a public API.

Comment 4

[Vladimir Vukicevic [:vlad] [:vladv] (needinfo me, slow to respond)]

So I checked this in, it doesn't seem to have been a Tp win on the tinderboxes, but I'm hoping it's a CPU usage win, esp. for bug 328380.

Comment 5

[Peter van der Woude [:Peter6]]

this seems to have done more bad than good Vlad
I'll file some bugs later

Comment 6

[Ria Klaassen (not reading all bugmail)]

There's a hole in the Gmail reply box