[cairo-bugs] [Bug 88203] Replaying recording surfaces with OVER has far worse performance than when using the SOURCE operator

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=88203

--- Comment #4 from Emanuele Aina <emanuele.aina at collabora.com> ---
> Right, an OVER operator with no mask need no incur the replay onto an
> intermediate surface.

\o/

>      /* Are we just copying a recording surface? */
> -    if (inplace &&
> +    if ((inplace || (op == CAIRO_OPERATOR_OVER && no_mask)) &&

Is it safe to ignore need_clip_mask when the operator is OVER?

>         /* first clear the area about to be overwritten */
> -       if (! dst->is_clear) {
> +       if (! dst->is_clear && op_is_source) {
>             status = compositor->fill_boxes (dst,
>                                              CAIRO_OPERATOR_CLEAR,
>                                              CAIRO_COLOR_TRANSPARENT,


If I got the code right, using op_is_source instead of checking for
CAIRO_OPERATOR_SOURCE here would produce some redundant clears in case we're
using CAIRO_OPERATOR_OVER or CAIRO_OPERATOR_ADD with an opaque pattern.

> or even replace that op_is_source with inplace for consistency.

I'm not sure what you meant here, sorry. :(

> Can you please check that the recording_surface_over test hits the new path
> and check that it missed the optimisation before.

I'm having some issue with the testsuite.

Even when using `Xvfb -screen 0 1680x1024x24 -ac -nolisten tcp :2` and
`DISPLAY=:2 CAIRO_TEST_TARGET=xlib ./cairo-test-suite recording-surface-over` I
only get xlib.argb32 and xlib-render-0_0.rgb24 to PASS, while xlib.rgb24,
xlib-window.rgb24 and xlib-fallback.rgb24 FAIL (with or without the patch
applied).

The two tests using CAIRO_TEST_TARGET=image PASS cleanly, without limiting
CAIRO_TEST_TARGET I get other failures.

But I also get lots of failures when I run the full testsuite, so I'm quite at
a loss here. :(

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cairographics.org/archives/cairo-bugs/attachments/20150108/8c52323b/attachment.html>