On Wed, Jan 12, 2011 at 10:42 AM, Behdad Esfahbod <<a href="mailto:behdad@behdad.org">behdad@behdad.org</a>> wrote: <div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"> <div class="im">On 01/10/11 17:47, Robert O'Callahan wrote: > > How about we enable per-channel masks, so you can represent the operation > truthfully? Sure, you need to keep an RGB mask and an ARGB source, but at > least it's not impossible. > > > I guess so, but won't it be slow? </div>Don't know. That's what's happening whe you render subpixel text. </blockquote> </div>I mean we'd have to carry around an RGB mask for lots of surfaces; we'd have to alter that mask when we draw *anything*, not just text, into those surfaces; and the cost of compositing those surfaces would increase. We're actually already doing something like this for our GPU-accelerated layers backends (which don't use cairo for compositing), but there it's not so bad because at least the composition is relatively cheap. I'm rather scared of doing it when we can only use the CPU. Rob -- "Now the Bereans were of more noble character than the Thessalonians, for they received the message with great eagerness and examined the Scriptures every day to see if what Paul said was true." [Acts 17:11]