[Vm-dev] Re: unit code for new faster bitblt

[tim Rowledge]

On 18-06-2013, at 7:14 PM, Casey Ransberger <casey.obrien.r at gmail.com> wrote:

> Wow, as much as 10x? As soon as I can get time I'll try this out on both my Mac and my Pi with a few different images. I bet Cuis breaks the sound barrier:)

It'll vary enormously; certainly no 10x for x86 machines I think. On Raspbian Pi some cases are 10x, many are 5x but a few are actually slower. The key thing is that there is a decently intelligible framework into which more special case can be bult as needed and available. I suspect the code to search for a matching fast path might be improvable too, which will speed up the important small area blts.

And of course, a lot of drawing really ought to be via vector code anyway.

tim
--
tim Rowledge; tim at rowledge.org; http://www.rowledge.org/tim
Implementation is the sincerest form of flattery.