[cairo] gallium surface still maintained ?

[Enrico Weigelt, metux IT consult]

On 05.08.2016 04:06, Lawrence D'Oliveiro wrote:

>> My point was that if QuickDraw could do that on a 25MHz
>> Motorola 68040, then there's no reason why Cairo shouldn't work just
>> fine with today's embedded CPUs (especially if hardware floating
>> point is available), even if a GPU is not available/not usable.
> 
> Given the experience that Enrico Weigelt is reporting, I’m not so sure.

Actually, I'm yet just trying on a plain notebook w/ an i915 - dont
have an SoC board here yet.

The funny part: using xcb surface, it's way faster - probably because
the X server will be using hw acceleration (haven't got cairo's drm+i915
backend up and running yet).

Interestingly, the most costly operations seem to be the accessing the
framebuffer memory, not the actual rendering (not surprising w/ an 4
core i7 cpu). Maybe my problem is the access granularity - should I
add some double buffering and just write the changed areas in bigger
chunks to the framebuffer ?


--mtx