Code review comment for lp:~mir-team/mir/touchspot-renderable

> You mean something like a generic version of the mesa shm_buffer? At one point
> I was thinking about something like this...but this way should be faster on
> platforms like android with HWC which allow generic buffer overlay.

That's interesting. GBM also allows overlays (bypass) of software buffers. However when I enabled it, the performance was worse, which is quite reasonable to expect but is driver-dependent too. Maybe Android would also benefit from intentionally disallowing overlaying of buffers that reside in CPU memory... Although if the CPU/GPU memory is shared then theoretically there should be no performance disadvantage to overlaying software buffers.