[Mesa-dev] [RFC] preparation for qxl sw wrapper driver

[Dave Airlie]

On Tue, Mar 19, 2013 at 9:21 PM, Dave Airlie <airlied at gmail.com> wrote:
>> I have no particular objection with the above in principle, as this seems a perfectly valid use of llvmpipe.
>>
>> I just have some remarks on implementation details done separately.
>>
>>
>> For my curiosity sake, how much these changes matter in practice? (For example, what is the % effect of gloss FPS in normal window and fullscreen?)
>
> Well its all about reducing guest CPU usage to increase density, and I
> haven't done a massive test yet, other than glxgears -fullscreen
> inside a VM went from 90FPS to 135FPS, so I expect a CPU usage drop
> for texture from pixmap will be quite decent as well though I don't
> have a good benchmark for it.
>

Thanks guys for discussion, it cleared things up a bit for me, and
made me realised the futility of bothering to do what I'm doing with
the qxl virt gpu. So I've resolved the upside-downness for TFP by
flipping in the server and only doing sw rendering from that point
onwards, which also solves the other problem of syncing to the hw. So
I have something that works without these hacks, and wraps llvmpipe.

Though I'd still like to see GLX_Y_INVERTED_EXT support, it would
clean up a couple of copies for me and I might investigate that,
though since I've no idea if any the compositors support it I'd
probably need to look at that as well.

Dave.