On Wed, Feb 3, 2016 at 9:11 AM, Ilia Mirkin <imir...@alum.mit.edu> wrote:
> On Wed, Feb 3, 2016 at 5:11 AM, Marek Olšák <mar...@gmail.com> wrote:
>> On Sun, Jan 31, 2016 at 9:56 PM, Ilia Mirkin <imir...@alum.mit.edu> wrote:
>>> On Sun, Jan 31, 2016 at 3:32 PM, Ilia Mirkin <imir...@alum.mit.edu> wrote:
>>>> This is supposed to return GL_TRUE or GL_FALSE, not the full counter
>>>> value.
>>>>
>>>> Signed-off-by: Ilia Mirkin <imir...@alum.mit.edu>
>>>> ---
>>>
>>> Errr... hm. Just realized many drivers will fail at this. Oh well,
>>> I'll drop it for now.
>>
>> Are you sure?
>
> Pretty sure. At the very least, nv30 doesn't support it. Definitely fixable.
>
>>
>> We might have to enable this eventually, because if apps use it, it
>> allows faster rendering than with OCCLUSION_COUNTER, because depth
>> test with depth-only rendering and no writes can run at HiZ rate
>> instead of pixel rate.
>
> Yeah, that makes sense. I think we just need to do a driver audit
> first, before enabling it.
>
> -ilia
And of course this came back to bite me - one of the piglit tests is
failing, it's checking for true/false. I'm dropping this hunk back in
for the non-qbo case at the mesa/main level.
if (q->Target == GL_ANY_SAMPLES_PASSED ||
q->Target == GL_ANY_SAMPLES_PASSED_CONSERVATIVE)
value = !!value;
I'm guessing other drivers might need it too. Ideally we should teach
all the drivers that support occlusion queries about the predicate
variant, and remove this wart.
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
