[Mesa-dev] [PATCH 2/2] r600g, radeonsi: use fences provided by the winsys

Christian König deathsimple at vodafone.de
Wed Oct 9 01:57:02 PDT 2013


Haven't looked in detail at the patch, but I'm pretty sure that this 
goes into the right direction.

Christian.

Am 08.10.2013 23:53, schrieb Marek Olšák:
> Sorry, I didn't realize this. You are right. My bad. Regardless of my
> comment, I still wonder what people's opinion on this patch is.
>
> Marek
>
> On Tue, Oct 8, 2013 at 11:28 PM, Fredrik Höglund <fredrik at kde.org> wrote:
>> On Tuesday 08 October 2013, Marek Olšák wrote:
>>> From: Marek Olšák <marek.olsak at amd.com>
>>>
>>> This was horribly, horribly broken. The limit was 1024 fences created
>>> from the start of the application and as you probably know, pipe fences are
>>> not reusable. If you wanted to use one fence per frame, you could only do
>>> that for 1024 frames, whish was pretty bad. The error message "too many
>>> concurrent fences" was misleading.
>> Not true. r600g has a pool of 1024 internal fences, and when a pipe_fence
>> is deleted, the internal fence is returned to the pool so it can be reused.
>>
>> If you see that error message, it really means that the application or the
>> state tracker has created 1024 fences without waiting on them or
>> deleting them.  Note that calling glClientWaitSync() or quering the state
>> of a signalled sync object also causes the state tracker to delete the fence.
>>
>> This is not an objection to the patch, but I don't believe the ultimate cause
>> of the problem is the r600g implementation.
>>
>> Fredrik
>>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/mesa-dev



More information about the mesa-dev mailing list