[Mesa-dev] [PATCH v2 4/8] i965: Set the maximum number of GS URB entries on Sandybridge.

[Paul Berry]

On 7 December 2011 15:02, Paul Berry <stereotype441 at gmail.com> wrote:

> On 7 December 2011 13:16, Eric Anholt <eric at anholt.net> wrote:
>
>> On Wed,  7 Dec 2011 11:09:12 -0800, Paul Berry <stereotype441 at gmail.com>
>> wrote:
>> > From: Kenneth Graunke <kenneth at whitecape.org>
>> >
>> > We never filled this in before because we didn't care.
>> >
>> > I'm skeptical these are correct; my sources indicate that both the VS
>> > and GS # of entries are 256 on both GT1 and GT2.
>> >
>> > I'm also loathe to change it and break stuff.
>>
>> My stale specs say
>>
>> GT2 VS: [24, 256]
>> GT1 VS: [24, 128]
>> GT2 GS: [0, 256]
>> GT1 GS: [0, 254].
>>
>> So these look like the docs our old VS numbers came from.  The current
>> docs I see say:
>>
>> GT1 VS: [24, 256]
>> GT1 GS: [0, 256]
>>
>> with the mention of gt2 removed.  The changelog doesn't show anything
>> obvious about "update DevSNB 3DSTATE_URB", though there are things for
>> other hardware that look like they might have stomped the old
>> information in the way it looks stomped.
>>
>> So, my guess is we should stick with our old values for VS (as is done
>> in this change), and use the old values for GS (GT2 256, GT1 254), and go
>> complain to doc authors for clarification.
>>
>
> I will try this and rerun piglit tests to see if anything breaks.
>

Argh, scratch that.  I have a GT2 Sandy Bridge machine, so I can't validate
this right now.

Since both your old docs and the current docs say 254 should be safe, and
Ken's 128 value was simply a guess, I'm going to go ahead and push these
patches with 254, and tomorrow I'll try to find a GT1 Sandy Bridge machine
to test on.  Even in the unlikely event that we're wrong and 254 isn't
safe, it shouldn't cause problems for anyone, since transform feedback
support isn't switched on yet.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/mesa-dev/attachments/20111207/fa3e3d61/attachment.html>