[Bug 83508] [UBO] Assertion for array of blocks

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=83508

--- Comment #3 from Antia Puentes <apuentes at igalia.com> ---
(In reply to Timothy Arceri from comment #2)
> (In reply to Samuel Iglesias from comment #1)
> > This is going to be fixed by Antía's patch:
> > 
> > http://lists.freedesktop.org/archives/mesa-dev/2015-March/079102.html
> > 
> > We are planning to send an updated version of this patch in the coming days.

Hi! Timothy,

I have a limited knowledge about the linking code but I did take a look at how
the link_uniform_block_active_visitor works.

> That will work around the issue for 'shared' and 'std140' but if a shader
> has a ubo with a 'packed' layout and is unused in the shader it will still
> hit this same problem right?

I think it will not. The 'shared' and 'std140' ubo arrays are always marked as
active, regardless of they being referenced or not in the code. However, for
'packed' ubos I can see that they are marked as active by the
link_uniform_block_active_visitor *only if* there is a reference to them. By
marking them as active, I mean that they will be added by the visitor to the
'block_hash' hashtable  declared in the "link_uniform_blocks" method.

> The problem is the linking code always expects there to be a least one
> active use of the ubo array in the shader, I attempted to fix this but it
> REALLY expects there to be a use so its not quite as simple as you would
> expect.

Could you point out where that is expected?. Thanks.
My understanding is that the linking process would only care about what we have
inside that 'block_hash' active uniform blocks hashtable and, in the case of
'packed' ubos, they only be there if they have been referenced.

> I may give it another go when I have some time as I'm playing around with
> this code for AoA support currently.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20150922/e5c2f2db/attachment.html>