[Mesa-dev] [PATCH 00/59] Initial arb_gpu_shader_fp64 support to the i965 scalar backend

[Samuel Iglesias Gonsálvez]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256



On 04/05/16 21:28, Francisco Jerez wrote:
> Samuel Iglesias Gonsálvez <siglesias at igalia.com> writes:
> 
> 
> 
> On 03/05/16 20:59, Kenneth Graunke wrote:
>>>> Other than patches 37, 56, and ones you agreed to drop, the
>>>> series is: Reviewed-by: Kenneth Graunke
>>>> <kenneth at whitecape.org>
>>>> 
>>>> I think you can go ahead and land all except those, and we
>>>> can land new solutions for those problems afterwards.
>>>> 
>>>> We still need to fix the horiz_offset problem, and I think
>>>> Curro's subscript() helper is a great way to do that.  I
>>>> would be fine with fixing that after this series lands - it
>>>> would probably mean less rebasing :)
>>>> 
> 
> We prepared a patch series without patches 11, 22 (both were
> dropped), 37, 40 (it has a NACK from Curro) and 56. Our idea was to
> push the patch series once we check there is no piglit
> regressions.
> 
> We got 0 piglit regressions on ILK, SNB, IVB, HSW and BDW, the
> latter with the extension enabled with the environment variable.
> 
> Sadly, on SKL we had regressions (including GPU hangs) only if we 
> *enable* the arb_gpu_shader_fp64 extension. Those regressions are 
> fixed by patch 40 but we need still to investigate why they are
> happening.
> 
>> That sounds like PATCH 40 was most likely working around
>> regs_read() being interpreted with incorrect units in some other
>> place.  The use of regs_read() in PATCH 36 is obviously wrong
>> which might explain the problem.  Something along the lines of
>> 'inst->components_read(i) * type_sz(inst->src[i]) / 4' would do
>> what you want -- With that changed PATCH 36 is:
> 
>> Reviewed-by: Francisco Jerez <currojerez at riseup.net>

This change fixes the SKL regressions. Thanks Curro!

Sam

> 
> For that reason, we have not pushed anything yet. Tomorrow we will 
> investigate why we have those regressions only on SKL and which
> tests are affected.
> 
>> No worries, you still have a couple of weeks until the merge
>> deadline, I don't see the need to rush in any changes that still
>> have outstanding comments.
> 
> JFTR, if we disable the extension on SKL (i.e. we don't set the 
> environment variable), we have 0 piglit regressions on that
> generation.
> 
> Thanks,
> 
> Sam
> 
> 
>> _______________________________________________ mesa-dev mailing
>> list mesa-dev at lists.freedesktop.org 
>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
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=G05c
-----END PGP SIGNATURE-----