[Mesa-dev] [PATCH 00/47] WIP: fp64 support for r600g
Marc Dietrich
marvin24 at gmx.de
Fri Aug 25 09:25:02 UTC 2017
Am Donnerstag, 24. August 2017, 16:39:58 CEST schrieben Sie:
> On 24 August 2017 at 15:26, Marc Dietrich <marvin24 at gmx.de> wrote:
> > Hi Elie,
> >
> > Am Mittwoch, 23. August 2017, 13:07:30 CEST schrieb Elie Tournier:
> >> From: Elie Tournier <elie.tournier at collabora.com>
> >>
> >> TL;DR
> >> This series is a "status update" of my work done for adding fp64 support
> >> on
> >> r600g. One of the biggest issue is due to a lake of accuracy on the rcp
> >> implementation. Divide relay on rcp.
> >>
> >> A branch is available on
> >> https://github.com/Hopetech/mesa/tree/glsl_arb_gpu_shader_fp64_v3
> >> Comments
> >> and reviews are welcome.
> >
> > what are the hardware requirements for testing this? R600g is a wide range
> > ...
> Hello Marc,
>
> My code targets all the cards who doesn't support double but who
> support GLSL 1.30:
> https://github.com/Hopetech/mesa/blob/glsl_arb_gpu_shader_fp64_v3/src/mesa/s
> tate_tracker/st_glsl_to_tgsi.cpp#L7059
>
> So iirc, all r600g cards.
sorry, I was confused by the glewinfo which shows
GL_ARB_gpu_shader_fp64: OK [MISSING]
with and without your patches, but glxinfo gets it right.
This is on an RS780 (HD3200?) integrated gpu.
running:
# ./piglit-run.py --dmesg --abort-on-monitored-error -s -l verbose -1 -t
arb_gpu_shader_fp64 all results/fp64
gives ~70% crashes until the system hard locks (or reboots) at some point.
Maybe it had also problems before running some shaders, so it is not necessary
a result of your work.
Because this is a very old platform which will never be able to see GL4, I
don't take this result too serious.
I will redo the tests until it hard locks, so you can get at least a partial
result.
Marc
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: This is a digitally signed message part.
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20170825/2638a721/attachment.sig>
More information about the mesa-dev
mailing list