[Piglit] [PATCH] add tests for using immediates in common operations

Eric Anholt eric at anholt.net
Tue Apr 24 00:37:47 UTC 2018

Ilia Mirkin <imirkin at alum.mit.edu> writes:

> On Mon, Apr 23, 2018 at 4:49 PM, Eric Anholt <eric at anholt.net> wrote:
>> Ilia Mirkin <imirkin at alum.mit.edu> writes:
>>> NVIDIA hardware has "short" immediates, and some instructions have both
>>> long and short immediate variants. Nouveau was picking the wrong one
>>> under certain conditions. This adds tests which are likely to cause
>>> these kinds of issues to surface in a backend.
>> My only concern is the integer multiply tests on non-native-integer
>> hardware or things where the number of integer bits is too small.  That
>> said, whoever tries to support one of those will have plenty of problems
>> already and need to define some shader_runner precision requirement
>> annotations.
>> Reviewed-by: Eric Anholt <eric at anholt.net>
> Hmmm ... I didn't think of that. Should I bump these up to GLSL 1.30?
> I don't think you could do GLSL 1.30 without proper integers.

Looks like 1.30 does give you exactly 32-bit ints, so that would cover
my concern.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 832 bytes
Desc: not available
URL: <https://lists.freedesktop.org/archives/piglit/attachments/20180423/de4e6228/attachment.sig>

More information about the Piglit mailing list