[Pixman] [PATCH] test: Adjust for clang's removal of __builtin_shuffle

Matt Turner mattst88 at gmail.com
Mon Jun 4 20:46:25 UTC 2018


On Mon, Jun 4, 2018 at 10:37 AM, Adam Jackson <ajax at redhat.com> wrote:
> On Mon, 2018-06-04 at 10:04 -0700, Matt Turner wrote:
>
>> #ifdef HAVE_GCC_VECTOR_EXTENSIONS
>> -            const uint8x16 bswap_shufflemask =
>> +# if __has_builtin(__builtin_shufflevector)
>> +            randdata.vb =
>> +                __builtin_shufflevector (randdata.vb, randdata.vb,
>> +                                          3,  2,  1,  0,  7,  6 , 5,  4,
>> +                                         11, 10,  9,  8, 15, 14, 13, 12);
>> +# else
>> +            static const uint8x16 bswap_shufflemask =
>   ^            ^^^^^^
>
> Seems superfluous, though I guess it doesn't change semantics. With or
> without that bit:

Oh, I think I added that when I was trying to consolidate the
constants between the two paths. I'll remove that.

> Reviewed-by: Adam Jackson <ajax at redhat.com>
>
> I think we're starting to be well overdue for an 0.36 release, but I'd
> like to take the opportunity to suggest moving to fdo's gitlab as we do
> that. I already have a copy imported personally and have CI working:
>
> https://gitlab.freedesktop.org/ajax/pixman/-/jobs/986

Agreed.

I would like to make 0.36 pass the test suite with clang, so if you
have any time or interest I'd appreciate a second set of eyes. I'll
filed https://bugs.freedesktop.org/show_bug.cgi?id=106818 so we can
track it.

I guess it's possible it's a clang bug.

I also need to take some time to look into the Loongson3 patch. If
you're not in a particular hurry, it would be nice to get that in.


More information about the Pixman mailing list