[Pixman] Pixman not building on MacOS X 10.11

Siarhei Siamashka siarhei.siamashka at gmail.com
Sat Oct 10 20:30:08 PDT 2015

On Sun, 11 Oct 2015 04:53:08 +0300
Siarhei Siamashka <siarhei.siamashka at gmail.com> wrote:

> On Sat, 10 Oct 2015 16:03:53 -0700
> Jeremy Huddleston Sequoia <jeremyhu at freedesktop.org> wrote:
> > > On Oct 10, 2015, at 13:48, Andrea Canciani <ranma42 at gmail.com> wrote:
> > > The attached hack gets the code to compile on modern clang, but I
> > > believe first of all we should improve the configure.ac detection code
> > > so that pixman can actually build both on old and on new clang
> > > versions (possibly with mmx disabled, if the asm constraints we need
> > > are not implemented).
> This workaround looks reasonable to me. We should probably just drop
> the whole "ifdef __OPTIMIZE__" part in
>     http://cgit.freedesktop.org/pixman/tree/pixman/pixman-mmx.c?id=pixman-0.32.8#n92
> I don't quite like the fact that this way of returning results from
> a macro is a GNU C specific extension. But as you said, the configure
> test can be updated to better match the code and also check if the
> compiler supports this particular construct.
> Could you please submit the final variant of your patch in a
> "git format-patch" format with a commit message and your
> Signed-off-by tag?

After looking at this issue a bit more, I realized that we are
about to add a second layer of workarounds on top of the existing
old workarounds :-)

The original problem is that certain compilers (just GCC?) did not
support some intrinsics when compiling MMX code (_mm_movemask_pi8,
_mm_mulhi_pu16, _mm_shuffle_pi16) and we got the following code:

In fact, these instructions were not available as part of the original
MMX, but only got introduced later with AMD Extended 3DNow! and Intel
SSE1. This is mentioned in the commit messages:

These extra instructions are unofficially known as MMX2. But GCC does
not have a separate option for "-mmmx2". Instead the GCC manual says
that these intrinsics are available when either "-msse" or a
combination of "-m3dnow -march=athlon" is used:

Now I wonder if the comment "We have to compile with -msse to use
xmmintrin.h" is still valid. I tried to tweak the following ifdef to
use the part of code, which includes <xmmintrin.h> and the it compiled
fine for me with CFLAGS="-O2 -m32" using recent versions of GCC and

I believe that this might be somehow related to the new __ALL_ISA__
define, which had been mentioned in 2013:

So what about just dropping this ugly stuff and adding a configure
check, which would verify if the MMX code can include <xmmintrin.h>?
And if this configure check fails (because of using an old compiler),
then simply disable MMX support when building pixman.

Best regards,
Siarhei Siamashka

More information about the Pixman mailing list