On Wed, Oct 30, 2019 at 08:01:28AM +0100, Uros Bizjak wrote:
> > While working on the OpenMP isa patch, I've noticed most of the x86 ISA
> > options imply in the help text that it enables MMX, when they do not.
> > We now have the mmx in sse2, so to some extent most of the built-in
> > functions
> > are enabled with sse2 on 64-bit targets, but we do not enable MMX code
> > generation and I think listing MMX in there does more harm than good,
> > it is something that should better be explained in the documentation.
>
> But e.g. -msse does enable MMX built-in functions. The fact that these
> builtins are implemented using SSE instructions on x86_64 is just an
> implementation detail. OTOH, on 32bit targets -msse still enables MMX
> buitlins *and* code generation.
>
> So, this test:
>
> --cut here--
> typedef int __m64 __attribute__ ((__vector_size__ (8), __may_alias__));
>
> __m64 test (__m64 __m1, __m64 __m2)
> {
> return __builtin_ia32_paddd (__m1, __m2);
> }
> --cut here--
>
> compiles MMX built-in function with -msse option to what the
> documentation claims.
Ok, patch withdrawn then.
Jakub
