Matt Turner <[email protected]> writes: > +/* We have to compile with -msse to use xmmintrin.h, but that causes SSE > + * instructions to be generated that we don't want. Just duplicate the > + * functions we want to use. */ > +extern __inline __m64 __attribute__((__gnu_inline__, __always_inline__, > __artificial__)) > +_mm_mulhi_pu16 (__m64 __A, __m64 __B) > +{ > + return (__m64) __builtin_ia32_pmulhuw ((__v4hi)__A, (__v4hi)__B); > +}
Shouldn't this be bracketed by #ifdef USE_X86_MMX? Or does __builtin_ia32_pmulhuw() work on iwMMX too? Also, in principle __builtin_ia32_pmulhuw() should be checked for in the MMX section of configure.ac; it might not be available on Solaris compilers for example. Similar things applies to the pshufw patch. Soren _______________________________________________ Pixman mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/pixman
