Andrea Pretto <andrea_pretto@xxxxxxxx> writes: > I have examined the GCC manual and I seen the built-in > functions for SIMD extensions. > For SSE there are functions like > > |v4sf __builtin_ia32_loadaps (float *) > > ex: > > foo = ||__builtin_ia32_loadaps ( array )| > | > which move tha data from array to foo, and foo is a > type v4fs which become a XMM register for SSE > instruction. > But MMX??? > For MMX there aren't built-in function for move the > data from array to mm* register (why??), therefore I > must use inline asm. You shouldn't need to use any inline assembler. gcc should handle everything behind your back. If you have tried it, and it failed, then what did you do, and how did it fail? Ian
