On Mon, 21 Jan 2013, Peng Yu wrote:
The following commands showed that -msse2 has to be used with -O (or its variants -O1, -O2, etc.) But I don't find this documented in the gcc manual. Does anybody know where this is documented? Thanks!
It is just luck that makes it sometimes work.
#include <xmmintrin.h> #include <stdio.h> int main() { float p[4]={1,2,3,4}; __m128 a = _mm_load_ps(p);
_mm_load_ps requires a 16-byte alignment.
float p1[4]; _mm_store_ps(p1, a); for(int i=0; i < 4; i++) { printf("%g\n", p1[i]); } }
-- Marc Glisse