asm_sse_test.c

struct fl4{ float x, y, z, w; };

void asm_test(void)
{
   struct fl4 v1, v2, v3;

   v1.x = 0.1;
   v1.y = 0.2;
   v1.z = 0.4;
   v1.w = 0.3;
    
   v2.x = 0.11;
   v2.y = 0.0;
   v2.z = 0.01;
   v2.w = 0.04;

   asm volatile (
       "movups %0, %%xmm0;"
       "movups %1, %%xmm1;"
       "addps %%xmm1, %%xmm0;"
       "movups %%xmm0, %2"
       :
       : "g" (v1), "g" (v2), "g" (v3)
       : "memory"
   );

   printf("sse fl4 add : %f %f %f %f\n", v3.x, v3.y, v3.z, v3.w);
   printf("expected : %f %f %f %f\n", v1.x+v2.x, v1.y+v2.y, v1.z+v2.z, v1.w+v2.w);
}

sse fl4 add : 0.210000 0.200000 0.410000 0.340000
expected : 0.210000 0.200000 0.410000 0.340000