SSE

i have a problem regarding SSE asm. code snippet is like this:

//det contains 4 single-pres floating-point no.
//andmask is 7fffffff 7fffffff 7fffffff 7fffffff
  .
  .
  movaps  xmm1,det;
  andps   xmm1,andmask;
  .
  .

basically i want to turn all the SP FP in xmm1 into positive by zero-ing the sign bit.but the code didn't give me what i want.
am i missing something here?

On a more serious note, isn't it a better idea to do SSE routines using the intrinsics? Just xmmintrin.h and you're good to go. Might want to get emmintrin.h and possibly pmmintrin.h too. No more byte order snags, and you get loop unrolling and register allocation courtesy of the compiler too.

Also, GCC 4.1 does _some_ (not much) vector optimization if you know how to tickle it.