- This is more reliable with regard to random compiler errors and deprecation.
- This also opens up the door for supporting SIMD on other architectures, or openCL
Conflicts:
src/vecmath.c
vec4f.m in common.h changed from "__m128" to "float __attribute__ ((vector_size (16)))"
All the built-ins use gcc names except for moveaps, which is intel's _mm_store_ps.
got some inline assmembly from http://www.cortstratton.org/articles/HugiCode.html#bm5 and hacked it into gcc extended-inline
gcc doesn't like it, so I will redo it with built-ins next commit. This is just for reference.
