The Intel Approximate
Math (AM) library has free SSE/SSE2 implementation for math
functions such as sin(), log(), exp()
; most interestingly,
it comes with documentation, relative error comparison, and a clock
cycle analysis for P3 and P4 CPUs.
gcc -mveclibabi=acml/svml
should be explored next...
posted at: 19:03 | path: /programming | permanent link