Add -X to allow bypassing the SIMD (SSE/MMX) routines.
Rewrite the SAD (Luma only) routines to use pointer arithmetic and the (builtin) abs() function. This speeds up processing almost 25% on a 1GHz G4/PPC
Authored by: sms00 2004-10-14
Parent: [r2448]
Child: [r2450]
