ppro asm
investigate MMX for pmmx and pii
investigate SSE for piii

fast little-endian loads on the sparc
sparc asm; note the stall on int conversion

fast little-endian loads on the alpha
21164 scheduling; shouldn't hurt on the 21264

tuning for other popular architectures

more documentation; what do people want to see?
more tests

filter? not useful for typical packet protocols
