Vectorized memcpy
Some benchmarks of memcpy function on my Compaq Presario S6000NX Celeron 2.7 GHz (Linux), PowerMac G4 400 MHz (Linux) and PowerBook G4 1.67 GHz (Linux and Mac OS X):
As we can see vectorized memcpy on PowerPC G4 1.67 GHz almost twice as fast as Celeron 2.7 GHz (which is sometimes similar to speed of vectorized memcpy on PowerPC G4 400 MHz). Also we see that Mac OS X has vectorized memcpy by default. See source code of this test: files/test_g4.c (2.5K)