Welcome to the future. Ain't that cool?

Before anyone asks about the dreaded why, it's not because of the larger addressable memory space that 64-bit apps provide, but because x86_64 (which the Mac Pro and all Core 2 Duo Macs have) has 16 general purpose registers and 16 SSE registers, doubling IA32's measly 8 general purpose registers and 8 SSE registers.

This means that there's considerably less register pressure, which means that gcc won't have to temporally store data in memory as often, since it might need to free up a register. Which, in turn, means that the code runs faster.