emotion said:
Not that I beleive 64 bit to be that much of an advantage (apart from marketing) the Core2 chips will be along and shipping in 3 or 4 months.
Speaking generally, no, 64 bit isn't an advantage, unless you need the extra memory space that you can access that way.
In the context of x86, though, it is a definite advantage: standard 16 bit x86, all the way through to 32 bit x86, have fourteen registers: AX, BX, CX, DX, CS, DS, SS, ES, DI, SI, SP, BP, IP, and FLAGS.
FLAGS is just a set of flags, set after various operations (such as compares), so isn't a general purpose register. IP is the instruction pointer, and again, isn't general purpose. CS, DS, SS, and ES are all segment registers, so can't be used for general purposes. So you're left with eight registers that can, possibly, be used in a general manner, although you can usually count on two of them (at least, possibly four) being used in other ways.
In comparison, PowerPC has 32 general purpose registers. Registers are the fastest form of memory you can access. You do the maths.
AMD, when they designed the 64 bit extensions to x86, introduced new registers as well, bringing the available registers to 16. These are only available to 64 bit code. More registers -- especially on a register-starved architecture like x86 -- mean that the code doesn't have to access main memory quite so much, so it will run that much faster.
That's why 64 bittedness (is that even a word?
) is a win on x86. Not because it's a bigger number, but because of architectural improvements under the hood. Put those extra registers into the 32 bit x86 CPUs, and 64 bittedness becomes a loss (relatively speaking), except when it's specifically needed for the extra RAM. You'd still need to recompile the code to use those extra registers, though, and such code would not run on the older CPUs.
