Calgarypuck Forums - The Unofficial Calgary Flames Fan Community - View Single Post

photon · 03-26-2010, 11:16 PM

I guess that would depend on how it's being measured and what's being multiplied and such.

One of the benefits of 64bit is memory addressing but a benchmark like this won't benefit from that I don't think.

But if you try to add or multiply numbers that are bigger than the 32 bit registers on a 32 bit processor I think it has to do some fancy juggling to get the result, whereas with a 64bit processor the result will fit in the registers.. so in that case rather than like 5 or 10 operations to get the result, you'd only need the one multiply.

But since the benchmark is testing relative processor speed and not specifically 64bit vs 32bit operations I would think the benchmark would focus on stuff that works best.

Or maybe there's a standard set of tests to test integer and floating performance.

03-26-2010, 11:16 PM	#83
photon The new goggles also do nothing. Join Date: Oct 2001 Location: Calgary Exp:	I guess that would depend on how it's being measured and what's being multiplied and such. One of the benefits of 64bit is memory addressing but a benchmark like this won't benefit from that I don't think. But if you try to add or multiply numbers that are bigger than the 32 bit registers on a 32 bit processor I think it has to do some fancy juggling to get the result, whereas with a 64bit processor the result will fit in the registers.. so in that case rather than like 5 or 10 operations to get the result, you'd only need the one multiply. But since the benchmark is testing relative processor speed and not specifically 64bit vs 32bit operations I would think the benchmark would focus on stuff that works best. Or maybe there's a standard set of tests to test integer and floating performance. __________________ Uncertainty is an uncomfortable position. But certainty is an absurd one.