Jan Engelhardt wrote:
> I repeat it again: glxgears is *NOT* an appropriate benchmark. All information is useful. We see a repeatable 30% difference, and you say it's meaningless. Pardon me if I'm skeptical, but you seem too quick to dismiss a result that is clearly telling us something. I've noticed that without exception, good video hardware plus good drivers gives high numbers, while low quality video hardware and/or poorly written drivers gives low numbers. Feel free to cite an "appropriate" graphics benchmark, and I'll be happy to run it. Obviously I won't be at all surprised if whatever benchmark you mention also shows better performance on a 64-bit OS. >> Being able to move memory around in bigger chunks can really help things like >> database performance too. *Having* to move data around in bigger chunks might >> not be as helpful for some scenarios, but I can't think of any. Got a >> specific >> example? > > Quite all libc functions profit from it, memcpy using rep movsq as a prime > example. Yes, that makes sense. Joe -- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]