Well, binaries optimised fpr StrongARM will in most cases be faster on XScales than binaries optimised for ARM6. But of course you could also make three binaries: one for ARM6, one for StrongARM and one for XScale. Since RISC OS programs are quite small anyway, this would not be a problem, you could still fit hundreds of them onto a CD (for commercial apps) and most users have fast internet connections anyway these days, so it would also not be problem for people who download the software.
And I never said "significantly faster".
I kind of get the feeling that you are trying to damage my reputation. You have no way of knowing wether I checked my facts. You only know that I didn't provide evidence to back up my *opignion*. And to be honest I have better things to do than to argue with you about such trivial things, so I can tell you right now that I will not take the trouble to back my claims up.
And concerning benchmarks: One can prove just about anyting with benchmarks. Just pick the right one and set the right compiler options and you should get the results you wanted.
And just in case you come round with the "this is technical discussion" argument again: I do not consider this level of discussion to be "of a technical nature". IMHO such terms are reserved for in-depth discussions of high-frequency electronic and digital IC technology (not sure these are the correct english terms for what I mean) and similar things.
Anyway, I've had just about enough of this kind of "discussions", so I have a feature request for drobe: Could you please implement some kind of ignore list list function, wich lets me create a set of users whos comments and forum articles will not be displayed to me?