"Nadav Har'El" <[EMAIL PROTECTED]> writes: > Maybe you're referring to "gentoo" that do such a compilation.
IIRC Mandrake used to compile for i586 in good old days, while RH compiled for i386. > > It's very hard to obtain more than 10%-15% speed increase from gcc just > by compiling for a specific processor (say 686) instead of 386. I believe, > though, that even if there were a 30% speed increase in desktop applications > you wouldn't be able to notice it... I agree with what you said in general, nadav, but there might be a wrinkle on this last argument. AFAIK, perception of interactive apps' performance is non-linear (as are the other 5 senses): in particular there is a response threshold above (or below) which the user gets annoyed. When I was working for a company that dealt with presenting the results of computations to the user on a screen, the vague rule was that the computation and the rendering had to finish within 1.5 seconds - above that it was noticeable and annoying. And that's when users were perfectly aware that the system was doing heavy number-crunching behind the scenes. > Of course, a distribution can also contain statically-linked executables > for improved performance - but I've yet to see any of those (other than > Embedded Linux distributions). I doubt this will lead to significant performance difference either, but again, it may be important in some hard RT systems. -- Oleg Goldshmidt | [EMAIL PROTECTED] ================================================================= To unsubscribe, send mail to [EMAIL PROTECTED] with the word "unsubscribe" in the message body, e.g., run the command echo unsubscribe | mail [EMAIL PROTECTED]