On Thu, 12 Dec 2002, Tesla 13 wrote: > > > Even if I don't gain anything, I just don't want my binaries to run on a > > > bloody 386 SX. Period. > > > >If procinfo shows 91% idle (as it does on my Athlon), then how much is > >there to gain? 10% of 9% is bugger-all. > > I CPU used is about 3%. But that really is not the point as I explained > above. What's it matter if your binaries _can_ run on a 386? From the information you've provided, I can see no way you will recoup the time spent rebuilding the software. However, if you wish to persist with building everything yourself, the the project for you is "Linux from Scratch." You _do_ get to build everything, and you don't even get coerced into using Red Hat's suboptimal (for you) configuration choices. > Also are you suggesting that P4 has no difference from a 386 except higher > clock speeds (this holds true between P3 and P4 though)? Not at all. However, if your machines are idle 97% of the time as you said, the best you will achieve is having them idle 98% of the time. > > >In my experience, that might mean that there's a problem that hasn't > >bitten you yet. > > I do not understand why anything would go wrong. I am not here to try all > available GCC flags. A simple march=pentium4 looks sufficient. Oh, the problems of youth. Mostly, software has bugs. Bugs are most likely to be round code that's been changed most recently. Guess where some of the most recent changes to gcc are? I'd guess code generation and optimisation for the Pentium IV. How do you tell whether software is buggy? one way is to look at its recent history. If there have been a lot of bugs found recently, there's a good chance there's a lot more to find. After all, the more there are the easier it is to find some. Judging from the number of iterations of gcc 2.96 (Red Hat's pocket version to be sure) and the short life of gcc 3.0, a lot have been found and there are more to be found. Probably those that remain are fairly well hidden. _______________________________________________ Redhat-devel-list mailing list Redhat-devel-list@redhat.com https://listman.redhat.com/mailman/listinfo/redhat-devel-list
