Stephan M|ller stephan_mueller@gmx.net wrote:
cool, will you add support for AMD processors? ie k5 and k6 etc?
What does a processor in sockel7-architecture has to do with the chipset? OpenBIOS is a x86-bios, so every x86-cpu (depending on of there's support for your particular chipset) can be used. The k5/k6 are basically the same as normal pentium-cpus, they will, of course, run with this code.
There should be some AMD (and other vendor) specific code, to enable features, MTRRs, setup cpuid to return correct values etc..
Read the BIOS writers guide for pretty much any non-Intel CPU vendor, and you'll get the idea.
To answer the original posters question, I think we should support as many different CPUs as possible. It's not easy to do such stuff in anything other than asm though, which it appears (to me at least) that Ron is trying to steer OpenBIOS away from.
Some of these things are already present in the kernel (check out arch/i386/kernel/setup.c). The question is, would it be better in the kernelspace, or BIOSspace. Which is more precious? :) The other question, is that as not everyone will be using OpenBIOS, so maybe it would be better in kernel. Better still would be a generic interface to MSRs in kernel space, that a userspace app could then turn on caches etc... Many people (myself included) have sent Linus drivers for MSR interfaces in the past, but he seems reluctant to include them. Maybe now that devfs is in 2.3.x, someone will have more luck creating a /dev/cpu/msr/ interface. In which case, this whole discussion is moot.
regards,