>"Andrew R. Baker" wrote:
>> There are also Kevin's patches that do this. I have integrated them into
>> my working 2.2 tree while incorporating his floating point work. I can
>> pull just those together and see how well they will integrate into the 2.3
>OK. The only note, please, keep cpu probing _before_ prom_init. I'm
>making another my old patch that adds r3081 detection and now I have to
>determine machine the kernel is running on by cpu type: old Baget has
>r3000a while new one has r3081.
Don't worry, it happens at the same place in the flow as the old
in-line stuff, before prom_init and indeed before the bss area is
The whole overhaul started with the fact that I needed one bit
of information - whether an FPU was both present and enabled
(in the configuration sense, not the Status.CU1 sense) -
for the FPU emulator code. In the end I killed two birds with
one stone and got rid of a lot of the spaghetti switch statements
in the kernel based on CPU type and replaced them with
checks of flag bits representing capabilities and options.
It makes thing much easier to read and maintain IMHO.
Many thanks to Andrew for working to merge this stuff into
the SGI repository! It's something I've wanted to do since
the 2_2 branch became quiescent, but I have not had