Matthew Dillon wrote:
Similarly, some bits of pc32 can be moved to the cpu hierarchy... but
ONLY those bits that are also needed by the virtual kernel (otherwise
they ARE platform and not cpu specific bits and have to stay in pc32).
I think we need to have something in between: a CPU specific section of the
platform code.
So: ACPI, APIC, 90% of all headers in platform/pc32/include, stuff like that,
all of this is the same between pc32/i386 and pc64/amd64.
Now. Different are bits like exception handling, etc. Basically everything in
a .s file or containing inline assembler.
So where does us leave this? We need an intermediate section, which should be "cpu
specific platform code". I think this does make sense. If somebody has a better
idea on how to structure it while avoiding duplication, I'm all ears. For example, I
think having platform specific cpu code seems kind of wrong.
There will be some near duplication of code but the purpose of having
pc64 is to provide a place where work can progress without causing the
pc32 builds to start failing and crashing.
Nobody intends to break or crash the pc32/i386 code base :) But still I am
sure that we need to restructure to split the pc32 code to make it really
represent the ibm pc architecture. And I think this is very much possible and
clean to do.
cheers
simon
