PATCH: gcc/config/mips/mips.h reorg

[Eric Christopher]

> 
> invoke.texi will need a patch describing -march= for MIPS.

Ok.


> While you are not introducing a regression, I think this is wrong.
> This means the compiler will emit branch-likely instructions for the
> TX39 even if the user has specified -mips1, which is supposed to
> restrict code generation to the MIPS I ISA.
> 

Currently, yes.  However, my patch doesn't change any behavior.

> I suggest my unreviewed patch for this...
> 
> http://gcc.gnu.org/ml/gcc-patches/2001-04/msg01024.html

I don't like this way long term.  I was planning on doing it a different
way, but submit patches piecemeal, not one giant patch.

Basically, I plan on having the -mipsX options correspond to code
generation in -march=.  -mcpu will be for scheduling only, unless it is
the only option on the line - at which point for compatability purposes
it will be used for code generation as well.

Any thoughts on this plan?

-eric

-- 
Pain is not the cleanser