Re: [mips-rewrite] New target hooks & vr4130 support

[Richard Sandiford <rsandifo at redhat dot com>]

Richard Henderson <rth@redhat.com> writes:
> Ordinarily, I'd have said to look at the alpha port, which 
> has the same problem wrt ev4 and ev5.  We fix this up in
> machine_dependent_reorg.
>
> This is harder for you, since you've got that nasty dbr_sched
> albatross.  I'd might support having another pass of
> machine_dependent_reorg after dbr_sched.

Thanks for the alpha pointer.  I guess the eventual aim should
be to use the same sort of thing to fix up 4130 code as well.
But it does seem to assume that alignment can be accurately
predicted in gcc, and with the current state of the MIPS port,
that isn't true even after dbr_sched.  I think Eric's aim is
to have everything in '.set noreorder' eventually, but it's
going to be a lot of work...

Until then, the idea was to describe the kind of thing we
need in reorderable assembly code.

Richard