RFA: patch to solve PR37535

[Jeff Law]

John David Anglin wrote:
>>
>
> I would like to refer you to PR 37608,
> <http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37608>.
>
> While PR 37535 was a case with an input hard register and output
> clobber, PR 37608 is an early clobber intended to reserve a specific
> hard register to save and restore the PIC register across a call.
>
> Richard's change didn't fix PR 37608.  From the discussion, I'm
> somewhat hopeful that Vlad's latest proposal will handle this case.
> I'm going to give it a try.
>   
I don't offhand see how Vlad's change is likely to fix this problem.   
But I haven't actually looked the debugging dumps to be sure.

I guess my first question would be what code ultimately selects r4 to 
hold something other than the saved PIC register?  Was it allocation or 
reloading?

I have vague memories of code which prevented allocation/reloading from 
using explicitly mentioned hard registers to hold values, but I couldn't 
find it after a quick scan today.

Jeff