Patch 2/9: Split up and reorganize some functions

[Bernd Schmidt]

On 06/21/2010 05:54 PM, Jeff Law wrote:
> I wasn't disagreeing with whether or not this change is more correct --
> in fact, I'm in total agreement that it's more correct.  I merely had a
> concern that if we had a port where the number of hard regs needed to
> represent a particular mode varied within the class was going to cause a
> problem.

Yeah, but I don't think that this part of the code is affected at all by
that particular issue.  It's an estimate of register pressure, and it
always uses the maximum nregs that could be needed by a class/mode
combination.  What am I missing?
The issue of variable nregs inside one class is something I have to deal
with in the final patch (which I expect to post sometime later this
evening, test results look good so far).

> I'd say, give them another 48hrs or so to object, and if none occur, go
> forward with the patch.

Will do.  Thanks.


Bernd