[PATCH] Regression caused by reload_cse_simplify_operands bugs

[Ulrich Weigand]

Ian Lance Taylor wrote:

> This is OK.  Thanks.

Committed, thanks!

> I guess nobody noticed since it most be rather rare that there would
> be two different alternatives available anyhow.  It's particularly
> hard to imagine it happening on a RISC chip.

Well, it's worse: it will sort *all* alternatives, even those that
aren't available -- and those will have alternative_nregs equal to
zero, so due to the inverted sort order they will come first ...

Bye,
Ulrich

-- 
  Dr. Ulrich Weigand
  Linux on zSeries Development
  Ulrich.Weigand@de.ibm.com