Re: simplify_subreg aborts bad?

[Jan Hubicka <jh at suse dot cz>]

> 
> > You can accomplish that by having one matching pattern with VOIDmode
> > and other mechanizm to generate that insn in expanders, that makes it to have
> > mode.
> 
> Thought of that too.  The optimizers aren't flexible enough in a few
> cases to deal with non-CCmode conditional jumps.  I don't recall the
> specifics as it was many months ago when I last worked on those
> patterns, but it had something to do with doloop optimizations.
I don't follow what went wrong.
In setCC instruction we were speaking, you don't use conditional jump.
Can you try to be more specific?  Is the issue reproducible on some
port available in the mainline?

Honza