Bernd Edlinger <bernd.edlin...@hotmail.de> writes:
> On Sat, 4 Jul 2015 09:04:41, Richard Sandiford wrote:
> >
> > The final return here would also mishandle SEQUENCE PATTERNs.
> > The idea was that this function would only see "real" instructions,
> > so I think instead the FOR_EACH_SUBINSN should be here:
> >
> > static bool
> > mips_find_gp_ref (bool *cache, bool (*pred) (rtx_insn *))
> > {
> > rtx_insn *insn;
> >
> > if (!*cache)
> > {
> > push_topmost_sequence ();
> > for (insn = get_insns (); insn; insn = NEXT_INSN (insn))
> > ---->
> > if (USEFUL_INSN_P (insn) && pred (insn))
> >
> > Thanks,
> > Richard
>
> Yes, I agree.
>
> I have now updated my patch as suggested.
>
> A mips-img-linux-gnu cross compiler builds as expected.
>
> OK for trunk?
OK, thanks.
Matthew
