This is the mail archive of the
mailing list for the GCC project.
Re: reload question about unmet constraints
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: dj at redhat dot com (DJ Delorie)
- Cc: jim dot wilson at linaro dot org, gcc at gcc dot gnu dot org
- Date: Wed, 7 Oct 2015 17:21:35 +0200 (CEST)
- Subject: Re: reload question about unmet constraints
- Authentication-results: sourceware.org; auth=none
DJ Delorie wrote:
> For such constraints that are memory operands but not
> define_memory_constraint, you need to use '*' to keep reload from
> trying to guess a register class from them (it guesses wrong for
> I.e. use "*Wfr" instead of "Wfr".
Huh? That seems weird. It should make no difference for purposes
of register class preferences whether a constraint is marked as
memory constraint or extra constraint: neither contributes to
register class preferences at all.
If the '*' makes any difference, I guess this can only be because
IRA chooses another alternative for the insn to begin with.
Do you have an example that shows the problem?
Dr. Ulrich Weigand
GNU/Linux compilers and toolchain