On 11/24/2016 08:16 AM, Richard Biener wrote:
IMHO switching insn_rtx_cost to be based on not just set_src_cost is
a good idea, but will require re-tuning of all targets, so it is not
stage 3 material.
Agreed.
That we compare different kinds of costs (which really has no meaning at
all, it's a heuristic at best) in various places is a known problem, not
a regression.
But technically stage 3 is for general bugfixing, not only regression fixing.
I'd say be prepared to revert but wait to see who screams first.
Right. And I would claim that we're early enough in stage3 that
attempting to address this BZ is a good thing. The BZ also happens to
be a 6/7 regression.
So I'd say let's go with the patch, but be aware that there may be a
need to twiddle other ports. If we find a bunch of ports are
problematical than we might need to think about reversion.
jeff