Re: [0/3] Fix PR78120, in ifcvt/rtlanal/i386.

[Bernd Schmidt <bschmidt at redhat dot com>]

On 11/24/2016 03:53 PM, Segher Boessenkool wrote:
> Your own 2/3 shows my point: you needed fixes to the i386 port for it
> to behave sanely after this 1/3; what makes you think other ports are
> not in the same boat?

When I realized i386 was broken I had a look at aarch64 and it looked 
sane, and that was the basis of the idea for patch 2/3.

It's possible other targets may need to handle SETs as well. In theory, 
something like the block of code I added for i386 could just work when 
copied to other backends if they have no "case SET" yet. If you do run 
into problems please try at least that very simple fix.

> 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.

It leads to observable regressions however as PR78120 shows, when code 
like ifcvt tries to use cost calculations in apparently-sensible ways. 
And, as Richard mentioned, we're not in a regressions-only phase.


Bernd