Re: [PATCH] x86: Replace ne:CCC/ne:CCO with UNSPEC_CC_NE in neg patterns

[Segher Boessenkool <segher@kernel.crashing.org>]

On Fri, Oct 28, 2022 at 11:55:35PM +0200, Eric Botcazou wrote:
> > You mean in CCV?  That works yes, but only because (or if) the setter
> > and getter of the CC reg both use CCV (so never use any other flag at
> > the same time; CCV has an empty intersection with all other CC modes).
> 
> We're talking about CCC here AFAIK, i.e. the carry, not CCV.

Yes.  But it is all the same: neither signed overflow nor unsigned
overflow (of an addition, say) can be described as the result of an
RTL comparison.

The point is that all of this is put completely outside of all other
MODE_CC handling, and only works because of that.  And a small
modification to the backend, completely elsewhere, can make that house
of cards collapse.  It is much more robust to use a different relation,
not EQ, to decribe this.  Something with an unspec is fine.

But what the sparc backend does does work.


Segher