[Bug rtl-optimization/58295] [4.8/4.9 regression] Missed zero-extension elimination in the combiner

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58295

--- Comment #6 from H.J. Lu <hjl.tools at gmail dot com> ---
(In reply to Eric Botcazou from comment #5)
> Yes and, although x86 is the dominant architecture, it shouldn't be allowed
> to penalize all the others.  I think we should restrict the effect of
> r191928, in particular it makes little sense as-is if you also define
> PROMOTE_MODE as on most RISC architectures.

We can limit the transformation to (truncate:SI (op:DI (x:DI) (y:DI))).