https://gcc.gnu.org/bugzilla/show_bug.cgi?id=119872
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rguenth at gcc dot gnu.org
--- Comment #7 from Richard Biener <rguenth at gcc dot gnu.org> ---
Please make sure to not "fix" something where the input is already wrong - see
the various issues where SCEV produces an invalid CHREC - forming a chrec
is association.
Using twice wide types within the same context is wrong IMO, the intent of the
previous change was to identify IVs that end up with the same sequence of IV
values, and we should preserve that. I don't see whether this is actually a
case where the IV values are not the same? It seems we're using one in place
of the other in a context where the evolution expression itself is re-used (but
the one with undefined overflow) and that's the wrong thing to do?
