https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66726
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
See Also| |https://gcc.gnu.org/bugzill
| |a/show_bug.cgi?id=112418,
| |https://gcc.gnu.org/bugzill
| |a/show_bug.cgi?id=112324,
| |https://gcc.gnu.org/bugzill
| |a/show_bug.cgi?id=64700
--- Comment #22 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Jeffrey A. Law from comment #7)
> The other thing to keep in mind, sinking of this nature ought to be
> applicable to other unary ops and cases where we have multiple PHI args that
> are SSA_NAMEs. It shouldn't be structurally limited to just conversions
> where one arg is an SSA_NAME and the other a constant.
Just an FYI on the above comment; other unary ops was implemented with
r14-575-g6d6c17e45f62cf (and improved to some builtins with
r15-2986-gcd2f394418be0c).
Multiple phi args is recorded as PR 112418, I have a patch which needs
improvement on the cost model side.
Also we can handle binary ops too which is recorded as PR 112324 (and PR
64700), I have a patch there but has some regressions where other passes need
improvement still (PR 64700 has those linked).
