https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80520
Marc Glisse <glisse at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2017-04-26
Summary|Performance regression from |[7/8 Regression]
|missing if-conversion |Performance regression from
| |missing if-conversion
Ever confirmed|0 |1
--- Comment #1 from Marc Glisse <glisse at gcc dot gnu.org> ---
clang generates much slower code (and even worse with libc++).
With a combination of -march=native and -fprofile-{generate,use}, I can get
faster code from gcc-7 than from gcc-6, while the .optimized dump still has
branches and not cond_expr.
