https://gcc.gnu.org/bugzilla/show_bug.cgi?id=112678
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rguenth at gcc dot gnu.org,
| |sebastian.huber@embedded-br
| |ains.de
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Hmm, the -fprofile-update mode default didn't change, did it? I'm guessing
that for some reason the generated code is different for
-fprofile-update=single now?
Note IIRC "compile-time" for SPEC with PGO includes the training run itself,
so that might be also the reason for the slowdown (runtime performance
regression of the instrumented binary).