https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80838
Markus Trippelsdorf changed:
What|Removed |Added
Status|NEW |RESOLVED
Resolution|---
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80838
--- Comment #7 from Jan Hubicka ---
Author: hubicka
Date: Mon Jul 10 13:25:23 2017
New Revision: 250094
URL: https://gcc.gnu.org/viewcvs?rev=250094=gcc=rev
Log:
PR lto/80838
* lto-wrapper.c (remove_option): New function.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80838
--- Comment #6 from Markus Trippelsdorf ---
Nice! I now get a ~2% improvement on the tramp3d testcase with PGO/LTO.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80838
--- Comment #5 from Jan Hubicka ---
OK, found the problem. LTO binary is built with -fPIC because of the way we
merge options. I saw the same problem previously on Firefox. It is easy
mistake to do. I think we should
1) warn in this case
2)
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80838
Jan Hubicka changed:
What|Removed |Added
Status|UNCONFIRMED |NEW
Last reconfirmed|
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80838
--- Comment #3 from Markus Trippelsdorf ---
bootstrap-lto/PGO:
Performance counter stats for 'g++ -Ofast -w tramp3d-v4.cpp' (10 runs):
16381.906087 task-clock (msec) #0.998 CPUs utilized
( +- 0.32% )
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80838
--- Comment #2 from Markus Trippelsdorf ---
(In reply to Jan Hubicka from comment #1)
> If it is easy to do, can you attach profiles please?
> i will try to reproduce this...
Thanks.
(There is still some tooling missing (perf, etc.) on the
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80838
--- Comment #1 from Jan Hubicka ---
If it is easy to do, can you attach profiles please?
i will try to reproduce this...
Honza
