This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/78348] [7 REGRESSION] 15% performance drop for coremark-pro/nnet-test after r242038
- From: "rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Tue, 15 Nov 2016 08:49:00 +0000
- Subject: [Bug tree-optimization/78348] [7 REGRESSION] 15% performance drop for coremark-pro/nnet-test after r242038
- Auto-submitted: auto-generated
- References: <bug-78348-4@http.gcc.gnu.org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78348
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rguenth at gcc dot gnu.org
Target Milestone|--- |7.0
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Note that -O3 enables -floop-distribute-patterns which detects open-coded
memcpy/memmove/memset. The idea is that canonicalizing leads to optimal
inline expansion or library code.
I'm not sure I would count this as two vs. three memory streams as usually
improving things on the store side helps (this is how the current heuristic
works). CPUs are usually more store-bandwith limited. But yes, some
sort of target specific parameter might help here (number of store streams
we can handle).