This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug middle-end/84200] [8 Regression] r256888 causes 30% performance regression of 519.lbm_r at -Ofast generic tuning on Zen
- From: "marxin at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Mon, 05 Feb 2018 13:37:51 +0000
- Subject: [Bug middle-end/84200] [8 Regression] r256888 causes 30% performance regression of 519.lbm_r at -Ofast generic tuning on Zen
- Auto-submitted: auto-generated
- References: <bug-84200-4@http.gcc.gnu.org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84200
--- Comment #2 from Martin Liška <marxin at gcc dot gnu.org> ---
Confirmed, thank you Martin for reporting that. I was able to reproduce that on
Zen.
I see ~25% regression on train size and reverting following predictor helps for
me:
-DEF_PREDICTOR (PRED_LOOP_EXIT, "loop exit", HITRATE (89),
+DEF_PREDICTOR (PRED_LOOP_EXIT, "loop exit", HITRATE (85),
I'm going to measure also ref size and then I'll analyze which loop is
affected.