https://gcc.gnu.org/bugzilla/show_bug.cgi?id=125809
--- Comment #3 from ptomsich at gcc dot gnu.org --- (In reply to Drea Pinski from comment #2) > >Before ad3fb999a1b (and with our proposed fix) worker is not specialized: > >the one hot indirect call is not worth a clone under a guessed profile. > > I am not sure that is the issue causing the slow down though. > > > tree-ssa-sccvn.cc:process_bb (called from do_rpo_vn) is split into two > > context clones (iterate=0/iterate=1) instead of one. > > Are you running into more icache misses, more branch misses or does > something afterwards cause the slow down? > > Because for me cloning specifically for devirtualization seems like will > always be a win. > > I am not sure you are evulating the problem with the final code produced but > just getting at the symptom here. The .text section increases by ~1MB, so I expect issues on the frontend/i-cache side. I'll reconfirm with counters to see which event group(s) jump out. More, once the perf runs are in.
