https://gcc.gnu.org/bugzilla/show_bug.cgi?id=77468
--- Comment #12 from Richard Biener <rguenth at gcc dot gnu.org> --- (In reply to Richard Biener from comment #11) > The recent 2nd bump on x86_64 is from > > ------------------------------------------------------------------------ > r243995 | hubicka | 2017-01-01 16:40:29 +0100 (Sun, 01 Jan 2017) | 7 lines > > > PR middle-end/77484 > * predict.def (PRED_CALL): Update hitrate. > (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors. > * predict.c (tree_estimate_probability_bb): Split CALL predictor > into direct/indirect/polymorphic variants. > > build uses -O3 -ffast-math -funroll-loops -march=native (core-avx2 in this > case) And it was good again for a short time, improving between r244241:244289 and regressing again r244382:244454. The former adjusts PRED_CALL the latter has the partial DSE adjustments.
