On Tue, Apr 25, 2023 at 05:10:50PM +0200, Richard Biener via Gcc-patches wrote: > On Tue, Apr 25, 2023 at 11:34 AM Eric Botcazou via Gcc-patches > <gcc-patches@gcc.gnu.org> wrote: > > > > Hi, > > > > insert_debug_temp_for_var_def has some strange code whereby it creates debug > > temporaries for SINGLE_RHS (RHS for gimple_assign_single_p) but not for > > other > > RHS in the same situation. > > It indeed looks odd (likewise the GIMPLE_PHI handling should be > covered by is_gimple_reg ()). > > > Removing it saves 25% of compilation time at -g -O > > for a pathological testcase I have. > > > > Bootstrapped/regtested on x86-64/Linux, OK for the mainline? > > OK. > > probably also helps PR109612 and the other similar PR referenced therein.
Haven't looked into detail, but just saving compilation time shouldn't be the only factor when deciding about debug info stuff, another and perhaps even more important would be whether it affects the emitted debug info. One of ways to quantify that was running https://github.com/pmachata/dwlocstat before/after the change on something large, e.g. bootstrapped cc1plus (of course with the patch reverted and stage3 rebuilt so that it is the same code just with possibly different debug info). Jakub
