https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109612
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> --- We have > ./cc1 -quiet t.i -w -O3 -g -fdump-tree-optimized > wc -l t.i.254t.optimized 4375637 t.i.254t.optimized > ./cc1 -quiet t.i -w -O3 -fdump-tree-optimized > wc -l t.i.254t.optimized 1701 t.i.254t.optimized so a case with almost everything dead and optimized away but a few million debug stmts left around. It looks like we should really try harder to nuke some of them. But yes, similar to PR109237
