https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84204
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> --- The limit you set to scev-max-expr-size is quite low but I expect it just needs a more complicated testcase to trigger this with larger values. We apply this limit in tree-chrec.c very inconsequential so I really wonder if we need to limit it there at all. When scev-max-expr-size was introduced it was _only_ taking effect there, now we also check this limit in instantiate_scev_r and have an additional limit, scev-max-expr-complexity enforced during analyze_scalar_evolution. Ripping this limit out again from chrec_fold_plus_1 doesn't regress the original testcase for what it was introduced for.
