https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122608
--- Comment #4 from Jeffrey A. Law <law at gcc dot gnu.org> --- Given the code looks like this: b_2 = a_1(D) > 0; t_3 = b_2 ? 10 : 20; t_4 = t_3 + 20; ISTM that we could have a match.pd pattern which optimizes into the b ? 30 : 40 form. Essentially given X op C where X comes from a COND_EXPR where the true/false arms are constants we can compute trueval op C and falseval op C and use thoes results in the COND_EXPR.
