https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83338
--- Comment #5 from Marc Glisse <glisse at gcc dot gnu.org> --- (In reply to Jakub Jelinek from comment #4) > I believe the bug is that get_vectype_for_scalar_type for the operands of > POINTER_DIFF_EXPR is vector of unsigned elements, the result of > POINTER_DIFF_EXPR is a signed and thus we IMHO need to vectorize > POINTER_DIFF_EXPR as effectively MINUS_EXPR on the vector unsigned operands > producing vector unsigned result, and then VIEW_CONVERT_EXPR the result into > vector signed result. Yes, that's actually what I thought was already happening until I saw this PR... (I added the POINTER_DIFF_EXPR -> MINUS_EXPR replacement because of a missed optimization in the testsuite, so it must have at least partially worked) > The reason POINTER_PLUS_EXPR works by being mapped as > PLUS_EXPR is that both sizetype and pointers are TYPE_UNSIGNED and we get > the same vector types for both. That certainly simplifies things for POINTER_PLUS_EXPR.
