https://gcc.gnu.org/bugzilla/show_bug.cgi?id=104116
--- Comment #3 from Avinash Jayakar <avinashd at gcc dot gnu.org> --- I had a couple of questions here 1. The normal mod operation is not vectorized in powerpc64le. During analysis phase of the vectorization, the function vect_analyze_loop_costing returns 0 for powerpc, and hence the actual vectorization does not happen as 0 means there is no benefit. Was this always the case? Before going into detail I just wanted to know if this was expected. 2. For this particular case of floor mod mentioned in the test case in commit https://gcc.gnu.org/cgit/gcc/commit/?id=95f6eb7ae707482fdeed57b0906dacb8e675385d we have to handle the expansion of the vector operation right (by perhaps adding a define_expand for these operations) instead of adding the adjustment code in tree-vect-patterns.cc?
