https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113247
--- Comment #5 from JuzheZhong <juzhe.zhong at rivai dot ai> --- (In reply to Robin Dapp from comment #4) > The other option is to assert that all tune models have at least a vector > cost model rather than NULL... But not falling back to the builtin costs > still makes sense. I have talked with kito. He just want rocket sifive-series-7 those tune chip has no vector hardware so I think it makes more sense we adapt builtin_vectorize_cost to set cost to generic vector cost when there is no vector cost in tune info.