>> We are missing several vec_extract chances because the current autovec >> patterns are not comprehensive. In particular we don't extract from >> pseudo-VLA modes that are actually VLS modes (just VLA modes in name). >> >> Rather than add even more mode combinations to vec_extract, this patch >> removes the expanders and replaces them with a dynamic approach in > > Wait, this patch does not remove anything? have you forgotten to add > those changes into this patch?
Oh, right. I forgot to remove that part from the patch description. The dynamic approach is not comprehensive enough and we'd miss some vectorization opportunities (where the vectorizer explicitly ask for vec_extract). So I didn't end up doing it. We have discussed other approaches for the vectorizer but it's clear that a different approach is needed. -- Regards Robin
