david-arm added a comment. In D108138#2967100 <https://reviews.llvm.org/D108138#2967100>, @lebedev.ri wrote:
> IMO anything other than enhancing LV is wrong. Hi @lebedev.ri I personally disagree here. Adding support to LV for this is significantly more work (and IMO unnecessary) because there are cases when LV has to handle a lot more than just the obvious flattened vectorisation case using vector comparisons and select instructions. We will also need to add support for vectorisation factors of 1 (with interleaving) and cases where VF>1,but we have to scalarise the switch statement. These latter two cases require basically doing exactly the same thing as @kmclaughlin's patch does here, i.e. unswitching the switch statement into compares/branches and new blocks. It seems far simpler to have a small pass that runs prior to the vectoriser (when enabled) that unswitches. Not sure what others think here? Repository: rG LLVM Github Monorepo CHANGES SINCE LAST ACTION https://reviews.llvm.org/D108138/new/ https://reviews.llvm.org/D108138 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
