On Thu, Jan 12, 2017 at 9:26 AM, Robin Dapp <rd...@linux.vnet.ibm.com> wrote: >> Yes, for memset with larger element we could add an optab plus >> internal function combination and use that when the target wants. Or >> always use such IFN and fall back to loopy expansion. > > So, adding additional patterns in tree-loop-distribute.c (and mapping > them to dedicated optabs) is fine? Or does the yes refer to the > "else"/"or" part of my question (how would the backend recognize the > patterns then)?
Yes, enhancing tree-loop-distribution.c with extra patterns is fine (hey, there were supposed to be patterns for all of lapack & friends ... ;)) The question is only whether loop-distribution should always create the IFN or just if the backend has an optab so expansion is trivial (rather than needing to re-build a loop doing the operation). Richard. >> I'd say a multibyte memchr might make sense, but strlen specifically? >> Not sure. > > ok, memchr would also work for the snippet I have in mind. > > Regards > Robin >
