Yep. We'll make it better by limiting the size, but trivially copyable is an improvement, as there are orders of magnitude more loops over small copyable types than over large ones.
On Sat, Sep 26, 2015, 9:02 PM comex <com...@gmail.com> wrote: > On Thu, Sep 24, 2015 at 7:28 AM, Manuel Klimek via cfe-commits > <cfe-commits@lists.llvm.org> wrote: > > Fix loop-convert for trivially copyable types. > > If I'm not misunderstanding the patch, "trivially copyable" by itself > seems like a suboptimal test. After all, "trivial" can still include > an expensive memcpy of a large struct. >
_______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
