On Dec 17, 2013, at 12:39 PM, Reid Kleckner <[email protected]> wrote:
> On Tue, Dec 17, 2013 at 12:19 PM, David Blaikie <[email protected]> wrote: > b) -flimit-debug-info is worth, at a guess, somewhere between 1 and 5%. This > vtable optimization is worth closer to 20%. That's /serious/ bloat to > consider accepting. > > I think you could address the concerns that Apple has without losing the 20% > gain by emitting debug info for base classes in the TU with the key function > for the derived class, and not all TUs requiring a complete type. Does that > make sense? > > Somebody should implement that and compare object file size with gcc on some > non-clang project, because we don't actually use that many polymorphic > classes. This might be what is currently implemented. The problem here is that people are linking against a class from another shared library where there will never be a TU that contains the full definition. _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
