When first looking at the testcase in pr13844 my first idea was caching the visibility computation. I did some simpler fixes first, but decided to try caching too.
The attached patch implements it, but after the previous fixes we are now spending most of the time of that testcase just linking the decls, so the performance impact is in the noise. It might still be a good idea to do it since it provides an assert on what conditions can change the computed visibility. Cheers, Rafael
t.patch
Description: Binary data
_______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
