>>! In D6391#4, @jroelofs wrote: > These are still useful to make sure we don't regress too badly on typeinfo > lookup performance, so I think we should default this to 'on'. > > Then we can turn it off in the configurations where it makes for unnecessary > failures (like on my baremetal stuff where there isn't a monotonic clock).
Sounds good. I'll make the change. > It is a bit lame that we don't have a way to automatically catch changes in > the perf numbers that come out of this though. Any ideas on how to fix that? Yes. Once the suffix test patch gets through review I'm going to start working on something. I actually have an internship starting in January to solve exactly this problem. http://reviews.llvm.org/D6391 _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
