search for: isrealprofileavail

...data? At least, comments like the one I quoted need to be updated to > indicate that a static profile is considered precise enough. > > > I think the existing code that checks 'hasEntryCount()' for 'isProfileAvailable' information will need to be refactored to query 'isRealProfileAvailable'. > > Also, one other question: in the case of FullLTO, how do we ensure that > the synthetic profile data from static heuristics is consistent across > TU's? Do we need to recalculate the synthetic profile data? If we need to > recalculate it, presumably we need to kno...

...pdated to indicate that a static profile is considered precise >>> enough. >>> >>> >>> >> I think the existing code that checks 'hasEntryCount()' for >> 'isProfileAvailable' information will need to be refactored to query >> 'isRealProfileAvailable'. >> >> >>> >>> Also, one other question: in the case of FullLTO, how do we ensure that >>> the synthetic profile data from static heuristics is consistent across >>> TU's? Do we need to recalculate the synthetic profile data? If we need t...

On Fri, Dec 15, 2017 at 12:22 AM, Sean Silva <chisophugis at gmail.com> wrote: > IIUC, this proposal is just saying that we should infer a static profile > for entry counts just like we do for branch probabilities. In the case of > entry counts, we do not hide that information behind an analysis like BPI, > so currently just annotating synthetic PGO entry counts is a simple >