On 23 November 2015 at 16:31, staticx wrote: > Hi Andrew, > > I am not saying that. > > My findings are : number of default optimizations in O2 = number of default > optimizations in O3 < number of default optimizations in O1. How do you count "number of optimizations"? > This is true when we turn off all corresponding optimizations that are > controllable through CLI. > > I tested that by myself. I can give you a reference paper arguing that. > http://users.elis.ugent.be/~leeckhou/papers/cgo08.pdf > > Please check Section 6 in this paper. Section 6 doesn't seem to say anything about optimizations that aren't controlled by a -fxxx flag.
