> On 2017.11.05 at 11:55 +0100, Jan Hubicka wrote: > > > On 2017.11.03 at 16:48 +0100, Jan Hubicka wrote: > > > > this is updated patch which I have comitted after profiledbootstrapping > > > > x86-64 > > > > > > Unfortunately, compiling tramp3d-v4.cpp is 6-7% slower after this patch. > > > This happens with an LTO/PGO bootstrapped gcc using > > > --enable-checking=release. > > > > our periodic testers has also picked up the change and there is no compile > > time > > regression reported for tramp3d. > > https://gcc.opensuse.org/gcc-old/c++bench-czerny/tramp3d/ > > so I would conclude that it is regression in LTO+PGO bootstrap. I am > > fixing one checking > > bug that may cause it (where we mix local and global profiles) so perhaps > > it will go away > > afterwards. > > Just to confirm: pure PGO bootstrap is fine, e.g. on Ryzen: > (LTO/PGO) 17.65 sec ( +- 0.68% ) > (PGO) 15.74 sec ( +- 0.27% )
Thanks. I have comitted the patch for inlining profile update bug, so with some luck LTO/PGO may be fine again. I will look for other insanities after chasing out the ICEs. callgraph is currently still having frequencies which I plan to drop next. With that the profile updating will become bitmore consistent over whole program. IPA profile updates will be fun with LTO though - even for gcc we link libbackend into multiple binaries and thus the whole program profile is not really whole program after all. It is bit of the common libbackend profile combined with frontend profiles and thus it can easilly accumulate to false zeros. Honza > > -- > Markus
