Hi, This link covers exactly what I need, thanks. ---- With best regards, Konstantin On Tue, Jul 24, 2012 at 8:03 PM, Jonathan Wakely <jwakely.gcc@xxxxxxxxx> wrote: > On 24 July 2012 12:25, Konstantin Vladimirov wrote: >> Hi, >> >> Please point me to documentation where relation between specifying: >> >> flto=x >> >> and >> >> flto-partition=y >> >> described completely over the domain of possible (x, y) pairs. >> >> I experience troubles with deducing possible values and their effect >> from source code of gcc, google isn't helpful too. >> >> Now I am trying to solve one particular problem with inlining (lto >> compilation) in private backend (core compiler is gcc 4.6.3). >> >> In brief: there is some code A, containing functions a() and b(). >> Being compiled with lto, b() is inlined in a() as it supposed to be. >> Now consider larger code B. B have no references to both a() and b(), >> but if we compile A+B with lto, b() is no more inlined into a(). >> >> I suppose problem is in LTO partition -- in situation of more code, >> partitioned part may contain some other functions b() refers to, so >> b() now can not be inlined in the presence of inter-module references. >> But I am not sure what options I can specify to control partition and >> tune lto compilation to catch problem between options, not between >> sources. > > Did you look in the manual? > > http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-flto-873
