On quarta-feira, 24 de fevereiro de 2016 22:28:23 PST Jean-Michaël Celerier wrote: > On Wed, Feb 24, 2016 at 8:50 PM, Thiago Macieira <thiago.macie...@intel.com> > wrote: > > no compilers is currently able to merge implementations that > > produce the same code, if they come from different functions. > > Isn't this what MSVC's /OPT:ICF ( > https://msdn.microsoft.com/en-us/library/bxwfs976%28v=vs.140%29.aspx)
I've never seen that before. I'll need to take a better look. > and gcc's -ffunction-sections -fdata-sections --gc-sections No. > or gold's --icf Kinda. As the link you pointed out: > https://gcc.gnu.org/ml/gcc-patches/2014-06/msg01499.html Says, a compiler-based ICF is more efficient than a linker-based ICF. The linker can only do ICF if you've compiled with -ffunction-sections and can only do it at function granularity. The compiler can do better. > > do ? (although maybe not with a 100% hit rate). > > A quick research has shown that clang contains a "mergefunc" option but it > does not seem to be production-quality. It looks like the GCC option -fipa-icf is already enabled at -O2 level and has been for some time. That is good, but given that I am still seeing duplicate code where strict analysis says it isn't needed, it means there's still a long way to go for the compilers to do the heavy-lifting. Helping the compiler is still necessary. -- Thiago Macieira - thiago.macieira (AT) intel.com Software Architect - Intel Open Source Technology Center _______________________________________________ Interest mailing list Interest@qt-project.org http://lists.qt-project.org/mailman/listinfo/interest
