On Thu, 2005-06-16 at 18:37 +0400, Serge Belyshev wrote: > >bruteforcing, or I am missing something out? Is there a reason for such > >performance differences? > > Yes, You have used '-O3' flag which triggers -finline-functions > I've tried with -O2 flag before, and had similar results, gcc-4 soars! > Default inline heuristics just cannot fit well for all, and thus are subject to > change from time to time. See [1] and [2] for more information on this topic. > > To fix that, You have few choices: > o use -O2 > o tweak inlining heuristics yourself (see [3]) > o use -fprofile-generate/-fprofile-use, like this: > > $ cd john-1.6.38/src > $ make CFLAGS='-c -O3 -fprofile-generate' LDFLAGS=-fprofile-generate linux-x86-64 > $ ../run/john --test > $ make clean > $ make CFLAGS='-c -O3 -fprofile-use' LDFLAGS=-fprofile-use linux-x86-64 > doing the above still gives me similar results as with general compilation. > [1] http://gcc.gnu.org/ml/gcc-patches/2005-03/msg00333.html > [2] http://gcc.gnu.org/ml/gcc-patches/2005-03/msg01936.html > [3] http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html#index-param-571
Attachment:
signature.asc
Description: This is a digitally signed message part