Path ping

Jan Hubicka hubicka@ucw.cz
Tue Jun 30 13:39:00 GMT 2009


Hi,
I would like to ping:
http://gcc.gnu.org/ml/gcc-patches/2009-04/msg02077.html

Together with 
http://gcc.gnu.org/ml/gcc-patches/2009-05/msg00036.html
that I apparently forgot to commit and will do so after re-testing and
IPA-SRA merge it keeps need to have inline limits up in order to make
tramp3d/botan/wave happy.  I tried merging inline limit decrease from
pretty-ipa branch this weekend, but without these changes I still get a
performance regression on couple tests.
See recent results at http://gcc.opensuse.org/c++bench-frescobaldi/
and
http://gcc.opensuse.org/SPEC/CINT/sb-frescobaldi.suse.de-head-64/recent.html
http://gcc.opensuse.org/SPEC/CFP/sb-frescobaldi.suse.de-head-64/recent.html

In general those I think are already pretty good code size/compile time
improvements at almost no runtime cost in average, but it is better to
avoid regressions when not neccesary.

Honza



More information about the Gcc-patches mailing list