http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37734
Siarhei Siamashka <siarhei.siamashka at gmail dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rearnsha at gcc dot gnu.org
--- Comment #3 from Siarhei Siamashka <siarhei.siamashka at gmail dot com>
2010-10-04 23:19:26 UTC ---
So if I understand it correctly, there are 2 independent performance issues
here:
1. one in the middle-end (redundant comparison with -1) when -O2 optimization
is selected.
2. another in ARM target, because it fails to produce efficient code with -Os
optimizations, while x86 can.
I just remembered that Mozilla has been using -Os optimizations up until now
because it was providing the best performance for them:
http://gcc.gnu.org/ml/gcc/2010-06/msg00715.html
I wonder if this particular missed-optimization issue is contributing to the
occasional performance advantage of -Os over -O2 (other than smaller code size
and reduced pressure on the instructions cache). Anyway, when looking at any
code generated by gcc, simple loops and branches always tend to contain
redundant instructions.
