On 11/17/2014 10:27 AM, Richard Biener wrote:
The generated code for g++.dg/torture/pr37922.C is pretty different at -O2,
but it's hard for me to tell whether the changes are good, bad, or neutral.
There is always the possibility of running the C++ portion of SPEC CPU 2006...
I ran the tramp3d benchmark over 500 iterations before and after the
change and couldn't see any measurable difference in runtime. The
binary with my change was 0.4% smaller.
I'm going to go ahead and check it in; if a performance hit shows up on
the automated testing we can revisit the choice.
