On Wed, May 4, 2016 at 6:11 AM, Bernd Schmidt <bschm...@redhat.com> wrote: > On 05/04/2016 12:03 PM, Eric Botcazou wrote: >>> >>> The IBM LTC team has tested the benefit of -frename-registers at -O2 >>> and sees no net benefit for PowerPC -- some benchmarks improve >>> slightly but others degrade slightly (a few percent). You mentioned >>> no overall benefit for x86. Although you mentioned benefit for >>> Itanium, it is not a primary nor secondary architecture target for GCC >>> and continues to have limited adoption. Andreas also reported a >>> bootstrap comparison failure for Itanium due to the change. >> >> >> ...which had nothing to do with -frename-registers but was a latent issue >> in >> the speculation support of the scheduler, see the audit trail. > > > Yeah, I'd forgotten to add this to the list of issues found. Also, Alan's > morestack representation issue. > > Given how many latent bugs it has shown up I think that alone would make it > valuable to have enabled at -O2.
We don't enable optimizations by default to find latent bugs. That is a very nice byproduct, but does not justify enabling a seemingly ineffective optimization. If this change is beneficial, why can no one provide evidence of overall benefit? All of the responses have referred to other, secondary effects or anecdotal evidence. Does it really still help PR59173 now? If this is needed for PR59173 and PR38825, add it to x86 override_options. I simply am asking for the same justification that is asked of all other optimization patches and default optimization changes. Thanks, David
