This is the mail archive of the
mailing list for the GCC project.
Re: PING^(n+1) fwprop merge (ping 6 or 7 by now)
Steven Bosscher wrote:
These submissions claim to be mostly performance neutral on SPEC, with a
slight hit for PPC,
No slight hit on PPC. I appreciate you reply to my ping, but it would
have been helpful if you've read Paolo's complete story. The PPC
regression was fixed already even before these fwprop patches were
Correct me if I'm wrong, but the only actual numbers for the final patch
I recall seeing were for ix86. I asked for more (after all we're
deleting optimization code here and should make sure there aren't any
serious regressions), but I don't think anything was ever posted.
Providing numbers would be more useful than pinging the extra n times.