This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Enable SSA at -O0


But since in 4.4 there is no more any reason to use -O1 when alias analysis
blows up, why couldn't -O1 *be* -Og?


Well, do you suggest to "drop" -O1 in favor of simply -O2 and
"re-introduce" it with -Og behavior?

Yes, but just because...

I think this is more confusing
than adding -Og.  But we could certainly think of taking out some
optimizations from -O1 that hurt debugging, but I wouldn't go as far
as I suggested for -Og.

... I agree with the AdaCore guys' opinion was that in 3.x -O1 was -Og in practice, and it was then broken in 4.x. So the non-debuggability of -O1 in 4.x is in some sense a regression. If we can now declare to have a hold on tree-level compile-time problems, and that there is no need to keep -O1 as a workaround for those, it's time to fix that regression.

Paolo

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]