I dont remember where I got the -xO3 idea, so thanks for the correction.
As for non-optimized code, many developers strongly resist (read "refuse) the
idea of turning on the optimizer because they dont have access to local
variables for debugging. So except in the case of the simplest most rigorously
tested modules, optimization is a non-starter for production code. Thus, I
would wish to use hwcprof for finding ways to address performance at the source
code level.
I sympathize with them, actually. It would seem to me that at least in the
case of -xO1, simple peephole optimization shouldnt be incompatable with having
access to locals. However, it turns out that code optimized at -xO1 is actually
worse than non-optimized code in both size and runtime. At least for the sample
of cases I've seen.
Regards
This message posted from opensolaris.org
_______________________________________________
tools-discuss mailing list
[email protected]
