optimization flag confusion

[grs2103]

Looking at the reference guide, there are a lot of optimization flag options. Assuming I'm executing my program on the same system I'm compiling and linking it (OS X core2duo), is there, in general, anything other than the -fast flag I should try? Will it automatically optimize for the architecture, or should I be giving it a -mtune flag?

[TimP]

-fast does select the option tieing the code most closely to the build CPU.
As you have dual core, you can do a quick check for threading opportunities with -parallel.
The first priority should be to assure correctness, and to make sure that you don't lose any accuracy by not invoking '-assume protect-parens -prec-div -prec-sqrt' (or maybe -fp-model source, which includes those).

[grs2103]

Quoting - tim18

-fast does select the option tieing the code most closely to the build CPU.
As you have dual core, you can do a quick check for threading opportunities with -parallel.
The first priority should be to assure correctness, and to make sure that you don't lose any accuracy by not invoking '-assume protect-parens -prec-div -prec-sqrt' (or maybe -fp-model source, which includes those).

I gather compiling with the flag -fast -fp-model precise will optimize as much as possible while retaining floating point accuracy?