Re: Apple's -fast option for gcc

[Mike Stump <mrs at apple dot com>]

On Tuesday, September 16, 2003, at 11:15 PM, Andreas Jaeger wrote:
> The question is whether Apple likes to add -fast to GCC.  IMO -fast
> should enable all flags that help a specific set of benchmarks, the
> flags are target dependand.

In the end, we'd like -O2 or -O3 to subsume and contain most of it.  In 
the beginning, we implement interesting new things, usually control 
them via flags, so that we can tell what is working and what isn't, to 
be able to try a multitude of various options and see what combination 
gives what type of improvements, and then we put the ones that work out 
under fast.  Kinda like a hat hook, we just hang code there.  Also, 
SPEC doesn't want a command line that wraps 40 times...  :-)  I'd 
expect that the good ideas that don't have abi impact or language 
impacts to eventually be in O2 or O3.  Some of them won't ever go into 
-On.  An example of this would be -mdynamic-no-pic, it is used to 
specific that the code that we compile, isn't going to go into a shared 
library, the default being, the code _can_ go into a shared library.  
This isn't ever appropriate for -O2 or -O3, it is a different shape, 
like -fPIC, or -static.

-fast is very, very small.  It just turns on 5-25 other, existing 
flags.  There isn't a reason why it could not be contributed.  Do 
others want it?  It is _good_ for SPEC.  That is the intent and the 
primary purpose.  It may give people ideas on what things may help 
certain types of programs, but, like any benchmark, your mileage may 
vary.  I don't think we were going to submit it, rather, all the work 
under it, and hopefully, get that into other options, or new options, 
as appropriate.  Another option besides -fast, would be for the 
documentation for the compiler to give the current, best known SPEC 
flags.

Disclaimer, I don't work on -fast, so I may have some details wrong.  
I've been working on making the compiler fast.  :-)