darwin cross compiling breakage

[Ian Lance Taylor]

Mike Stump <mrs@apple.com> writes:

> On Jun 12, 2006, at 9:59 AM, Ian Lance Taylor wrote:
> > Seems like a promising approach to me.  Of course it will need some
> > work in config.host and config/i386/driver-i386.c.
> 
> Do you have a testcase that shows a regression from the status quo?
> I can't envision one.  Further, I don't know of anything else that
> needs doing, can you give us a hint?

The testcase would be to compile with -march=native and see what kind
of optimization you get.

Geoff proposed enabling -march=native when __i386__ is defined even if
CROSS_COMPILE is not defined.  That requires that %local_cpu_detect
work.  That requires that host_detect_local_cpu be compiled into the
driver program.  That requires that config.host enable that--but if
you search for driver-i386 in config.host, you will see that it is
only compiled in if host = target.  It also requires that
config/i386/driver-i386.c do something useful--but if you look at that
file you will see that it only defines host_detect_local_cpu if
CROSS_COMPILE is not defined.

Ian