On Thu, 3 Nov 2011, Joern Rennecke wrote:
> > Comment about the reason for this option?
>
> Would this be OK?
>
> /* { dg-options "-O2" } */
> +/* Using -mshort-calls avoids constant loading the function addresses in
> + registers and thus getting the counts wrong. */
> +/* { dg-options "-O2 -mshort-calls" { target epiphany-*-* } } */
Yes, seems reasonable.
> > > +/* We would like to use Posix for profiling, but the simulator interface
> > > + lacks mkdir. */
> > > +#undef TARGET_POSIX_IO
> >
> > This doesn't belong in a generic file for a given CPU. It might go in a
> > file for a (CPU, OS) combination - but it's the bare-metal default anyway.
>
> The original idea was to define TARGET_POSIX_IO, but as the comment states,
> we got to fix the simulator first.
> So are you saying I need a special file for the singular (CPU, no-OS)
> combination if I want to set TARGET_POSIX_IO ?
Given that there is only one target triplet for this CPU, an appropriately
commented definition could reasonably go here as long as no other OS ports
are added - but the #undef which is the default doesn't belong there.
--
Joseph S. Myers
[email protected]