This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [ColdFire 9/63] Add -mcpu, -march, -mtune, and related options


Richard Sandiford <richard@codesourcery.com> writes:
> Roman Zippel <zippel@linux-m68k.org> writes:
>>> I'm just hand-waving though.  You know what's appropriate for 680x0
>>> better than I do, so if you think they should be treated differently
>>> from ColdFire, I can take the default float setting for 680x0 targets
>>> from the float setting for the default CPU, rather than the chosen CPU.
>>
>> For Linux it doesn't really matter, what I have in mind is that a generic 
>> target like m68k-elf could be used with --with-float=soft.
>
> But what precisely do you think the semantics should be?
> Like I say, I think it's reasonable to use the device defaults
> for ColdFire, as -mhard-float does not change the calling conventions.
> Are you saying that you disagree, and that you think the default should
> be decided by the configuration regardless of -mcpu?  Or are you saying
> that the default should be decided in a different way for 680x0 and
> ColdFire targets?

Also bear in mind that targets like m68k-elf by default build separate
multilibs for 68000, 68020, etc.  Those libraries will be built with
the default as decided by this patch, so code compiled with -m68000
will be link-compatible with the -m68000 libraries.

If someone stepped forward to implement --with-float, it would still
work regardless of what we do here.  --with-soft would add an explicit
floating-point option to the command line, avoiding the inference code
altogether.

Richard


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]