Re: [PATCH] gas/arc: Don't rely on bfd list of cpu type for cpu selection

[Andrew Burgess <andrew dot burgess at embecosm dot com>]

* anonymous <johnandsara2@cox.net> [2016-07-11 19:44:08 -0400]:

> Claudiu Zissulescu wrote:
> > Hi,
> > 
> > > -  if (mach == -1)
> > > -    goto bad_cpu;
> > > +  if ((!strcmp ("ARC600", cpu))
> > > +      || (!strcmp ("ARC601", cpu))
> > > +      || (!strcmp ("A6", cpu)))
> > > +    cpu_name = "arc600";
> > > +  else if ((!strcmp ("ARC700", cpu))
> > > +           || (!strcmp ("A7", cpu)))
> > > +    cpu_name = "arc700";
> > > +  else if (!strcmp ("EM", cpu))
> > > +    cpu_name = "arcem";
> > > +  else if (!strcmp ("HS", cpu))
> > > +    cpu_name = "archs";
> > > +  else if (!strcmp ("NPS400", cpu))
> > > +    cpu_name = "nps400";
> > > +  else
> > > +    cpu_name = NULL;
> > 
> > I would prefer your (very) old proposal here than this. It is more flexible.
> > 
> > Thanks,
> > Claudiu
> > 
> 
> cpu have become hacks in silicon - adding removing altering features,
> claiming and revoking compatibility: and you/we may see this effecting base
> cpu without warning
> 
> HOWEVER that the example you cited is for existing CPU it works for so
> answer there is: leave it alone it works for those (or most of) :)  it the
> CPU that aren't those which may be a problem.
> 
> it's safe to say avoid dependance without specific need and check otherwise
> - and defer to user to set it.  i always have to select options anyhow since
> the span of features wished to support is is custom to release, and releases
> today are typically "one off", once off the shelf it is hardly supported
> further.

Thanks for taking the time to review my patch, I'm always grateful to
receive feedback.

Unfortunately, even after reading your feedback several times, I am
still not sure what point you are making, whether you are objecting to
the change I propose, suggesting a different approach, or supporting
my proposal.

Would you mind rephrasing your feedback please, in order to help me
understand.  Is there some particular aspect of the change you
disagree with (I guess the part you quote), and is there a better
solution that you'd like to see?

Thanks for your time,
Andrew

Attachment: signature.asc
Description: PGP signature