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: [PATCH] re-enable -mabi=no-altivec on powerpc64-linux


On Tue, 2008-02-05 at 11:49 +1030, Alan Modra wrote:
> On Mon, Feb 04, 2008 at 10:30:21AM -0800, Janis Johnson wrote:
> > Before this patch, rs6000_explicit_options.abi was explicitly NOT set
> > for -mabi=no-altivec (by the comment with "???").  Alan, please take a
> > look to see if I broke something for SPE.  Perhaps there should be
> > separate fields for altivec_abi and spec_abi?
> 
> Well, you've changed the behaviour.  I don't know whether that
> qualifies as broken..  I believe the thinking behind the old behaviour
> is that -mabi=no-altivec ought to have no effect on whether the spe
> abi is selected by default or not.  After all, if you're compiling for
> spe you don't have altivec hardware and -mabi=no-altivec doesn't
> change the default.  (You could argue that -mabi=no-spe similarly
> shouldn't have any effect when compiling for a non-spe processor but
> we don't do that for some reason.)
> 
> I do think the code would be simpler if we had separate abi fields.
> -mabi=altivec and -mabi=spe ought to count as explicitly setting both
> fields, but the no- versions ought to just set their own field.

When processing -mabi=no-altivec do we know whether we're generating
spe code so we can just ignore that option?  Similarly for
-mabi=no-spe.  If not I'll look at splitting the abi field.

Janis


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