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] libffi support for CRIS


V sob, 06.11.2004 ob 18:48 je Hans-Peter Nilsson napisal(a):
> Huh, again sent as an attachment, not in-text.  It takes a
> little longer to review it this way.  Though that's not the
> main reason for this delay, sorry.
It's my dam mailer. I told it, it should send the attachments in-line
but is still does not do it (copy&past makes a mess of the patch).
Sorry.

> > > > +ffi_status
> > > > +ffi_prep_cif_machdep (ffi_cif * cif,
> > >
> > > > +  if ((cif->rtype->size == 0)
> > > > +      && (initialize_aggregate (cif->rtype) != FFI_OK))
> > >
> > > Don't you mean initialize_aggregate_packed_struct here?
> > No, thats correct - this one works on aligned parameters, not packed.
> 
> What do you mean by that?  The return value doesn't have any
> different layout than parameters; it's "packed" too.  I see no
> reason to not call initialize_aggregate_packed_struct there.
There was a reason why I left that in, but i can not recall what it was.
Does the ffitest (or the test-suite) work if you comment it out? [I
would try it myself but i really do not have the time to do it right
now]

> > > > +	;; Used for stack pointer saving.
> > > > +	push $r8
> > >
> > > Still suboptimal register allocation, wrt. my earlier movem comment.
> > This is still open. I do not have the time (and my knowledge of CRIS asm
> > is not that good) to fix it right now. (I tried a quick fix but I am
> > getting just segfaults).
> 
> If you care to revisit some time, look at output from an -O2
> compilation.
OK, I will look in to it when I find the time.

			Regards Simon



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