Re: [RFA] gdbtypes.[ch] rs6000-tdep.c--AltiVec regs types

[Daniel Jacobowitz <drow at mvista dot com>]

On Fri, Jan 11, 2002 at 03:42:16PM -0800, Michael Snyder wrote:
> Daniel Jacobowitz wrote:
> > 
> > On Fri, Jan 11, 2002 at 05:32:45PM -0500, Elena Zannoni wrote:
> > >
> > > This patch provides a builtin union type for the AltiVec registers,
> > > so that the printing of such regs is a bit more enlightening, like:
> > >
> > >
> > > (gdb) p $vr0
> > > $2 = {uint128 = 0x00000064000000c80000012c00000190,
> > >       v4sf = {f = {1.40129846e-43, 2.80259693e-43, 4.20389539e-43, 5.60519386e-43}},
> > >       v4si = {f = {100, 200, 300, 400}},
> > >       v8hi = {f = {0, 100, 0, 200, 0, 300, 0, 400}},
> > >       v16qi = {f = "\0\0\0d\0\0\0È\0\0\001,\0\0\001\220"}}
> > > (gdb) p $vr0.v4si
> > > $3 = {f = {100, 200, 300, 400}}
> > > (gdb) p $vr0.v4si.f[2]
> > > $4 = 300
> > > (gdb) p $vr0.v4si.f[2]=444
> > > $5 = 444
> > 
> > Why is the .f necessary?  Why not make $vr0.v4si an array rather than
> > struct type?
> 
> I know this is historical (it's been done the same way for other
> targets), 
> and I'm guessing it's because a struct type can be passed by value,
> while
> an array type is always passed by reference.  You want a type that can
> be passed to a function.
> 
> I should know this (I've dealt with it before), but my 
> memory cache is shrinking with age and misuse.  ;-(

Hmm, could we accomplish this with an analogue to anonymous unions, I
wonder?... a question for another time.

union altivec_register {
  struct v4sf { float v4sf[4]; };
  ...
};

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer