Re: [Qemu-devel] [PATCH 06/14] Add v{max,min}fp instructions

[Aurelien Jarno <aurelien@aurel32.net>]

On Tue, Feb 03, 2009 at 08:51:13PM +0100, Aurelien Jarno wrote:
> On Thu, Jan 22, 2009 at 12:44:06PM -0800, Nathan Froyd wrote:
> > 
> > Signed-off-by: Nathan Froyd <froydnj@codesourcery.com>
> > ---
> >  target-ppc/helper.h    |    2 ++
> >  target-ppc/op_helper.c |   19 +++++++++++++++++++
> >  target-ppc/translate.c |    2 ++
> >  3 files changed, 23 insertions(+), 0 deletions(-)
> > 
> > diff --git a/target-ppc/helper.h b/target-ppc/helper.h
> > index 8c04ba7..ebdeabe 100644
> > --- a/target-ppc/helper.h
> > +++ b/target-ppc/helper.h
> > @@ -232,6 +232,8 @@ DEF_HELPER_3(vsum2sws, void, avr, avr, avr)
> >  DEF_HELPER_3(vsum4sbs, void, avr, avr, avr)
> >  DEF_HELPER_3(vsum4shs, void, avr, avr, avr)
> >  DEF_HELPER_3(vsum4ubs, void, avr, avr, avr)
> > +DEF_HELPER_3(vmaxfp, void, avr, avr, avr)
> > +DEF_HELPER_3(vminfp, void, avr, avr, avr)
> >  
> >  DEF_HELPER_1(efscfsi, i32, i32)
> >  DEF_HELPER_1(efscfui, i32, i32)
> > diff --git a/target-ppc/op_helper.c b/target-ppc/op_helper.c
> > index 107f977..717bbd4 100644
> > --- a/target-ppc/op_helper.c
> > +++ b/target-ppc/op_helper.c
> > @@ -2252,6 +2252,25 @@ VMINMAX(uw, u32)
> >  #undef VMINMAX_DO
> >  #undef VMINMAX
> >  
> > +#define VMINMAXFP(suffix, relation)                                     \
> > +    void helper_v##suffix (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b)    \
> > +    {                                                                   \
> > +        int i;                                                          \
> > +        for (i = 0; i < ARRAY_SIZE(r->f); i++) {                        \
> > +            HANDLE_NAN2(r->f[i], a->f[i], b->f[i]) {                    \
> > +                int rel = float32_compare_quiet(a->f[i], b->f[i], &env->vec_status);  \
> 
> Given that we only care about lower / greater, I wonder if it would be
> better to use float32_lt_quiet and reversing the arguments.
> 
> > +                if (rel == relation) {                                  \
> > +                    r->f[i] = a->f[i];                                  \
> > +                } else {                                                \
> > +                    r->f[i] = b->f[i];                                  \
> > +                }                                                       \
> > +            }                                                           \
> > +        }                                                               \
> > +    }
> > +VMINMAXFP(minfp, float_relation_less)
> > +VMINMAXFP(maxfp, float_relation_greater)
> > +#undef VMINMAXFP
> > +
> >  void helper_vmladduhm (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
> >  {
> >      int i;
> > diff --git a/target-ppc/translate.c b/target-ppc/translate.c
> > index 19abec1..35693e0 100644
> > --- a/target-ppc/translate.c
> > +++ b/target-ppc/translate.c
> > @@ -6395,6 +6395,8 @@ GEN_VXFORM(vsum4sbs, 4, 28);
> >  GEN_VXFORM(vsum4shs, 4, 25);
> >  GEN_VXFORM(vsum2sws, 4, 26);
> >  GEN_VXFORM(vsumsws, 4, 30);
> > +GEN_VXFORM(vmaxfp, 5, 10);
> > +GEN_VXFORM(vminfp, 5, 11);

Note also that those values are wrong. The correct values are 0x10 and
0x11.

> >  
> >  #define GEN_VXRFORM1(opname, name, str, opc2, opc3)                     \
> >      GEN_HANDLER2(name, str, 0x4, opc2, opc3, 0x00000000, PPC_ALTIVEC)   \
> > -- 
> > 1.6.0.5
> > 
> > 
> > 
> > 
> 
> -- 
> Aurelien Jarno	                        GPG: 1024D/F1BCDB73
> aurelien@aurel32.net                 http://www.aurel32.net
> 
> 
> 

-- 
Aurelien Jarno	                        GPG: 1024D/F1BCDB73
aurelien@aurel32.net                 http://www.aurel32.net