Re: [Qemu-devel] [PATCH 14/14] Add vrsqrtefp instruction

[Aurelien Jarno]

On Thu, Jan 22, 2009 at 12:44:14PM -0800, Nathan Froyd wrote:
> 
> Signed-off-by: Nathan Froyd <address@hidden>
> ---
>  target-ppc/helper.h    |    1 +
>  target-ppc/op_helper.c |   28 ++++++++++++++++++++++++++++
>  target-ppc/translate.c |    1 +
>  3 files changed, 30 insertions(+), 0 deletions(-)
> 
> diff --git a/target-ppc/helper.h b/target-ppc/helper.h
> index 516cee0..179f077 100644
> --- a/target-ppc/helper.h
> +++ b/target-ppc/helper.h
> @@ -245,6 +245,7 @@ DEF_HELPER_3(vsubfp, void, avr, avr, avr)
>  DEF_HELPER_3(vmaxfp, void, avr, avr, avr)
>  DEF_HELPER_3(vminfp, void, avr, avr, avr)
>  DEF_HELPER_2(vrefp, void, avr, avr)
> +DEF_HELPER_2(vrsqrtefp, void, avr, avr)
>  DEF_HELPER_2(vrfim, void, avr, avr)
>  DEF_HELPER_2(vrfin, void, avr, avr)
>  DEF_HELPER_2(vrfip, void, avr, avr)
> diff --git a/target-ppc/op_helper.c b/target-ppc/op_helper.c
> index 8055e1a..4ef0b1c 100644
> --- a/target-ppc/op_helper.c
> +++ b/target-ppc/op_helper.c
> @@ -2756,6 +2756,34 @@ VROTATE(h, u16)
>  VROTATE(w, u32)
>  #undef VROTATE
>  
> +void helper_vrsqrtefp (ppc_avr_t *r, ppc_avr_t *b)
> +{
> +    int i;
> +    for (i = 0; i < ARRAY_SIZE(r->f); i++) {
> +        HANDLE_NAN1(r->f[i], b->f[i]) {
> +            if (float32_is_zero(b->f[i])) {
> +                if (float32_is_neg(b->f[i])) {
> +                    /* Negative infinity */
> +                    r->u32[i] = 0xff800000;
> +                } else {
> +                    /* Positive infinity */
> +                    r->u32[i] = 0x7f800000;
> +                }
> +            } else if (float32_is_neg(b->f[i])) {
> +                /* Canonical QNaN */
> +                r->u32[i] = 0x7fc00000;
> +            } else if (float32_is_infinity(b->f[i])) {
> +                /* Negative infinity handled earlier */
> +                r->f[i] = float32_zero;
> +            } else {
> +                float32 one = int32_to_float32(1, &env->vec_status);
> +                float32 t = float32_sqrt(b->f[i], &env->vec_status);
> +                r->f[i] = float32_div(one, t, &env->vec_status);

The same comments as on the vrefp patch applies here.

> +            }
> +        }
> +    }
> +}
> +
>  void helper_vsel (ppc_avr_t *r, ppc_avr_t *a, ppc_avr_t *b, ppc_avr_t *c)
>  {
>      r->u64[0] = (a->u64[0] & ~c->u64[0]) | (b->u64[0] & c->u64[0]);
> diff --git a/target-ppc/translate.c b/target-ppc/translate.c
> index 01a4f11..1c44c59 100644
> --- a/target-ppc/translate.c
> +++ b/target-ppc/translate.c
> @@ -6477,6 +6477,7 @@ GEN_VXFORM_NOA(vupklsh, 7, 11);
>  GEN_VXFORM_NOA(vupkhpx, 7, 13);
>  GEN_VXFORM_NOA(vupklpx, 7, 15);
>  GEN_VXFORM_NOA(vrefp, 5, 4);
> +GEN_VXFORM_NOA(vrsqrtefp, 5, 5);
>  GEN_VXFORM_NOA(vrfim, 5, 8);
>  GEN_VXFORM_NOA(vrfin, 5, 9);
>  GEN_VXFORM_NOA(vrfip, 5, 10);
> -- 
> 1.6.0.5
> 
> 
> 
> 

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