Re: [Qemu-devel] [PATCH] target/ppc: rewrite f[n]m[add, sub] using float64_muladd

[Nikunj A Dadhania <nikunj@linux.vnet.ibm.com>]

Nikunj A Dadhania <nikunj@linux.vnet.ibm.com> writes:

> Bharata B Rao <bharata@linux.vnet.ibm.com> writes:
>
>> On Wed, Mar 01, 2017 at 01:14:11PM +0530, Nikunj A Dadhania wrote:
>>> Use the softfloat api for fused multiply-add. As we are using the fused
>>> multiply-add, the intermediate result for setting VXISI is not
>>> available.
>>
>> Isn't the behaviour of setting VXISI similar to vector muladd instructions ?
>> If so refer to VSX_MADD() to see when VXISI is set.
>
>                 if ((float64_is_infinity(xa.fld) ||
>                      float64_is_infinity(b->fld)) &&
>                      float64_is_infinity(c->fld))) {
>                    float_invalid_op_excp(env, POWERPC_EXCP_FP_VXISI, sfprf);
>                }
>                
> AFAIU, VXISI is set when (infinity - infinity). The above doesn't take
> into consideration about substraction.

To be precise:

"If the product of x and y is an Infinity and z is an Infinity of the
opposite sign, vxisi_flag is set to 1."

Regards
Nikunj