Re: [Qemu-devel] [PATCH 6/7] target-tricore: Add cmp.f instruction

[Richard Henderson <rth@twiddle.net>]

On 03/08/2016 10:11 AM, Bastian Koppelmann wrote:
> On 03/08/2016 03:42 PM, Richard Henderson wrote:
>> On 03/08/2016 05:20 AM, Bastian Koppelmann wrote:
>>> On 03/01/2016 07:21 PM, Richard Henderson wrote:
>>>> On 03/01/2016 08:24 AM, Bastian Koppelmann wrote:
>>>>> +
>>>>> +    lt = float32_lt_quiet(arg1, arg2, &env->fp_status);
>>>>> +    eq = float32_eq_quiet(arg1, arg2, &env->fp_status);
>>>>> +    uo = float32_unordered(arg1, arg2, &env->fp_status);
>>>>
>>>> Do it in one step with float32_compare.
>>>> You don't want _quiet; see below re psw_flags.
>>>>
>>> Unfortunately, if I don't use _quiet my tests fail against real
>>> hardware...
>>
>> Fail in what way?  I don't think _quiet computes all of the bits you need.
>
> Also as far as I see the softfloat implementation, _quiet only
> determines whether an invalid exception is raised if the inputs are any
> nans or just signaling nans.
>

You're right -- I mis-read the set_FI status at the bottom of the page.


r~