Re: [PATCH v2 04/12] Hexagon (target/hexagon) properly handle SNaN in dfmin/dfmax/sfmin/sfmax

[Richard Henderson]

On 2/10/22 13:15, Taylor Simpson wrote:
> The float??_minnum implementation differs from Hexagon for SNaN,
> it returns NaN, but Hexagon returns the other input.  So, we add
> checks for NaN before calling it.
>
> test cases added in a subsequent patch to more extensively test USR bits
>
> Signed-off-by: Taylor Simpson <tsimpson@quicinc.com>

This appears to be the same as the IEEE 754-2019 minimumNumber (as opposed to the earlier 
754-2008 minNum), which a recent RISC-V revision adopted.  We added support for that 
directly in softfloat: float32_minimum_number et al.


r~