[PATCH] benchtests: Add sinf traces

[Adhemerval Zanella Netto]

On 02/04/26 15:27, Wilco Dijkstra wrote:
> Hi Adhemerval,
> 
>>> The actual binary values matter very little unless they hit cornercases that are
>>> significantly slower. The range is never an issue, nobody ever does sin (DBL_MAX)!
>>
>> Right, but I now really following your point here. Are you saying a possible binary64
>> traced rounded to binary32 from specpu2017 would *not* be a good benchmark?
> 
> The opposite - what I'm saying the low bits of the mantissa don't actually matter at
> all since we're just computing a polynomial. What matters is the quadrant it falls into
> and special cases like which range reducer, early exits like close to zero or corner cases
> for rounding.

Ok we are on the same page then.  The main problem is for some benchmark there a *lot* of
sincos calls, so I will need to add some strategy to pull a representative range from
the trace.