Unity Mathf - Why Did Unity Make Epilson Volatile?

Unity Engine
1

I’m reviewing Unity Mathf operations (UnityCsReference/Runtime/Export/Math/Mathf.cs at master · Unity-Technologies/UnityCsReference · GitHub). Fun… and I see that Epilson is defined as follows:

public static volatile float FloatMinNormal = 1.17549435E-38f;
public static volatile float FloatMinDenormal = Single.Epsilon;

Is it correct to mark Epsilon as volatile? This keyword implies a high frequency of changes to the value. In reality, this static value never changes beyond initial creation. By marking this variable volatile it prevents the register/memory from residing in the CPU cache. Would removing the volatile keyword here improve performance of Mathf operations which use Epsilon?

2

Looks like there was this comment above that got lost 4.5 years ago:

    // Note that the following needs to be volatile in order to prevent
   // the condition in IsFlushToZeroEnabled being optimized out

C# compiler can apparently optimize it out, but you don’t actually know what the value will be there at runtime because it depends on the CPU you’re running on.

1 Like
3

Thanks @Tautvydas-Zilys ! I didn’t read the code right – I now see that value is eventually carried into Epsilon, which is not marked volatile. I’m sure it is challenging to maintain such a complex and storied code base. Thank you for your insights!

      // A tiny floating point value (RO).
        public static readonly float Epsilon =
            UnityEngineInternal.MathfInternal.IsFlushToZeroEnabled ? UnityEngineInternal.MathfInternal.FloatMinNormal
            : UnityEngineInternal.MathfInternal.FloatMinDenorma
1 Like