According to section 4.2.3 of JLS, particular Java implementation may support
1) float extended-exponent value set
2) double-extended-exponent value set
But this is not required.
Example:
If we have a product 2.3 * 3.7
and this product is located in non-strictfp context (i.e. in non-strictfp method or class) then before actual calculation of the product JVM has an option to choose between two Value Sets: double value set
and double-extended-exponent value set
.
Do you know particular Java implementation that support this feature?
Maybe this feature is supported in particular platform (Windows, Unix ,...)?