this saves a useless normalization operation.
Also this removes spurious changes of numerical behavior when one adapt a codebase to be compatible with both double and float from only one of them. This helps for instance in automatic noop testing on continuous integration system.
this saves a useless normalization operation. Also this removes spurious changes of numerical behavior when one adapt a codebase to be compatible with both double and float from only one of them. This helps for instance in automatic noop testing on continuous integration system.