The patch add FixedPoint64 class for unsigned fixed point arithmetic and corresponding tests.

The new class will be used to update https://reviews.llvm.org/D30527.

I've skipped signed arithmetic, because the class is supposed to be used to work with probabilities and costs.

For now point is fixed in the middle [32 bits].[32 bits], which makes calculations simple and fast.

# Details

Details

# Diff Detail

Diff Detail

### Event Timeline

Comment Actions

You mention "For now point is fixed in the middle [32 bits].[32 bits], which makes calculations simple and fast.". Perhaps this should be commented in the code as well.