It uses BigDecimal. My first goal with the library was to provide a well-designed API. So to keep it simple for myself, I relied on BigDecimal for the calculations.
I think you are right, API is the most important part; internal implementation can be optimized later.
What is the double hack used in OpenHFT? I tried looking this up and came up short.
I’ll byte. Perhaps you didn’t look long enough. I’m sure the answer will float by at some point.