And I had the misfortune for working for a few years with bond modeling software that used floats internally.
Turns out that the first thing a financial analyst does to compare two implementations of a bond securitization model is to look at the pennies. If those tie out, then you're good. But if your software predicts money off by a dime in a billion dollar model deal, the analyst WILL notice and WILL try to get to the bottom of it.
Yeah, floating point roundoff error seems like it should be immaterial. However it can matter a whole lot.
They need some way to validate things, and it turns out if the number isn't exactly the same when they run their samples they don't trust any of them. For at least some value of them.
The founder at the last place I worked at was famously difficult to satisfy because of this. You couldn't even begin to talk about how you were presenting data to him unless you were absolutely sure the numbers in your mockup were correct. He just couldn't see past an "incorrect" value when dealing with faked data ...
Turns out that the first thing a financial analyst does to compare two implementations of a bond securitization model is to look at the pennies. If those tie out, then you're good. But if your software predicts money off by a dime in a billion dollar model deal, the analyst WILL notice and WILL try to get to the bottom of it.
Yeah, floating point roundoff error seems like it should be immaterial. However it can matter a whole lot.