Conference | Live Schedule | Talks | Get your ticket | Set your timezone
Portable floating-point calculations
Guy Davidson
On Day 1 at 16:00 (CET/Berlin) in Track B [Opal Room and online]
The behaviour of floating point calculations is often unintuitive and hard to reason about, complicated by the non-associativity of floating point addition as well as rounding modes. If you need cross-platform portability, which you may if you are in domains like games or audio where it can be critically important, then you are in a whole heap of trouble. In this talk I'll tell you why you may need this portability, why it is so hard to solve, and what approach we took at Six Impossible Things Before Breakfast to achieve this critical feature.
Please login to comment