This week’s Gray Scott Thursdays webinar will focus on computing precision. Scheduled for 19 February 2026, it will explore how calculations behave when using primitive data types like floating-point numbers.
The computation is ideally fast when using simple objects known as primitive types, such as floating-point numbers. These include float or double, which are called real in some languages. The webinar will highlight some surprising aspects of their behavior. These numbers can be represented as floating-point, which may introduce errors or other issues. Additionally, different use cases will be presented to help identify pitfalls of floating-point calculations and to find an optimal balance between performance, precision, and accuracy. An important point is that performance varies across different hardware, especially when exotic numbers like denormals, NaN, or infinity appear during computation.
