<ul><li>In the early days of computing, mathematicians focused on understanding and mitigating the limitations of floating point arithmetic.</li><li>With the shift to double precision, the need to consider floating point computation has decreased, but it still remains relevant in some cases.</li><li>There is now a growing interest in using lower precision numbers, such as half precision, particularly in applications like neural networks.</li><li>However, using lower precision numbers requires revisiting issues like different ways of rounding, which were studied in the past.</li></ul>

Floating point: Everything old is new again

Discover more