I recently stumbled upon a very interesting blog post from Martin Fuller on the topic of normalized float compression[1]. What follows is a summary of Martin’s write up along with some additional material from relevant resources and some notes of my own. Some of the content below is copied verbatim from the sources referenced at the end. IEEE754 Single Precision Floats The IEEE754 standard for floating point arithmetic defines a single precision 32-bit float using three components: A sign b...
