Looking at the complexity and area of hardware floating point, I often wonder why we don't see more unified combined integer+floating point units, like done in

Looking at the complexity and area of hardware floating point, I often wonder why we don't see more unified combined integer+floating point units, like done in the R4200 [1], which reused most of the integer datapath while just adding a smaller extra smaller 12-bit datapath for the exponent. [1] https://en.wikipedia.org/wiki/R4200 https://en.wikipedia.org/wiki/R4200