An amazing letter that intersects with my two recent posts (https://research.swtch.com/fp-knuth and https://research.swtch.com/fp): Donald Knuth writing to the IEEE 754 committee in favor of subnormal numbers and gradual underflow:

“Thus I strongly hope that the IEEE standard will be based on the concept of gradual underflow. At last we have a chance to get floatingpoint arithmetic that will attract scientific minds, instead of repelling them. Maybe in ten years we will even be able to have portable programs that routinely give the same answers on different computers. Unfortunately I am now having to change any TEX program to do only integer arithmetic, because as things stand in 1980 I dare not use floating point.”

https://www.arithmazium.org/library/lib/knuth_to_delp_apr1980.pdf