New project: Fortran unums!
As mentioned in my weekly update, I have a new project up where I'm implementing the universal number (unum) in Fortran based on the prototype in John Gustafson's The End of Error. I touch on what makes a unum a unum and why I'm implementing it in Fortran on the project page, but briefly unums are a new proposal for how to store numbers and how to capture the fact that computers cannot represent every number exactly, something that is dealt with currently in computers by approximation. The Fortran is for speed. As a past coworker of mine would point out, any computer language can be made slow, so I'm trying to keep an eye out for ways to leverage Fortran's speed instead of just relying on it to magically make things efficient. And as always, the code is on Github for those who are interested.
Hopefully when I'm done I'll be able to try out some of Gustafson's claims for myself, as well as some of the criticisms he's received. And I'll have a Fortran module I can call from a bunch of other languages, so hopefully this will be an interesting one.