4 ms·

I think we've reached the point where saying "float" should mean IEEE754, and the burden of being more specific falls on those working with various legacy forma

I think we've reached the point where saying "float" should mean IEEE754, and the burden of being more specific falls on those working with various legacy formats.

buildbot 2y ago
There's recently been an expansion of floating point types, with BF16 and the MX datatypes - so many modern examples of float meaning different things too.
- samatman 2y ago
  I'm very much in favor of that! These should also be referred to using something more specific than "float".