...
[22:22:54 CDT(-0500)] <colinclark> but rather that there is no integer arithmetic
[22:23:01 CDT(-0500)] <thealphanerd> ahhhh
[22:23:03 CDT(-0500)] <thealphanerd> I did say that
[22:23:10 CDT(-0500)] <thealphanerd> all numbers are stored as 32 bit floating point
[22:23:21 CDT(-0500)] <colinclark> So architecture-specific optimizations like assuming 32-bit integers will wrap when they overflow don't work
[22:23:38 CDT(-0500)] <colinclark> Which, honestly, doesn't necessarily strike me as a problem
[22:23:53 CDT(-0500)] <colinclark> Since there have been, and will be again, architectures where assuming this behaviour isn't valid
[22:23:58 CDT(-0500)] <colinclark> such as a GPU