4 ms·
That sounds surprisingly reasonable idea for signeds. Less so for unsigneds though. Has there been any architecture doing anything like that?
by zokier 1y ago
That sounds surprisingly reasonable idea for signeds. Less so for unsigneds though. Has there been any architecture doing anything like that?
- newpavlov 1y agoI can not name an ISA with such instructions out of my head. As for unsigned integers, as I mentioned in the other comment, we probably need two separate instruction sets for "wrapping" and NaN-able operations on unsigned integers.