3 ms·
> I disliked the Alpha floating point, it was always signalling exceptions for underflow. Alpha floating point wasn't the problem. The problem was that a whol
by bsder 7y ago
> I disliked the Alpha floating point, it was always signalling exceptions for underflow.
Alpha floating point wasn't the problem.
The problem was that a whole bunch of "clever" folks used "underflow" for all manner of weird reasons on x86. So, whenever Alpha either ran ported or emulated x86 code, it ran into underflows with far greater frequency than any actual numerical applications ever would.