3 ms·yet it is still used for arithmetic with "long double"by enriquto 7y agoyet it is still used for arithmetic with "long double"josefx 7y agoWouldn't that resovle the issue? long double contains the whole 80 bit state, not much space for undefined bits left.enriquto 7y agoYes, it resolves the issue in practice. However, the issue is an implementation detail of the compiler, not of the floating point format itself.