bias is not consistent with the IEEE 754 standard

[1ucky2]

Have you genuinely reviewed the IEEE 754 standard documentation? The description of bias and subnormal numbers in your paper is completely inconsistent with the standard.

[TiRune]

Wow, thanks a lot for reading our paper and for this to be your first post on the account!

When writing the paper, we tried to make sure to stick to commonly used FP notation and standards. Would you be able to elaborate where you feel the description is 'completely inconsistent'?

Cross referencing the two, I can only see that b = 2^(e−1) is off by 1, and it should be b = 2^(e−1) - 1 . That was likely a type-o on our end, or we might have lifted that from some incorrect source material a couple of years ago.

[1ucky2]

You're right, my oversight on the subnormal cases. By the way, the schematic in Figure 4 of your paper could be clearer; while the idea comes across, a crisper, more intuitive graphic would let readers grasp it easily. Regardless, your work is very cool and has helped me a lot, thanks for sharing!