IEEE rules require that NaN not propagate through min and max unless both operands are NaN. See William Kahan’s remarks here. But also in that thread, the consensus was that the rule was surprising and perhaps should not be followed by Julia.
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Faster `min`,`max` for `Float64|32` | 38 | 2388 | August 17, 2021 | |
| `minimum()` 3x - 6x slower than `numpy.min()` | 16 | 619 | November 2, 2024 | |
| min(NaN, 1.0) should return NaN | 2 | 737 | March 17, 2017 | |
| NaN gates in julia? (from SIGBOBIK) | 2 | 466 | April 11, 2019 | |
| Treating NaN as error: Helping debugging | 19 | 4089 | October 29, 2024 |