Your argument might be worth considering if Intel didn't use the same "cripple AMD" check across a lot of other libraries and their own compiler too - where the

Your argument might be worth considering if Intel didn't use the same "cripple AMD" check across a lot of other libraries and their own compiler too - where there are no such "numerical approximation" concerns. They've been doing it long before MKL [1]. --- [1] https://www.agner.org/optimize/blog/read.php?i=49 https://www.agner.org/optimize/blog/read.php?i=49