My first gen threadripper occasionally deadlocks in futex code within libgomp (gnu implementation of omp). Eventually I gave up and concluded it was either a ha

My first gen threadripper occasionally deadlocks in futex code within libgomp (gnu implementation of omp). Eventually I gave up and concluded it was either a hardware bug or a bug that incorrectly relies on atomic behaviour of intel CPUs. I eventually switched to using clang with its own omp implementation and the problem magically disappeared.