I am calling getrf from one thread in my application, the other threads are waiting at a taskwait. Task manager looks like all threads are maxed out and getrf never returns.
The matrix size is quite small 2763x2763. But can be up to 10000 square.
If I call mkl_set_num_threads_local(1) before entry it fixes the issue but then getrf is then only using one thread.
I am using MKL 11.3 with mkl_intel_thread_dll.lib on Windows 64 bit