I’m trying to implement a shared numpy array for use in multiprocessing. It works fine on smaller arrays, but then failed when I tried to use a ~2.7GB numpy array (about 5.6Mx64 matrix). After a little tinkering, I found that the error occurs when writing the input matrix to the assigned shared memory, e.g. if dst is the buffered numpy array and test_matrix is the large matrix being copied over, then dst[:] = test_matrix[:] throws the following system error: Process finished with exit code 135 (interrupted by signal 7:SIGBUS).