matmuls aren't a syscall though. Little fun fact: Numpy doesn't even come with an efficient, blocked matmul procedure. It has to be linked against a BLAS imple

matmuls aren't a syscall though. Little fun fact: Numpy doesn't even come with an efficient, blocked matmul procedure. It has to be linked against a BLAS implement to really provide any decent performance. This also explains why Numpy performance can vary from distribution to distribution. Anaconda ships it with a different BLAS than Pypi.