Memory layout of NumPy arrays: why one axis sums faster than the other

Afterwards you can explain how NumPy arrays sit in memory, read their strides, predict which access order is fast, and tell a view from a copy.