Rohit Beriwal

Notes from 20mins lecture (Basics)

Few key points:

• Compiling the jit function first time, takes A LOT of time. Subsequent runs are faster.
• If the return type, is a collection of elements with different data types, Numba will be SLOW.
• Generally, use njit, or jit(nopython=true), mostly, DONT use jit. (njit is basically jit with nopython).
• DONT use python list() , wiht numba, because it will use reflection() to figure out the datatype in the list, which takes a lot of time.
• Use vectorized() decorator for scalar computation.