Speeding up NumPy with parallelism

"If your NumPy code is too slow, what next? One option is taking advantage of the multiple cores on your CPU: using a thread pool to do work in parallel. Another option is to tune your code so it's less wasteful. Or, since these are two different sources of speed, you can do both. In this article I'll cover: A simple example of making a NumPy algorithm parallel. A separate kind of optimization, making a more efficient implementation in Numba."

"From single-threaded to parallelism: an example Let's say you want to calculate the sum of the squared difference between two arrays. Here's how you'd do it with single-threaded NumPy: Because this function is summing values across the array, it can easily be parallelized by running the function on chunks of the array, and then summing those partial results. Here's one way to do it, using only the Python standard library:"

Parallel execution across multiple CPU cores speeds NumPy reductions by chunking arrays, running work concurrently, and summing partial results. Using a thread pool lowers wall-clock time and reduces temporary-array memory usage, which improves cache behavior. Algorithmic optimization via Numba compiles elementwise loops to machine code and can eliminate temporary arrays by accumulating values one at a time. Parallelism and algorithmic efficiency are distinct but complementary: parallelism increases throughput while better algorithms reduce work and memory traffic. Combining both approaches produces greater speedups. Hardware limits on parallelism and trade-offs in Numba's parallel features should be considered.

Read at PythonSpeed