Python is a shockingly slow language. A test on a raspberry pi of simply “turn this pin on and off as fast as you can” gave the results below.
| System | Library | Speed |
|---|---|---|
| Shell | /proc/mem access | 2.8 kHz |
| Shell / gpio utility | WiringPi gpio utility | 40 Hz |
| Python | RPI.GPIO | 70 kHz |
| Python | wiringPi2 bindings | 28 kHz |
| Ruby | wiringPi bindings | 21 kHz |
| C | Native library | 22 MHz |
| C | BCM 2835 | 5.4 MHz |
| C | wiringPi | 4.1 – 4.6 MHz |
| Perl | BCM 2835 | 48 kHz |
Python is a prominent language in the ML and scientific computing space, and for good reason. Python is easy-to-learn and readable, and it offers a vast selection of libraries such as NumPy for numerical computation, Pandas for data manipulation, SciPy for scientific computing, TensorFlow, and PyTorch for deep learning, along with RDKit and Open Babel for cheminformatics. It is understandably an appealing choice for developers and researchers alike. However, a closer look at many common Python libraries reveals their foundations in C++.
Revisiting C++ Advantages
Many of Python libraries including TensorFlow, PyTorch, and RDKit are all heavily-reliant on C++. C++ allows developers to manage memory and CPU resources more effectively than Python, making it a good choice when handling large volumes of data at a fast pace. A previous post on this blog discusses C++’s speed, its utility in GPU programming through CUDA, and the complexities of managing its libraries. Despite the steeper learning curve and verbosity compared to Python, the performance benefits of C++ are undeniable, especially in contexts where execution speed and memory management are critical.
Rust: A New Contender for High-Performance Computing
Continue reading