GPU Accelerated Computing with Python

Python is one of the most popular programming languages today for science, engineering, data analytics and deep learning applications. However, as an interpreted language, it has been considered too slow for high-performance computing. That has changed with CUDA Python from Continuum Analytics.

With CUDA Python, using the Numba Python compiler, you get the best of both worlds: rapid iterative development with Python combined with the speed of a compiled language targeting both CPUs and NVIDIA GPUs.

1
SETUP CUDA PYTHON

To run CUDA Python, you will need the CUDA Toolkit installed on a system with CUDA capable GPUs.

If you do not have a CUDA-capable GPU, you can access one of the thousands of GPUs available from cloud service providers including Amazon AWS, Microsoft Azure and IBM SoftLayer. The NVIDIA-maintained CUDA Amazon Machine Image (AMI) on AWS, for example, comes pre-installed with CUDA and is available for use today.

Use this guide for easy steps to install CUDA. To setup CUDA Python, first install the Anaconda python distribution. Then install the latest version of the Numba package. You can find detailed installation instructions in the Numba documentation.

Or, watch the short video below and follow along.

2
YOUR FIRST CUDA PYTHON PROGRAM

You are now ready for your first python program on the GPU. The video below walks through a simple example that adds two vectors for you to follow along.

If you are new to Python, explore the beginner section of the Python website for some excellent getting started resources. The blog, An Even Easier Introduction to CUDA, introduces key CUDA concepts through simple examples.

In the Numba documentation you will find information about how to vectorize functions to accelerate them automatically as well as how to write CUDA code in Python. Download and execute Jupyter Notebooks for the Mandelbrot and Monte Carlo Option Pricer examples on your local machine.

3
PRACTICE

Check out Numbas github repository for additional examples to practice.

NVIDIA also provides hands-on training through a collection of self-paced labs . The labs guide you step-by-step through editing and execution of code, and even interaction with visual tools is all woven together into a simple immersive experience. Practice the techniques you learned in the materials above through hands-on labs.

For a more formal,instructor-led introduction to CUDA, explore the Introduction to Parallel Programming on UDACITY. The course covers a series of image processing algorithms such as you might find in Photoshop or Instagram. You'll be able to program and run your assignments on high-end GPUs, even if you don't have one yourself.

Availability

The Numba package is available as a Continuum Analytics sponsored open-source project.

The CUDA Toolkit is a free download from NVIDIA and is supported on Windows, Mac, and most standard Linux distributions.

So, now youre ready to deploy your application?

Register today for free access to NVIDIA TESLA GPUs in the cloud.

Latest News

Generating Photorealistic Images of Fake Celebrities with Artificial Intelligence

Researchers from NVIDIA recently published a paper detailing their new methodology for generative adversarial networks (GANs) that generated photorealistic pictures of fake celebrities.

High-Performance GPU Computing in the Julia Programming Language

Julia is a high-level programming language for mathematical computing that is as easy to use as Python, but as fast as C.

Build, Train and Deploy AI in the Cloud with NVIDIA GPU Cloud

NVIDIA just announced the NVIDIA GPU Cloud (NGC) — a GPU-accelerated cloud platform that makes it easy to get started with the top deep learning frameworks on-premises or on Amazon Web Services.

Programming Tensor Cores in CUDA 9

A defining feature of the new Volta GPU Architecture is its Tensor Cores, which give the Tesla V100 accelerator a peak throughput 12 times the 32-bit floating point throughput of the previous-generation Tesla P100.

Blogs: Parallel ForAll

Pro Tip: Pinpointing Runtime Errors in CUDA Fortran

We’ve all been there. Your CUDA Fortran code is humming along and suddenly you get a runtime error: copyin, copyout, usually accompanied by FAILED in all caps.

High-Performance GPU Computing in the Julia Programming Language

Julia is a high-level programming language for mathematical computing that is as easy to use as Python, but as fast as C.

DeepStream: Next-Generation Video Analytics for Smart Cities

Imagine every parent’s worst nightmare: a child lost in a crowded mall.

Programming Tensor Cores in CUDA 9

A defining feature of the new Volta GPU Architecture is its Tensor Cores, which give the Tesla V100 accelerator a peak throughput 12 times the 32-bit floating point throughput of the previous-generation Tesla P100.