Hands-On GPU Computing with Python电子书

Dedication

About Packt

Why subscribe?

Packt.com

Contributors

About the author

About the reviewer

Packt is searching for authors like you

Preface

Who this book is for

What this book covers

To get the most out of this book

Download the example code files

Download the color images

Code in Action

Conventions used

Get in touch

Reviews

Section 1: Computing with GPUs Introduction, Fundamental Concepts, and Hardware

Introducing GPU Computing

The world of GPU computing beyond PC gaming

What is a GPU?

Conventional CPU computing – before the advent of GPUs

How the gaming industry made GPU computing affordable for individuals

The emergence of full-fledged GPU computing

The rise of AI and the need for GPUs

The simplicity of Python code and the power of GPUs – a dual advantage

The C language – a short prologue

From C to Python

The simplicity of Python as a programming language – why many researchers and scientists prefer it

The power of GPUs

Ray tracing

Artificial intelligence (AI)

Programmable shading

RTX-OPS

Latest GPUs at the time of writing this book (can be subject to change)

NVIDIA GeForce RTX 2070

NVIDIA GeForce RTX 2080

NVIDIA GeForce RTX 2080 Ti

NVIDIA Titan RTX

Radeon RX Vega 56

Radeon RX Vega 64

Radeon VII

Significance of FP64 in GPU computing

The dual advantage – Python and GPUs, a powerful combination

How GPUs empower science and AI in current times

Bioinformatics workflow management

Magnetic Resonance Imaging (MRI) reconstruction techniques

Digital-signal processing for communication receivers

Studies on the brain – neuroscience research

Large-scale molecular dynamics simulations

GPU-powered AI and self-driving cars

Research work posited by AI scientists

Deep learning on commodity Android devices

Motif discovery with deep learning

Structural biology meets data science

Heart-rate estimation on modern wearable devices

Drug target discovery

Deep learning for computational chemistry

The social impact of GPUs

Archaeological restoration/reconstruction

Numerical weather prediction

Composing music

Real-time segmentation of sports players

Creating art

Security

Agriculture

Economics

Summary

Further reading

Designing a GPU Computing Strategy

Getting started with the hardware

The significance of compatible hardware for your GPU

Beginners

Intermediate users

Advanced users

Motherboard

Case

Power supply unit (PSU)

CPU

RAM

Hard-disk drive (HDD)

Solid-state drive (SSD)

Monitor

Building your first GPU-enabled parallel computer – minimum system requirements

Scope of hardware scalability

Branded desktops

Do it yourself (DIY) desktops

Beginner range

Mid range

High-end range

Liquid cooling – should you consider it?

The temperature factor

Airflow management

Thermal paste

Conventional air cooling

Stock coolers

Overclocking

So, what are custom/aftermarket coolers?

Liquid cooling

The specific heat capacity of cooling agents

Why is water the best liquid coolant?

Branded GPU-enabled PCs

Purpose

Feasibility

Upgradeability

Refining an effective budget

Warranty

Bundled monitors

Ready-to-deploy GPU systems

GPU solutions for individuals

Branded solutions in liquid cooling

Why not DIY?

GPU

CPU

Motherboard

RAM

Storage

PSU

Uninterrupted power supply (UPS)

Thermal paste

Heat sink

Radiator

Types of cooling fans

Bottlenecking

Estimating the build and performing compatibility checks

Purpose

Feasibility

Upgradeability

Refining an effective budget

Warranty for individual components

DIY solutions in liquid cooling

Assembling your system

Connecting all the power and case cables in place

Installing CUDA on a fresh Ubuntu installation

Entry-level budget

Mid-range budget

High-end budget

Summary

Further reading

Setting Up a GPU Computing Platform with NVIDIA and AMD

GPU manufacturers

First generation

Second generation

Third generation

Fourth generation

Fifth generation

Sixth generation

Seventh generation and beyond

Computing on NVIDIA GPUs

GeForce platforms

Quadro platforms

Tesla platforms

GPUDirect

SXM and NVLink

NVIDIA CUDA

Computing on AMD APUs and GPUs

Accelerated processing units (APUs)

The GPU in the APU – the significance of APU design

AMD GPUs – programmable platforms

Radeon platforms

Radeon Pro platforms

Radeon Instinct platforms

AMD ROCm

Comparing GPU programmable platforms on NVIDIA and AMD

GPUOpen

The significance of double precision in scientific computing from a GPU perspective

Current models from both brands that are ideal for GPU computing

AMD Radeon VII GPU – the new people's champion

NVIDIA Titan V GPU – raw compute power

An enthusiast's guide to GPU computing hardware

Summary

Further reading

Section 2: Hands-On Development with GPU Programming

Fundamentals of GPU Programming

GPU-programmable platforms

Basic CUDA concepts

Installing and testing

Compute capability

Threads, blocks, and grids

Threads

Blocks

Grids

Managing memory

Unified Memory Access (UMA)

Dynamic parallelism

Predefined libraries

OpenCL

Basic ROCm concepts

Installation procedure and testing

Official deprecation notice for HCC from AMD

Generating chips

ROCm components (APIs), including OpenCL

CUDA-like memory management with HIP

hipify

Predefined libraries

OpenCL

The Anaconda Python distribution for package management and deployment

Installing the Anaconda Python distribution on Ubuntu 18.04

Application-specific usage

GPU-enabled Python programming

The dual advantage

PyCUDA

PyOpenCL

CuPy

Numba (formerly Accelerate)

Summary

Further reading

Setting Up Your Environment for GPU Programming

Choosing a suitable IDE for your Python code

PyCharm – an IDE exclusively made for Python

Different versions of PyCharm

The Community edition

The Professional edition

The Educational edition – PyCharm Edu

Features for learners

Features for educators

PyCharm for Anaconda

Installing PyCharm

First run

EduTools plugin for existing PyCharm users

Alternative IDEs for Python – PyDev and Jupyter

Installing the PyDev Python IDE for Eclipse

Installing Jupyter Notebook and Jupyter Lab

Summary

Further reading

Working with CUDA and PyCUDA

Technical requirements

Understanding how CUDA-C/C++ works via a simple example

Installing PyCUDA for Python within an existing CUDA environment

Anaconda-based installation of PyCUDA

pip – system-wide Python-based installation of PyCUDA

Configuring PyCUDA on your Python IDE

Conda-based virtual environment

pip-based system-wide environment

How computing in PyCUDA works on Python

Comparing PyCUDA to CUDA – an introductory perspective on reduction

What is reduction?

Writing your first PyCUDA programs to compute a general-purpose solution

Useful exercise on computational problem solving

Exercise

Summary

Further reading

Working with ROCm and PyOpenCL

Technical requirements

Understanding how ROCm-C/C++ works with hipify, HIP, and OpenCL

Converting CUDA code into cross-platform HIP code with hipify

Understanding how ROCm-C/C++ works with HIP

Output on an NVIDIA platform

Output on an AMD platform

Understanding how OpenCL works

Installing PyOpenCL for Python (AMD and NVIDIA)

Anaconda-based installation of PyOpenCL

pip – system-wide Python base installation of PyOpenCL

Configuring PyOpenCL on your Python IDE

Conda-based virtual environment

pip-based system-wide environment

How computing in PyOpenCL works on Python

Comparing PyOpenCL to HIP and OpenCL – revisiting the reduction perspective

Reduction with HIP, OpenCL, and PyOpenCL

Writing your first PyOpenCL programs to compute a general-purpose solution

Useful exercise on computational problem solving

Solution assistance

Summary

Further reading

Working with Anaconda, CuPy, and Numba for GPUs

Technical requirements

Understanding how Anaconda works with CuPy and Numba

Conda

CuPy

Numba

GPU-accelerated Numba on Python

Installing CuPy and Numba for Python within an existing Anaconda environment

Coupling Python with CuPy

Conda-based installation of CuPy

pip-based installation of CuPy

Coupling Python with Numba for CUDA and ROCm

Installing Numba with Conda for NVIDIA CUDA GPUs

Installing Numba with Conda for AMD ROC GPUs

System-wide installation of Numba with pip (optional)

Configuring CuPy on your Python IDE

How computing in CuPy works on Python

Implementing multiple GPUs with CuPy

Configuring Numba on your Python IDE

How computing in Numba works on Python

Using vectorize

Explicit kernels

Writing your first CuPy and Numba enabled accelerated programs to compute GPGPU solutions

Interoperability between CuPy and Numba within a single Python program

Comparing CuPy to NumPy and CUDA

Comparing Numba to NumPy, ROCm, and CUDA

Useful exercise on computational problem solving

Summary

Further reading

Section 3: Containerization and Machine Learning with GPU-Powered Python

Containerization on GPU-Enabled Platforms

Programmable environments

Programmable environments – system-wide and virtual

Specific situations of usage

Preferring virtual over system-wide

Preferring system-wide over virtual

System-wide (open) environments

$HOME directory

System directories

Advantages of open environments

Disadvantages of open environments

Virtual (closed) environments

$HOME directory

Virtual system directories

Advantages of closed environments

Disadvantages of closed environments

Virtualization

Virtualenv

Installing virtualenv on Ubuntu Linux system

Using Virtualenv to create and manage a virtual environment

Key benefits of using Virtualenv

VirtualBox

Installing VirtualBox

GPU passthrough

Local containers

Docker

Installing Docker Community Edition (CE) on Ubuntu 18.04

NVIDIA Docker

Installing NVIDIA Docker

ROCm Docker

Kubernetes

Cloud containers

An overview on GPU computing with Google Colab

Summary

Further reading

Accelerated Machine Learning on GPUs

Technical requirements

The significance of Python in AI – the dual advantage

The need for big data management

Using Python for machine learning

Exploring machine learning training modules

The advent of deep learning

Introducing machine learning frameworks

Tensors by example

Introducing TensorFlow

Dataflow programming

Differentiable programming

TensorFlow on GPUs

Introducing PyTorch

The two primary features of PyTorch

Installing TensorFlow and PyTorch for GPUs

Installing cuDNN

Coupling Python with TensorFlow for GPUs

Coupling Python with PyTorch for GPUs

Configuring TensorFlow on PyCharm and Google Colab

Using TensorFlow on PyCharm

Using TensorFlow on Google Colab

Configuring PyTorch on PyCharm and Google Colab

Using PyTorch on PyCharm

Using PyTorch on Google Colab

Machine learning with TensorFlow and PyTorch

MNIST

Fashion-MNIST

CIFAR-10

Keras

Dataset downloads

Downloading Fashion-MNIST with Keras

Downloading CIFAR-10 with PyTorch

Writing your first GPU-accelerated machine learning programs

Fashion-MNIST prediction with TensorFlow

TensorFlow output on the PyCharm console

Training Fashion-MNIST for 100 epochs

CIFAR-10 prediction with PyTorch

PyTorch output on a PyCharm console

Revisiting our computational exercises with a machine learning approach

Solution assistance

Summary

Further reading

GPU Acceleration for Scientific Applications Using DeepChem

Technical requirements

Decoding scientific concepts for DeepChem

Atom

Molecule

Protein molecule

Biological cell

Medicinal drug – a small molecule

Ki

Crystallographic structures

Assays

Histogram

Open Source Drug Discovery (OSDD)

Convolution

Ensemble

Random Forest (RF)

Graph convolutional neural networks (GCN)

One-shot learning

Multiple ways to install DeepChem

Installing Google Colab

Conda on your local PyCharm IDE

NVIDIA Docker-based deployment

Configuring DeepChem on PyCharm

Testing an example from the DeepChem repository

How medicines reach their targets in our body

Alzheimer's disease

IC50

The Beta-Site APP-Cleaving Enzyme (BACE)

A DeepChem programming example

Output on the PyCharm console

Developing your own deep learning framework like DeepChem – a brief outlook

Summary

Final thoughts

References

Appendix A

GPU-accelerated machine learning in Python – benchmark research

GPU-accelerated machine learning with Python applied to cancer research

Deep Learning with GPU-accelerated Python for applied computer vision – Pavement Distress

Other Books You May Enjoy

Leave a review - let other readers know what you think