Title Page -- Preface -- Contents -- Introduction -- Existing Standard Parallel Programming Systems -- MPI -- OpenMP -- Two Parallelization Strategies: Data Parallelism and Task Parallelism -- Data Parallelism -- Task Parallelism -- Example -- History and Goals of OpenCL -- Origins of Using GPU in General Purpose Computing -- Short History of OpenCL -- Heterogeneous Computer Memories and Data Transfer -- Heterogeneous Computer Memories -- Data Transfer -- The Fourth Generation CUDA -- Host Code -- Phase a. Initialization and Creating Context -- Phase b. Kernel Creation, Compilation and Preparations for Kernel Execution -- Phase c. Creating Command Queues and Kernel Execution -- Finalization and Releasing Resource -- Applications of Heterogeneous Computing -- Accelerating Scientific/Engineering Applications -- Conjugate Gradient Method -- Jacobi Method -- Power Method -- Monte Carlo Methods -- Conclusions -- Benchmarking CGM -- Introduction -- Additional CGM Description -- Heterogeneous Machine -- Algorithm Implementation and Timing Results -- Conclusions -- OpenCL Fundamentals -- OpenCL Overview -- What is OpenCL -- CPU + Accelerators -- Massive Parallelism Idea -- Work Items and Workgroups -- OpenCL Execution Model -- OpenCL Memory Structure -- OpenCL C Language for Programming Kernels -- Queues, Events and Context -- Host Program and Kernel -- Data Parallelism in OpenCL -- Task Parallelism in OpenCL -- How to Start Using OpenCL -- Header Files -- Libraries -- Compilation -- Platforms and Devices -- OpenCL Platform Properties -- Devices Provided by Platform -- OpenCL Platforms - C++ -- OpenCL Context to Manage Devices -- Different Types of Devices -- CPU Device Type -- GPU Device Type -- Accelerator -- Different Device Types - Summary -- Context Initialization - by Device Type -- Context Initialization - Selecting Particular Device.

Getting Information about Context -- OpenCL Context to Manage Devices - C++ -- Error Handling -- Checking Error Codes -- Using Exceptions - Available in C++ -- Using Custom Error Messages -- Command Queues -- In-order Command Queue -- Out-of-order Command Queue -- Command Queue Control -- Profiling Basics -- Profiling Using Events - C example -- Profiling Using Events - C++ example -- Work-Items and Work-Groups -- Information About Index Space from a Kernel -- NDRange Kernel Execution -- Task Execution -- Using Work Offset -- OpenCL Memory -- Different Memory Regions - the Kernel Perspective -- Relaxed Memory Consistency -- Global and Constant Memory Allocation - Host Code -- Memory Transfers - the Host Code -- Programming and Calling Kernel -- Loading and Compilation of an OpenCL Program -- Kernel Invocation and Arguments -- Kernel Declaration -- Supported Scalar Data Types -- Vector Data Types and Common Functions -- Synchronization Functions -- Counting Parallel Sum -- Parallel Sum - Kernel -- Parallel Sum - Host Program -- Structure of the OpenCL Host Program -- Initialization -- Preparation of OpenCL Programs -- Using Binary OpenCL Programs -- Computation -- Release of Resources -- Structure of OpenCL host Programs in C++ -- Initialization -- Preparation of OpenCL Programs -- Using Binary OpenCL Programs -- Computation -- Release of Resources -- The SAXPY Example -- Kernel -- The Example SAXPY Application - C Language -- The example SAXPY application - C++ language -- Step by Step Conversion of an Ordinary C Program to OpenCL -- Sequential Version -- OpenCL Initialization -- Data Allocation on the Device -- Sequential Function to OpenCL Kernel -- Loading and Executing a Kernel -- Gathering Results -- Matrix by Vector Multiplication Example -- The Program Calculating matrix times vector -- Performance -- Experiment -- Conclusions.

Advanced OpenCL -- OpenCL Extensions -- Different Classes of Extensions -- Detecting Available Extensions from API -- Using Runtime Extension Functions -- Using Extensions from OpenCL Program -- Debugging OpenCL codes -- Printf -- Using GDB -- Performance and Double Precision -- Floating Point Arithmetics -- Arithmetics Precision - Practical Approach -- Profiling OpenCL Application -- Using the Internal Profiler -- Using External Profiler -- Effective Use of Memories - Memory Access Patterns -- Matrix Multiplication - Optimization Issues -- OpenCL and OpenGL -- Extensions Used -- Libraries -- Header Files -- Common Actions -- OpenGL Initialization -- OpenCL Initialization -- Creating Buffer for OpenGL and OpenCL -- Kernel -- Generating Effect -- Running Kernel that Operates on Shared Buffer -- Results Display -- Message Handling -- Cleanup -- Notes and Further Reading -- Case Study - Genetic Algorithm -- Historical Notes -- Terminology -- Genetic Algorithm -- Example Problem Definition -- Genetic Algorithm Implementation Overview -- OpenCL Program -- Most Important Elements of Host Code -- Summary -- Experiment Results -- Comparing CUDA with OpenCL -- Introduction to CUDA -- Short CUDA Overview -- CUDA 4.0 Release and Compatibility -- CUDA Versions and Device Capability -- CUDA Runtime API Example -- CUDA Program Explained -- Blocks and Threads Indexing Formulas -- Runtime Error Handling -- CUDA Driver API Example -- Theoretical Foundations of Heterogeneous Computing -- Parallel Computer Architectures -- Clusters and SMP -- DSM and ccNUMA -- Parallel Chip Computer -- Performance of OpenCL Programs -- Combining MPI with OpenCL -- Matrix Multiplication - Algorithm and Implementation -- Matrix Multiplication -- Implementation -- OpenCL Kernel -- Initialization and Setup -- Kernel Arguments -- Executing Kernel -- Using Examples Attached to the Book.

Compilation and Setup -- Linux -- Windows -- Bibliography and References.