Parallel and Concurrent Programming in Haskell: Techniques for Multicore and Multithreaded Programming

Description

If you have a working knowledge of Haskell, this hands-on book shows you how to use the language’s many APIs and frameworks for writing both parallel and concurrent programs. You’ll learn how parallelism exploits multicore processors to speed up computation-heavy programs, and how concurrency enables you to write programs with threads for multiple interactions. Author Simon Marlow walks you through the process with lots of code examples that you can run, experiment with, and extend. Divided into separate sections on Parallel and Concurrent Haskell, this book also includes exercises to help you become familiar with the concepts presented: Express parallelism in Haskell with the Eval monad and Evaluation Strategies Parallelize ordinary Haskell code with the Par monad Build parallel array-based computations, using the Repa library Use the Accelerate library to run computations directly on the GPU Work with basic interfaces for writing concurrent code Build trees of threads for larger and more complex programs Learn how to build high-speed concurrent network servers Write distributed programs that run on multiple machines in a network Table of Contents Preface Audience How to Read This Book Conventions Used in This Book Using Sample Code Safari® Books Online How to Contact Us Acknowledgments Chapter 1. Introduction Terminology: Parallelism and Concurrency Tools and Resources Sample Code Part I. Parallel Haskell Chapter 2. Basic Parallelism: The Eval Monad Lazy Evaluation and Weak Head Normal Form The Eval Monad, rpar, and rseq Example: Parallelizing a Sudoku Solver Deepseq Chapter 3. Evaluation Strategies Parameterized Strategies A Strategy for Evaluating a List in Parallel Example: The K-Means Problem Parallelizing K-Means Performance and Analysis Visualizing Spark Activity Granularity GC’d Sparks and Speculative Parallelism Parallelizing Lazy Streams with parBuffer Chunking Strategies The Identity Property Chapter 4. Dataflow Parallelism: The Par Monad Example: Shortest Paths in a Graph Pipeline Parallelism Rate-Limiting the Producer Limitations of Pipeline Parallelism Example: A Conference Timetable Adding Parallelism Example: A Parallel Type Inferencer Using Different Schedulers The Par Monad Compared to Strategies Chapter 5. Data Parallel Programming with Repa Arrays, Shapes, and Indices Operations on Arrays Example: Computing Shortest Paths Parallelizing the Program Folding and Shape-Polymorphism Example: Image Rotation Summary Chapter 6. GPU Programming with Accelerate Overview Arrays and Indices Running a Simple Accelerate Computation Scalar Arrays Indexing Arrays Creating Arrays Inside Acc Zipping Two Arrays Constants Example: Shortest Paths Running on the GPU Debugging the CUDA Backend Example: A Mandelbrot Set Generator Part II. Concurrent Haskell Chapter 7. Basic Concurrency: Threads and MVars A Simple Example: Reminders Communication: MVars MVar as a Simple Channel: A Logging Service MVar as a Container for Shared State MVar as a Building Block: Unbounded Channels Fairness Chapter 8. Overlapping Input/Output Exceptions in Haskell Error Handling with Async Merging Chapter 9. Cancellation and Timeouts Asynchronous Exceptions Masking Asynchronous Exceptions The bracket Operation Asynchronous Exception Safety for Channels Timeouts Catching Asynchronous Exceptions mask and forkIO Asynchronous Exceptions: Discussion Chapter 10. Software Transactional Memory Running Example: Managing Windows Blocking Blocking Until Something Changes Merging with STM Async Revisited Implementing Channels with STM More Operations Are Possible Composition of Blocking Operations Asynchronous Exception Safety An Alternative Channel Implementation Bounded Channels What Can We Not Do with STM? Performance Summary Chapter 11. Higher-Level Concurrency Abstractions Avoiding Thread Leakage Symmetric Concurrency Combinators Timeouts Using race Adding a Functor Instance Summary: The Async API Chapter 12. Concurrent Network Servers A Trivial Server Extending the Simple Server with State Design One: One Giant Lock Design Two: One Chan Per Server Thread Design Three: Use a Broadcast Chan Design Four: Use STM The Implementation A Chat Server Architecture Client Data Server Data The Server Setting Up a New Client Running the Client Recap Chapter 13. Parallel Programming Using Threads How to Achieve Parallelism with Concurrency Example: Searching for Files Sequential Version Parallel Version Performance and Scaling Limiting the Number of Threads with a Semaphore The ParIO monad Chapter 14. Distributed Programming The Distributed-Process Family of Packages Distributed Concurrency or Parallelism? A First Example: Pings Processes and the Process Monad Defining a Message Type The Ping Server Process The Master Process The main Function Summing Up the Ping Example Multi-Node Ping Running with Multiple Nodes on One Machine Running on Multiple Machines Typed Channels Merging Channels Handling Failure The Philosophy of Distributed Failure A Distributed Chat Server Data Types Sending Messages Broadcasting Distribution Testing the Server Failure and Adding/Removing Nodes Exercise: A Distributed Key-Value Store Chapter 15. Debugging, Tuning, and Interfacing with Foreign Code Debugging Concurrent Programs Inspecting the Status of a Thread Event Logging and ThreadScope Detecting Deadlock Tuning Concurrent (and Parallel) Programs Thread Creation and MVar Operations Shared Concurrent Data Structures RTS Options to Tweak Concurrency and the Foreign Function Interface Threads and Foreign Out-Calls Asynchronous Exceptions and Foreign Calls Threads and Foreign In-Calls Index