Work stealing

As explained in Learn Haskell by Example

Work stealing is a scheduling technique in which an idle execution worker takes available work from a shared pool.

Work stealing schedules parallel work by letting an idle Haskell execution context take a spark from a shared spark pool and evaluate it. section 13.4.3

The runtime handles the main scheduling details:

• rpar places a computation, represented by a spark, into the spark pool without waiting for it to finish. section 13.4.3
• An available Haskell execution context steals a spark and evaluates its underlying thunk in parallel. section 13.4.3
• Sparks can become duds, fizzle, overflow, or be garbage collected; the runtime manages these cases. section 13.4.3

Parallel list evaluation

Control.Parallel.Strategies can apply rpar across a list with evalList, while withStrategy starts the chosen evaluation strategy. Chunking work reduces excessive spark creation for large vectors.

Figure 13.5 Haskell execution contexts and spark pools

A figure showing Haskell execution contexts taking sparks from shared spark pools. section 13.4.3

How Learn Haskell by Example covers it

How deep it goes

Explains spark pools, execution contexts, and failure states, while deferring scheduling internals to the runtime.

Examples

Uses parallel list mapping, pixel-vector processing, chunked execution, and runtime spark statistics.

What it assumes

Assumes familiarity with Haskell computations and introduces sparks, thunks, execution contexts, and evaluation strategies.

What you can do after

Apply it via Control.Parallel.Strategies to parallelize list or vector computations and inspect spark behavior.

How it connects

Connects sparks, thunks, Haskell execution contexts, spark pools, rpar, rseq, and parallel evaluation strategies.

liveBookmanning.com

Read Learn Haskell by Example on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Clojure: The Essential ReferenceWork-stealing is a parallel scheduling technique in which an idle worker takes unprocessed tasks from another worker's queue.
• 

  liveBookmanning.com
  Learn Concurrent Programming with GoWork stealing is a scheduling technique in which an execution worker takes tasks from another worker's queue to improve utilization.
• 

  liveBookmanning.com
  C++ Concurrency in Action, Second EditionWork stealing is a scheduling technique in which an idle worker takes tasks from another worker’s queue.
• 

  liveBookmanning.com
  100 Go Mistakes and How to Avoid ThemWork stealing is a scheduling technique in which an underutilized processor obtains runnable tasks from another processor's queue.

Compare how 5 books explain work stealing →

Related concepts

• green threadused by · A green thread is a lightweight user-space execution unit whose scheduling and context switching are managed by a runtime and multiplexed onto operating-system threads.
• reduction operationused by · A reduction operation combines multiple input values or partial results according to a combining rule to produce fewer results, often a single aggregate.
• parallel streamused by · A parallel stream is a Java stream configured to allow its operations to process elements concurrently across multiple execution threads.
• parallelismimplements · Parallelism is the concurrent execution of multiple computational operations, often across multiple processing resources.
• work queue patternuses · The work queue pattern stores pending units of work for workers or processors to retrieve and process, often asynchronously.

This explainer was generated by Manning's knowledge graph from the text of Learn Haskell by Example. The section links point to the parts of the book it draws on.