m (clear some confusion, using '+RTS' wasn't working, '+RTS -N' is needed (ghc 6.12.1 Linux x86_64))
(→Multicore GHC: use transclusion so we can share practical tips)
|Line 38:||Line 38:|
=== Multicore GHC ===
=== Multicore GHC ===
=== Related work ===
=== Related work ===
Revision as of 14:17, 16 March 2011
1 Parallel and Concurrent Programming in GHC
This page contains notes and information about how to write concurrent and/or parallel programs in GHC.
GHC provides multi-scale support for parallel programming, from very fine-grained, small "sparks", to coarse-grained explicit threads and locks, along with other models of concurrent and parallel programming, including actors, CSP-style concurrency, nested data parallelism and Intel Concurrent Collections. Synchronization between tasks is possible via messages, regular Haskell variables, MVar shared state or transactional memory.
- See "Real World Haskell" chapter 24, for an introduction to the most common forms of concurrent and parallel programming in GHC.
- A reading list for parallelism in Haskell.
- The status of parallel and concurrent programming in Haskell.
The concurrent and parallel programming models in GHC can be divided into the following forms:
- Very fine grained: parallel sparks and futures, as described in the paper "Runtime Support for Multicore Haskell"
- Fine grained: lightweight Haskell threads, explicit synchronization with STM or MVars. See the paper "Tackling the Awkward Squad" below.
- Nested data parallelism: a parallel programming model based on bulk data parallelism, in the form of the DPH and Repa libraries for transparently parallel arrays.
- Intel Concurrent Collections for Haskell: a graph-oriented parallel programming model.
- CHP: CSP-style concurrency for Haskell.
The most important (as of 2010) to get to know are the basic "concurrent Haskell" model of threads using forkIO and MVars, the use of transactional memory via STM, implicit parallelism via sparks and, if you're interested in scientific programming specifically, nested data parallelism in Haskell.
1.1 Starting points
- Basic concurrency: forkIO and MVars.
- Software Transactional Memory (STM) is a new way to coordinate concurrent threads. There's a separate Wiki page devoted to STM.
- STM was added to GHC 6.4, and is described in the paper Composable memory transactions. The paper Lock-free data structures using Software Transactional Memory in Haskell gives further examples of concurrent programming using STM.
- Foreign function interface. If you are calling foreign functions in a concurrent program, you need to know about bound threads. They are described in a Haskell workshop paper, Extending the Haskell Foreign Function Interface with Concurrency. The GHC Commentary Supporting multi-threaded interoperation contains more detailed explanation of cooperation between FFI calls and multi-threaded runtime.
- Nested Data Parallelism. For an approach to exploiting the implicit parallelism in array programs for multiprocessors, see Data Parallel Haskell (work in progress).
1.2 Using concurrency in GHC
- You get access to concurrency operations by importing the library Control.Concurrent.
- The GHC manual gives a few useful flags that control scheduling (not usually necessary) RTS options.
1.3 Multicore GHC
Since 2004, GHC supports running programs in parallel on an SMP or multi-core machine. How to do it:
- Compile your program using the -threaded switch.
- Run the program with +RTS -N2 to use 2 threads, for example (RTS stands for runtime system; see the GHC users' guide). You should use a -N value equal to the number of CPU cores on your machine (not including Hyper-threading cores). As of GHC v6.12, you can leave off the number of cores and all available cores will be used (you still need to pass -N however, like so: +RTS -N).
- Concurrent threads (forkIO) will run in parallel, and you can also use the par combinator and Strategies from the Control.Parallel.Strategies module to create parallelism.
- Use +RTS -sstderr for timing stats.
- To debug parallel program performance, use ThreadScope.
1.4 Related work
- The Sun project to improve http://ghcsparc.blogspot.com/ GHC performance on Sparc]
- A Microsoft project to improve industrial applications of GHC parallelism.
- Simon Marlow's publications on parallelism and GHC
- Glasgow Parallel Haskell
- Glasgow Distributed Haskell