barrucadu/dejafu
1
1
Fork 0
mirror of https://github.com/barrucadu/dejafu.git synced 2025-01-07 06:17:21 +03:00
Code Issues Projects Releases Wiki Activity
e8bbb191ed
dejafu/dejafu
History
Michael Walker e8bbb191ed Add compare-and-swap primitives to MonadConc
2015-11-15 15:42:45 +00:00
..
Control Add compare-and-swap primitives to MonadConc 2015-11-15 15:42:45 +00:00
Data/List Stick everything in a dejafu subdirectory 2015-10-05 16:44:14 +01:00
Test Add compare-and-swap primitives to MonadConc 2015-11-15 15:42:45 +00:00
tests Fix dejafu testsuite 2015-10-27 17:59:00 +00:00
dejafu.cabal Add compare-and-swap primitives to MonadConc 2015-11-15 15:42:45 +00:00
LICENSE Stick everything in a dejafu subdirectory 2015-10-05 16:44:14 +01:00
README.markdown Clarify default memory model 2015-11-15 15:39:07 +00:00
Setup.hs Stick everything in a dejafu subdirectory 2015-10-05 16:44:14 +01:00

README.markdown

dejafu Build Status

[Déjà Fu is] A martial art in which the user's limbs move in time as well as space, […] It is best described as "the feeling that you have been kicked in the head this way before"

-- Terry Pratchett, Thief of Time

Concurrency is nice, deadlocks and race conditions not so much. The Par monad family, as defined in abstract-par provides deterministic parallelism, but sometimes we can tolerate a bit of nondeterminism.

This package provides a class of monads for potentially nondeterministic concurrency, with an interface in the spirit of GHC's normal concurrency abstraction.

The documentation of the latest developmental version is available online.

Note on running the test suite: I can't seem to figure out how to have cabal or stack nicely build both dejafu and hunit-dejafu, and then run the dejafu test suite without getting conflicting package IDs (dejafu gets built at least twice, it seems). However, using runhaskell to execute Tests.hs from within the tests directory works. I expected this to just work, so I've filed an issue with stack in the hope that this workflow becomes supported.

MonadConc and IO

The intention of the MonadConc class is to provide concurrency where any apparent nondeterminism arises purely from the scheduling behaviour. To put it another way, a given computation, parametrised with a fixed set of scheduling decisions, is deterministic. This assumption is used by the testing functionality provided by Control.Monad.Conc.SCT.

Whilst this assumption may not hold in general when IO is involved, you should strive to produce test cases where it does.

Memory Model

The testing functionality supports a few different memory models, for computations which use non-synchronised CRef operations. The supported models are:

  • Sequential Consistency: A program behaves as a simple interleaving of the actions in different threads. When a CRef is written to, that write is immediately visible to all threads.

  • Total Store Order (TSO): Each thread has a write buffer. A thread sees its writes immediately, but other threads will only see writes when they are committed, which may happen later. Writes are committed in the same order that they are created.

  • Partial Store Order (PSO): Each CRef has a write buffer. A thread sees its writes immediately, but other threads will only see writes when they are committed, which may happen later. Writes to different CRefs are not necessarily committed in the same order that they are created.

If a testing function does not take the memory model as a parameter, it uses TSO.

Contributing

Bug reports, pull requests, and comments are very welcome!

Feel free to contact me on GitHub, through IRC (#haskell on freenode), or email (mike@barrucadu.co.uk).