speculation: A framework for safe, programmable, speculative parallelism

[ bsd3, concurrency, library ] [ Propose Tags ]

A framework for safe, programmable, speculative parallelism, loosely based on:

This package provides speculative function application and speculative folds. Speculative STM transactions take the place of the transactional rollback machinery from the paper.

For example:

spec g f a evaluates f g while forcing a, if g == a then f g is returned, otherwise f a is evaluated and returned. Furthermore, if the argument has already been evaluated, we skip the f g computation entirely. If a good guess at the value of a is available, this is one way to induce parallelism in an otherwise sequential task. However, if the guess isn't available more cheaply than the actual answer, then this saves no work and if the guess is wrong, you risk evaluating the function twice. Under high load, since 'f g' is computed via the spark queue, the speculation will be skipped and you will obtain the same answer as 'f $! a'.

The best-case timeline looks like:

foreground: [----- a -----]
foreground:               [-]    (check g == a)
spark:         [----- f g -----]
overall:    [--- spec g f a ---]

The worst-case timeline looks like:

foreground: [----- a -----]
foreground:               [-]               (check g == a)
foreground:                 [---- f a ----]
spark:         [----- f g -----]
overall:    [-------- spec g f a ---------]

Note that, if f g takes longer than a to compute, in the HEAD release of GHC, f g will be collected and killed during garbage collection.

foreground: [----- a -----]
foreground:               [-]               (check g == a)
foreground:                 [---- f a ----]
spark:         [---- f g ----######         (#'s mark when this spark is collectable)
overall:    [--------- spec g f a --------]

Under high load:

foreground: [----- a -----]
foreground:               [-]               (check g == a)
foreground:                 [---- f a ----]
overall:    [-------- spec g f a ---------]

Compare these to the timeline of f $! a:

foreground: [----- a -----]
foreground:               [---- f a ----]
orverall:   [---------- f $! a ---------]

specSTM provides a similar time table for STM actions, but also rolls back side-effects. The one unfortunate operational distinction is that it is forced to compute a in the background thread and therefore degrades slightly less gracefully under load, although we mitigate this effect by only enqueuing if the number of sparks for the current capability is lower than the total number of capabilities, to try to avoid wasting time when all computational resources are in use.


[Skip to Readme]

Flags

Automatic Flags
NameDescriptionDefault
optimize

Enable optimizations for the library and benchmarks

Enabled

Use -f <flag> to enable a flag, or -f -<flag> to disable that flag. More info

Downloads

Maintainer's Corner

Package maintainers

For package maintainers and hackage trustees

Candidates

  • No Candidates
Versions [RSS] 0.0.0, 0.0.1, 0.0.2, 0.1.0, 0.2.0, 0.3.0, 0.4.0, 0.5.0, 0.5.1, 0.6.0, 0.7.0, 0.8.0, 0.8.0.1, 0.8.0.2, 0.8.1.0, 0.8.2.0, 0.9.0.0, 1.0.0.0, 1.1.0.0, 1.2.0.0, 1.2.0.1, 1.2.0.2, 1.3, 1.4, 1.4.1, 1.4.1.1, 1.4.1.2, 1.5, 1.5.0.1, 1.5.0.2, 1.5.0.3
Change log CHANGELOG.markdown
Dependencies base (>=4.3 && <6), ghc-prim, stm (>=2.1 && <2.5), tag-bits (>=0.1 && <0.2), transformers (>=0.2.2.0 && <0.4) [details]
License BSD-3-Clause
Copyright (c) 2010 Edward A. Kmett
Author Edward A. Kmett
Maintainer Edward A. Kmett <ekmett@gmail.com>
Category Concurrency
Home page http://github.com/ekmett/speculation
Bug tracker http://github.com/ekmett/speculation/issues
Source repo head: git clone http://github.com/ekmett/speculation.git -b master
Uploaded by EdwardKmett at 2012-09-10T21:21:43Z
Distributions
Reverse Dependencies 3 direct, 7788 indirect [details]
Downloads 23944 total (70 in the last 30 days)
Rating 2.0 (votes: 1) [estimated by Bayesian average]
Your Rating
  • λ
  • λ
  • λ
Status Docs uploaded by user
Build status unknown [no reports yet]

Readme for speculation-1.4.1.2

[back to package description]

speculation

A framework for safe, programmable, speculative parallelism, loosely based on:

This package provides speculative function application and speculative folds. Speculative STM transactions take the place of the transactional rollback machinery from the paper.

You can download it using cabal install speculation, if you have the Haskell Platform installed.

Speculative Function Application (Control.Concurrent.Speculation)

Various speculative function application combinators are provided. Two fairly canonical samples are described here.

spec

spec :: Eq a => a -> (a -> b) -> a -> b

spec g f a evaluates f g while forcing a, if g == a then f g is returned. Otherwise f a is evaluated.

Furthermore, if the argument has already been evaluated, we avoid sparking the parallel computation at all.

If g is a good guess at the value of a, this is one way to induce parallelism in an otherwise sequential task.

However, if g isn't available more cheaply than a, then this saves no work, and if g is wrong, you risk evaluating the function twice. spec a f a = f $! a

The best-case timeline looks like: [---- f g ----] [----- a -----] [-- spec g f a --]

The worst-case timeline looks like: [---- f g ----] [----- a -----] [---- f a ----] [------- spec g f a -----------]

Compare these to the timeline of f $! a: [---- a -----] [---- f a ----]

specSTM

specSTM provides a similar compressed timeline for speculated STM actions, but also rolls back side-effects.

Speculative Folds (Data.Foldable.Speculation)

A speculative version of the combinators from Data.Foldable is provided as Data.Foldable.Speculation.

Each combinator therein takes an extra argument that is used to speculate on the value of the list.

foldr

foldr :: (Foldable f, Eq b) => (Int -> b) -> (a -> b -> b) -> b -> f a -> b

Given a valid estimator g, foldr g f z xs yields the same answer as Foldable.foldr' f z xs.

g n should supply an estimate of the value returned from folding over the last n elements of the container.

As with spec, if the guess g n is accurate a reasonable percentage of the time and faster to compute than the ensuing fold, then this can provide increased opportunities for parallelism.

foldl

foldl :: (Foldable f, Eq b) => (Int -> b) -> (b -> a -> b) -> b -> f a -> b

foldl works similarly to Foldable.foldl', except that g n should provide an estimate for the first n elements.

Contact Information

Contributions and bug reports are welcome!

I can be reached through the user ekmett on github, as edwardk on irc.freenode.net #haskell channel, or by email at ekmett@gmail.com.

-Edward Kmett