Project

pmap

0.26
No commit activity in last 3 years
No release in over 3 years
Add parallel methods into Enumerable: pmap and peach
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
 Dependencies

Development

 Project Readme

pmap

Build status Contributor Covenant Apache License

This Ruby gem adds three methods to any Enumerable (notably including any Array). The three added methods are:

  • pmap parallel map
  • peach parallel each
  • peach_with_index parallel each_with_index

Threading in Ruby has limitations

Matz Ruby 1.8.* uses green threads. All Ruby threads are run within a single thread in a single process. A single Ruby program will never use more than a single core of a multi-core machine.

Matz Ruby 1.9.* uses native threads. Each Ruby thread maps directly to a thread in the underlying operating system. In theory, a single Ruby program can use multiple cores. Unfortunately, there is a global interpreter lock GIL that causes single-threaded behavior.

JRuby also uses native threads. JRuby avoids the global interpreter lock, allowing a single Ruby program to really use multiple CPU cores.

Threading useful for remote IO, such as HTTP

Despite the Matz Ruby threading limitations, IO bound actions can greatly benefit from multi-threading. A very typical use is making multiple HTTP requests in parallel. Issuing those requests in separate Ruby threads means the requests will be issued very quickly, well before the responses start coming back. As responses come back, they will be processed as they arrive.

Example

Suppose that we have a function get_quote that calls out to a stock quote service to get a current stock price. The response time for get_quote ranges averages 0.5 seconds.

stock_symbols = [:ibm, :goog, :appl, :msft, :hp, :orcl]

# This will take about three seconds;
# an eternity if you want to render a web page.
stock_quotes = stock_symbols.map {|s| get_quote(s)}

# Replacing "map" with "pmap" speeds it up.
# This will take about half a second;
# however long the single slowest response took.
stock_quotes = stock_symbols.pmap {|s| get_quote(s)}

Thread Count

The thread count defaults to 64 and is set based on $pmap_default_thread_count.

You can also set the thread count per call by passing it as an argument to the pmap and peach methods.

# Use the default thread count (64)
(1..128).peach { |i| sleep 1 } # Takes 2 seconds

# Use a thread count of 128
(1..128).peach(128) { |i| sleep 1 } # Takes 1 second

# Use a thread count of 2
(1..128).peach(2) { |i| sleep 1 } # Takes 64 seconds