High Scale Lib

A fork of Cliff Click's High Scale Library. Improved with bug fixes and a real build system.
Alternatives To High Scale Lib
Project NameStarsDownloadsRepos Using ThisPackages Using ThisMost Recent CommitTotal ReleasesLatest ReleaseOpen IssuesLicenseLanguage
High Scale Lib35918618 years ago1May 13, 20146Java
A fork of Cliff Click's High Scale Library. Improved with bug fixes and a real build system.
High Scale Lib150
5 years ago6November 28, 20114otherJava
A Mavenized fork of http://high-scale-lib.sourceforge.net/
4 months ago6mitRust
Simple, fast, standalone pastebin service
4 years ago1Java
支持并发的实时排行、排名、排行榜、排序工具(Support for concurrent real-time ranking\sorting tools)-java
Cwalk22432 years ago1January 25, 20211mitGo
cwalk = Concurrent filepath.Walk replacement (Go)
10 years ago25otherPython
Concurrent FreeBSD port builder
The Computer7
5 years agoJavaScript
Concurrent computations in node.js made easy.
Spmc Buffer6
5 years ago8February 11, 20186mpl-2.0Rust
An extension of triple-buffering to multiple-consumer scenarios
9 years agomit
Highly concurrent programming language tuned for raw performance
5 years agomitObjective-C
wrapper basic on YYDispatchQueuePool, avoid thread count too many !
Alternatives To High Scale Lib
Select To Compare

Alternative Project Comparisons

High Scale Lib

This is Boundary's fork of Cliff Click's high scale lib. We will be maintaining this fork with bug fixes, improvements and versioned builds in maven central.

To use high scale lib from ivy declare the dependency in ivy.xml:

  rev="1.0.3" />

From maven simply add the dependency to the dependencies element:


Original README

IF YOU ARE LOOKING for the drop-in replacement for java.util.Hashtable, it's in the lib directory, lib/java_util_hashtable.jar. It needs to be in your bootclasspath. Example:

java -Xbootclasspath/p:lib/java_util_hashtable.jar my_java_app_goes_here

A collection of Concurrent and Highly Scalable Utilities. These are intended as direct replacements for the java.util.* or java.util.concurrent.* collections but with better performance when many CPUs are using the collection concurrently. Single-threaded performance may be slightly lower.

The direct replacements match the API - but not all behaviors are covered by the API, and so they may not work for your program. In particular, the replacement for java.util.Hashtable is NOT synchronized (that is the point!), although it is multi-threaded safe. If you rely on the undocumented synchronization behavior of the JDK Hashtable, your program may not work. Similarly, the iteration order is different between this version and the JDK version (this exact issue broke the SpecJBB benchmark when the iteration order was changed slightly (via using a slightly different hash function) between JDK rev's).

If you want to drop-in the non-blocking versions of Hashtable, HashMap or ConcurrentHashMap, you'll need to alter your bootclasspath - these classes come directly from your JDK and so are found via the System loader before any class-path hacks can be done.

To replace the JDK implementation of Hashtable with a non-blocking version of Hashtable, add java_util_hashtable.jar to your java launch line:

java -Xbootclasspath/p:lib/java_util_hashtable.jar my_app_goes_here

Similarly for ConcurrentHashMap, add java_util_concurrent_chm.jar:

java -Xbootclasspath/p:lib/java_util_concurrent_chm.jar my_app_goes_here

The other utilities do not have direct JDK replacements; you need to call them out directly and place high_scale_lib.jar in your classpath:

  • NonBlockingHashMap - Fast, concurrent, lock-free HashMap. Linear scaling to 768 CPUs.
  • NonBlockingHashMapLong - Same as above, but using primitive 'long' keys
  • NonBlockingHashSet - A Set version of NBHM
  • NonBlockingSetInt - A fast fully concurrent BitVector
  • Counter - A simple counter that scales linearly even when extremely hot. Most simple counters are either unsynchronized (hence drop counts, generally really badly beyond 2 cpus), or are normally lock'd (hence bottleneck in the 5-10 cpu range), or might use Atomic's (hence bottleneck in the 25-50 cpu range). This version scales linearly to 768 CPUs.

Cliff Click

Popular Cpu Projects
Popular Concurrent Projects
Popular Hardware Categories

Get A Weekly Email With Trending Projects For These Categories
No Spam. Unsubscribe easily at any time.