A System/Technical Code Library by Axon7

The Jibu software library provides a high-level approach to concurrent and parallel programming that greatly simplifies programming of multi-core machines. By abstracting low-level constructs like threads and various locking and synchronization mechanisms, Jibu enables

* Better program design
* Simple transition from design to implementation
* Faster development
* Less error prone programs
* Easier debugging
* Easier communication and synchronization

Abstracting most of the difficulties inherent in traditional thread programming - such as locking, thread communication and synchronization - makes it much easier to detect and avoid deadlocks, livelocks and race conditions, which is some of the most frequent errors in concurrent and parallel programming.

Jibu contains a few, but very powerful, constructs that allows programmers to define and implement concurrent programs without worrying about low-level technicalities. The main constructs in Jibu are

* Task
* Mailbox
* Channel
* Parallel
* Choice
* Scheduler

Tasks are user defined units of work and are either Futures that return a value or Asyncs that do not return a value. Tasks are run explicit by calling a Start method or implicit by various calls to the static Parallel class.

The Parallel class also features parallel versions of the for-loop, foreach-loop and reduce constructs.

Concurrent programming often involves communication and synchronization between different tasks, and therefore Jibu provides two different methods of communication: mailboxes and channels. They both ensure that communication and synchronization are done safely and efficiently eliminating the dangers of race conditions.

The Choice class makes it easy to coordinate communication between different tasks through channels. A Choice object can listen for data in multiple channels and notify the programmer when tasks write to or read from the channels.
Jibu features an advanced task scheduler that controls scheduling and execution of concurrent tasks. The scheduler automatically adjust the number of operating systems threads in order to maintain the optimal concurrency level. Based on the number of cores in the machine, the number of running tasks and the number of blocked tasks, new threads are created and destroyed dynamically in order to keep the number of working executing threads as close to the number of cores as possible.

Licensing

Free non-commercial version - no support, no bug fixes. Commercial version - includes support, upgrades & bugfixes

System Requirements

Windows XP or later, 32-bit One of the supported C++ compilers: * Microsoft Visual Studio 2005 or 2008 * Intel C++ Compiler 10.1 * Borland C++ Builder 200