The goals of this project
- To continue evolution of the API specification as an open standard, such that it meets the needs of the parallel graphics community.
- To develop an optimized cross-platform reference implementation of the API.
- To encourage development of parallel graphics applications based on this API.
The current library implementation has the following features:
- Depth compositing, Ordered alpha compositing
- Screen space compositing
- Native support for InfiniBand
- Fast compositing code for Xeon/Opteron CPUs(x86_64)
- OpenGL contexts as source of pixels
Release 1.2 of the library is available for download from SourceForge.
A set of sample programs is included in the source tree of the library. Please look at the "samples" directory after downloading.
The project uses Subversion, and the sources can be viewed here. If the release version does not work for you, then please download the latest stable snapshot using anonymous SVN, as follows:
# svn co https://paracomp.svn.sourceforge.net/svnroot/paracomp/trunk paracomp
Reference Guide: A user guide for the library can be downloaded from this link. This document describes the concepts that an application programmer needs to understand, describes the samples, and has coding tips too. The document is primarily aimed at users of HP's Scalable Visualization Array product, but most of the content is applicable on any generic cluster.
API Specification: The complete specification for v1.01 of the API is available as a PDF document
Whitepaper: A whitepaper about the benefits of the library and how to use it is available from this link
Benchmark Framework: A framework to benchmark the performance of any implementation of the API is available from this webpage. This benchmark was contributed by BME Innovation and Knowledge Centre of Information Technology.
Who is using the library?
The University of Manchester has developed versions of AVS/Express for multi CPU and GPU systems. Parallel volume rendering and sort last compositing of geometry is supported. The project is supported by KGT.
Image shows the Armadillo dataset (from the UTCT Data Archive). The 252x256x953 dataset is distributed over 4 nodes.
The Technical University of Budapest has developed several projects using the library:
Mailing ListsThe following mailing lists are available from the SourceForge project page
- paracomp-announce - Major announcements related to the project
- paracomp-users - For the users of the API
- paracomp-devel - For the developers of the API
- paracomp-spec - Discussions related to the Parallel Compositing library specification, including change proposals
- paracomp-commits - SVN commit notifications