The Rootbeer GPU Compiler makes it easy to use Graphics Processing Units from within Java.
Rootbeer is more advanced than CUDA or OpenCL Java Language Bindings. With bindings the developer must serialize complex graphs of objects into arrays of primitive types. With Rootbeer this is done automatically. Also with language bindings, the developer must write the GPU kernel in CUDA or OpenCL. With Rootbeer a static analysis of the Java Bytecode is done (using Soot) and CUDA code is automatically generated.
doc/hpcc_rootbeer.pdf for the conference slides from HPCC-2012.
doc/rootbeer1_paper.pdf for the conference paper from HPCC-2012.
Rootbeer is licensed under the MIT license.
Rootbeer was created using Test Driven Development and testing is essentially important in Rootbeer. Rootbeer is 20k lines of product code and 7k of test code and all tests pass on both Windows and Linux. The Rootbeer test case suite covers every aspect of the Java Programming language except:
- native methods
- dynamic method invocation
- sleeping while inside a monitor.
This means that all of the familar Java code you have been writing can be executed on the GPU
The original publication for Rootbeer was in HPCC-2012.
"Rootbeer: Seamlessly using GPUs from Java"
Philip C. Pratt-Szeliga, James W. Fawcett, Roy D. Welch.
To appear in HPCC-2012.
This work is supported by the National Science Foundation.
- Clone the github repo to
- Use the
You need to have the CUDA Toolkit and CUDA Driver installed to use Rootbeer. Download it from http://www.nvidia.com/content/cuda/cuda-downloads.html
We are looking for people to provide test machines accessible over the internet. We would like to be able to run all the test cases every time a push is made to github on Windows, Linux and Mac on a variety of GPU models. Please email firstname.lastname@example.org if you are have resources and are interested in helping out.
Information about Rootbeer
Rootbeer announcements are pushed to @rootbeer_gpu
Rootbeer mailing list: http://chirrup.org/pipermail/rootbeer-dev/
Rootbeer does NOT automatically parallelize code. It provides help serializing state and writing CUDA code, but you still must specify what each GPU core is going to do.
Phil Pratt-Szeliga is still a PhD student at Syracuse University. All the code was written by Phil. Dr. Jim Fawcett and Dr. Roy Welch are Phil's PhD advisors.
Rootbeer is written by: