Java Matrix Benchmark

Java Matrix Benchmark (JMatBench) is a tool for evaluating Java linear algebra libraries for speed and stability. This tool can be used by users to select the best library for their application and by developers for identifying bugs and weaknesses. Its goal is to help improve the state of high performance computing on the Java platform.

JMatBench rigorously tests performance across a wide range of matrix sizes and linear algebra operations. The runtime performance benchmark measures how fast each library can run under optimal conditions. It generates accurate results by taking in account JavaVM runtime optimizations, dynamically adjusts to the platform it is run on, and uses well established good benchmarking techniques. The stability benchmark evaluates several standard linear algebra operations for their accuracy, sensitivity, and ability to handle overflow/underflow conditions. For more information see the approach wikipage.

To understand the benchmarks please take a look at the following wikipages:

• RuntimePerformanceBenchmark
• MemoryBenchmark
• StabilityBenchmark

For instructions on running the benchmark click on manual. To understand how the runtime benchmark computes its results click here Methodology_RuntimeBenchmark.

Runtime Performance Results:

Memory Usage Results:

Stability Results:

For older benchmark results see HistoryOfResults.

Tested Libraries:

• Colt
• Commons Math
• Efficient Java Matrix Library (EJML)
• Jama
• jblas
• JScience (Older benchmarks only)
• Matrix Toolkit Java (MTJ)
• OjAlgo
• Parallel Colt
• Universal Java Matrix Package (UJMP)
• Elegant Linear Algebra for java (la4j)

If you wish to ask a question consider posting to the discussion group, or if it is specific to a particular wikipage post a comment on the wikipage itself.

For another benchmark take a look at http://www.ujmp.org/java-matrix/benchmark/.