Java Matrix Benchmark (JMatBench) is a tool used to evaluate the runtime performance, stability, and memory efficiency of linear algebra libraries at performing dense real operations.
WARNING The latest version is a major modification from the past and isn't fully functional yet
JMatBench is designed to be run as a command line application. As much of the process has been automated as possible, however a few libraries aren't on Maven central so you need to download their jars manually and place them in the libraries "libs/" directory. Instructions for where to download the latest jar is typically provided in the "libs/source.txt". After those jars have been downloaded gradle will be able to compile the project.
- Install dependencies required by tested libraries
sudo apt-get install libblas-dev liblapack-dev
- Manual download jars for libraries which haven't been posted on Maven central
- jeigen: See its readme.md
- ./gradlew createRuntimeJars; ./gradlew benchmarkJar
- java -jar benchmark.jar
- Follow printed instructions for the benchmark you wish to run
Create a file called "email_login.txt". DO NOT ADD TO GIT.
- First line is your login.
- Second line is your password.
- Third line is the destination e-mail.
This is configured for gmail only right now. You need to turn on "Allow less secure apps" by going to https://myaccount.google.com/security
- Create new directory in external
- Copy "src" and "build.gradle" from another project in external into your directory
- Modify the source code, build.gradle, and rename classes
- Create your own "TestSetInfo.txt" file using CreateTestSetInfo.
- Add the library to "external/DefaultTestSet.txt" if you wish it to be included by default
- Add your library to settings.gradle so that Gradle will compile it
This project and all its source code is released under the GNU GENERAL PUBLIC LICENSE version 3 license.