Benchmarking platform

The benchmarking platform allows to submit results for different client analyses. Functionality is there, usability and performance need fine-tuning. But it's worth checking it out.

The test setup of the benchmarking platform is online here.

You can basically do two things: Upload results for a points-to analysis (that's the main purpose), and run microbenchmark tests, which tell you approximately what your P2A does and what not.

For the latter, click on Upload Microbenchmark Data and follow the instructions. You can find some sample files to upload in the SVN repository here.

For the former, you can do the following to try out the platform without actually having to perform own experiments: Register a new project, say, "antlr-test" with version "1.2.3" and specify name filter as "antlr", everything else can be left blank. Then you can upload files as found here. 2_... and 5_... are static result files, the other are dynamic files. Upload the static files first to avoid a current bug (ooops), then the dynamic result files. Note that after clicking "Submit uploaded data", nothing seems to be happening for a few seconds. That's a pure usability issue due to an AJAX request, it's being performed after all. After a few seconds, you will be forwarded to the "results" page. Click on "Back to project" to submit more results.

Note that user management (and with it, management of absence proofs as well) on the platform is currently removed. I was not at all happy with the implementation, and it makes it much easier for potential users to test the platform quickly.