Introducing QVAL, a desktop command-line tool for large-scale, massive-parallel parameter space exploration.

QVAL is designed to run millions of evaluations simultaneously, harnessing the raw parallel power of your GPU to deliver results in seconds. You define your objective using a simple YAML configuration file—no code changes or programming expertise required.

You just have to specify your cost function to evaluate. It could be a simple math expression—or even complex OpenCL code.

It works directly with your data. Point QVAL to columns in your existing CSV or Excel (XLSX) files, and it will use that data as input for its calculations. Whether you’re a developer, analyst, or just someone trying to find the best answer hidden in your spreadsheet—QVAL helps you test thousands or millions of combinations automatically.

Examples? Think finding the optimal combination of loan term and interest rate that minimizes your total repayment, or figuring out the best schedule of shifts (evaluated as whole numbers) that covers your weekly needs while minimizing cost. QVAL handles the heavy lifting so you don’t have to guess or manually try endless combinations. It generates variations using flexible methods (Random, Sobol, LHS, Halton), compiles an optimized OpenCL kernel for your expression, and evaluates everything on your hardware. The output is a clear, ranked report in CSV or XLSX format, complete with summary statistics and the generated kernel code for full traceability.

QVAL is free for personal and academic use.

Download for Mac (ARM) or Windows (x64):
https://mdx1358.github.io/qval

If you have any questions, feel free to contact Denis at mdx1358@fastmail.com.