Key Features

Simple configs

Describe your formula and parameters in a small YAML file. No code changes needed.

Your data: CSV/XLSX

Read inputs from CSV or Excel (XLSX) by column name or index. Get clear CSV/XLSX reports out.

Fast on your machine

Runs on your GPU when available; can run on CPU when an OpenCL CPU runtime (e.g., POCL) is installed.

Clear results

Top results table, summary stats, and saved kernel/code for traceability (CSV and XLSX outputs).

Flexible sampling

Random, Sobol, LHS, or Halton — choose what works best without complexity.

How it works

1

Define a config

Write a small YAML: expr, goal (min/max), samples, top_k, and params.

2

Load your data

Point to CSV/XLSX columns by name or index, or use constants. Per-parameter or global.

3

Sample variations

Generate candidates using RNG, Sobol, LHS, or Halton to explore the parameter space.

4

Compute

QVAL builds an OpenCL kernel for your expression and evaluates all variations on the device.

5

Get results

QVAL ranks all results, saves the best ones, and generates detailed reports with analysis.

Quick Start

1

Download & unzip

Grab the build for your OS and unzip.

2

Run an example

Show commands
# macOS/Linux
./run_basic_example.sh
# Windows (PowerShell)
.\run_basic_example.ps1
3

Open the results

Where to look
output/tutorial/00_minimal/report/report.txt
output/tutorial/00_minimal/code/eval.cl

CPU execution requires an OpenCL CPU runtime (e.g., POCL). Without it, only GPU devices are used.

Contact

Get in Touch

For inquiries, email: mdx1358@fastmail.com

GitHub

📚 View on GitHub

Documentation, releases, issues, and example configs