This repository contains a mini application that reproduces the gradient computation code inside AVBP. It serves as a testbed for optimizing and benchmarking different versions of the gradient computation with OpenACC, OpenMP, and CUDA.
The repository is structured as follows:
results/: Contains CSV files with execution times for different runs.donnees/: Contains input data for performing gradient operations.versions/: Includes different versions of the code modified for OpenACC, OpenMP, and CUDA.
- A compatible compiler (e.g., NVIDIA HPC SDK for OpenACC, GCC for OpenMP, or CUDA Toolkit for CUDA)
- Make utility
Depending on the compiler and programming model you want to use, modify the Makefile accordingly and then compile the code:
makeThis will generate the test executable.
Once compiled, execute the miniapp as follows:
./testThis will run the gradient computation using the provided data.
This repository is private and restricted to authorized users who have been granted access. The following conditions apply:
- Non-commercial usage only: The software and its contents cannot be used for commercial purposes.
- Restricted distribution: Only users with access to this repository are allowed to use and modify the code.
- No public sharing: Redistribution or public dissemination of this code is not permitted.
This project is licensed under a restricted-use license. Permission is granted solely to authorized users within the repository. Unauthorized distribution, modification, or commercial use is strictly prohibited.
Contributions are welcome from authorized users. If you would like to contribute:
- Fork the repository (if permitted).
- Create a new branch.
- Make your modifications.
- Submit a merge request.
For major changes, please discuss with the maintainers before making modifications.
For any questions or issues, please reach out to the repository maintainers via GitLab.