README.md

SPECFEM++ benchmarks

This repository contains workflows to generate performance benchmarks for SPECFEM++ against the original SPECFEM solvers (SPECFEM2D).

The benchmarks are generated using Snakemake.

Benchmarks

Benchmarks for 2D forward simulations :

Reproducing the benchmarks

To reproduce the benchmarks, you need to have the following dependencies installed:

1. SPECFEM2D with a working cpu and gpu installation.
2. SPECFEM++ with a working cpu and gpu installation.
3. Poetry for managing python dependencies.

Once you have installed the dependencies, you can run the benchmarks using the following steps:

1. Clone the repository:

git clone 

2. Change to the repository directory:

cd specfempp_benchmarks

3. Install the python dependencies:

poetry install

4. Update the config.yaml(located within forward_simulations directory) file with the paths to the SPECFEM2D and SPECFEM++ executables.

5. Run the benchmarks:

Note: Make sure you have access to a GPU to run GPU benchmarks.

cd forward_simulations
snakemake -j 4

If you are using a SLURM cluster, you can schedule the benchmark runs as slurm jobs using the following command:

cd forward_simulations
snakemake -j 4 --executor slurm

Warning: The benchmarks can take a long time to run depending on the number of processors.

License

SPECFEM++ benchmarks is distributed under the GPL v3 license