Developing scalable software that leverages exascale computing is critically important for first-principles calculations. Perturbo enables an efficient GPU parallel acceleration of electron-phonon algorithms for both transport and nonequilibrium dynamics calculations in the Boltzmann equation formalism employing data structures and code optimized for GPUs.
Usage
'trans'
and 'dynamics-run'
supports GPU.Overall, it just takes two steps to invoke GPU capability in Perturbo. First, you should compile Perturbo with OpenACC enabled (refer to Download and installation). Second, just one additional parameter needs to be specified in the pert.in. This input parameter will lead the code to utilitze the well-designed target data structures.
&perturbo
scat_impl = 'tgt'
/
Finally, you just need to launch the job using the GPU version of Perturbo executable in the usual way. Please refer to individual job management system for how to submit a job to GPU architectures.