You can read more about the release on the EPW website:
https://docs.epw-code.org/doc/Releases.html
For instructions on downloading, installing, and running EPW v6.1, please refer to the updated Download and Install page:
https://docs.epw-code.org/DownloadAndInstall.html
EPW v6.1 introduces several major new features and performance improvements:
- GPU offloading support through CUDA, OpenACC, and OpenMP
EPW can now take advantage of GPU-accelerated computing platforms using CUDA, OpenACC, and OpenMP target offloading. Installation and execution instructions for GPU-enabled systems are available on the updated Download and Install page. - Two-level parallelization across EPW
Two-level MPI parallelization using image and pool parallelism has been extended across EPW. The implementation supports parallelization of calculations involving both coarse and fine electron and phonon grids.
The pool level distributes electronic k-points, while the image level distributes phonon q-points and related workloads.
A calculation using both levels of parallelization can be launched using:
where:Code: Select all
mpirun -np <nprocs> epw.x -nk <npool> -ni <nimage> -in epw.in > epw.out
nprocs = npool × nimage - Support for DFPT+U calculations
EPW v6.1 supports electron–phonon calculations based on density-functional perturbation theory with Hubbard corrections (DFPT+U). This enables the study of electron–phonon interactions and related properties in materials containing localized electronic states. - Support for LSDA calculations
EPW v6.1 supports calculations using the local spin-density approximation (LSDA), enabling electron–phonon and related calculations for spin-polarized systems.