Batchification of the Lanczos exponential method. This heavily optimize this path in the code as it allows to
- apply the phase in batch mode
- apply the Hamiltonian over batches
- perform Gram-Schmidt orthogonalization in batches
- use GPU acceleration
The new version will be used for most of the propagators, with the exception of ETRS and ETRSsc, as this would require a bit more work to have the combined two exponentials working efficiently.
Closes #163 (closed).
Batchification of the Lanczos exponential method for improved performances.
- I have checked that my code follows the Octopus coding standards
- I have added tests for all the new features added in this request.