QUANTUM ESPRESSO is an open-source distribution of computer codes for quantum-mechanical materials modeling, based on density-functional theory, pseu-
dopotentials, and plane waves, and renowned for its performance on a wide range of hardware architectures, from laptops to massively parallel computers, as well
as for the breadth of its applications.
The goal of this thesis is to describe recent and ongoing work to port QUANTUM ESPRESSO software distribution to heterogeneous architectures based on hard-
ware accelerators, which will overcome the energy constraints that are currently hindering the way toward exascale computing. We explore the diagonalization
methods used in the PWscf (Plane-Wave Self Consistent Field), a key component of the Quantum ESPRESSO open-source suite of codes for materials modelling.
For the high performance of the iterative diagonalization solvers several solutions are proposed and Porting of existing solvers to GPU systems using CUDA Fortran and Kernel loop directives (CUF kernels).We focus, in particular, on the implementation and the GPU porting of a new algorithm for the iterative diagonalization methods call RMM-DIIS which is based on the the minimization of the norm of the residual vector.
Thesis not available.