Speaker
Description
In the last decade, advances in theoretical and algorithmic methods together with the growth of computational resources have made lattice QCD a key tool at the precision frontier of particle physics, enabling systematically improvable predictions for observables measured with increasing experimental accuracy. Within this context, we present the development of software for lattice QCD and QCD+QED simulations using the open-source lattice QCD libraries NISSA and openQ*D. In particular, GPU kernels for NVIDIA architectures were developed, together with an interface between the CPU code and the highly optimized QUDA library, developed by the lattice QCD community with the support of NVIDIA to exploit state-of-the-art GPU solvers. At the same time, the CPU codebase was optimized to ensure portability and efficient execution across heterogeneous architectures, preventing non-accelerated components from becoming performance bottlenecks.