Accelerating Numerical Relativity with NRPy: GPU-Optimized Code for Consumer Hardware

Numerical relativity (NR) provides essential tools for modeling and interpreting gravitational wave signals; however, solving the initial value problem and performing time evolutions remain computationally challenging. Despite over five decades of advancements in mathematical techniques and significant improvements in computing power (CPUs, GPUs, hybrid systems), NR simulations are still highly taxing on modern hardware, often making them impractical for consumer-grade devices. I present ongoing efforts to extend NRPy, a Python-based NR code generation framework, to produce highly optimized GPU code, enabling robust and efficient NR applications on consumer PCs. The effectiveness of this extension is demonstrated through initial data generation via NRPyElliptic and simulations of binary black holes using BlackHoles@Home.