superB: a task-based parallelized version of the BlackHoles@Home numerical relativity code for simulating extreme binary black hole scenarios

Next-generation gravitational wave (GW) detectors will require a tenfold increase in numerical relativity (NR) waveform accuracy to model high mass ratio inspirals and commensurate efficiency boosts to enable rapid follow-ups to GW detections. BlackHoles@Home (BH@H) facilitates binary black hole simulations on consumer hardware but takes months and is limited to moderate mass ratios. We present superB, a supercomputer-ready version of BH@H that aims to unlock high mass ratios and rapid GW followups by integrating task-based parallelism with the Charm++ framework for high-performance clusters. Over the past year, we developed a single-grid approach optimized for black hole spectroscopy of merged binaries sharing an event horizon. This method uses spherical-like coordinates to concentrate grid points where needed, eliminating grid refinement levels and avoiding spurious reflections at boundaries. We performed simulations of merged spinning black holes, successfully extracting quasinormal modes and validating superB through rigorous convergence tests, achieving excellent scalability on supercomputing clusters. Our results demonstrate that superB enhances NR waveform accuracy and computational efficiency, paving the way for extreme binary black hole simulations and rapid follow-ups.