Hi all,
Here are the minutes from today’s CarpetX Call:
Present: Liwei, Roland, Steve, Lucas, Cheng-Hsin, Yosef, Allen, Helvi, Swapnil, Zach, Erik
Liwei Optimized the subcycling implementation in CarpetX, acchieving better performance compared to Carpet
Roland Working on adding a Z4c test
Lucas Tuning the parameter within the hybrid RK integrator to achieve larger CFL factors (7x faster already compared to the traditional RK method) Trying to evolve BBH simulations on Multipatch grids using CanudaX PR on NewRadX is almost ready
Cheng-Hsin Successfully compiled CanudaX on Bridges-2, which features V100 GPUs. Next steps include testing and profiling its performance Based on the tests on Delta, CanudaX is slightly faster then Z4c, but still comparable (~1.5 M/h on 1 GPU node, with qc0).
Jay Installing Fuka on Frontier to test the optimized subcycling implementation with BNS simulations
Swapnil A colleague is interested in porting their nucleosynthesis code to CarpetX
Zach BHaHAHA is working with ETK, doing some final performance and accuracy tests
Steve Progress on generating BSSN with NRPy+ for CarpetX
Erik Investigating an issue where the compiler struggles with accessing grid functions using foo(p.I), particularly when an object file is provided.
Regards, Liwei