Bulletin of the American Physical Society
APS April Meeting 2021
Volume 66, Number 5
Saturday–Tuesday, April 17–20, 2021; Virtual; Time Zone: Central Daylight Time, USA
Session T05: Applications of Code Generation and High Performance Computing to Astrophysical SystemsInvited Live
|
Hide Abstracts |
Sponsoring Units: DCOMP Chair: Adam Peterson, Lawrence Berkeley Laboratory |
Monday, April 19, 2021 3:45PM - 4:21PM Live |
T05.00001: NRPy+, a Python-Based Code Generation Package for Numerical Relativity... and Beyond! Invited Speaker: Zach Etienne Encoding partial differential equations (PDEs) into high-performance software by hand can be a frustrating, error-prone process. NRPy+ is a SymPy-based Python package aimed at accelerating this process. Users may input complex expressions in a compact, human-friendly mathematical form (e.g., Einstein notation). NRPy+ then processes this input and outputs highly optimized C code kernels with CSE and/or SIMD compiler intrinsics if desired, converting partial derivatives to finite-difference expressions. NRPy+ has been used to generate kernels for solving Einstein's equations of general relativity on HPC systems in a variety of singular coordinate systems, and it forms the heart of BlackHoles@Home, a proposed volunteer computing project that brings numerical relativity to the consumer-grade desktop computer for the benefit of gravitational wave astronomy. Although originally designed for numerical relativity, NRPy+ has been used to quickly and efficiently solve a number of hyperbolic PDEs, and is pedagogically documented in a collection of Jupyter notebooks. I will report on NRPy+'s capabilities and new features, which include a new \LaTeX\ lexer/parser for inputting expressions. http://nrpyplus.net/ [Preview Abstract] |
Monday, April 19, 2021 4:21PM - 4:57PM Live |
T05.00002: Neutrino Flavor Transformations with Emu: A New Particle-in-Cell Code for Quantum Kinetics Invited Speaker: Don Willcox Neutrino flavor transformations occurring in neutron star mergers and core collapse supernovae may significantly modify the amount of electron flavor content and thus the nucleosynthesis outcomes of these events. Especially important to determining the neutrino flavor in such events is the fast-growing neutrino flavor instability driven by neutrino-neutrino interactions. However, solving the equations of neutrino quantum kinetics has remained a challenge because capturing the neutrino self-interaction potential requires resolving small length, time, and angular scales in the neutrino distribution. We present Emu, a particle-in-cell simulation code implementing the neutrino quantum kinetics equations that enables arbitrary angular resolution by representing the neutrino distribution with a set of computational particles, each with unique position, momentum, and quantum state. Emu’s C++ kernels for computing the evolution of each particle’s quantum state are all symbolically generated from the quantum kinetics equations using Sympy. Emu is based on the AMReX computational framework for its scalable, domain-distributed particle-mesh routines and is performance portable on CPUs and GPUs. Emu thus enables detailed multidimensional studies of the neutrino fast flavor instability, resolved in space, angle, and time. We will discuss Emu’s design and its scalability on modern supercomputing platforms along with our ongoing simulation explorations of the fast flavor instability and its astrophysical implications. [Preview Abstract] |
Monday, April 19, 2021 4:57PM - 5:33PM Live |
T05.00003: Incorporating Relativity into FLASH Multiphysics Simulations Invited Speaker: Michael Pajkos From creating unique compact objects to evolving the chemical composition of the universe, better understanding the explosive ending to massive stars, core collapse supernovae (CCSNe), is beneficial to the field of astrophysics as a whole. Proper numerical models of CCSNe must incorporate a wide range of length scales and a variety of physics. The resulting immense computational cost demands the use of high performance computing resources. We present new physics features that incorporate general relativistic magnetohydrodynamics (GRMHD) into the FLASH code framework to support CCSN modeling. Likewise, we show simulation results from highly energetic, dynamical systems that rely on GRMHD. [Preview Abstract] |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2024 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700