Computer Science – Performance
Scientific paper
Jan 1998
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=1998baas...30q.765r&link_type=abstract
American Astronomical Society Meeting, 191, #128.11
Computer Science
Performance
Scientific paper
We outline the current status of our project to perform direct simulations of planet formation using N ~ 10(6) planetesimals over timescales of t ~ 10(6) yr (1997, BAAS 29, 1027). These simulations will address fundamental questions regarding the origin of our Solar System: the primordial mass distribution, the extent of radial mixing, the cause of mass depletion in the asteroid belt, the origin of planetary spins and obliquities, the likelihood that the Moon was formed by a late massive impact, and the role of giant planets in terrestrial planet formation. The unprecedented numerical resolution means that we will follow the transition from runaway growth to the final accumulation of protoplanets into planets for the first time. We achieve this by modifying a spatially adaptive cosmology code (Stadel & Quinn, in preparation). Scaling is nearly linear and the code currently sustains over 1 Gflop using 32 nodes of a Cray T3E. We anticipate matching the 30 Gflop performance of the original cosmology code on 512 nodes. For testing, we use a local cluster of four DEC Alphas, with a sustained performance of about 300 Mflop (for ~ \12,000 in equipment that is a mere \40/Mflop!). The integrator is a second-order leapfrog and hence symplectic. Particle collisions are detected by linear extrapolation of each ``drift'' step; colliding particles can either agglomerate or bounce, depending on input criteria. An auxiliary analysis code monitors the global properties of the system and tracks individual particles as they evolve. Comparison with recent state-of-the-art simulations by Kokubo and Ida (1997, submitted to Icarus) shows good agreement. We are working on a multi-stepping algorithm to improve the efficiency of the code. We present the latest results of our simulations along with a plan for future work.
Lake George
Quinn Terry
Richardson Chris D.
Stadel Joachim
No associations
LandOfFree
Direct Simulation of Planet Formation With a Million Planetesimals: A Progress Report does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Direct Simulation of Planet Formation With a Million Planetesimals: A Progress Report, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Direct Simulation of Planet Formation With a Million Planetesimals: A Progress Report will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1654473