arXiv subjects

Jeroen Bedorf

Publications and source records attributed to Jeroen Bedorf.

4 recordsLinked to original sources

arXiv2021

Resolving local and global kinematic signatures of satellite mergers with billion particle simulations

In this work we present two new $\sim10^9$ particle self-consistent simulations of the merger of a Sagittarius-like dwarf galaxy with a Milky Way-like disc galaxy. One model is a violent merger creating a thick disc, and a Gaia-Enceladus/Sausage like remnant. The other is a highly stable disc which we use to illustrate how the improved phase space resolution allows us to better examine the formation and evolution of structures that have been observed in small, local volumes in the Milky Way, such as the $z-v_z$ phase spiral and clustering in the $v_{\mathrm{R}}-v_ϕ$ plane when compared to previous works. The local $z-v_z$ phase spirals are clearly linked to the global asymmetry across the disc: we find both 2-armed and 1-armed phase spirals, which are related to breathing and bending behaviors respectively. Hercules-like moving groups are common, clustered in $v_{\mathrm{R}}-v_ϕ$ in local data samples in the simulation. These groups migrate outwards from the inner galaxy, matching observed metallicity trends even in the absence of a galactic bar. We currently release the best fitting `present day' merger snapshots along with the unperturbed galaxies for comparison.

astro-ph.GA↗

arXiv2018

Scalable Distributed DNN Training using TensorFlow and CUDA-Aware MPI: Characterization, Designs, and Performance Evaluation

TensorFlow has been the most widely adopted Machine/Deep Learning framework. However, little exists in the literature that provides a thorough understanding of the capabilities which TensorFlow offers for the distributed training of large ML/DL models that need computation and communication at scale. Most commonly used distributed training approaches for TF can be categorized as follows: 1) Google Remote Procedure Call (gRPC), 2) gRPC+X: X=(InfiniBand Verbs, Message Passing Interface, and GPUDirect RDMA), and 3) No-gRPC: Baidu Allreduce with MPI, Horovod with MPI, and Horovod with NVIDIA NCCL. In this paper, we provide an in-depth performance characterization and analysis of these distributed training approaches on various GPU clusters including the Piz Daint system (6 on Top500). We perform experiments to gain novel insights along the following vectors: 1) Application-level scalability of DNN training, 2) Effect of Batch Size on scaling efficiency, 3) Impact of the MPI library used for no-gRPC approaches, and 4) Type and size of DNN architectures. Based on these experiments, we present two key insights: 1) Overall, No-gRPC designs achieve better performance compared to gRPC-based approaches for most configurations, and 2) The performance of No-gRPC is heavily influenced by the gradient aggregation using Allreduce. Finally, we propose a truly CUDA-Aware MPI Allreduce design that exploits CUDA kernels and pointer caching to perform large reductions efficiently. Our proposed designs offer 5-17X better performance than NCCL2 for small and medium messages, and reduces latency by 29% for large messages. The proposed optimizations help Horovod-MPI to achieve approximately 90% scaling efficiency for ResNet-50 training on 64 GPUs. Further, Horovod-MPI achieves 1.8X and 3.2X higher throughput than the native gRPC method for ResNet-50 and MobileNet, respectively, on the Piz Daint cluster.

cs.DC↗

arXiv2018

The origin of interstellar asteroidal objects like 1I/2017 U1 'Oumuamua

We study the origin of the interstellar object 1I/2017 U1 'Oumuamua by juxtaposing estimates based on the observations with simulations. We speculate that objects like 'Oumuamua are formed in the debris disc as left over from the star and planet formation process, and subsequently liberated. The liberation process is mediated either by interaction with other stars in the parental star-cluster, by resonant interactions within the planetesimal disc or by the relatively sudden mass loss when the host star becomes a compact object. Integrating backward in time in the Galactic potential together with stars from the Gaia-TGAS catalogue we find that about 1.3Myr ago 'Oumuamua passed the nearby star HIP 17288 within a mean distance of $1.3$pc. By comparing nearby observed L-dwarfs with simulations of the Galaxy we conclude that the kinematics of 'Oumuamua is consistent with relatively young objects of $1.1$--$1.7$Gyr. We just met 'Oumuamua by chance, and with a derived mean Galactic density of $\sim 3\times 10^{5}$ similarly sized objects within 100\,au from the Sun or $\sim 10^{14}$ per cubic parsec we expect about 2 to 12 such visitors per year within 1au from the Sun.

astro-ph.EP↗

arXiv2007

High Performance Direct Gravitational N-body Simulations on Graphics Processing Units -- II: An implementation in CUDA

We present the results of gravitational direct $N$-body simulations using the Graphics Processing Unit (GPU) on a commercial NVIDIA GeForce 8800GTX designed for gaming computers. The force evaluation of the $N$-body problem is implemented in ``Compute Unified Device Architecture'' (CUDA) using the GPU to speed-up the calculations. We tested the implementation on three different $N$-body codes: two direct $N$-body integration codes, using the 4th order predictor-corrector Hermite integrator with block time-steps, and one Barnes-Hut treecode, which uses a 2nd order leapfrog integration scheme. The integration of the equations of motions for all codes is performed on the host CPU. We find that for $N > 512$ particles the GPU outperforms the GRAPE-6Af, if some softening in the force calculation is accepted. Without softening and for very small integration time steps the GRAPE still outperforms the GPU. We conclude that modern GPUs offer an attractive alternative to GRAPE-6Af special purpose hardware. Using the same time-step criterion, the total energy of the $N$-body system was conserved better than to one in $10^6$ on the GPU, only about an order of magnitude worse than obtained with GRAPE-6Af. For $N \apgt 10^5$ the 8800GTX outperforms the host CPU by a factor of about 100 and runs at about the same speed as the GRAPE-6Af.

astro-ph↗