Researcher profile

Xuebin Chi

Xuebin Chi contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate

astro-ph.CO Distributed, Parallel, and Cluster Computing Graphics Mathematical Software Numerical Analysis

Trust snapshot

Quick read

Trust 15 - Baseline

3works

0followers

5topics

4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

preprint2016arXiv

An efficient hybrid tridiagonal divide-and-conquer algorithm on distributed memory architectures

In this paper, an efficient divide-and-conquer (DC) algorithm is proposed for the symmetric tridiagonal matrices based on ScaLAPACK and the hierarchically semiseparable (HSS) matrices. HSS is an important type of rank-structured matrices.Most time of the DC algorithm is cost by computing the eigenvectors via the matrix-matrix multiplications (MMM). In our parallel hybrid DC (PHDC) algorithm, MMM is accelerated by using the HSS matrix techniques when the intermediate matrix is large. All the HSS algorithms are done via the package STRUMPACK. PHDC has been tested by using many different matrices. Compared with the DC implementation in MKL, PHDC can be faster for some matrices with few deflations when using hundreds of processes. However, the gains decrease as the number of processes increases. The comparisons of PHDC with ELPA (the Eigenvalue soLvers for Petascale Applications library) are similar. PHDC is usually slower than MKL and ELPA when using 300 or more processes on Tianhe-2 supercomputer.

preprint2014arXiv

Interactive Visual Exploration of Halos in Large Scale Cosmology Simulation

Halo is one of the most important basic elements in cosmology simulation, which merges from small clumps to ever larger objects. The processes of the birth and merging of the halos play a fundamental role in studying the evolution of large scale cosmological structures. In this paper, a visual analysis system is developed to interactively identify and explore the evolution histories of thousands of halos. In this system, an intelligent structure-aware selection method in What You See Is What You Get manner is designed to efficiently define the interesting region in 3D space with 2D hand-drawn lasso input. Then the exact information of halos within this 3D region is identified by data mining in the merger tree files. To avoid visual clutter, all the halos are projected in 2D space with a MDS method. Through the linked view of 3D View and 2D graph, Users can interactively explore these halos, including the tracing path and evolution history tree.

preprint2012arXiv

Bulk flow of halos in ΛCDM simulation

Analysis of the Pangu N-body simulation validates that the bulk flow of halos follows a Maxwellian distribution which variance is consistent with the prediction of the linear theory of structure formation. We propose that the consistency between the observed bulk velocity and theories should be examined at the effective scale of the radius of a spherical top-hat window function yielding the same smoothed velocity variance in linear theory as the sample window function does. We compared some recently estimated bulk flows from observational samples with the prediction of the ΛCDM model we used; some results deviate from expectation at a level of ~ 3σbut the discrepancy is not as severe as previously claimed. We show that bulk flow is only weakly correlated with the dipole of the internal mass distribution, the alignment angle between the mass dipole and the bulk flow has a broad distribution peaked at ~ 30-50 deg., and also that the bulk flow shows little dependence on the mass of the halos used in the estimation. In a simulation of box size 1Gpc/h, for a cell of radius 100 Mpc/h the maximal bulk velocity is >500 km/s, dipoles of the environmental mass outside the cell are not tightly aligned with the bulk flow, but are rather located randomly around it with separation angles ~ 20-40 deg. In the fastest cell there is a slightly smaller number of low-mass halos; however halos inside are clustered more strongly at scales > ~ 20 Mpc/h, which might be a significant feature since the correlation between bulk flow and halo clustering actually increases in significance beyond such scales.