00.00.00 GENERAL
Refine
Year of publication
Document Type
- ZIB-Report (17)
- Bachelor's Thesis (1)
- Master's Thesis (1)
Keywords
大規模二次割当問題への挑戦
(2022)
Many scientific applications deal with data from a multitude of different sources, e.g., measurements, imaging and simulations. Each source provides an additional perspective on the phenomenon of interest, but also comes with specific limitations, e.g. regarding accuracy, spatial and temporal availability. Effectively combining and analyzing such multimodal and partially incomplete data of limited accuracy in an integrated way is challenging. In this work, we outline an approach for an integrated analysis and visualization of the atmospheric impact of volcano eruptions. The data sets comprise observation and imaging data from satellites as well as results from numerical particle simulations. To analyze the clouds from the volcano eruption in the spatiotemporal domain we apply topological methods. Extremal structures reveal structures in the data that support clustering and comparison. We further discuss the robustness of those methods with respect to different properties of the data and different parameter setups. Finally we outline open challenges for the effective integrated visualization using topological methods.
Lagrangian Coherent Structures (LCS) have become a widespread and powerful method to describe dynamic motion patterns in time-dependent flow fields. The standard way to extract LCS is to compute height ridges in the Finite Time Lyapunov Exponent (FTLE) field. In this work, we present an alternative method to approximate Lagrangian features for 2D unsteady flow fields that achieves subgrid accuracy without additional particle sampling. We obtain this by a geometric reconstruction of the flow map using additional material constraints for the available samples. In comparison to the standard method, this allows for a more accurate global approximation of LCS on sparse grids and for long integration intervals. The proposed algorithm works directly on a set of given particle trajectories and without additional flow map derivatives. We demonstrate its application for a set of computational fluid dynamic examples, as well as trajectories acquired by Lagrangian methods, and discuss its
benefits and limitations.
Mixed integer linear programming (MIP) is a general form to model combinatorial optimization problems and has many industrial applications. The performance of MIP solvers has improved tremendously in the last two decades and these solvers have been used to solve many real-word problems. However, against the backdrop of modern computer technology, parallelization is of pivotal importance. In this way, ParaSCIP is the most successful parallel MIP solver in terms of solving previously unsolvable instances from the well-known benchmark instance set MIPLIB by using supercomputers. It solved two instances from MIPLIB2003 and 12 from MIPLIB2010 for the first time to optimality by using up to 80,000 cores on supercomputers. ParaSCIP has been developed by using the Ubiquity Generator (UG) framework, which is a general software package to parallelize any state-of-the-art branch-and-bound based solver. This paper discusses 7 years of progress in parallelizing branch-and-bound solvers with UG.
With the growing number of hardware components and the increasing software complexity in the upcoming exascale computers, system failures will become the norm rather than an exception for long-running applications. Fault-tolerance can be achieved by the creation of checkpoints during the execution of a parallel program. Checkpoint/Restart (C/R) mechanisms allow for both task migration (even if there were no hardware faults) and restarting of tasks after the occurrence of hardware faults. Affected tasks are then migrated to other nodes which may result in unfortunate process placement and/or oversubscription of compute resources. In this paper we analyze the impact of unfortunate process placement and oversubscription of compute resources on the performance and scalability of two typical HPC application workloads, CP2K and MOM5. Results are given for a Cray XC30/40 with Aries dragonfly topology. Our results indicate that unfortunate process placement has only little negative impact while oversubscription substantially degrades the performance. The latter might be only (partially) beneficial when placing multiple applications with different computational characteristics on the same node.
To achieve high floating point compute performance, modern processors draw on short vector SIMD units, as found e.g. in Intel CPUs (SSE, AVX1, AVX2 as well as AVX-512 on the roadmap) and the Intel Xeon Phi coprocessor, to operate an increasingly larger number of operands simultaneously. Making use of SIMD vector operations therefore is essential to get close to the processor’s floating point peak performance.
Two approaches are typically used by programmers to utilize the vector units: compiler driven vectorization via directives and code annotations, and manual vectorization by means of SIMD intrinsic operations or assembly.
In this paper, we investigate the capabilities of the current Intel compiler (version 15 and later) to generate vector code for non-trivial coding patterns within loops. Beside the more or less uniform data-parallel standard loops or loop nests, which are typical candidates for SIMDfication, the occurrence of e.g. (conditional) function calls including branching, and early returns from functions may pose difficulties regarding the effective use of vector operations. Recent improvements of the compiler's capabilities involve the generation of SIMD-enabled functions. We will study the effectiveness of the vector code generated by the compiler by comparing it against hand-coded intrinsics versions of different kinds of functions that are invoked within innermost loops.
PIPS-SBB is a distributed-memory parallel solver with a scalable data distribution paradigm. It is designed to solve MIPs with a dual-block angular structure, which is characteristic of deterministic-equivalent Stochastic Mixed-Integer Programs (SMIPs). In this paper, we present two different parallelizations of Branch & Bound (B&B), implementing both as extensions of PIPS-SBB, thus adding an additional layer of parallelism. In the first of the proposed frameworks, PIPS-PSBB, the coordination and load-balancing of the different optimization workers is done in a decentralized fashion. This new framework is designed to ensure all available cores are processing the most promising parts of the B&B tree. The second, ug[PIPS-SBB,MPI], is a parallel implementation using the Ubiquity Generator (UG), a universal framework for parallelizing B&B tree search that has been successfully applied to other MIP solvers. We show the effects of leveraging multiple levels of parallelism in potentially improving scaling performance beyond thousands of cores.
Traditionally, Lagrangian fields such as finite-time Lyapunov exponents (FTLE)
are precomputed on a discrete grid and are ray casted afterwards. This, however,
introduces both grid discretization errors and sampling errors during ray marching.
In this work, we apply a progressive, view-dependent Monte Carlo-based approach
for the visualization of such Lagrangian fields in time-dependent flows. Our ap-
proach avoids grid discretization and ray marching errors completely, is consistent,
and has a low memory consumption. The system provides noisy previews that con-
verge over time to an accurate high-quality visualization. Compared to traditional
approaches, the proposed system avoids explicitly predefined fieldline seeding
structures, and uses a Monte Carlo sampling strategy named Woodcock tracking
to distribute samples along the view ray. An acceleration of this sampling strategy
requires local upper bounds for the FTLE values, which we progressively acquire
during the rendering. Our approach is tailored for high-quality visualizations of
complex FTLE fields and is guaranteed to faithfully represent detailed ridge surface
structures as indicators for Lagrangian coherent structures (LCS). We demonstrate
the effectiveness of our approach by using a set of analytic test cases and real-world numerical simulations.
For cryptanalysis in lattice-based schemes, the performance evaluation of lattice basis reduction using high-performance computers is becoming increasingly important for the determination of the security level. We propose a distributed and asynchronous parallel reduction algorithm based on randomization and DeepBKZ, which is an improved variant of the block Korkine-Zolotarev (BKZ) reduction algorithm. Randomized copies of a lattice basis are distributed to up to 103,680 cores and independently reduced in parallel, while some basis vectors are shared asynchronously among all processes via MPI. There is a trade-off between randomization and information sharing; if a substantial amount of information is shared, all processes will work on the same problem, thereby diminishing the benefit of parallelization. To monitor this balance between randomness and sharing, we propose a metric to quantify the variety of lattice bases. We empirically find an optimal parameter of sharing for high-dimensional lattices. We demonstrate the efficacy of our proposed parallel algorithm and implementation with respect to both performance and scalability through our experiments.