A5017241944- Parallel Computing and Optimization Techniques
- Distributed and Parallel Computing Systems
- Advanced Data Storage Technologies
- Interconnection Networks and Systems
- Embedded Systems Design Techniques
- Distributed systems and fault tolerance
- Matrix Theory and Algorithms
- Advanced Numerical Methods in Computational Mathematics
- Cloud Computing and Resource Management
- Scientific Computing and Data Management
- Numerical methods for differential equations
- Electromagnetic Scattering and Analysis
- Computational Fluid Dynamics and Aerodynamics
- Formal Methods in Verification
- Numerical Methods and Algorithms
- Computational Physics and Python Applications
- Real-Time Systems Scheduling
- Electromagnetic Simulation and Numerical Methods
- Stochastic Gradient Optimization Techniques
- Network Traffic and Congestion Control
- Meteorological Phenomena and Simulations
- Algorithms and Data Compression
- Numerical methods in engineering
- Model Reduction and Neural Networks
- Software System Performance and Reliability
University of Illinois Urbana-Champaign
2015-2024
Argonne National Laboratory
2009-2024
Yale University
1987-2023
IEEE Computer Society
2018-2023
Stanford University
2023
Machine Science
2023
UCLouvain
2023
UC Irvine Health
2023
Hôpital Saint-Julien
2022
Institute of Electrical and Electronics Engineers
2019-2022
We describe our work on improving the performance of collective communication operations in MPICH for clusters connected by switched networks. For each operation, we use multiple algorithms depending message size, with goal minimizing latency short messages and bandwidth long messages. Although have implemented new all MPI (Message Passing Interface) operations, because limited space only allgather, broadcast, all-to-all, reduce-scatter, reduce, allreduce. Performance results a...
Over the last 20 years, open-source community has provided more and software on which world’s high-performance computing systems depend for performance productivity. The invested millions of dollars years effort to build key components. However, although investments in these separate elements have been tremendously valuable, a great deal productivity also lost because lack planning, coordination, integration technologies necessary make them work together smoothly efficiently, both within...
The I/O access patterns of parallel programs often consist accesses to a large number small, noncontiguous pieces data. If an application's needs are met by making many distinct requests, however, the performance degrades drastically. To avoid this problem, MPI-IO allows users data set with single function call. This feature provides implementations opportunity optimize access. We describe how our implementation, ROMIO, delivers high in presence requests. explain detail two key optimizations...
I. Parallelism 1. Introduction 2. Parallel Computer Architectures 3. Programming Considerations II. Applications 4. General Application Issues 5. Computing in CFD 6. Environment and Energy 7. Computational Chemistry 8. Overviews III. Software technologies 9. Technologies 10. Message Passing Threads 11. I/O 12. Languages Compilers 13. Object-Oriented Libraries 14. Problem-Solving Environments 15. Tools for Performance Tuning Debugging 16. The 2-D Poisson Problem IV. Enabling Algorithms 17....
Dataset storage, exchange, and access play a critical role in scientific applications. For such purposes netCDF serves as portable, efficient file format programming interface, which is popular numerous application domains. However, the original interface does not provide an mechanism for parallel data storage access. In this work, we present new writing reading datasets. This derived with minimal changes from serial but defines semantics tailored high performance. The underlying I/O...
Over the past few years resilience has became a major issue for high-performance computing (HPC) systems, in particular perspective of large petascale systems and future exascale systems. These will typically gather from half million to several millions central processing unit (CPU) cores running up billion threads. From current knowledge observations existing it is anticipated that experience various kind faults many times per day. It also approach resilience, which relies on automatic or...
We consider multiphysics applications from algorithmic and architectural perspectives, where “algorithmic” includes both mathematical analysis computational complexity, “architectural” software hardware environments. Many diverse can be reduced, en route to their simulation, a common algebraic coupling paradigm. Mathematical of in this form is not always practical for realistic applications, but model problems representative discussed herein provide insight. A variety frameworks have been...
This paper presents an analytical model to predict the performance of
Resilience is a major roadblock for HPC executions on future exascale systems. These systems will typically gather millions of CPU cores running up to billion threads. Projections from current large and technology evolution predict errors happen in many times per day. propagate generate various kinds malfunctions, simple process crashes result corruptions.The past five years have seen extraordinary technical progress domains related resilience. Several options, initially considered...
Article On implementing MPI-IO portably and with high performance Share on Authors: Rajeev Thakur Mathematics Computer Science Division, Argonne National Laboratory, Argonne, IL ILView Profile , William Gropp Ewing Lusk Authors Info & Claims IOPADS '99: Proceedings of the sixth workshop I/O in parallel distributed systemsMay 1999 Pages 23–32https://doi.org/10.1145/301816.301826Published:01 May 256citation678DownloadsMetricsTotal Citations256Total Downloads678Last 12 Months17Last 6 weeks6 Get...