Stanford High Performance Computing Conference III

Tuesday, August 21st, 2007
8:00am - 5:00pm

Wallenberg Hall

The focus this year is software development and research computing, bringing together system managers, researchers, developers, computational scientists and industry affiliates to discuss recent developments and future advancement in High Performance Computing.

Registration now open. Be sure to sign up early!
http://www.clustercorp.com/conferences/index.html
* The conference is free to academic, research and government  affiliates. Open to affiliates of conference sponsors

 
Main Theatre Session
Lab Sessions (Room 120)
Lab Sessions (Room 127)
8:00 am - 8:30 am
***Registration & Continental Breakfast***
8:30 am - 9:15 am
Keynote Speaker
Prof. Friedrich B. Prinz
ME Department Chair
   
9:20 am - 10:05 am
 
break
***Break***
10:25 am - 11:10 am
Numerical Modeling of Chemically Reacting Flows Guillaume Blanquart & Olivier Desjardins
 
11:15 am - 12:00 pm
 
Lunch
***Lunch & Visit Vendor Booths***
1:00 pm - 2:00 pm
OpenMPI
Nikhil Kelshikar, Cisco Systems
High Performance Computing with PGI Compilers and Tools
Part I
Doug Miles and Don Breazeal,
The Portland Group
2:00 pm - 3:00 pm
High Performance Computing with PGI Compilers and Tools
Part II
Doug Miles and Don Breazeal,
The Portland Group
 
Break
***Break***
3:20 pm - 4:20 pm

MPI-IO In Practice
Jeff Layton, Panasas

4:20 pm - 5:00 pm
Running Fluent in Parallel
Gorazd Medic, Stanford FPCE
 

 

*** Theatre Sessions ***

9:20 am -10:05 am

Large-scale simulation approaches for viral entry and membrane dynamics

Peter Kasson
A mechanistic understanding of biomolecular systems often requires both large tightly-coupled simulations in molecular detail and good statistical sampling.  We have recently made a number of advances in simulating large molecular systems via high-performance computing and in combining HPC with distributed-computing approaches to make optimal use of each.  We have also developed novel computing platforms that blur the boundaries between HPC and DC.  To illustrate these methods, I will discuss how we employ them to address challenges in understanding how the influenza virus infects cells and how lipid membrane properties control the lipid vesicle fusion.

10:25 am -11:10 am

Numerical Modeling of Chemically Reacting Flows

Guillaume Blanquart & Olivier Desjardins
In recent years, the numerical modeling of chemically reacting flows has become profoundly important*. It is required, for example, in the design of combustion engines such as gas-turbines or reciprocating engines, as well as the development of fuel cells, process controls for chemical reactors, and the understanding of accidental fires. The physical complexity and intricate coupling of fluid dynamics, chemical reactions, evaporation, and radiation in such simple cases are typical of the more complex processes occurring in combustion or energy conversion in fuel cells. Numerical modeling of all kinds of chemically reacting flows requires multi-disciplinary research.

11:15 am -12:00 pm

Extremely High Resolution Simulations of Early Universe Cosmological Phenomena

Matthew J. Turk
In order to study the formation of the first stars and galaxies, we must consider physical processes on a wide variety of length and temporal scales, ranging from scales larger than the galaxy down to scales within the radius of the Sun.  To do so effectively, we utilize a technique called Adaptive Mesh Refinement, wherein resolution is added in the regions of greatest interest, enabling us to conduct fully self-consistent simulations spanning a dynamic range of 10^15.  I will discuss simulations of the formation of the first stars and galaxies, as well as the physical processes included in these simulations, and the techniques used for visualization and management of data.  Additionally, I will present on efforts to manage a large-scale astrophysical simulation code with Python, utilizing the latest generation of free and open source software.

2:00 pm - 3:00 pm

Effective Use of Matlab on HPC Clusters

Michael Vitus
Interested in how to run Matlab on a cluster?  During this lab session, we will provide an introduction on how to effectively use Matlab on a cluster.  We will also provide an introduction to parallel programming with Matlab with an in-depth look at MatlabMPI.  As motivation, a distributed collision avoidance algorithm developed in Stanford's Hybrid Systems Lab will be presented.

4:20 pm - 5:00 pm

APC Power and Cooling high density solutions; Followed by walking tour of the Mechanical Engineering Cluster APC Hot Aisle Containment System

Luca Melluso, APC
Hot aisle containment for high-density configurations. In-row precision air conditioning for medium to large data centers including high density applications. Management platform that increases visibility of datacenters by outlining the health and status of all APC devices. Comprehensive power distribution systems for Network-Critical Physical Infrastructure needs.

 

 

*** Lab Session (Room 120) ***

9:20 am -10:05 am

Hands-on Cluster Building

Interested in building a Linux cluster? We'll have our technology partners provide hands-on demonstrations building multi-node  compute clusters. Equipment provided by Dell, Cisco and Panasas.  Rocks+ software and demonstrations provided by Clustercorp.

1:00 pm - 2:00 pm

High Performance Computing with PGI Compilers and Tools

Doug Miles and Don Breazeal, The Portland Group Part I
Optimizing performance of HPC applications on Intel and AMD x64 processor-based systems depends on maximizing SSE vectorization, ensuring alignment of vectors, minimizing the number of cycles processors are stalled waiting on data from main memory, and using multiple cores effectively.  The PGI F95/C/C++ compilers support a number of options and directives that allow programmers to control and guide optimizations including vectorization, parallelization, function inlining, memory prefetching, interprocedural optimization, and others.  The PGI debugger and profiler are parallel-enabled to support development and tuning of parallel and cluster applications that use both the OpenMP and MPI programming models.  In this two part tutorial, we will present detailed examples of how to extract maximum performance from x64 processors using PGI compilers, and provide an introduction and overview of the PGDBG OpenMP/MPI parallel debugger.

2:00 pm - 3:00 pm

High Performance Computing with PGI Compilers and Tools

Doug Miles and Don Breazeal, The Portland Group Part II
Optimizing performance of HPC applications on Intel and AMD x64 processor-based systems depends on maximizing SSE vectorization, ensuring alignment of vectors, minimizing the number of cycles processors are stalled waiting on data from main memory, and using multiple cores effectively.  The PGI F95/C/C++ compilers support a number of options and directives that allow programmers to control and guide optimizations including vectorization, parallelization, function inlining, memory prefetching, interprocedural optimization, and others.  The PGI debugger and profiler are parallel-enabled to support development and tuning of parallel and cluster applications that use both the OpenMP and MPI programming models.  In this two part tutorial, we will present detailed examples of how to extract maximum performance from x64 processors using PGI compilers, and provide an introduction and overview of the PGDBG OpenMP/MPI parallel debugger.

3:20 pm - 4:20 pm

Sun Studio compilers, tools and more...
Darryl Gove

Sun Studio compilers and tools overview
[What's available and what it runs on.]

CMT Developer Tools overview
[Recently released tools to help developers]

Parallelisation strategies
[Covering OpenMP, Autopar, Pthreads, MPI...]

Performance Tuning Parallel applications
[Using the performance analyzer to do optimisation.]

 

 

*** Lab Session (Room 127) ***

9:20 am -10:05 am

TBA

10:25 am -11:10 am

TBA

11:15 am -12:00 pm

TBA

1:30 pm - 3:00 pm

TBA

3:20 pm - 5:00 pm

TBA

© Stanford University. All Rights Reserved. Stanford, CA 94305. (650) 723-2300. Terms of Use | Copyright Complaints