Difference between revisions of "Sig HPC"

From OpenFOAMWiki
(Created page with 'During the BoF sessions of the OF Workshop in Montreal a session on High Performance Computing (HPC) was held. Results of this session were:<br> - OF is often used in a HPC envi...')
 
(Added to the category of incomplete pages)
 
(14 intermediate revisions by 2 users not shown)
Line 1: Line 1:
During the BoF sessions of the OF Workshop in Montreal a session on High Performance Computing (HPC) was held.
+
== General ==
 +
 
 +
During the [http://en.wikipedia.org/wiki/Birds_of_a_Feather_(computing) BoF] sessions of the [http://www.openfoamworkshop.org/2009/index.php5?title=Main_Page OF Workshop 2009] in Montreal a session on High Performance Computing ([http://en.wikipedia.org/wiki/High-performance_computing HPC]) was held.
  
 
Results of this session were:<br>
 
Results of this session were:<br>
- OF is often used in a HPC environment<br>
+
* OF is often used in a HPC environment<br>
- A collection of benchmarking results of OF on various hardware should be started to form a basis for<br>
+
* A collection of benchmarking results of OF on various hardware should be started to form a basis for<br>
-- performance optimisation in the OF source code<br>
+
** performance optimisation in the OF source code<br>
-- hardware recommondations<br>
+
** Investigation of influence of compiler optimisation switches / different compilers
 +
** hardware recommendations <br>
 +
** best practice guides
 +
 
 +
=== Benchmark ===
 +
Criteria for a universal benchmarking case need to be defined and the benchmarking case made available. Furthermore a questionnaire describing the performance crucial hardware components should be created to form a basis for analysis of benchmark results.
 +
==== Benchmark case ====
 +
TODO: Find a more or less universal benchmarking case. Preferably the case does not need a complex mesh / solver.
 +
The icoFoam cavity tutorial could be a first starting point.
 +
 
 +
==== Hardware description ====
 +
TODO: Define a list of key hardware components that sufficiently precise describe the benchmarked system.
 +
* CPU
 +
* Memory
 +
* Inter-connect
 +
 
 +
==== Compiler description ====
 +
* Type of compiler e.g. [http://gcc.gnu.org/ GCC], [http://developer.amd.com/cpu/open64/Pages/default.aspx AMD], [http://de.wikipedia.org/wiki/Intel_C%2B%2B_Compiler Intel CC], [http://www.pgroup.com/ PGI]
 +
* Compiler flags
 +
 
 +
=== Links ===
 +
[http://www.eneagrid.enea.it/papers_presentations/papers/NapoliEScience2008_09_CRESCO.pdf Cross checking OpenFOAM and Fluent results of CFD simulations in ENEA-GRID environment]<br>
 +
[http://techreport.com/articles.x/17005/8 AMD's 'Istanbul' six-core Opteron processors - STARS Euler3d computational fluid dynamics] <br>
 +
[http://support.amd.com/us/Processor_TechDocs/32035.pdf Compiler Usage Guidelines for 64-Bit Operating Systems on AMD64]
 +
 
 +
 
 +
-- [[User:DSpreitz|DSpreitz]] 13:20, 13 June 2009 (UTC)
  
As a short term goal, criteria for a universal benchmarking case need to be defined and the benchmarking case made available. Furthermore a questionnaire describing the performance crucial hardware components should be created to form a basis for analysis of benchmark results.
+
[[Category:Special Interest Groups]]
 +
[[Category:Incomplete pages]]

Latest revision as of 11:21, 2 November 2013

1 General

During the BoF sessions of the OF Workshop 2009 in Montreal a session on High Performance Computing (HPC) was held.

Results of this session were:

  • OF is often used in a HPC environment
  • A collection of benchmarking results of OF on various hardware should be started to form a basis for
    • performance optimisation in the OF source code
    • Investigation of influence of compiler optimisation switches / different compilers
    • hardware recommendations
    • best practice guides

1.1 Benchmark

Criteria for a universal benchmarking case need to be defined and the benchmarking case made available. Furthermore a questionnaire describing the performance crucial hardware components should be created to form a basis for analysis of benchmark results.

1.1.1 Benchmark case

TODO: Find a more or less universal benchmarking case. Preferably the case does not need a complex mesh / solver. The icoFoam cavity tutorial could be a first starting point.

1.1.2 Hardware description

TODO: Define a list of key hardware components that sufficiently precise describe the benchmarked system.

  • CPU
  • Memory
  • Inter-connect

1.1.3 Compiler description

1.2 Links

Cross checking OpenFOAM and Fluent results of CFD simulations in ENEA-GRID environment
AMD's 'Istanbul' six-core Opteron processors - STARS Euler3d computational fluid dynamics
Compiler Usage Guidelines for 64-Bit Operating Systems on AMD64


-- DSpreitz 13:20, 13 June 2009 (UTC)