Welcome, Guest
Username: Password: Remember me

TOPIC: telemac2d v7p2 : malpasset large test case on PC Ubuntu 16.04

telemac2d v7p2 : malpasset large test case on PC Ubuntu 16.04 7 years 8 months ago #25395

  • stef
  • stef's Avatar
Hello, i made a step by step guide to install telemac in ubuntu 16.04, here.

Here are published my results with telemac2d and malpasset large test case on PC Ubuntu 16.04 with i7-2600k (specs), 8Go RAM DDR3 1333, with or without virtual box solution.

With the virtual box solution I limit to 7 processors to leave 1 processors for the host system

I compile telemac with :
- gfortran and mpich
- intel and mpich
- full intel (intel mpi)

For details see my post under linux installation category.

Not pretending my results are "real" benchmarks, I post them here for the discussion. I can re-do some scenarios if questions.


diag1.png


Capturedu2017-02-2510-14-10.png


diag2.png


Capturedu2017-02-2510-14-59.png
The administrator has disabled public write access.

telemac2d v7p2 : malpasset large test case on PC Ubuntu 16.04 7 years 8 months ago #25398

  • huyquangtran
  • huyquangtran's Avatar
  • OFFLINE
  • Expert Boarder
  • Posts: 271
  • Thank you received: 23
Hi Stef,

Thanks for testing!

Do you mean that you did some tests on a desktop with different scenarios (1 core, 4 cores, 7 cores) or on a powerful computer (a node) with different processors (1, 4, 7, 8 processors)?

Regards
Huy
The administrator has disabled public write access.

telemac2d v7p2 : malpasset large test case on PC Ubuntu 16.04 7 years 8 months ago #25399

  • stef
  • stef's Avatar
Hi, I did it on the same desktop personal computer with a single i7 processor, that has 4 cores and 8 threads. I test differents scenarii with --ncsize=1,4,7,8.
The administrator has disabled public write access.

telemac2d v7p2 : malpasset large test case on PC Ubuntu 16.04 7 years 8 months ago #25622

  • jmhervouet
  • jmhervouet's Avatar
Hello,

I did similar tests on the same processor and got similar results. It means that only the number of cores is relevant. When it comes to threads, there is seemingly some problem, probably with the access to memory, and we do not have more efficiency, while we know that with real massively parallel machines the efficiency can be close to 1 even up to 4000 processors.

With best regards,

Jean-Michel Hervouet
The administrator has disabled public write access.

The open TELEMAC-MASCARET template for Joomla!2.5, the HTML 4 version.