OpenFOAM compiled with Intel C++ Compiler?

af631717 · June 1, 2009, 12:50

Hi,
has anybody compiled OpenFOAM with the Intel Compiler and has compared the difference of the performance for a specific solver?
Especially with the new Intel Nehalem architecture should be a significant speedup.

Regards
Achim

bjj · June 5, 2009, 05:24

Achim,
As part of a benchmark test series for parallel computing we made a dambreak test case with icoFoam on the Intel Nehalem processor (on up to 32 cores). We tried to compile with both Gcc and Icc. In general we reduced the computational time with about 15% when using Icc.

Regards,
Bjarne

af631717 · June 5, 2009, 06:19

Bjarne,

thanks for this information. It is about what I expected and mean, that it is sometimes better to buy a good compiler for 500 Euro than buy a new hardware.

How long did the calculation take for n iterations? Is the orig. tutorial case not to small for 32 cores?

Is it an advantage to use HyperThreading with the Nehalem? And use 64 "theoretical" cores?

Regards
Achim

af631717 · June 5, 2009, 06:32

Was it difficult to compile OF with the intel compiler?
Achim

bjj · June 5, 2009, 06:41

With the Nehalem tests we had a run time on 1 core at 3350s while the run time on 32 cores were 170s. So a speed-up at about 19-20 on 32 cores.

The original tutorial case is to small for a test on 32 cores. We used a maximum at 4M computational point.

We did not use HyperThreading in this test so I can't say if this is an advantage here.

Regarding the compilation the switch from Gcc to Icc is fairly straight forward. Edit the compiler settings in your etc/bashrc-file.

Regards,
Bjarne

June 1, 2009, 12:50	OpenFOAM compiled with Intel C++ Compiler?	#1
af631717 New Member Achim Join Date: May 2009 Posts: 22 Rep Power: 16	Hi, has anybody compiled OpenFOAM with the Intel Compiler and has compared the difference of the performance for a specific solver? Especially with the new Intel Nehalem architecture should be a significant speedup. Regards Achim

June 5, 2009, 06:19	OpenFOAM compiled with Intel C++ Compiler?	#3
af631717 New Member Achim Join Date: May 2009 Posts: 22 Rep Power: 16	Bjarne, thanks for this information. It is about what I expected and mean, that it is sometimes better to buy a good compiler for 500 Euro than buy a new hardware. How long did the calculation take for n iterations? Is the orig. tutorial case not to small for 32 cores? Is it an advantage to use HyperThreading with the Nehalem? And use 64 "theoretical" cores? Regards Achim

June 5, 2009, 06:32	OpenFOAM compiled with Intel C++ Compiler?	#4
af631717 New Member Achim Join Date: May 2009 Posts: 22 Rep Power: 16	Was it difficult to compile OF with the intel compiler? Achim

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search
Display Modes
Linear Mode Switch to Hybrid Mode Switch to Threaded Mode

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
CFX11 + Fortran compiler ?	Mohan	CFX	20	March 30, 2011 18:56
Intel Fortran Compiler & Tutorial 17	Young	CFX	1	March 30, 2011 01:39
64bitrhel5 OF installation instructions	mirko	OpenFOAM Installation	2	August 12, 2008 18:07
Compiler in OpenFoam	gulp	OpenFOAM Installation	4	October 27, 2007 11:36
Intel compiler for linux x86	marcus	Siemens	1	November 9, 2006 16:43

June 5, 2009, 05:24		#2
bjj New Member Bjarne Jensen Join Date: Mar 2009 Location: Denmark Posts: 7 Rep Power: 17	Achim, As part of a benchmark test series for parallel computing we made a dambreak test case with icoFoam on the Intel Nehalem processor (on up to 32 cores). We tried to compile with both Gcc and Icc. In general we reduced the computational time with about 15% when using Icc. Regards, Bjarne

June 5, 2009, 06:41		#5
bjj New Member Bjarne Jensen Join Date: Mar 2009 Location: Denmark Posts: 7 Rep Power: 17	With the Nehalem tests we had a run time on 1 core at 3350s while the run time on 32 cores were 170s. So a speed-up at about 19-20 on 32 cores. The original tutorial case is to small for a test on 32 cores. We used a maximum at 4M computational point. We did not use HyperThreading in this test so I can't say if this is an advantage here. Regarding the compilation the switch from Gcc to Icc is fairly straight forward. Edit the compiler settings in your etc/bashrc-file. Regards, Bjarne