multiple processor

[opethian]

Hi i would like to ask if it is possible to use multiple processors for modal analysis in nastran ? and if yes how ? i have a PC with 2 double core processors. Can i use mutiple processor with sol 103 ?

 

[ThomasH]

Hi

What kind of Nastran do you use (NEi/MSC/NX) and is it 32 or 64 bit versions?

There is also a question od licensing, do you have a licens for multiprocessor use. Some versions might require that.

Other that that. As far as I know it's a setting in the bulkdata file or inte the initiation file.

Regards

Thomas

 

[opethian]

Hi Thomas,
I am using MSC Nastran 2005.1.2 version it is 32 bit i am running nastran in windows XP. I think i have licence for multiple processor use could you please let me know how to make those settings for bulk data file or the command line for initation file.
Regards

 

[ThomasH]

Hi

Well, I have never run MSC.Nastran in a parallell environment and I doubt it will do much for the performance in 32 bit OS.

Nevertheless, there is a command PARALLELL (Table 1-1 in my old QRG) that might help you. There is also a statement in Femap in the Executive control section regarding "Number of processors". It might work but I can't test it since I don't use MSC.

Good Luck anyway

Thomas

 

[opethian]

Hi Thomas i couldnt find table 1-1 what do you mean by QRG ? i am new to this subject so may be i dont know it yet ?
Thanks a lot for your help

 

[ThomasH]

Hi

QRG is the Quick Reference Guide. It should be in your documentation

I hope this helps

Thomas

 

[silvershadow]

use this

nastran filename.bdf parallel=n

where n is the number of processors you want to address to the solution's process.

N.B.
You can have speedups only in normal modes and transient solutions. Linear and non linear solutions do not benefit too much of the secon processor, at least with MSC (i'm using 2005.5)

 

[opethian]

Thanks silver.
I got i queston this code you gave has to be written in dat file isn´t it and why do we have to add bdf as an extension ?
thanks for your replies

 

[wvillers]

bdf stands for bulk data file, it's just an extension... As far as nastran is concerned, the extension doesn't matter as long as the input file is in the proper nastran bulk data format...

Basically, the syntax to manually submitt a job with nastran is: nastran <filename> <option1> <option2>, ... (e.g. nastran myfile.dat memory=1.5gb parallel=2).

If you want to "hardcode" the SMP solve (i.e. number of processors used for solving), you can always add a SYSTEM call to your bulk data deck... Essentially, you can place a NASTRAN SYSTEM(107)=2 at the begining of your bulk data deck to perform an SMP solve with 2 CPU. SYSTEM(107) gives you actually more flexibility as you can specify which matrix operations will be done using parallel routines by choosing a number greater than 1023 (take a look at the QRG for more info). If SYSTEM(107) is set to the number of CPUs, then everything will be solved used the parallel routines...

 

[R1100S]

I usually get the best performance from multiple processors by running multiple load or geometry cases simultaneously. Each instance of Nastran runs on one processor. The other advantage with multiple processors is that you can run one load case and still have processing power for drafting, preprocessing, post-processing or playing solitaire.