OpenSeesSP and OpenSeesMP questions.

[mtsonicc]

Hi,

I'm studying OpenSees parallel computing with the document 'TN-2007-16' ( http://opensees.berkeley.edu/OpenSees/p ... essing.pdf ).

and I would like to ask a question about OpenSeesSP and OpenSeesMP.

For OpenSeesSP examples(Section 4.2 of TN-2007-16), 'model.tcl' is divided into many pieces(using substucture analysis or other methods) depending on the number of CPU cores and then subjected to static or dynamic loads.
(In other words, CPU01 is analyzing top-half of 'model.tcl' subjected to given loads, CPU02 is analyzing bottom-half of 'model.tcl' subjected to given loads, and so on. and then the result of CPU01, CPU02 is combined at the end of analysis.)

And for OpenSeesMP examples(Section 5.3.i and 5.3.ii of TN-2007-16), same 'model.tcl' is subjected to 'n' numbers of ground motions and analyzed in each separate CPU cores.
(In other words, CPU01 is analyzing 'model.tcl' subjected to GM1, CPU02 is analyzing 'model.tcl' subjected to GM2... and so on.)

Am I understanding it right?

[fmk]

1. OpenSeesSP are not independent analysis; the results cannot be simply added, instead while the element state determination is being done on the seperate processors, there is one big global matrix equation being set up for and solved at each step by all proccessors in commmincation with each other.

OpenSeesMP: yes you have it

no the element state determination is being done for some on different processors.

i suggest looking at the material on following page:
http://opensees.berkeley.edu/wiki/index ... _Computing

[mtsonicc]

Dear fmk,

Thanks for your kind reply.

I will study the material you suggested!

Thank you again.