open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I want the CRITICAL EROSION SHEAR STRESS OF THE MUD to be variable in the zone.

So I use the FORMATTED DATA FILE 1 to input the nodes and theirs values of CRITICAL EROSION SHEAR STRESS OF THE MUD. And then modified the SUBROUTINE SUSPENSION_EROSION_COH.

But I notice that it cost more time on muti-processors(parallel mode) than on a single processor. All the tests run on the same computer.

I find out that it would spend more time on reading the FORMATTED DATA FILE 1 in the parallel computation.

In my case ,is there any solution to improve the speed in parallel ?

Thank you in advance!

Best regards

He Jiawei

(PS：希望更多在中国使用TELEMAC的朋友进QQ群219764703互相交流学习)

Hi

as already wrote in some previous post, programming in parallel is more complex and not so easy if you really want to be efficient for the computation time.

In your case, you should separate the works in two main step:

1. read the file and find the correspondence between local and global
2. use the values of the file

Of course, the first step is done only once at the first pass

This will probably help you to reduce the computation time

Regards

Hello c.coulet，

Thank you very much!

I just read the FORMATTED DATA FILE once and use the SAVE to define my variables.
Now the parallel computation is faster than single-processor computation.

Best regards

He Jiawei