Point data in parallel runs [SOLVED]

General discussion about Elmer
Post Reply
fjimenez
Posts: 63
Joined: 27 Sep 2021, 23:40
Antispam: Yes

Point data in parallel runs [SOLVED]

Post by fjimenez »

Hi all,

There seems to be a problem during parallel execution when trying to save point data using savescalars. The solver crashes when searching for an element close to the point within partitions that do not contain the specified point. I have added a very simply case for testing. If I do not use parallel reduce, the file op.dat.0 (the specified point is in partition 0) is created but it crashes afterwards. Is this a bug or am I missing something?

Cheers.
case.tgz
(69.4 KiB) Downloaded 19 times
Last edited by fjimenez on 20 Feb 2024, 20:04, edited 1 time in total.
kevinarden
Posts: 2290
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Point data in parallel runs

Post by kevinarden »

It isn't the parallel crashing it it is

Exact Coordinates = Logical True

it runs in parallel with

Exact Coordinates = Logical False

with this option you get an op.dat.0 and an op.dat.1, but they have different values
without parallel it runs with Exact = True
you get an op.dat file, the values in op.dat do not match either of the previous two files.
kevinarden
Posts: 2290
Joined: 25 Jan 2019, 01:28
Antispam: Yes

Re: Point data in parallel runs

Post by kevinarden »

There is not a node at 0,0,0 the closet node is 449 so Save Points(1) = 449 works in parallel
however the answer is again different from all of the other options
It appears the code is crashing when trying to perform the linear combinations of the node values of the element that the point belongs to
It may be because the partition runs through node 449, i.e. elements using node 449 are in both partitions.

Save Points(n) Integer
Save the specified degrees of freedom in the n nodes specified.
Save Coordinates(n,DIM) Real
Save the degrees of freedom in the nodes nearest to the given n coordinates.
Exact Coordinates Logical
When this keyword is true the coordinates will be looked in an exact manner. Then the degrees
of freedom are linear combinations of the node values of the element that the point belongs to.
raback
Site Admin
Posts: 4821
Joined: 22 Aug 2009, 11:57
Antispam: Yes
Location: Espoo, Finland
Contact:

Re: Point data in parallel runs

Post by raback »

Hi

I tried to fix this. At one point the saving of point data was made more generic (e.g. it can handle Hcurl elements at the node) but it may have broken this parallel operation which was not guarded by tests.

Thanx for reporting.

It is good idea to use the "exact" mode that find the element closest. Based on elements all types of data may be studied, when using just nodes interpolations are not available.

-Peter
fjimenez
Posts: 63
Joined: 27 Sep 2021, 23:40
Antispam: Yes

Re: Point data in parallel runs [SOLVED]

Post by fjimenez »

Hi,

Thank you Kevin for properly diagnosing the problem and thank you Peter for fixing it. I usually extract point data using paraview but I wanted
to compared it with Elmer since you can directly interpolate from the edge elements.

Cheers,
Post Reply