Random error NIV2_FLOPS #15910
Replies: 4 comments 2 replies
-
I'm very happy to see you using this forum @srinath-chakravarthy. It is going to take me a little bit to get my workflow setup so that I see discussions quickly. Right now you have to watch the whole repository to get new discussions posts notifications, and if I watch the whole repo I'm going to have to setup a filter because I don't care about every new issue and PR that gets opened. I have never seen this error before. Some googling shows some hits for other FEM codes (I saw Elmer) and also a PETSc post. Most of these seem to be concerned with MUMPS, although apparently they submitted a fix for that. There was also this PETSc thread. It's unclear to me from the message whether this is a CPU or memory error. What is the "current avatar"? Perhaps @fdkong has some ideas about this |
Beta Was this translation helpful? Give feedback.
-
What is the problem size? If the problem is large, it would be an out of integer error. |
Beta Was this translation helpful? Give feedback.
-
Alex and fande,
Problem size is relatively small, while I'm not in front of my computer
now, it has about 3000 quad elements, 3600 nodes with 5 dof per node. I
checked mem usage and it is small 126 MB for each processor with replicated
mesh. I have not faced this issue before... I'm more than happy to share my
repo with you as well as the input file.
Alex, I also Googled before posting and found the same things you did.
Cheers
Srinath
P.S: I just thought of giving discussions a try since it is being touted as
an alternative to the moose users forum. I am not sure about workflow on
my side either ...
…On Wed, Oct 7, 2020, 8:18 PM Fande Kong ***@***.***> wrote:
What is the problem size? If the problem is large, it would be an out of
integer error.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#15910 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACYC6LJ6YQAPORTBP5OJGC3SJUAMLANCNFSM4SHY4LEA>
.
|
Beta Was this translation helpful? Give feedback.
-
He is using superlu_dist
…On Wed, Oct 7, 2020 at 5:32 PM Fande Kong ***@***.***> wrote:
It is interesting. Are you using a direct solver, eg, LU?
I was wondering what will happen if you try -pc_factor_mat_solver_type
mumps and -pc_factor_mat_solver_type superlu_dist, respectively? The
reason is that we want to know where the error message comes from.
If you could reproduce this using a moose example, it would be a big help.
—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
<#15910 (reply in thread)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACOGA4CB7UYTVKZ5HMWO67TSJUCBZANCNFSM4SHY4LEA>
.
|
Beta Was this translation helpful? Give feedback.
-
Hi all,
I am having some trouble with simulations randomly stopping due to NIV2_FLOPS error.
The problem is a contact problem with AD and i am using superlu_dist with Newton solver. After a random period of time, the simulation aborts with "problem with NIV2_FLOPS message -5165450.0000000000 0 -5165450.0000000000". In the current avatar i have used 16 cores and there does not seem to be any memory issues for this problem.
Any help will be appreciated.
Cheers
Srinath
Beta Was this translation helpful? Give feedback.
All reactions