Skip to content

GitLab

  • Projects
  • Groups
  • Snippets
  • Help
    • Loading...
  • Help
    • Help
    • Support
    • Community forum
    • Submit feedback
  • Sign in
S src
  • Project overview
    • Project overview
    • Details
    • Activity
    • Releases
  • Repository
    • Repository
    • Files
    • Commits
    • Branches
    • Tags
    • Contributors
    • Graph
    • Compare
    • Locked Files
  • Issues 59
    • Issues 59
    • List
    • Boards
    • Labels
    • Service Desk
    • Milestones
    • Iterations
  • Merge requests 3
    • Merge requests 3
  • CI/CD
    • CI/CD
    • Pipelines
    • Jobs
    • Schedules
  • Operations
    • Operations
    • Incidents
    • Environments
  • Analytics
    • Analytics
    • CI/CD
    • Code Review
    • Issue
    • Repository
    • Value Stream
  • Wiki
    • Wiki
  • Snippets
    • Snippets
  • Members
    • Members
  • Activity
  • Graph
  • Create a new issue
  • Jobs
  • Commits
  • Issue Boards
Collapse sidebar
  • OPAL
  • src
  • Issues
  • #625

Closed
Open
Created Dec 01, 2020 by ext-calvo_p@ext-calvo_pDeveloper0 of 1 task completed0/1 task

Fix exceptions in parallel

The following discussion from !458 (merged) should be addressed:

  • @snuverink_j started a discussion: (+5 comments)

    If I understand correctly the file is only read by node 0, so this check can be done by node 0 only (as it was before).

    But to be honest, I don't understand why it was not working before in the parallel environment. Can you elaborate a bit?

@ext-calvo_p

I thought in the same way, but when OPAL is run in the parallel environment and the distribution file doesn't exist, OpalException is not thrown.

I think that checking if the file exists before opening it does not have to be done exclusively by node 0.

@snuverink_j

I can reproduce it: the simulation hangs.

But this seems something more fundamental to me, because I would also have expected that a throw by a single node would be enough to stop the simulation. @gsell: Should that not be the case?

@kraus

In Main.cpp we catch the exception and then call MPI_Abort on MPI_COMM_WORLD. I thought that this should also stop the other nodes but this isn't the case. So we try to throw the exception on all nodes.

Edited Dec 01, 2020 by snuverink_j
Assignee
Assign to
None
Milestone
None
Assign milestone
Time tracking