Skip to content

GitLab

  • Projects
  • Groups
  • Snippets
  • Help
    • Loading...
  • Help
    • Help
    • Support
    • Community forum
    • Submit feedback
    • Contribute to GitLab
  • Sign in / Register
S
simgrid
  • Project overview
    • Project overview
    • Details
    • Activity
    • Releases
  • Repository
    • Repository
    • Files
    • Commits
    • Branches
    • Tags
    • Contributors
    • Graph
    • Compare
  • Issues 14
    • Issues 14
    • List
    • Boards
    • Labels
    • Service Desk
    • Milestones
  • Merge Requests 0
    • Merge Requests 0
  • CI / CD
    • CI / CD
    • Pipelines
    • Jobs
    • Schedules
  • Operations
    • Operations
    • Incidents
    • Environments
  • Analytics
    • Analytics
    • CI / CD
    • Repository
    • Value Stream
  • Members
    • Members
  • Activity
  • Graph
  • Create a new issue
  • Jobs
  • Commits
  • Issue Boards
Collapse sidebar
  • simgrid
  • simgrid
  • Issues
  • #44

Closed
Open
Opened Nov 29, 2019 by Adrien Faure@adfaureContributor

Uncategorized tracing broken with smpi

Hello, I made a simple program in which 4 mpi ranks send data a two different intervals of time. Here description of what each of the ranks is programmed to do.

0 (host-0) => sleeps for 25 sec => sends X bytes to 1
1 (host-1) => sleeps for 25 sec => recv  X bytes from 0
2 (host-0) => sleeps for 50 sec => sends X bytes to 3
3 (host-1) => sleeps for 50 sec => recv  X bytes from 2

Depending on the configuration of the execution (speed of the links, and the amount of data exchanged) the communication between 0 <-> 1 and 2 <-> 3 can overlap, resulting in two communication using the same link.

The issue concerns the tracing during this kind of configuration.

I am interested in plotting the communication of the application but at the link level. Fortunately, simgrid has an option to do that --cfg=tracing/uncategorized:yes.

However, the paje trace fails to render the communications correctly. The attached tarball contains two paje trace, one with overlapping trace_buggy.paje and one without trace_ok.paje. We can see that the paje trace with communication overlap fails to render the first communication happening at 25.000 of the simulation.

The tarball also contains everything needed to run the code.

simgrid_uncategorized_issue.tar

Edited Jan 20, 2020 by Adrien Faure
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information
Assignee
Assign to
None
Milestone
None
Assign milestone
Time tracking
None
Due date
None
Reference: simgrid/simgrid#44