Probe seems to consume the CPU

by

Yes; most MPI implementations, for the sake of performance, busy-wait on blocking operations. The assumption is that the MPI job is the only thing going on that we care about on the processor, and if the task is blocked waiting for communications, the best thing to do is to continually poll for that communication to reduce latency; so that there’s virtually no delay between when the message arrives and when it’s handed off to the MPI task. This typically means that CPU is pegged at 100% even when nothing “real” is being done.

That’s probably the best default behaviour for most MPI users, but it isn’t always what you want. Typically MPI implementations allow turning this off; with OpenMPI, you can turn this behaviour off with an MCA parameter,

mpirun -np N --mca mpi_yield_when_idle 1 ./a.out

More Related Contents:

  1. MPI_Rank return same process number for all process
  2. Can MPI_Publish_name be used for two separately started applications?
  3. MPI hangs on MPI_Send for large messages
  4. sending blocks of 2D array in C using MPI
  5. Sending and receiving 2D array over MPI
  6. MPI_Recv overwrites parts of memory it should not access
  7. Ordering Output in MPI
  8. struct serialization in C and transfer over MPI
  9. How do I debug an MPI program?
  10. Sending 2D arrays in Fortran with MPI_Gather
  11. MPI_Bcast a dynamic 2d array
  12. MPI: blocking vs non-blocking
  13. Using MPI_Bcast for MPI communication
  14. Tools to measure MPI communication costs [closed]
  15. Trouble Understanding MPI_Type_create_struct
  16. Creating a counter that stays synchronized across MPI processes
  17. Good profiler for Fortran and MPI
  18. Understanding concurrent file writes from multiple processes
  19. MPI partition matrix into blocks
  20. Scatter Matrix Blocks of Different Sizes using MPI
  21. Sending columns of a matrix using MPI_Scatter
  22. MPI Reading from a text file
  23. How to compile an MPI included c program using cmake
  24. Parallel Algorithms for Generating Prime Numbers (possibly using Hadoop’s map reduce)
  25. MPI_Type_create_subarray and MPI_Gather
  26. writing a matrix into a single txt file with mpi
  27. assign two MPI processes per core
  28. thread safety of MPI send using threads created with std::async
  29. Creating an MPI_Datatype for a structure containing pointers
  30. MPI_Gather gives seg fault in the most basic code

Leave a Comment