Draft: SMPI network calibration: tutorial
Presents how to do a proper SMPI calibration using SimGrid. Extract models for communications from real measures of dahu cluster. The models are created automatically using pycewise and ckmeans/dhist.
Implemented support for dynamic cost in SMPI operations.
Main chainges:
- New tutorial: includes a dockerfile and jupyter notebook explaining how to do the SMPI calibration
- SMPI dynamic costs: added API to allow user to add a callback to set
dynamic costs for MPI_Send, MPI_Isend and MPI_Recv operations.
- ortogonal with other costs configs, e.g. smpi/os, smpi/or.
- New example: examples/smpi/comm_dynamic_costs
Edited by Bruno Donassolo