Parconnect

Last week

  • Run testcases with mvapich2                  

  • Complete the graphs in paper

  • Prepare the run, parse, plot script
    https://github.com/afg984/sc16

  • Test the scalability of parconnect

Crash reason

Subtitle

Mvapich2

  • Bigger test range in  ap, ap_lb

  • Slower than intel mpi

  • More unstable than intel mpi                  

Graph in paper

  • There are 3 graphs in paper                    

  • Computation and communication with iteration

  • The tuple count with each iteration

  • The benchmark

Graph with intel

Graph with mvapich

group update

TODO

  • Write script to estimate execution time

  • Rerun the entire workflow

  • Support other ap

Made with Slides.com