none
Why is MS R slower? RRS feed

  • Question

  • Platform: SunFire x4800

    Testing newest Cran R and MS R with a taxometric R program:

    MAXEIG(paranoid[,c(2:8)],Supplied.P = 0.049, Ind.Triplets=T, File.Output = T, Gr.Rows = 5, Gr.Cols = 2)
    > system.time(MAXEIG(paranoid[,c(2:8)],Supplied.P = 0.049, Ind.Triplets=T, File.Output = T, Gr.Rows = 5, Gr.Cols = 2))
         user    system   elapsed
    Microsoft R
    10972.475    18.196 10995.596
    > system.time(MAXEIG(paranoid[,c(2:8)],Supplied.P = 0.049, Ind.Triplets=T, File.Output = T, Gr.Rows = 5, Gr.Cols = 2))
         user    system   elapsed
    13251.465    35.814 13276.047

    MS R is actually slower!  I set it for 8 threads.

    Roger

    Wednesday, October 25, 2017 4:32 PM

All replies

  • It's hard to know without knowing what MAXEIG is doing, but the fact that the user and elapsed times are so close indicates that there's no benefit being accrued from multi-threading. Which platform are you on?

    You might also want to try re-running the example a couple of times on each platform, and taking an average while discarding the first result (to try and eliminate cache effects and things like that).

    Wednesday, October 25, 2017 5:12 PM
  • Platform: SunFire x4800
    Thursday, October 26, 2017 3:27 PM
  • I can't include maxeig - too long.
    Thursday, October 26, 2017 3:32 PM
  • Can you tell us what R package you are using here, that contains the 'MAXEIG' function?

    Thanks.

    Stephen Weller

    Microsoft R Product Team

    Monday, October 30, 2017 3:38 PM