Menu

Hardware requirements to run the model of Portuguese economy

dp118m
2018-10-17
2021-08-22
  • dp118m

    dp118m - 2018-10-17

    Hi!

    How much CPU power and RAM do I need in order to be able to run the model of Portuguese economy ( https://github.com/highperformancecoder/minsky-models/tree/master/PedroPratasThesis ) on a Windows machine (more or less smoothly)?

    Thanks

    Dmitri Pisarenko

     
    • High Performance Coder

      Just for kickers, I loaded up Pedro's model on my antidiluvian laptop
      - a 1.6GHz Intel Atom based machine from 2010, with 2GB memory.

      It took about 20 seconds to redraw the canvas after (say) zooming, but
      this dropped to a more managable 5 seconds or so by turning off the
      panopticon. Panning works instantaneously, but I wouldn't recommend
      trying to build a model of that complexity on such a low powered
      machine.

      When running the simulation, it took about 12 seconds per year of
      simulation time.

      One thing, though - my laptop runs OpenSUSE Tumbleweed Linux - I don't
      think it would be powerful enough to run any modern version of Windows
      anyway. However, I gather that Minsky performance on Windows can be up
      to a factor of two slower than running it on Linux. And the MacOSX version
      significantly slower again. No apparent reason why.

      --


      Dr Russell Standish Phone 0425 253119 (mobile)
      Principal, High Performance Coders
      Visiting Senior Research Fellow hpcoder@hpcoders.com.au
      Economics, Kingston University http://www.hpcoders.com.au


       
      • dp118m

        dp118m - 2018-10-18

        Thanks for your answer. Switching off the panopticon actually helped.

         
  • Kresimir

    Kresimir - 2018-10-17

    Try first to disable panopticon in preferences.

     
    • dp118m

      dp118m - 2018-10-18

      Thanks!

       
  • Ivan Johnstone

    Ivan Johnstone - 2019-11-19

    I have downloaded Minsky 2.17 and for the first time had a look at a larger Minsky file PratasModelPortugal. In equation mode I find that scrolling through the equations to be extremely slow. Under Preferences I have turned off Wrap long equations in LaTeX export, Enable panopticon, and both with no change in scrolling speed. My desktop computer grunt is an Intel i7 quad core first generation CPU, 16 GB RAM, and my software is on an SSD. Lack of scrolling speed might be related to LaTeX and from memory somewhere in documentaion there is reference to a separate licence. This is the first time I have come across software that runs sluggishly on my desltop computer. How do I resolve the inadequate scrolling speed?

     
  • High Performance Coder

    Scrolling speed of the equation tab is a known issue. Please see https://sourceforge.net/p/minsky/tickets/914/. As a workaround, you could export the Canvas as LaTeX, and compile the LaTeX to pdf (via pdflatex) and then use a PDF viewer like okular to scroll around the equations.

     
  • gas ase

    gas ase - 2021-08-22

    I'm playing with predator-prey structures with low populations, so integer values are crucial
    Also the documentation mentions sliders for changing value page.

     
    • High Performance Coder

      On Sun, Aug 22, 2021 at 06:44:20PM -0000, gas ase wrote:

      I'm playing with predator-prey structures with low populations, so integer
      values are crucial
      Also the documentation mentions sliders for changing value page.

      Hi Gas - I'm not sure what you're questions are here.

      Minsky does not support integer types - everything is double precision
      floating point.

      Re Lotke-Volterra equations, you can simulate extinction via adding a
      small random value, and feeding the result into a floor operation.

      x' = ⌊x+r⌋

      where r is a random number from [0,1). This is the way it is done in
      EcoLab. The other way it is sometimes done is to deploy an arbitrary
      extinction threshold, below which the species is considered extinct.

      In Minsky you can generate random numbers by creating a parameter to
      hold the random values for the simulation. Bear in mind, the random
      value need to be generated ahead of time, not of the fly, as the
      equations need to be deterministic. So if you want 1000 random values,
      spread over timesteps of 0.1, you would create a parameter r
      initialised with rand(1000). Then to get the value at t, take r[10*t],
      where [] is the gather operator (right-most in Minsky).

      Re sliders, they only apply to parameters. Edit the parameter, and set
      the min/max values to a range, and the slider will appear (a small dot
      on the top of the parameter icon). You can adjust the simulation as it
      proceeds.

      --


      Dr Russell Standish Phone 0425 253119 (mobile)
      Principal, High Performance Coders hpcoder@hpcoders.com.au
      http://www.hpcoders.com.au


       

Log in to post a comment.