Chronological Thread 

< Chronological >      < Thread >

---

David Nickerson wrote:
> Thats good to see Andrew. Have you tried running the same experiment
> with one of the GSL integrators?
No, but I have compared other models (e.g. the Zhang S.A.N. model from
Alan), and it is significantly faster.
> The next step will be to see how the
> simulation goes when runs within PCEnv (which is essentially a GUI
> around the CIS, right? or is CIS not using the cellml_corba_bridge?)
>
CIS is a shared library, which can interact with the CellML DOM API
either across CORBA or directly, without going through CORBA.
cellml_corba_server loads CISService (which links to CIS), and sets it
up so CIS interacts directly with the CellML DOM API. Calls to CIS can
then come across CORBA.

The command line RunCellML does not need to stream points across CORBA.
Obviously, we would get performance issues if we tried to stream every
point across CORBA and draw them on the canvas, but PCEnv applies a
density limit on points (see below) so at the default density, a maximum
of 1000 points get sent across CORBA (for other models of similar size,
this seems to add a few seconds maximum. Once you get past 1000 steps,
this will not be affected by the length of time you run the simulation
over).
> The simulation time with CIS is pretty close to that with my code with
> the small difference probably due to the slightly tighter tolerances I
> used (1.0e-6 relative and 1.0e-8 absolute) and that I'm running on a
> 3GHz P4.
>
> I suspect the difference in time with COR might be due to what is being
> done with the results. At least for my code, I am streaming the full
> results set (just over 90 variables) to disk every 1ms - which probably
> takes as much time as the numerical integration (should benchmark that
> also I guess). And you're probably doing the same thing using CIS,
> right?
No, I set the maximum density to 1 point (CIS uses point density as
follows: Compute gap = (endPoint - startPoint) / density. If an output
point is less than gap after the last point to be sent back to the user,
ignore that point). The one point I do generate gets written to the console.
> The question is what COR is doing with those results, Alan? (This
> is assuming the the two different version of the TNNP model are actually
> equivalent :)
>
It would be good to compare them.

Best regards,
Andrew

---

• [cellml-discussion] pcenv development priorities, (continued)
  • [cellml-discussion] pcenv development priorities, David Nickerson, 10/30/2006
    • [cellml-discussion] pcenv development priorities, Alan Garny, 10/30/2006
  • [cellml-discussion] pcenv development priorities, Poul Nielsen, 10/30/2006
    • [cellml-discussion] pcenv development priorities, Andrew Miller, 10/30/2006
  • [cellml-discussion] pcenv development priorities, David Nickerson, 10/30/2006
    • [cellml-discussion] pcenv development priorities, Alan Garny, 10/30/2006
      • [cellml-discussion] pcenv development priorities, David Nickerson, 10/30/2006
        • [cellml-discussion] pcenv development priorities, Alan Garny, 10/31/2006
        • [cellml-discussion] pcenv development priorities, Andrew Miller, 10/31/2006
          • [cellml-discussion] pcenv development priorities, David Nickerson, 10/31/2006
            • [cellml-discussion] pcenv development priorities, Andrew Miller, 10/31/2006
            • [cellml-discussion] pcenv development priorities, Alan Garny, 10/31/2006

---

Archive powered by MHonArc 2.6.18.