Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.
Particle tracing: timing issues
Posted 9 gen 2014, 09:17 GMT-5 2 Replies
Please login with a confirmed email address before reporting spam
Hello,
I have created a simulation suite that calculates the electrostatic field of a given device and traces particles through this field. Since none of the available particle release nodes suited my needs (I wanted to generate particles in the (x,dx/dz) and (y,dy/dz) 2-dimensional phase spaces according to a gaussian distribution described by the Twiss-parameters), I followed the suggestion of the COMSOL experts and generated a ReleaseFromGrid node for each individual particle, where I could set the particle coordinates explicitely. It runs fine, but as I increase the number of particles, the simulation time seems to show an exponential increase (see attached file, simulation time [s] vs. number of particles). This would become prohibitively slow very quickly.
The physical and virtual memory usage of the simulation process was not significantly different among the studied cases (around 1+1 GB). My notebook has 4 GB memory and a Core-i7 processor.
As far as I know inter-particle interactions are not taken into account in COMSOL by default, so I would naively expect a roughly linear behaviour (with an offset corresponding to the electrostatics calculation which is independent of the number of particles). I have COMSOL 4.3b
Can somebody comment on this issue? What is the best way to simulate a large (500-1000) number of particles with input coordinates specified individually?
Thank you
Daniel
I have created a simulation suite that calculates the electrostatic field of a given device and traces particles through this field. Since none of the available particle release nodes suited my needs (I wanted to generate particles in the (x,dx/dz) and (y,dy/dz) 2-dimensional phase spaces according to a gaussian distribution described by the Twiss-parameters), I followed the suggestion of the COMSOL experts and generated a ReleaseFromGrid node for each individual particle, where I could set the particle coordinates explicitely. It runs fine, but as I increase the number of particles, the simulation time seems to show an exponential increase (see attached file, simulation time [s] vs. number of particles). This would become prohibitively slow very quickly.
The physical and virtual memory usage of the simulation process was not significantly different among the studied cases (around 1+1 GB). My notebook has 4 GB memory and a Core-i7 processor.
As far as I know inter-particle interactions are not taken into account in COMSOL by default, so I would naively expect a roughly linear behaviour (with an offset corresponding to the electrostatics calculation which is independent of the number of particles). I have COMSOL 4.3b
Can somebody comment on this issue? What is the best way to simulate a large (500-1000) number of particles with input coordinates specified individually?
Thank you
Daniel
Attachments:
2 Replies Last Post 13 gen 2014, 05:44 GMT-5
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
HI Daniel,
in case the community does not provide an answer in that case, feel free to contact us in support@comsol.com.
Sven
in case the community does not provide an answer in that case, feel free to contact us in support@comsol.com.
Sven
Please login with a confirmed email address before reporting spam
Posted:
1 decade ago
Hello
Thank you, I have just done so.
Daniel
Thank you, I have just done so.
Daniel