Hello,
I noticed heterogeneous medium leads to longer computation time than homogenous(in the same way, adding void content in the medium leads to further time).
Could you enlighten me why this is occuring? Is this simply due to calculation regarding boundary interaction?
Thanks,
Sheun
If you are using the maketime() function then the time step is calculated in accordance with the formula for the CFL number in the k-wave manual.
As the time step is dependent on the speed of sound specified in the grid and the CFL number, changing the media to a heterogeneous one will result in a different time step being calculated, and consequently a change in the time for the simulation.
I believe the best way to specify this is that the behavior of the wave at the boundary requires a smaller time step to accurately model, and a longer simulation time results because of this.
Hi Sheun,
In addition to what ouelletn has pointed out above, you may also see a small performance increase when using a homogeneous medium because some element-wise matrix multiplications are replaced with matrix-constant multiplications. The homogeneous medium parameters can also be easily cached, which reduces the required memory bandwidth. However, from experience, the difference in performance because of these factors is only on the order of 5-10%.
Brad.
You must log in to post.