k-Wave

1. bISHOP
   

   Member
   Joined: Jul '10
   Posts: 6

   Hi.

   I´m sorry about being first to post... an error, but I´m glad to help with this fantastic tool you have developed. I´ve followed the steps to install CUDA, GPUmat and k-Wave and everything seemed to be allright until I tried the homogeneus propagation medium (the first one, at http://www.k-wave.org/documentation/example_ps_homogeneous_medium.php ) with 'Datacast', 'GPUsingle' in the kspaceFirstOrder2D call. MATLAB is not happy. MATLAB disagrees with my code this way:

   Running k-space simulation...
     dt: 20ns, t_end: 12.06us, time steps: 604
     input grid size: 128 by 128 pixels (12.8 by 12.8mm)
     maximum supported frequency: 7.5MHz
     smoothing p0 distribution...
     calculating Delaunay triangulation...
     casting variables to kWaveGPUsingle type...
     precomputation completed in 2.2222s
     starting time loop...
   ??? Error using ==> times
   Operations between Matlab arrays and GPU arrays are not supported.
   (ERROR code 5)
   
   Error in ==> gridDataFast at 96
   zi = sum(z(tri) .* w,2);
   
   Error in ==> kspaceFirstOrder2D at 1194
                   sensor_data(:, t_index) = gridDataFast(kgrid.z, kgrid.x, p,
   sensor_z, sensor_x, del_tri);
   
   Error in ==> ej211 at 40
   sensor_data = kspaceFirstOrder2D(kgrid, medium, source, sensor,
   'DataCast', 'GPUsingle');

   The only modifications are two addpath lines at the beggining and the DataCast parameter.

   I´m using a Intel Core2 Quad (Q6600, 2.4Ghz) with 2GB RAM, Windows XP SP3, MATLAB 7.7.0 (R2008b), CUDA 3.1, GPUmat 0.251 and k-Wave Version B.0.2

   Thank you and I hope it is an error of mine.

   Daniel.

   Posted 14 years ago #

2. Bradley Treeby
   

   Developer
   Joined: Mar '10
   Posts: 1,643

   Hi Daniel,

   Thanks for your feedback and post. Unfortunately, the function which calculates the pressure at sensor points which are defined in Cartesian space using linear interpolation (gridDataFast) does not currently support execution on the GPU. Hopefully this will be fixed by the next release. To get around this in the meantime, you can either set the interpolation mode to nearest neighbour, e.g.,

   sensor_data = kspaceFirstOrder2D(kgrid, medium, source, sensor, 'DataCast', 'GPUsingle', 'CartInterp', 'nearest');

   or you can use a binary sensor mask as in the second photoacoustic example.

   Posted 14 years ago #

3. bISHOP
   

   Member
   Joined: Jul '10
   Posts: 6

   Thank you. Interpolating to nearest has worked.

   I think I´m going to be a regular user of this forum with exclusive attention.

   Great job!

   Daniel.

   Posted 14 years ago #

RSS feed for this topic

Reply

You must log in to post.