http://www.k-wave.org/forum/topic/simulate-axisymmetric-acoustic-field-using-cuda-binary-got-wrong-result Support for the k-Wave MATLAB toolbox en-US Wed, 13 May 2026 02:30:13 +0000 http://bbpress.org/?v=1.0.2 q http://www.k-wave.org/forum/search.php http://www.k-wave.org/forum/topic/simulate-axisymmetric-acoustic-field-using-cuda-binary-got-wrong-result#post-8250 Wed, 21 Jul 2021 14:34:42 +0000 Bradley Treeby 8250@http://www.k-wave.org/forum/ <p>Hi vbtraz,</p> <p>There is no GPU version of the axisymmetric code (CUDA has no native support for the required DTTs). It looks the CUDA code has incorrectly interpreted your simulations as 2D simulations, which is why they are incorrect.</p> <p>Brad. </p> http://www.k-wave.org/forum/topic/simulate-axisymmetric-acoustic-field-using-cuda-binary-got-wrong-result#post-8237 Wed, 14 Jul 2021 07:08:26 +0000 vbtraz 8237@http://www.k-wave.org/forum/ <p>Hi, </p> <p>I want to calculate a AS field using a line source(disk source in 3D) with time varying signal.<br /> I got same result using kspaceFirstOrderAS and kspaceFirstOrderASC as <img src="https://i.ibb.co/7zMhnZz/kspace-First-Order-AS.jpg" alt="kspace-First-Order-AS" /><br /> But the result seems wrong using kspaceFirstOrder-CUDA.exe. (I change kspaceFirstOrderASC to kspaceFirstOderASG)<br /> <img src="https://i.ibb.co/5B70XK4/kspace-First-Order-CUDA-exe.jpg" alt="kspace-First-Order-CUDA-exe" /></p> <p>My code is here: </p> <pre><code>clear; %% signal parameters freq = 100e3; %[Hz] p0 = 10; %[Pa] %% medium (water / linear) medium.sound_speed = 1500; %[m/s] medium.density = 1000; %[kg/m^3] %% grids dx = medium.sound_speed / (freq * 10); Nx = 512; Ny = 128; xMax = Nx * dx; yMax = Ny * dx; kgrid = kWaveGrid(Nx, dx, Ny, dx); kgrid.makeTime(medium.sound_speed, 0.3, (sqrt(xMax^2 + yMax^2)/medium.sound_speed + 1e-3) * 1.01); %% source source.p_mask = zeros(Nx, Ny); source.p_mask(21, 1:round(0.1/dx + 1)) = 1; source.p = p0 * sin(2*pi*freq * kgrid.t_array); %% sensor sensor.mask = zeros(Nx, Ny); sensor.mask(21:Nx-20, 1) = 1; %% RUN sensor_data = kspaceFirstOrderAS(kgrid, medium, source, sensor, &#39;DataCast&#39;, &#39;gpuArray-single&#39;, &#39;DataRecast&#39;, true, &#39;PlotScale&#39;, [-p0, p0]); %sensor_data = kspaceFirstOrderASG(kgrid, medium, source, sensor, &#39;DataCast&#39;, &#39;gpuArray-single&#39;, &#39;DataRecast&#39;, true, &#39;PlotScale&#39;, [-p0, p0]); %% plot figure; data = sensor_data(:, end-round(1e-3/kgrid.dt-1) : end); semilogy((0:Nx-41)*dx, rms(data, 2));</code></pre> <p>I change functon kspaceFirstOrderASC to kspaceFirstOrderASG to call kspaceFirstOrder-CUDA.exe<br /> <pre><code>function sensor_data = kspaceFirstOrderASG(varargin) if ~any(strcmp(&#39;FunctionName&#39;, varargin)) varargin = [varargin {&#39;FunctionName&#39;, &#39;kspaceFirstOrderAS&#39;}]; end % Check for the binary name input. If not defined, set the default name of % the GPU binary if ~any(strcmp(&#39;BinaryName&#39;, varargin)) if isunix varargin = [varargin {&#39;BinaryName&#39;, &#39;kspaceFirstOrder-CUDA&#39;}]; else varargin = [varargin {&#39;BinaryName&#39;, &#39;kspaceFirstOrder-CUDA.exe&#39;}]; end end</code></pre>