k-Wave User Forum » Topic: `Fatal error: Could not write into "" dataset.: iostream stream error`

k-Wave User Forum » Topic: `Fatal error: Could not write into "" dataset.: iostream stream error` http://www.k-wave.org/forum/topic/fatal-error-could-not-write-into-dataset-iostream-stream-error Support for the k-Wave MATLAB toolbox en-US Tue, 12 May 2026 22:24:41 +0000 http://bbpress.org/?v=1.0.2 <![CDATA[Search]]> q http://www.k-wave.org/forum/search.php Jiri Jaros on "`Fatal error: Could not write into "" dataset.: iostream stream error`" http://www.k-wave.org/forum/topic/fatal-error-could-not-write-into-dataset-iostream-stream-error#post-8792 Tue, 06 Jun 2023 12:51:25 +0000 Jiri Jaros 8792@http://www.k-wave.org/forum/ It is the problem of HDF5 incompatibility in new Matlab versions, we released a fix in version v1.4 that saves HDF5Atributes using system encoding. subhadip627@ on "`Fatal error: Could not write into "" dataset.: iostream stream error`" http://www.k-wave.org/forum/topic/fatal-error-could-not-write-into-dataset-iostream-stream-error#post-8715 Wed, 08 Mar 2023 07:28:57 +0000 subhadip627@ 8715@http://www.k-wave.org/forum/ Hi I am facing a compilation problem with the following code, i can't understand the solution as i am a new user for cpp code. clear all run([getkWavePath 'private/getH5Literals']); filename = 'input12.h5'; % create the computational grid Nx = 64; % number of grid points in the x direction Ny = 64; % number of grid points in the y direction Nz = 64; % number of grid points in the z direction dx = 0.1e-3; % grid point spacing in the x direction [m] dy = 0.1e-3; % grid point spacing in the y direction [m] dz = 0.1e-3; % grid point spacing in the z direction [m] kgrid = kWaveGrid(Nx, dx, Ny, dy, Nz, dz); % define the properties of the propagation medium medium.sound_speed = 1500 * ones(Nx, Ny, Nz); % [m/s] % medium.sound_speed(1:Nx/2, :, :) = 1800; % [m/s] medium.density = 1000 * ones(Nx, Ny, Nz); % [kg/m^3] % medium.density(:, Ny/4:end, :) = 1200; % [kg/m^3] c0=medium.sound_speed; % make sure the input is in the correct data format eval(['c0 = ' MATRIX_DATA_TYPE_MATLAB '(c0);']); % save the sound speed matrix writeMatrix([filename], c0, 'c0'); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % % % % create initial pressure distribution using makeBall ball_magnitude = 10; % [Pa] ball_x_pos = 38; % [grid points] ball_y_pos = 32; % [grid points] ball_z_pos = 32; % [grid points] ball_radius = 5; % [grid points] p_mask = makeBall(Nx, Ny, Nz, ball_x_pos, ball_y_pos, ball_z_pos, ball_radius); ball_1 = ball_magnitude *p_mask ; source.p0 = ball_1; %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % define a square source mask facing in the x-direction using the % find linear source indices p_source_index = find(p_mask == 1); p_source_index = reshape(p_source_index, [], 1); % make sure the input is in the correct data format eval(['p_source_index = ' INTEGER_DATA_TYPE_MATLAB '(p_source_index);']); % save the source index matrix writeMatrix([filename], p_source_index, 'p_source_index'); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % cast matrix to single precision eval(['ball_1 = ' MATRIX_DATA_TYPE_MATLAB '(ball_1);']); % save the input signal writeMatrix([filename], ball_1, 'ball_1'); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % define a sensor mask through the central plane sensor_mask = false(Nx, Ny, Nz); sensor_mask(:, :, Nz/2) = 1; % extract the indices of the active sensor mask elements sensor_mask_index = find(sensor_mask); sensor_mask_index = reshape(sensor_mask_index, [], 1); % make sure the input is in the correct data format eval(['sensor_mask_index = ' INTEGER_DATA_TYPE_MATLAB '(sensor_mask_index);']); sensor.mask=sensor_mask; % save the sensor mask writeMatrix([filename], sensor_mask_index, 'sensor_mask_index'); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% % write grid parameters writeGrid([filename], [Nx, Ny, Nz], [dx, dy, dz], ... [10, 10, 10], [10, 10,10], ... Nt, dt, 1500); % write flags writeFlags([filename]); % set additional file attributes writeAttributes([filename]); kspaceFirstOrder3D(kgrid, medium, source, sensor, 'SaveToDisk', filename); %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% here i am able to make .h5 file but while compiling the .h5 file with kspaceFirstOrder-OMP i got ---- +---------------------------------------------------------------+ | kspaceFirstOrder-OMP v1.3 | +---------------------------------------------------------------+ | Reading simulation configuration: Failed | +---------------------------------------------------------------+ +---------------------------------------------------------------+ | !!! K-Wave experienced a fatal error !!! | +---------------------------------------------------------------+ | Error: Invalid format of the input file: iostream stream | | error | +---------------------------------------------------------------+ | Execution terminated | +---------------------------------------------------------------+ Jiri Jaros on "`Fatal error: Could not write into "" dataset.: iostream stream error`" http://www.k-wave.org/forum/topic/fatal-error-could-not-write-into-dataset-iostream-stream-error#post-6813 Fri, 29 Mar 2019 09:35:49 +0000 Jiri Jaros 6813@http://www.k-wave.org/forum/ Hi shak360, 1. I will send you the codes with a readme via email. 2. You can see a detail log using a parameter --verbose 2 3. The execution time is inversely proportional to the number of cores. Typically, when you double the number of cores, you decrease the simulation time by a factor of 1.6. But keep in mind, the highest number of cores used is limited by Nz and Ny (precisely by the highest common denominator). With 1024 cores you can get down to 2 days. Best Jiri shak360 on "`Fatal error: Could not write into "" dataset.: iostream stream error`" http://www.k-wave.org/forum/topic/fatal-error-could-not-write-into-dataset-iostream-stream-error#post-6808 Thu, 28 Mar 2019 17:47:09 +0000 shak360 6808@http://www.k-wave.org/forum/ (The above post should say 690 kHz not 690 Hz.) Sorry, but I just wanted to add this briefly: In this paper: <a href="https://asa.scitation.org/doi/pdf/10.1121/1.4976339?class=pdf" rel="nofollow">https://asa.scitation.org/doi/pdf/10.1121/1.4976339?class=pdf</a>, the following specs were used for "3D simulations [on head models at 500 kHz]: - on the IT4I Salomon supercomputing cluster. - Each simulation was carried out on Intel Xeon E5-4627v2, 3.3 GHz, 8 cores and - 256 GB of RAM per simulation. The largest 3D simulations had a domain size of 1024^3 including the PML and comprised 22,718 time steps, with a total runtime of 112.3 hours (or ~5 days)." This is similar to what I'm trying to do. Granted this was most likely the longest simulation you had to test, how feasible was it when you tried to simulate this? I'm just concerned because, what would you do in the scenario where you wanted to change one aspect of the simulation (such as the placement or the orientation of the transducer, or some simulation parameters), but you would have to wait ~5 days to reconfigure? shak360 on "`Fatal error: Could not write into "" dataset.: iostream stream error`" http://www.k-wave.org/forum/topic/fatal-error-could-not-write-into-dataset-iostream-stream-error#post-6807 Thu, 28 Mar 2019 17:08:47 +0000 shak360 6807@http://www.k-wave.org/forum/ Hi Dr. Jaros, Thanks for the quick reply. I see, so even storing the entire field can lead to issues with memory (I'm assuming the other 30 matrices are similar to the ones like <code>medium.density</code> and <code>medium.sound_speed</code>, which are why they would have to be the same size as the grid). I need the high resolution spatial grid, however, to sample my transducer frequency of 690 Hz accordingly. The smallest amount of time I want to run the simulation for is 5e-4 seconds, which is still relatively long. I'd appreciate if you send me the distributed simulation code because I do have access to a cluster, but I'm seeing now that I'll have to approach this problem differently. I'm trying to simulate ultrasound on head-models from CT scans, and low-resolution, short time-scale simulations are possible in 2D, but I want to be able to do more. I'm still relatively new to simulation and ultrasound in general but k-Wave has helped me a lot; do you have any advice on how to move forward? Also, where I can find the whole simulation log? Jiri Jaros on "`Fatal error: Could not write into "" dataset.: iostream stream error`" http://www.k-wave.org/forum/topic/fatal-error-could-not-write-into-dataset-iostream-stream-error#post-6769 Fri, 15 Mar 2019 13:55:57 +0000 Jiri Jaros 6769@http://www.k-wave.org/forum/ Hi shak360, Could you please send me the whole simulation log and what machine specs. This simple calculation indicates it is a memory problem. If the simulation is absorbing, nonlinear and heterogeneous, you need about 30 matrices of the grid size. In this case 1024*1024*512 points x 4B = 60GB of memory. There's no GPU on the market able to running this simulation and a sever with 64GB of RAM could go into troubles with swapping. Storing the whole field rises up another problem - One time step of pressure consumes 2GB. Sampling even 512 time steps takes 1TB of data - opening this dataset in matlab at once is virtually impossible. There's not much we can do about it. I'd definitely reduce the area you're sampling as well as the number of time steps. If you had access to a distributed cluster/supercomputer, we could send you a prototype of the distributed simulation code. But keep in mind, at least 256 CPU cores would be necessary and the simulation output can easily become intractable. shak360 on "`Fatal error: Could not write into "" dataset.: iostream stream error`" http://www.k-wave.org/forum/topic/fatal-error-could-not-write-into-dataset-iostream-stream-error#post-6766 Fri, 08 Mar 2019 22:17:42 +0000 shak360 6766@http://www.k-wave.org/forum/ Hi, I'm trying to use the C++ 3D simulations after running into some memory errors using the MATLAB version. I received the error: <code>Fatal error: Could not write into "" dataset.: iostream stream error</code> I'm pretty sure I have everything set up right, though... Is this a memory issue? I'm asking it to sense the entire field: <code>sensor.mask = [1, 1, 1, Nx, Ny, Nz].';</code> Is this not a good way of going about that? I've spent a lot of time with k-Wave so intuitively I think some options would be to reduce that <code>sensor.mask</code> to some volume (or maybe even 2D area) around the focus and only record a certain portion of time at steady-state. Is this correct? Would this help with any memory issues? Is it reasonable to measure one or two periods before the end of the simulation as input to <code>sensor.record_start_index</code>? I have a GPU and I managed to get both the CUDA (MATLAB) and CUDA (C++) modules that you have supplied also working, but I'm running into GPU memory issues there as well. I don't think I can make any more changes to the grid size or the domain dimensions (1024 x 1024 x 512) (0.243 mm x 0.243 mm x 0.25 mm) as I want to be able to support a frequency of 690kHz. Thank you in advance for the help. k-Wave is an excellent tool.