http://www.k-wave.org/forum/topic/libmvec Support for the k-Wave MATLAB toolbox en-US Tue, 12 May 2026 23:27:53 +0000 http://bbpress.org/?v=1.0.2 q http://www.k-wave.org/forum/search.php http://www.k-wave.org/forum/topic/libmvec#post-6386 Fri, 06 Apr 2018 08:54:51 +0000 Jiri Jaros 6386@http://www.k-wave.org/forum/ <p>Hi<br /> I did see this problem on a couple of systems. libmvec is a C++ Math library with SSE/AVX vectorization, however, I'm not sure why the code cannot find it, it may have been replaced by Matlab. It also might be causes by the fact the code was compiled with GCC 5.4 and CUDA 8.0</p> <p>To solve it, I'd try one one of these possibilities:<br /> 1. Try to run ./kspaceFirstOrder3D-CUDA to see what happens if the CUDA code is launched as a standalone application.<br /> 1. Try to load proper modules.<br /> 2. Find the libm.so file, typically it can be found in /lib/x86_64-linux-gnu and change the LD_LIBRARY_PATH in the system command.<br /> 3. Try to download the libs from here: <a href="http://www.fit.vutbr.cz/~jarosjir/download/libm.tgz" rel="nofollow">http://www.fit.vutbr.cz/~jarosjir/download/libm.tgz</a><br /> 4. Recompile the code from sources. It should be easy, once you find the HDF5 module and set the paths to CUDA and HDF5 in the makefile.</p> <p>Best,<br /> Jiri </p> http://www.k-wave.org/forum/topic/libmvec#post-6384 Thu, 05 Apr 2018 18:30:03 +0000 mcdannold 6384@http://www.k-wave.org/forum/ <p>I tried running the gpu compiled code and am getting the following error:</p> <p>“./kspaceFirstOrder3D-CUDA: error while loading shared libraries: libmvec.so.1: cannot open shared object file: No such file or directory”</p> <p>I can't find this library, and my google searches are coming up empty. Does anyone have an idea where I might find it?</p> <p>I am also not sure I am doing this properly.</p> <p>Here is how I log into a node with a gpu:</p> <p>[njm7@login02 ~]$ srun -n 1 --pty -t 1:00:00 -p gpu --gres=gpu:1 bash<br /> srun: job 11701595 queued and waiting for resources<br /> srun: job 11701595 has been allocated resources<br /> [njm7@compute-g-16-175 ~]$ module load gcc/6.2.0<br /> [njm7@compute-g-16-175 ~]$ module load cuda/9.0<br /> [njm7@compute-g-16-175 ~]$ module load matlab/2017a<br /> [njm7@compute-g-16-175 ~]$ LD_PRELOAD='/n/app/gcc/6.2.0/lib64/libstdc++.so.6.0.22' matlab -nodesktop</p> <p>Here is what I get when I try to run the simulation using the compiled gpu binary:</p> <p>Running k-Wave simulation...<br /> start time: 05-Apr-2018 10:59:25<br /> reference sound speed: 3274.9813m/s<br /> dt: 7.6889ns, t_end: 119.708us, time steps: 15570<br /> input grid size: 44 by 44 by 492 grid points (14.5055 by 14.5055 by 162.1978mm)<br /> maximum supported frequency: 2.275MHz<br /> expanding computational grid...<br /> computational grid size: 64 by 64 by 512 grid points<br /> precomputation completed in 1.14s<br /> saving input files to disk...<br /> completed in 2.1249s<br /> system('export LD_LIBRARY_PATH=; cd /home/njm7/MFILES/k-Wave/binaries/; ./kspaceFirstOrder3D-CUDA -i /home/njm7/DATA/TMPat1Element0054a_105925.mat_input.h5 -o /home/njm7/DATA/TMPat1Element0054a_105925.mat_output.h5 --p_raw -s 14542' ,'-echo');<br /> ./kspaceFirstOrder3D-CUDA: error while loading shared libraries: libmvec.so.1: cannot open shared object file: No such file or directory</p> <p>Is this the right way to do this? </p>