karthik-nagabhushana Support for the k-Wave MATLAB toolbox en-US Tue, 12 May 2026 23:02:23 +0000 http://bbpress.org/?v=1.0.2 q http://www.k-wave.org/forum/search.php http://www.k-wave.org/forum/topic/multiple-gpu-operation#post-8310 Tue, 21 Sep 2021 16:23:42 +0000 Jiri Jaros 8310@http://www.k-wave.org/forum/ <p>Hi Karthik ,<br /> 1) Yes, you can compile the code for A100. Just install CUDA 11 and add SM architecture 8.0 and 8.6. Check this post<br /> <a href="http://www.k-wave.org/forum/topic/is-nvidia-rtx-3060-compatible-with-k-wave#post-8267" rel="nofollow">http://www.k-wave.org/forum/topic/is-nvidia-rtx-3060-compatible-with-k-wave#post-8267</a><br /> 2) This is still not possible. k-Wave is so much communication bound that the using multiple GPUs don't bring any performance improvement. This is given by the cudaFFT library which doesn't scale. </p> http://www.k-wave.org/forum/topic/multiple-gpu-operation#post-8241 Tue, 20 Jul 2021 03:26:49 +0000 Karthik Nagabhushana 8241@http://www.k-wave.org/forum/ <p>Our team did get access to two HPC clusters. Cluster 1 is NVIDIA DGX A100 (GPU - 8x NVIDIA A100 (40GB/unit), CPU - Dual AMD Rome 7742 (128 cores @ 2.25GHz)). Cluster 2 is made up of IBM AC922 nodes (GPU - 4x NVIDIA V100 (16GB/unit), CPU - 2x 20-core IBM POWER9 @ 2.4GHz). So, I have two questions - 1) Can I compile the single-GPU kWave source files to create binaries for the above CPU architectures? 2) This is a repeat of the multi-GPU question in the previous comment. We want to make use of the multiple GPUs in a given node to support bigger domains. So, is the multi-GPU source files or binaries available for public use? Thanks!</p> <p>Regards,<br /> Karthik </p> http://www.k-wave.org/forum/topic/multiple-gpu-operation#post-8234 Mon, 12 Jul 2021 06:07:00 +0000 Karthik Nagabhushana 8234@http://www.k-wave.org/forum/ <p>Hi, </p> <p>I have the same question as Alex. Our team is exploring to use a multi-GPU system such as NVIDIA DGX A100 (8x NVIDIA A100, 40GB memory per GPU) to simulate large domains (1024x1024x1024 and higher). Representing such a domain requires &gt;100GB memory which can only be achieved with multiple GPUs. So, I was wondering if the multi-GPU functionality mentioned in this paper (<a href="http://bug.medphys.ucl.ac.uk/papers/2018-Treeby-ISNA.pdf" rel="nofollow">http://bug.medphys.ucl.ac.uk/papers/2018-Treeby-ISNA.pdf</a>) is supported in the 1.3 version (C++ GPU code)? If not, do you plan to release this functionality in the future? </p> <p>In the past, we had used CPU-based compute nodes (128GB memory) to simulate such large domains but want to explore multi-GPU systems to improve our simulation time. </p> <p>Thanks </p> <p>Karthik </p> http://www.k-wave.org/forum/topic/power-law-absorption-in-axisymmetric-simulations#post-7964 Sun, 29 Nov 2020 03:23:47 +0000 Karthik Nagabhushana 7964@http://www.k-wave.org/forum/ <p>Hi Professor Treeby, </p> <p>Thanks for your comment. </p> <p>Thanks for suggesting an approximate workaround as well; I will try to use the axisymmetric simulations in a piecewise f^2 manner. </p> <p>Karthik </p> http://www.k-wave.org/forum/topic/power-law-absorption-in-axisymmetric-simulations#post-7938 Thu, 26 Nov 2020 11:23:28 +0000 Bradley Treeby 7938@http://www.k-wave.org/forum/ <p>Hi Karthik,</p> <p>&gt;&gt; Is this a fundamental limitation of the axisymmetric approach?</p> <p>In short, yes. It's because the generalisation of the Laplacian in cylindrical coordinates to the equivalent fractional version is not straightforward to solve.</p> <p>If your frequency content is centred around a particular frequency, you can get approximately the right absorption behaviour by adjusting the values of <code>alpha_coeff</code> so that you get the same absorption at that frequency, e.g., </p> <p><code>alpha_coeff_adjusted * f^2 = alpha_coeff * f^y</code></p> <p>No current plans to extend the code to support power law absorption I'm afraid.</p> <p>Hope that helps,</p> <p>Brad </p> http://www.k-wave.org/forum/topic/power-law-absorption-in-axisymmetric-simulations#post-7912 Tue, 03 Nov 2020 02:31:11 +0000 Karthik Nagabhushana 7912@http://www.k-wave.org/forum/ <p>Hi, </p> <p>I am currently trying to model imaging with a spherically focused transducer. Compared to the conventional 3D representation (kspaceFirstOrder3DC), I have found the axisymmetric representation (kspaceFirstOrderAS) to be extremely useful in terms of compute time and memory requirements for large grid sizes (&gt;1024^3). </p> <p>However, one constraint in the axisymmetric representation is that the full range of acoustic power law is not supported (or medium.alpha_power is set to 2). Is this a fundamental limitation of the axisymmetric approach? If not, do you have plans in the future to support the acoustic power law? </p> <p>Thanks in advance </p> <p>Regards<br /> Karthik </p> http://www.k-wave.org/forum/topic/multiple-gpu-operation#post-6792 Tue, 19 Mar 2019 12:17:01 +0000 alexalex811 6792@http://www.k-wave.org/forum/ <p>Thanks Jiri, thats very helpful.<br /> Alex. </p> http://www.k-wave.org/forum/topic/multiple-gpu-operation#post-6770 Fri, 15 Mar 2019 13:58:51 +0000 Jiri Jaros 6770@http://www.k-wave.org/forum/ <p>Hi alexalex811,<br /> the release of this code is planned for the Q3/Q4 of 2019. We still need to investigate the stability of the new code.</p> <p>Best<br /> Jiri </p> http://www.k-wave.org/forum/topic/multiple-gpu-operation#post-6767 Mon, 11 Mar 2019 16:07:42 +0000 alexalex811 6767@http://www.k-wave.org/forum/ <p>Hi,</p> <p>My lab is looking at getting a new high performance PC, on which we are likely to run a number of k-wave simulations.</p> <p>I was wondering if it is now possible to run simulations across multiple GPU's? </p> <p>Alex </p>