http://www.k-wave.org/forum/topic/fft-plans-creation Support for the k-Wave MATLAB toolbox en-US Tue, 12 May 2026 22:32:33 +0000 http://bbpress.org/?v=1.0.2 q http://www.k-wave.org/forum/search.php http://www.k-wave.org/forum/topic/fft-plans-creation#post-6943 Mon, 01 Jul 2019 10:34:00 +0000 Jiri Jaros 6943@http://www.k-wave.org/forum/ <p>Hi Tony,<br /> from the point of viewt the C++ code, every scan line is an independent simulation. That's why it generates a new plan for FFT. (In general, every simulation can be of different sizes, that's why the code generates a new one every time).</p> <p>In the provided binary, the MKL is used and the planing phase should be instant. If you compile it with FFTW, you can change the level of planning and speed the pre-processing phase up.<br /> 1. Open the source file MatrixClasses/FftwComplexMatrix.h<br /> 2. change this line<br /> <pre><code>static const unsigned kFftMeasureFlag = FFTW_MEASURE;</code></pre> <p>to<br /> <pre><code>static const unsigned kFftMeasureFlag = FFTW_ESTIMATE;</code></pre> <p>In the future, I may add a flag to import system-wise wisdom or the one stored in file.<br /> Best wishes<br /> Jiri </p> http://www.k-wave.org/forum/topic/fft-plans-creation#post-6872 Wed, 05 Jun 2019 16:33:55 +0000 guns2roses 6872@http://www.k-wave.org/forum/ <p>Hi Brad,</p> <p>I use kspaceFirstOrder3DG(...) for phased array B-mode ultrasound imaging simulation. However, when computing each scan line, the initialisation step always appears before the actual simulation. In the initialisation process, the "FFT plans creation" takes most of the computational time (about 3 - 4 mins for grid size 900*900*25 points, time steps: 6307). I understand that this depends on the throughput of the CPU, but I wonder why the simulation has to do it for every single scan line calculation as the medium is always the same in each iteration. Would it be possible to just do the initialisation at the very beginning of the simulation, saved it, and used it for the rest of the scan line simulations? If this is possible, we would save quite significant amount of time for the simulation (eg. 4 mins*65 scan lines = 4.33 hrs). Thanks a million :)</p> <p>Best wishes,</p> <p>Toni </p>