https://bugs.kde.org/show_bug.cgi?id=478426
--- Comment #19 from Niklāvs Koļesņikovs <89q1r1...@relay.firefox.com> --- I'm pretty sure that the nice documentation is the computer graphics book that CImg developers want their users to buy. However, being 2000' academic software, it does have the Java style API documentation that was all the rage back in those days. The main link is slightly broken, however by clicking around a working one can be obtained (this one should lead straight to the resize function that allows picking the kernel: https://cimg.eu/reference/structcimg__library_1_1CImg.html#a6a668c8b3f9d756264d1fb31b7a915fc ). There's also examples ( https://github.com/GreycLab/CImg/tree/master/examples ) to see how the CImg API comes together. That being said, I'm not aware of CImg having any kind of GPU acceleration and any CPU acceleration or optimization is likely handled by the optional `-Dcimg_use_openmp -Dcimg_use_lapack` which will pull in OpenMP and LAPACK respectively. Neither should be as much of a hassle as OpenCV, so an improvement overall. When I did play around with C++ graphics on my own time, I always used C libraries wrapped in custom RAII C++ code. I have not used it myself but there is Intel's oneAPI ( https://spec.oneapi.io/oneipl/0.6/transform/resize_lanczos.html ) and specifically its DPC++ language/compiler ( https://intel.github.io/llvm-docs/GetStartedGuide.html ). I would not expect it to actually use GPU acceleration on the average Linux system right now, however the same might actually be true with OpenCV as well, because all my searching indicates it only supports CUDA for GPU acceleration. Overall, maybe best to just lower the expectations and pick something CPU based like CImg and open a feature request for KF6 or Qt6 to implement better scaling methods (and ideally ability to use libplacebo compatible shaders, because there's quite a large selection of awesome shaders for mpv). -- You are receiving this mail because: You are watching all bug changes.
