nihui
nihui
The fp16 arithemetic produces unwanted artifacts for waifu2x models iirc https://github.com/nihui/waifu2x-ncnn-vulkan/blob/b152fc7ee703ea46575fb0d20fef18cb7237236c/src/waifu2x.cpp#L51
what is the android system version ？
``` Waifu2x::Waifu2x(int gpuid, bool _tta_mode)) { net.opt.use_image_storage = false; } ``` A potential workaround for some old buggy driver, but it will probably slow down speed a lot on qcom...
> > conversion order is not random if you follow my instructions > > I may have found out why.because of the parameter " -j 4:4:4", and if it is...
https://www.techpowerup.com/gpu-specs/tesla-k80.c2616 Tesla K80 supports vulkan, so try re-installing your graphics driver, make sure all the driver components are installed properly, especially opengl.
``` if (gpuid.empty()) { gpuid.push_back(ncnn::get_default_gpu_index()); } ``` In xxx-ncnn-vulkan, we use the default gpu if not specified, that should be discrete gpu first, integrated gpu next and others last.
The latest master branch supports compiling gpu functions without installing vulkan-sdk. You can give it a try. If it doesn’t work, feel free to reopen the issue and post your...
fixed in https://github.com/Tencent/ncnn/commit/6647396667426405cee4da8da79f723148f60dd5
最新版本优化了许多 intel cpu 的性能，欢迎尝试，如有问题，可以 reopen issue 继续交流～～