Package Details: waifu2x-converter-cpp-cuda-git 5.3.4.2.g57520b2-2

Git Clone URL: https://aur.archlinux.org/waifu2x-converter-cpp-cuda-git.git (read-only, click to copy)
Package Base: waifu2x-converter-cpp-cuda-git
Description: Image Super-Resolution for Anime-Style-Art. (re-implementation in C++ using OpenCV). with CUDA support (GIT Version)
Upstream URL: https://github.com/DeadSix27/waifu2x-converter-cpp
Licenses: MIT
Conflicts: waifu2x-converter-cpp
Provides: libw2xc.so, waifu2x-converter-cpp
Submitter: YamashitaRen
Maintainer: sl1pkn07
Last Packager: sl1pkn07
Votes: 14
Popularity: 0.000000
First Submitted: 2015-09-24 22:02 (UTC)
Last Updated: 2023-11-01 00:27 (UTC)

Latest Comments

« First ‹ Previous 1 2 3 4 5 6 Next › Last »

fitu996 commented on 2018-01-12 10:46 (UTC)

Upstream fixed cuda9 and your patch cause build fails now.

I have tested that simply remove it will be ok.

sl1pkn07 commented on 2017-11-17 05:37 (UTC)

Talk with upstream Greetings

Vitruvius commented on 2017-11-17 00:56 (UTC)

>nvcc fatal : Value 'sm_20' is not defined for option 'gpu-architecture' I got it to compile by removing references to "sm_20" and "ptx20" from the code. This has something to do with what cuda virtual architectures it's compiled to be compatible with, but I have no idea what that actually means. https://gist.github.com/gekkey/1740a1d57e015036b14d7067285777cf

pingplug commented on 2017-11-16 11:04 (UTC)

nvcc fatal : Value 'sm_20' is not defined for option 'gpu-architecture'

sl1pkn07 commented on 2017-02-04 12:37 (UTC)

moving to fork. try build now seems the original don't work anymore with opencv 3.2.0

eigengrau commented on 2017-02-04 08:25 (UTC) (edited on 2017-02-04 08:28 (UTC) by eigengrau)

For some reason, when building this now, neither the binary nor the shared library are installed. Also, which is odd, the symbol table of the shared library in the build directory is empty, and no w2xconv binary is built.

sl1pkn07 commented on 2017-01-05 01:50 (UTC)

this variable is part of cmake, and still exist in the file FindCUDA.cmake in /usr/share/cmake-3.7/Modules if see the CMakeCache.txt in the build directory, you can see the CUDA_TOOLKIT_ROOT_DIR and CUDA_SDK_ROOT_DIR is set in /opt/cuda is more, seems CUDA_SDK_ROOT_DIR or CUDA_TOOLKIT_ROOT_DIR no need set in this project (see output if add VERBOSE=1 in the make rule greetings

Mitame commented on 2017-01-04 22:22 (UTC)

I think in the latest git versions, 'CUDA_SDK_ROOT_DIR' has been changed to 'CUDA_TOOLKIT_ROOT_DIR'. Changing the defines in this PKGBUILD allowed it to build with CUDA.

sl1pkn07 commented on 2016-11-10 14:33 (UTC)

done

<deleted-account> commented on 2016-11-09 15:19 (UTC)

Installing opencl-headers also works. It should definitely be added as a makedepend.

« First ‹ Previous 1 2 3 4 5 6 Next › Last »