一 尝试在intel+tesla平台上安装opencl,遇到很多问题。不再尝试。换amd平台
1 strerror.cpp:56:8: error: ‘CL_INVALID_PROPERTY’ was not declared in this scope
解答: 在 strerror.cpp中注释掉对应行
Hi Evan, thanks for your email. The issue is only partly due to OpenCL 1.1. Recent NVIDIA drivers do not ship their own OpenCL header files any longer, so one has to manually get them from the Khronos group. If you are updating to recent NVIDIA drivers, the old header files might still be in place. It seems that the earlier NVIDIA header files are not 100 percent compatible, e.g. CL_INVALID_PROPERTY is not defined. Anyway, it should be sufficient if you just comment that particular line(s) in error.hpp.
2 undefined reference problem in clAmdFft
Has anyone ever seen this error when linking libclAmdFft.Runtime?
I'm sure that my opencl environment is setup properly and other opencl codes run perfectly in my computer. It seems that the opencl installed in my computer is version 1.1 while clAmdFft library requires version 1.0.
Does anyone know how to solve it? Thanks in advance.
/home/jason/Dropbox/clAmdFft/lib32/libclAmdFft.Runtime.so: undefined reference to `clCreateBuffer@OPENCL_1.0'
Solved.
Apparently I referenced the libOpenCL.so provided by nvidia, in /usr/local/cuda/lib. After I changed it to /opt/AMDAPP/lib/x86, it works fine.
However, I still think there is a bug in here. Different applications should have been compatible with the standard OpenCL library. If AMD's application only runs on AMD's OpenCL library and nvidia's application only runs on nvidia's OpenCL library. What can I do if I want to use both of them?
http://devgurus.amd.com/thread/157052
进展:编译错误。
可能是opencl库版本不对 或者 如上边所说需要连接 amd提供的opencl库
二 AMD平台上安装clmagma