I am new to OpenCL. I have written a vector addition code in OpenCL with help from Internet. I have included one header file i.e. CL/cl.h using #include.
I am using NVIDIA graphic card and the OpenCL implementation is NVIDIA_GPU_Computing_SDK. My OpenCL header files are residing at this path /opt/NVIDIA_GPU_Computing_SDK/OpenCL/common/inc. I can run OpenCL programs through linux terminal by adding this path when compiling my code. But now I want to write CMake file for this code. CMake files are working fine for C programs, but not OpenCL programs because of this Path problem. In terminal, I used to enter $cmake ., after this $make, it will search for a Makefile which is created by cmake, now my error is after entering command make
fatal error: CL/cl.h: No such file or directory!
Now tell me how can I include this header file into CMake file?
You will need to put these lines into CMakeLists.txt:
include_directories(/opt/NVIDIA_GPU_Computing_SDK/OpenCL/common/inc)
link_directories(/opt/NVIDIA_GPU_Computing_SDK/OpenCL/common/<lib or something similar>)
add_executable(yourexe src1.c ...)
target_link_libraries(yourexe OpenCL)
But beware that this is not portable, because OpenCL SDK can be somewhere else on another machine. The proper way to do this is to use FindOpenCL.cmake module.
Maybe you can use a CMake "find" script like:
http://gitorious.org/findopencl/findopencl/blobs/master/FindOpenCL.cmake
http://code.google.com/p/opencl-book-samples/source/browse/trunk/cmake/FindOpenCL.cmake?r=14
CMake file example from OpenCL Programming Guide Book: http://code.google.com/p/opencl-book-samples/source/browse/trunk/CMakeLists.txt?r=14
I was looking for FindOpenCL.cmake macro which would work well on Windows, OSX and Linux... I couldn't find any which did work well on every platform, so I wrote new one which I use in couple of projects (webcl-validator and opencl-testsuite).
https://github.com/elhigu/cmake-findopencl
Especially Windows support is improved in this one.
In Windows it checks if 64bit or 32bit lib should be used and it also tries to find libraries from according to environment variables set by Nvidia, Intel and AMD OpenCL SDKs.
It also tries to find .lib in Cygwin, which didn't work with other scripts I tried.
Related
I've just started working with CMake and I noticed that they have both a find_package and a find_library. And this confuses me. Can somebody explain the difference between a package and a library in the world of programming? Or, in the world of CMake?
Appreciate it, guys!
Imagine you want to use zlib in your project, you need to find the header file zlib.h, and the library libz.so (on Linux). You can use the low-level cmake commands find_path and find_library to find them, or you can use find_package(ZLIB). The later command will try to find out all what is necessary to use zlib. It can be extra macro definitions, or dependencies.
Update, more detail about find_package: when the CMake command find_package(SomeThing) is called, as in the documentation, there are two possible modes that cmake can run:
the module mode (that searches for a file FindSomeThing.cmake)
or the config mode (that searches for a file named SomeThingConfig.cmake)
For ZLIB, there is a module named FindZLIB, shipped with CMake itself (on my Linux machine that is the file /usr/share/cmake/Modules/FindZLIB.cmake). That module is a CMake script that uses the CMake API to search for ZLIB files in default locations, or ask the user for the location if it cannot be found automatically.
I am using CMake to build a Qt5 project on OS X. I need to create a build process that is as simple as possible for others.
By default Qt5 installs to the home folder on OS X. However, it then places its files within a directory named after the exact version number, e.g. 5.2.1.
At the moment I am using these lines in my CMake file:
set(QT5_PATH $ENV{HOME}/Qt5.2.1/5.2.1/clang_64/ CACHE PATH "Path to Qt5")
set(QT5_MODULE_PATH ${QT5_PATH}/lib/cmake)
set(CMAKE_PREFIX_PATH ${CMAKE_PREFIX_PATH} ${QT5_MODULE_PATH})
This works, but unfortunately breaks with each minor update of Qt as the 5.2.1 needs to be changed to 5.2.2, etc.
In Windows there are environment variables that can be used to find Qt. Is there anything similar in OS X that I can use within CMake to find a Qt installation?
Use find_package instead of juggling with the paths yourself. Then your users can rely on standard CMake mechanism instead of figuring out your own CMake code.
Use find_package(Qt5Widgets) to get targets like Qt::Widgets to link against and for the includes.
Similar for Qt5Core and whatever part of Qt 5 you need.
See Qt's documentation: http://doc.qt.io/qt-5/cmake-manual.html
I'm cross-compiling for VxWorks using cmake. When I run cmake the first time I have to provide informations about compiler, target OS etc..
In the cross-compile dialogue there are three target system settings I set:
Operating System
Version
Processor
(followed by compiler etc.)
While I can retrieve the first one using CMAKE_SYSTEM_NAME, i can't get the version and the processor.
Both return an empty string.
Here's an example:
MESSAGE("CMAKE_SYSTEM_PROCESSOR: ${CMAKE_SYSTEM_PROCESSOR}")
MESSAGE("CMAKE_SYSTEM_VERSION: ${CMAKE_SYSTEM_VERSION}")
Output:
CMAKE_SYSTEM_PROCESSOR:
CMAKE_SYSTEM_VERSION:
My Cmake Version is 2.8.10.2 and target OS is VxWorks (if this matters - compiler are WindRiver GNU).
How can I get the version and processor I've set in the beginning? Or is this impossible if I cross-compile to an OS that's unknown to cmake?
(Btw. Compiling works fine)
It seems this is not possible so far. I'm getting empty strings all the time.
However, there's a working solution, and i guess it's the better way:
Before:
I specified cross-compile settings (Compiler and target system, see question), then it runs over VxWorks specific parts in the CMake list (checked with if( VxWorks ) to ensure it's not executed when other systems are used).
Now (Solution):
I wrote a toolchain file and platform files for VxWorks and required processors.
Cons:
I have to write some extra files:
Toolchain file
Platform file for VxWorks
Further Platform files for each Processor (and processor type, Gnu and Diab)
Pros:
CMake list is much cleaner now
Separate Project and Target settings
Separate System and processor settings - easy to add new Processors in a very clear way but keep System settings
I write some settings in the toolchain file and CMake loads related system / processor settings
...
I'm currently building a compiler/interpreter in C/C++.
When I noticed LLVM I thought it would fit greatly to what I needed and so I'm trying to integrate LLVM in my existing build system (I use CMake).
I read this bout integration of LLVM in CMake. I copy and pasted the example CMakeLists.txt, changed the LLVM_ROOT to ~/.llvm/ (that's where I downloaded and build LLVM and clang) and it says it isn't a valid LLVM-install. Best result I could achieve was the error message "Can't find LLVMConfig" by changing LLVM_ROOT to ~/.llvm/llvm.
My ~/.llvm/ folder looks like this:
~/.llvm/llvm # this folder contains source files
~/.llvm/build # this folder contains object, executable and library files
I downloaded LLVM and clang via SVN. I did not build it with CMake.
Is it just me or is something wrong with the CMakeLists.txt?
This CMake documentation page got rotted, but setting up CMake for LLVM developing isn't different from any other project. If your headers/libs are installed into non-standard prefix, there is no way for CMake to guess it.
You need to set CMAKE_PREFIX_PATH to the LLVM installation prefix or CMAKE_MODULE_PATH to prefix/share/llvm/cmake to make it work.
And yes, use the second code snippet from documentation (under Alternativaly, you can utilize CMake’s find_package functionality. line).
When it comes to using the terminal to build libraries manually and such I unfortunately do not have much experience and I'm stuck a bit here.
I've downloaded a library for objective-c which came with makefiles and such.
I can see that the folder also contains an executable file called "libtool", I did some searching and I suppose this is the program I have to use to build the neccessary .a files? Unfortunately I couldn't really find any useful article for this that seemed to work.
The folder for the library contains some .sh files, .pc files and also some .la files, but I'm a bit unsure of which ones I have to use as input to the libtool program to compile them into a .a file.
So my question is what files do you have to input into libtool to compile them into the necessary .a file? And what commands do you use exactly to accomplish this?
Thank you all for your time :)
First a little introduction to static libraries:
Static libraries in Unix environments (like Mac OSX, and Linux too) are actually just an archive of object files created by the ar command line program.
That is what the .a extension stands for: Archive.
To create a static library with some object files you can use the command like this:
ar crv libmy_library.a objectfile1.o objectfile2.o
As for your actual question, libtool should be called automatically from the makefile, creating the library, which is the file ending in .la. However, this is not the real library, the real library is in a hidden directory. You can find it by doing e.g.
find . -name '*.a'
But like I said, the makefile should already take care of everything, including installing the correct library in the correct place when you do e.g. make install.
For information about libtool, see this site.