I need to edit the Visual Studio command line for the NVidia CUDA compiler nvcc.exe build step, however it can not be edited.
I'm sure this question has been asked but it is tricky because if I search for anything with "Command Line" it returns hundreds of articles on how to just use external command line compiling, not from inside VS.
NVidia CUDA *.cu files have a different command lines for both the compile and link steps, and if I go to my Visual Studio project properties, Configuration Properties, CUDA C/C++, Command Line, neither can be edited.
I searched for some of the command line text in the NVidia toolkit folder and found this.
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\extras\visual_studio_integration\MSBuildExtensions\CUDA 10.1.props
This file seems to define the build step, but if I edit it and restart Visual Studio, nothing changes. It seems to have been used only during the CUDA installation. I've tried opening various VS files such as project.user but the nvcc command line is not defined anywhere in the solution files. Also, it's no good changing the start up project definition because I need to be able to add CUDA GPU processing to my existing project.
Have any of you Visual Studio guru's been able to edit the Command Line for third party compilers and if so, how? Thanks so much!
You just edit it like you do with any other build tool in Visual Studio: by editing the properties. The command line category in the project properties is really just there to show you what the command is going to look like based on the properties you set. The whole point of having properties is that you just set properties and have the IDE translate those properties into the correct command line for the specific tool rather than specifying a command line directly…
The .props file of a build customization sets up the default properties to use for the item types associated with that build customization. The one actually being used should be installed in the MSBuild directory, e.g., C:\Program Files (x86)\MSBuild\Microsoft.Cpp\v4.0\BuildCustomizations. I would advise against editing that file directly though, as it sets up system-global defaults. If you want to change the defaults for your project, simply edit the CUDA C/C++ properties of the project (i.e., select the project in the solution explorer and edit its properties rather than the properties of an individual .cu file).
Properties in Visual Studio are built up in layers (strictly speaking, we're talking about item metadata here rather than properties, but that's a minor technicality which should be irrelevant here). At the bottom, you have the default property definitions which the project imports for all its build tools. On top of those, you have the project-wide property definitions. And each item (file) within the project can have its own property definitions on top of that. The set of properties for each item is the combination of all properties defined for this item type (e.g., a CUDA C/C++ file) as well as any item-specific definition, where the last definition overwrites previous ones. When you see a property rendered in bold font in the Visual Studio properties dialog, that means that that particular property value is overwritten at that level. Non-bold properties are inherited from layers beneath. So, for example, a bold property in the project properties means that setting is overwritten on top of the default property coming from the build integration. A bold property at the item level means that property is overwritten for this item on top of the project-wide properties. Further customization is possible, e.g., by defining your own property sheets, which allow you to add your own layers of property definitions. Property sheets are also standalone files which can be imported into multiple projects, e.g., to share common settings. Simply open the Property Manager (View > Property Manager) to see all the individual property sheets in your project at the moment and, e.g., add your own…
I'm posting an answer to my own question here after resolving it so if others have the same problem they will be able to find a simple solution here.
As you can see above the nvcc.exe compiler command line is not editable because it is simply a function of various other editable properties.
However, the whole reason I wanted to edit the nvcc command line in the first place was because of the following compiler error:
1>nvcc fatal : A single input file is required for a non-link phase when an outputfile is specified
The NVidia CUDA nvcc.exe compiler was working correctly for my Visual Studio Debug build but failed every time with the above error for any Release build.
I anticipated that to solve this I would have to edit the nvcc.exe command line, however this turned out not to be the case, for the following reason.
When I deleted all spaces in my Visual Studio Configuration Build names the problem completely disappeared!
Apparently the nvcc.exe compiler can not handle spaces in configuration build names despite the use of double quotes around them in all command line references. This problem therefore is probably something internal to the nvcc compiler and was never found by NVidia simply because most people are smart enough not to use spaces in their build names.
Bottom line, do not use spaces in your VS configuration build names and you will not have this problem with the NVidia CUDA compiler.
Related
I'm trying to generate a VxWorks 6.9.4.8 kernel configuration that is identical to another kernel workbench project. The Workbench 3.3.6 only allows GUI configuration.
Is there an underlying kernel configuration file, produced by the GUI, which can be replaced?
After updating the kernel configuration using the Workbench GUI, I see the following files have changed:
linkSyms.c,
prjComps.h,
prjConfig.c, and
prjParams.h
I guess my question is, which one, if any uniquely identifies the kernel as built?
prjComps.h will contain all the component's names, as you have chosen in your kernel configuration GUI.
First step to create new Kernel configuration based on some other Kernel configuration is to use GUI configurator and add the missing component in prjComps.h, Better use some diff tool like 'beyond compare', and keep reducing the differences by adding/removing the components. Remember not to edit this file directly, but via GUI configurator only. As the tool calculates the dependent component and adds/removes them.
Second step is to create the new prjParams.h as above.
The Workbench actually allows to use command line to edit Kernel configuration via vxprj tool in vxworks 6.9(this tool has been replaced by "wrtool" in vxworks 7), you can right click on the Image project and chose 'Open Wind River vxWorks 6.9 Developement Shell'.
If you want to add a component for e.g. telnet client (INCLUDE_TELNET_CLIENT)
, you can use the following command
vxprj component add INCLUDE_TELNET_CLIENT
To remove a component
vxprj component remove INCLUDE_TELNET_CLIENT
For more of vxprj tool, you can look up the documentation in the workbench itself.
The project configuration is held in a handful of files in the kernel project directory.
These are:
.project
.cproject
.wrproject
projectname.wpj
Files such as prjComps.h, prjParams.h prjConfig.c are all generated by the configuration tool, however these are not configuration files themselves. Instead, this is generated C code that contains, amongst other things, a list of selected components.
These files are also re-generated, I believe, when you rebuild the project.
As such, these are not really the authoritative source you are interested in.
For this, you need to look at the project files. In terms of a list of components, the most interesting is the .wpj file, which contains amongst other things a list of explicitly and implicitly included components.
The explicitly included components are those you manually selected in the Kernel Configuration GUI, the implicitly included are those that were then included to satisfy dependencies.
This distinction can sometimes make comparing kernel configurations tricky, then you may want to fall back on the generated files eg prjComps.h, however you should always remember that this is a representation of the configuration, not the source.
The .project etc configuration files are big and complex, but a decent diff tool, such as BeyondCompare can make comparisons of the project directories fairly easy
Thanks for the clue, #endTunnel. I looked at that file, and noticed that a few files get modified when I save my GUI selections.
prjComps.h - all the components #included in the kernel build
prjParams.h - the additional parameters set for the enabled components
prjConfig.c - the configuration and initialization calls for each module included.
'linkSyms.c' also gets modified. Not sure how that is used, yet.
I can now use diff to compare kernel configurations, and perhaps even duplicate a configuration (haven't tried that yet).
I am trying to compile a cuda program which uses cusparse library. I am getting linking error:
kernel.cu.obj : error LNK2019: unresolved external symbol _cusparseSetMatIndexBase#8 referenced in function _main
And a lot of error of same kind related to cusparse library. I have included "cusparse_v2.h".
How do i link cusparse library in visual studio 2010 during compilation?
The general instructions for linking libraries is not specific to CUDA. So you may want to learn more about using MS VS. Anyway, the steps are like this:
Make sure you have opened the project that you want to work on.
Select View...Property Pages (from the menu) A new dialog box will open up.
On the left hand side of this dialog box, select Linker to open up it's sub-menu
Under linker, select Input
Now, on the pane on the right, observe the first item which is "Additional Dependencies". I believe cudart.lib should already be present there.
Click to the right of cudart.lib You can now type in new libraries to be added. Type a space (to separate from cudart.lib) and type cusparse.lib
Now click "Apply" in the lower right corner of the dialog box.
That should be all that's needed, if your project/solution file is already set up using a cuda template. If cudart.lib is not present, or your project/solution files do not already comprehend cuda, that is a separate issue. In that case I would recommend starting over, by cloning a project from the Samples, and building your project using that as a starting point. It will then pick up all the proper directories to search as well as the cuda build rules. Since all the main cuda libraries (cudart, cublas, cufft, cusparse, etc.) are all in the same location, the same search path should pick any of them up as needed.
If you wanted to link another library, such as cublas.lib, for example, you could follow a similar sequence, replacing cusparse.lib above with cublas.lib
Note that with newer versions of CUDA (e.g. CUDA 7.5) it will be necessary to build a 64-bit project only (follow the above steps when modifying the x64 project properties.) CUDA 7.5 and beyond are dropping support for 32-bit projects.
I fixed it by following steps:
Add cuda path:
Go: "Configuration Properties->Linker->General->Additional Libary Directories" and add $(CudaToolkitLibDir) to the list.
Add cuda realtime library:
Go: "Solution Properties->Linker->Input->Additional Dependencies" and add cudart.lib to the list.
Add cublas library:
Go: "Solution Properties->Linker->Input->Additional Dependencies" and add cublas.lib to the list.
Changing platform to x64:
Go: "Configuration Properties->Platform" and set it to x64.
Run cmd.exe as administrator.
Type in and run the following two lines of command:
netsh winsock reset catalog
netsh int ip reset reset.log hit
It may say that a reboot is required, but actually that is not necessary.
Try to debug your application again, the problem should be solved.
I am trying to compile a cuda program which uses cusparse library. I am getting linking error:
kernel.cu.obj : error LNK2019: unresolved external symbol _cusparseSetMatIndexBase#8 referenced in function _main
And a lot of error of same kind related to cusparse library. I have included "cusparse_v2.h".
How do i link cusparse library in visual studio 2010 during compilation?
The general instructions for linking libraries is not specific to CUDA. So you may want to learn more about using MS VS. Anyway, the steps are like this:
Make sure you have opened the project that you want to work on.
Select View...Property Pages (from the menu) A new dialog box will open up.
On the left hand side of this dialog box, select Linker to open up it's sub-menu
Under linker, select Input
Now, on the pane on the right, observe the first item which is "Additional Dependencies". I believe cudart.lib should already be present there.
Click to the right of cudart.lib You can now type in new libraries to be added. Type a space (to separate from cudart.lib) and type cusparse.lib
Now click "Apply" in the lower right corner of the dialog box.
That should be all that's needed, if your project/solution file is already set up using a cuda template. If cudart.lib is not present, or your project/solution files do not already comprehend cuda, that is a separate issue. In that case I would recommend starting over, by cloning a project from the Samples, and building your project using that as a starting point. It will then pick up all the proper directories to search as well as the cuda build rules. Since all the main cuda libraries (cudart, cublas, cufft, cusparse, etc.) are all in the same location, the same search path should pick any of them up as needed.
If you wanted to link another library, such as cublas.lib, for example, you could follow a similar sequence, replacing cusparse.lib above with cublas.lib
Note that with newer versions of CUDA (e.g. CUDA 7.5) it will be necessary to build a 64-bit project only (follow the above steps when modifying the x64 project properties.) CUDA 7.5 and beyond are dropping support for 32-bit projects.
I fixed it by following steps:
Add cuda path:
Go: "Configuration Properties->Linker->General->Additional Libary Directories" and add $(CudaToolkitLibDir) to the list.
Add cuda realtime library:
Go: "Solution Properties->Linker->Input->Additional Dependencies" and add cudart.lib to the list.
Add cublas library:
Go: "Solution Properties->Linker->Input->Additional Dependencies" and add cublas.lib to the list.
Changing platform to x64:
Go: "Configuration Properties->Platform" and set it to x64.
Run cmd.exe as administrator.
Type in and run the following two lines of command:
netsh winsock reset catalog
netsh int ip reset reset.log hit
It may say that a reboot is required, but actually that is not necessary.
Try to debug your application again, the problem should be solved.
I'm using CUDA 5.5 and my compiler is VC2012. My projects consists of 2 .cu files. I need to use dynamic parallelism so I have enabled "Generate relocatable device code" option. As soon as this option is enabled I get the following linking error:
error LNK2001: unresolved external symbol __fatbinwrap_54_tmpxft_0000110c_00000000_8_cuda_device_runtime_cpp1_ii_5f6993ef
If I turn that option off the error disappears. The error does not depend on the contents of the files - if I comment out everything in them (#if 0 .... #endif so that they are essentially empty files) I still get the same error.
Update: I installed VC2010, and I still get the exact same error message!
Update 2: I got really tired with this whole thing so I took a dynamic parallelism sample project from VC2010 from the SDK samples (cdpSimplePrint), replaced the files in it and compiled. The issue is gone. Tried to compare the settings between the two projects, but I couldn't find anything. I don't know what caused it, but frankly, at this stage I don't care already, as long as I have a workaround.
Check that you are linking with the cudadevrt library, see the "Compiling and Linking" section in the CUDA Dynamic Parallelism Programming Guide for more info.
We have a C++ Builder application that has lived through many versions of Embarcadero RAD Studio. Now I was trying to get it to compile in XE2 and ran into a problem where a class called TGroup became ambiguos with the TGroup class of Winsock.
To my anger I found that the initial architects of the system didn't use any namespaces, which would have solved this problem. I decided to add all the project code to a single namespace to start with. But pretty soon it dawned on me that I would have to go through about 1400 .h and .cpp files.
Please, for the love of god... Is there a good tool out there to quickly encapsulate an entire project within a namespace?
Most of the files differ a bit in layout. Some have defines after the includes and some do not etc. But to just insert the namespace declaration after the includes should probably work ok to start with and then close it just before the last #endif, which is the inclusion guard #define.
Check if the offending header files already use namespaces. All VCL-based header files DO use namespaces, but they also have using namespace statements at the bottom to dump everything into the global namespace by default. Most versions have special #define values you can set to disable that behavior.