I am currently writing a CUDA application and want to use the boost::program_options library to get the required parameters and user input.
The trouble I am having is that NVCC cannot handle compiling the boost file any.hpp giving errors such as
1>C:\boost_1_47_0\boost/any.hpp(68): error C3857: 'boost::any': multiple template parameter lists are not allowed
I searched online and found it is because NVCC cannot handle the certain constructs used in the boost code but that NVCC should delegate compilation of host code to the C++ compiler. In my case I am using Visual Studio 2010 so host code should be passed to cl.
Since NVCC seemed to be getting confused I even wrote a simple wrapper around the boost stuff and stuck it in a separate .cpp (instead of .cu) file but I am still getting build errors. Weirdly the error is thrown upon compiling my main.cu instead of the wrapper.cpp but still is caused by boost even though main.cu doesn't include any boost code.
Does anybody know of a solution or even workaround for this problem?
Dan, I have written a CUDA code using boost::program_options in the past, and looked back to it to see how I dealt with your problem. There are certainly some quirks in the nvcc compile chain. I believe you can generally deal with this if you've decomposed your classes appropriately, and realize that often NVCC can't handle C++ code/headers, but your C++ compiler can handle the CUDA-related headers just fine.
I essentially have main.cpp which includes my program_options header, and the parsing stuff dictating what to do with the options. The program_options header then includes the CUDA-related headers/class prototypes. The important part (as I think you've seen) is to just not have the CUDA code and accompanying headers include that options header. Pass your objects to an options function and have that fill in relevant info. Something like an ugly version of a Strategy Pattern. Concatenated:
main.cpp:
#include "myprogramoptionsparser.hpp"
(...)
CudaObject* MyCudaObj = new CudaObject;
GetCommandLineOptions(argc,argv,MyCudaObj);
myprogramoptionsparser.hpp:
#include <boost/program_options.hpp>
#include "CudaObject.hpp"
void GetCommandLineOptions(int argc,char **argv,CudaObject* obj){
(do stuff to cuda object) }
CudaObject.hpp:
(do not include myprogramoptionsparser.hpp)
CudaObject.cu:
#include "CudaObject.hpp"
It can be a bit annoying, but the nvcc compiler seems to be getting better at handling more C++ code. This has worked fine for me in VC2008/2010, and linux/g++.
You have to split the code in two parts:
the kernel have to be compiled by nvcc
the program that invokes the kernel has to be compiled by g++.
Then link the two objects together and everything should be working.
nvcc is required only to compile the CUDA kernel code.
Thanks to #ronag's comment I realised I was still (indirectly) including boost/program_options.hpp indirectly in my header since I had some member variables in my wrapper class definition which needed it.
To get around this I moved these variables outside the class and thus could move them outside the class defintion and into the .cpp file. They are no longer member variables and now global inside wrapper.cpp
This seems to work but it is ugly and I have the feeling nvcc should handle this gracefully; if anybody else has a proper solution please still post it :)
Another option is to wrap cpp only code in
#ifndef __CUDACC__
Related
I'm trying to integrate CUDA to an existing aplication wich uses boost::spirit.
Isolating the problem, I've found out that the following code does not copile with nvcc:
main.cu:
#include <boost/spirit/include/qi.hpp>
int main(){
exit(0);
}
Compiling with nvcc -o cudaTest main.cu I get a lot of errors that can be seen here.
But if I change the filename to main.cpp, and compile again using nvcc, it works. What is happening here and how can I fix it?
nvcc sometimes has trouble compiling complex template code such as is found in Boost, even if the code is only used in __host__ functions.
When a file's extension is .cpp, nvcc performs no parsing itself and instead forwards the code to the host compiler, which is why you observe different behavior depending on the file extension.
If possible, try to quarantine code which depends on Boost into .cpp files which needn't be parsed by nvcc.
I'd also make sure to try the nvcc which ships with the recent CUDA 4.1. nvcc's template support improves with each release.
I am a beginner was trying to do some C++ programming on Xcode. It works fine, but when I try to compile the same c++ file on my windows pc using VS, there were some errors. After I look at my code closely, there are really some stupid mistakes that I have made which caused the errors, but Xcode seemed to have ignored them...
My question is that is there any setting that I need to change to prevent Xcode from being so smart?
For example, the following code can actually compile in xcode:
#include <iostream>
using namespace std;
int main() {
if (true or false){
cout << "How is this possible? \n";
}
return 0;
}
There are also other cases where the code is actually wrong, but it can compile just fine is Xcode which is the annoying part and I want to disable that.
As far as I can see there is nothing wrong with your code.
The ISO C++ standard does not specify which standard headers are included by other standard headers. So, it is entirely possible that the version of iostream used by Xcode directly or indirectly includes ciso646. Whereas Visual Studio's version of iostream does not include ciso646. There are many similar cases with other headers. You just need to read the error messages and realize that your error (when you move your file to a different platform) is due to a missing header file.
It would be nice if writing portable code meant writing code in accordance with the C++ standard specification, but unfortunately that's not the case. Although there are various compiler options on various implementations which can help bring different implementations closer together, in general you will just have to bring the code into the target environment and actually test it there.
So ultimately writing portable code means you'll have to learn some subset of C++ that is accepted by all the implementations you want to target.
or is an 'alternative token' in C++, and VS is incorrect to reject it. There's no option in Xcode to disable support for alternative tokens. However VS has non-standard support for or as a macro using the header <ciso646>, and Xcode does have a header <ciso646> which does nothing (as the standard specifies). So you can write code which uses or and which works in both Xcode and VS by including this header.
#include <iostream>
#include <ciso646> // does nothing in Xcode, allows `or` in VS
using namespace std;
int main() {
if (true or false){
cout << "How is this possible? \n";
}
return 0;
}
Unfortunately VS can't support all of the alternative tokens through macros and so Xcode will still support some that VS doesn't.
There are also other cases where the code is actually wrong, but it can compile just fine is Xcode which is the annoying part and I want to disable that.
If you give specific examples then I can provide additional advice on how to write portable code.
Rather than changing your Xcode settings, I suggest cross-checking your code using another development environment.
If you're looking for something cheap and full-proof. Download a VirtualBox Windows VM, and run download Dev C++ (bloodhshed)
VS does not support or: you need to use || instead.
You can include some special files but it doesn't inject or sufficiently well into the language for it to work in all instances.
If you want to suppress use of or (and your compiler supports no better way)
#define it to something that emits a compiler error, for example
#define or OR
This at least means that the nature of the compilation errors will be identical on Xcode and VC.
I've been searching for a workaround for days. So far no luck.
What I use:
STM32F103VET6
J-Link
RealView MDK-ARM v4.12
Both C and C++ code in my program
Before I included STL in my code, everything works fine. I can retarget printf() and scanf() to USART without a problem. This is done by including Retarget.c that came with RV-MDK and writing my own sendchar() and getkey(). Retarget.c has a line that says #pragma import(__use_no_semihosting_swi) which demands the linker to use the retargeting version of <cstdio>, instead invoking the semihosting implementation.
But when I started to #include <deque> and deque<int> buffer;, it doesn't work anymore. It seems that the linker had decided that <deque> depends on the the semihosting <cstdio>, which conflicts with the retargeting code I wrote just now.
What I tried and failed:
Get rid of __no_semihosting_swi and instead try to override _sys_open(). It doesn't work because the semihosting <cstdio> depends on an object file which already exports this symbol.
Use $super$$ and $sub$$ syntax suggested here (I think I've tried everything they provided!)
Use "MicroLIB". It's EPIC FAIL.
Digging into the supplied STL headers. I found NO reference to any stdio. And, since I can #include <cstdio> and still have retargeting, I don't think even an reference would make any difference.
I suspect the solution has something to do with directing the linker, but so far I haven't figured out... This is getting so frustrating!
I'm trying to integrate CUDA to an existing aplication wich uses boost::spirit.
Isolating the problem, I've found out that the following code does not copile with nvcc:
main.cu:
#include <boost/spirit/include/qi.hpp>
int main(){
exit(0);
}
Compiling with nvcc -o cudaTest main.cu I get a lot of errors that can be seen here.
But if I change the filename to main.cpp, and compile again using nvcc, it works. What is happening here and how can I fix it?
nvcc sometimes has trouble compiling complex template code such as is found in Boost, even if the code is only used in __host__ functions.
When a file's extension is .cpp, nvcc performs no parsing itself and instead forwards the code to the host compiler, which is why you observe different behavior depending on the file extension.
If possible, try to quarantine code which depends on Boost into .cpp files which needn't be parsed by nvcc.
I'd also make sure to try the nvcc which ships with the recent CUDA 4.1. nvcc's template support improves with each release.
I have the following very simple application that compiles and runs fine:
EDIT: changed the example to be simpilar to end confusion of the real issue
int main() {
return 0;
}
As soon as I add #include <string> (and not even reference std::string), it fails to compile and I get the following error:
/usr/include/c++/4.1.2/bits/allocator.h:82 error: expected template-name before '<' token
Along with about 456 other, similar errors.
Any ideas? Thanks!
UPDATE:
Line 82 of /usr/include/c++/4.1.2/bits/allocator.h references the template __glibcxx_base_allocator at the location of the error. That template is defined in bits/c++allocator.h. When I search the system for that file, I get 3 hits, but none of them are in /usr/include/c++/4.1.2/bits/ as one would expect.
I have version 3.1.6, 4.1.1, and 4.3.2, but not 4.1.2 as the rest of the includes I am using. I am not sure which one is being used (if any, however, I don't get any error for an unknown file), but it seems the problem may stem from this.
The problem appears to be the installed development packages are not correct or incomplete (not to be confused with corrupt). Forcing g++ to use different include versions corrects that:
g++ -nostdic++ hello.cc -o hello -I/usr/include/c++/3.4.6
All the alternative directories (4.1.1, 4.1.2 and 4.3.2) are incomplete causing inappropriate files to be included causing the unusually errors. For example:
/usr/include/c++/4.1.2/bits/allocator.h requires __glibcxx_base_allocator located in bits/c++allocator.h which is being included from either /usr/include/c++/4.1.1 or /usr/include/c++/4.3.2 and appear to be incompatible. Forcing the compiler to use the only complete set of includes rectifies this.
Almost certainly g++ is detecting .cc as a C source file, not C++ and passes it through to gcc instead of compiling as C++. You can easily test by renaming your file to hello.C. There's also a language parameter to g++ you can use.
EDIT: This seems to work fine in g++ 4.2 with a .cc extension so that might not be it. Do you have any other headers included you aren't showing us? They could be interfering with <string>.
EDIT2: Alternatively your headers might not be set up right. Does this work:
#include <string>
int main()
{
return 0;
}
Errors like this have been heard of to occur when the C++ standard library headers are corrupted/not fully installed – maybe there is even a message referring to a missing include among your 456 other errors.
In any case, make sure that libstdc++-devel, resp. the package containing the C++ standard library header files of your distribution, is properly installed.
Check your include path. The paths can be specified as environment variables or specified on the command line. You could be using an include file from a different compiler or different version of the same compiler.
Also, try using <cstdio> rather than <stdio.h>.
Another suggestion: change <> to "".
This could be error caused at preprocess stage. Just preprocess your cpp file by passing flag -E to gcc and Look at the place the compiler complains.