I have some code that is heavily dependent on Eigen. I would like to optimize it with CUDA, but when I am compiling I get:
[tcai4#golubh4 Try1]$ nvcc conv_parallel.cu -I /home/tcai4/project-cse/Try1 -lfftw3 -o conv.o
In file included from Eigen/Dense:1,
from Eigen/Eigen:1,
from functions.h:8,
from conv_parallel.cu:10:
Eigen/Core:44:34: error: math_functions.hpp: No such file or directory
I think math_functions.hpp is a file from CUDA. Can someone help me figure out why nvcc cannot find it?
edit: I am using CUDA 5.5 and Eigen 3.3, except from linking Eigen and fftw3 library, I did not use any other flags(as you can see from my code).
I encountered this issue while building TensorFlow 1.4.1 with Cuda 9.1, and strangely math_functions.hpp existed only in include/crt.
Creating a symlink from cuda/include/math_functions.hpp to cuda/include/crt/math_functions.hpp fixed the issue:
ln -s /usr/local/cuda/include/crt/math_functions.hpp /usr/local/cuda/include/math_functions.hpp
The reason nvcc cannot find the file in question is because that file is part of the CUDA Math library, which was introduced in CUDA 6. Your almost 4 year old version of CUDA predates the release of the Math library. Your CUDA version doesn't contain said file.
You should, therefore, assume that what you are trying to do cannot work without first updating to a newer version of the CUDA toolkit.
Creating symlink sometimes causes other complication.
You can try replacing
// We need math_functions.hpp to ensure that that EIGEN_USING_STD_MATH macro
// works properly on the device side
#include <math_functions.hpp>
with
// We need cuda_runtime.h to ensure that that EIGEN_USING_STD_MATH macro
// works properly on the device side
#include <cuda_runtime.h>
in
/usr/include/eigen3/Eigen/Core,
which works for me.
The reason why "math_functions.hpp" cannot be found is because "math_functions.hpp" has been renamed to "math_functions.h". So you just need to go to
/usr/include/eigen3/Eigen/Core
and change "math_functions.hpp" to "math_functions.h"
Related
I recently installed include-what-you-use (IWYU) to try it out. I am on version include-what-you-use 0.12 based on clang version 9.0.1-10. (My OS is Ubuntu 20.04 on WSL1)
I tried to do a very simple example on a source code file, but it quickly comes back as an error on the very first include? (Which is just a c standard library file)
After running
include-what-you-use myutils.cpp:
In file included from myutils.cpp:8:
In file included from ./myutils.h:11:
/usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/cfloat:42:10: fatal error: 'float.h' file not found
#include <float.h>
^~~~~~~~~
Why is it error'ing out on a c standard library include?
I cannot post the entire source here, and without extensive testing I am not sure what minimal example would cause this problem.
Am I just mis-using IWYU? I installed it via apt rather than building it and llvm/clang/etc from scratch. Is that fine?
Sort briefing what I am trying to do:
I want to use vnode-lp on my windows PC. I’ve installed MinGW. I’ve installed the necessary dependencies like LAPACK & BLAS libraries as well as Profil/BIAS. The installation of all libraries passed the make and make install process without errors. I hope (!) I’ve managed to install it correctly.
Now the Problem:
Now I’ve tried to get a simple program compiled with basically northing in it just an #include “vnode.h”. First I tried it with Microsoft Visual Studio. Since this gave me several errors I tried to compile it with g++ using MinGW. This gives me the same errors. It starts with
./matrix.w:90:17: error: ‘v_bias’ is not a namespace-name
The Question:
How to include vnodelp into an c++ program and compile it without errors under windows, am I missing something?
Off-topic:
I am trying to get this running for over a week now and don’t know what to do anymore.
C++ is most definitely not C, and packages designed for C++ will never compile and run as C code. So, what you're trying to do really can't be done unless you do some fancy stuff by creating a .dll or something like that, and even then I think you wouldn't get the functionality you want. Why not write your code in C++ and compile with g++ or a similar compiler?
I have managed to finally solve this issue. In case someone has the same problem here is the solution.
I have missed something in the call. Here is the full call that has worked for me:
g++ -o2 –Wall –Wno-deprecated –DNDEBUG – DPROFIL_VNODE – DMAXORDER=50 –I(path to profil bias)/include –I(path to profil bias)/include/BIAS –I(path to profil bias)/src/Base –I(path to vnodelp)/FADBAD++ -I../include –DNDEBUG –c –o (filename).o (filename).cc
g++ -L(path to profil bias)/lib –L(path to lapack)/lib –L../lib –o (filename) (filename).o –lvnode –lProfil –lBias –llr (path to lapack)/lib/liblapack.lib (path to lapack)/lib/libblas.lib –lstd++
funfact:
This also complies with gcc instead of g++
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 was wondering if someone could help me with this.
I've retrieved the source code for the gnuplot-iostream interface from http://www.stahlke.org/dan/gnuplot-iostream/. However, when I attempt to compile the code using the command:
]$ cmake .; make
I get the following compiler error
/.../gnuplot-iostream.h: In constructor ‘Gnuplot::Gnuplot(const std::string&)’:
/.../gnuplot-iostream.h:427: error: ‘never_close_handle’ is not a member of ‘boost::iostreams’
I'm using Scientific Linux 6.2 (kernal 2.6.32-220.23.1.el6.x86_64), g++ 4.4.6, and have boost libraries installed (/usr/include/boost/iostreams/ exists).
Any assistance would be very much appreciated.
D
enum file_descriptor_flags was added in boost::iostreams only in 1.44.0.
enum file_descriptor_flags
{
never_close_handle = 0,
close_handle = 3
};
So, the solution is simply update the boost library (thanks ForEveR).
If however, like me, you do not have the access to update the libraries on your system you should note that I was able to force compilation and obtain basic functionality by simply replacing the two occurrences of boost::iostreams::never_close_handle in the gnuplot-iostream.h file with 0.
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.