error: cuda_runtime.h: No such file or directory

Question

How can I force gcc to look in /usr/cuda/local/include for cuda_runtime.h?

I'm attempting to compile a CUDA application with a C wrapper. I'm running Ubuntu 10.04.

I've successfully compiled my CUDA application into a .so with the following command:

nvcc -arch=sm_11 -o libtest.so --shared -Xcompiler -fPIC main.cu

When I try and compile my c wrapper file with the following command:

gcc -std=c99 -o main -L. -ltest main.c

I receive the error:

error: cuda_runtime.h: No such file or directory

I've verified that cuda_runtime.h is in fact present in /usr/local/cuda/include

I assume you mean /usr/local/cuda/include. Did you try adding a -I switch to your gcc command line? For example: gcc -std=c99 -I/usr/local/cuda/include -o main -L. -ltest main.c — Robert Crovella, Oct 31 '12 at 21:22
If you wouldn't mind, please post it as an answer to your question. Then accept it or I will upvote the answer. Thank you. — Robert Crovella, Oct 31 '12 at 21:38
Is there any option to make it work without adding -l? A library gives me this error — Keloo, Sep 06 '16 at 10:45

score 16 · Answer 1 · answered Apr 13 '17 at 10:00

16

If you are using CMake

find_package(CUDA  REQUIRED)
include_directories("${CUDA_INCLUDE_DIRS}")

answered Apr 13 '17 at 10:00

Alex Punnen

5,287
3
59
71

This is depecrated in modern CMake, see my answer https://stackoverflow.com/a/75559127/1150303, below – DomTomCat Feb 28 '23 at 07:43

score 12 · Answer 2 · edited Nov 02 '12 at 12:10

12

Using an -I switch allowed gcc to find the cuda_runtime.h file:

gcc -std=c99 -I/usr/local/cuda/include -o main -L. -ltest main.c

edited Nov 02 '12 at 12:10

talonmies

70,661
34
192
269

answered Oct 31 '12 at 21:45

skrieder

379
1
2
8

score 10 · Answer 3 · answered Nov 22 '19 at 14:17

We were using CMake but it still wasn't able to find the header files (maybe it is the CMake version that couldn't find the directory ./targets/x86_64-linux/include or because we have multiple CUDA versions). Setting CPATH and LD_LIBRARY_PATH fixed it for us:

export CPATH=/usr/local/cuda-10.1/targets/x86_64-linux/include:$CPATH
export LD_LIBRARY_PATH=/usr/local/cuda-10.1/targets/x86_64-linux/lib:$LD_LIBRARY_PATH
export PATH=/usr/local/cuda-10.1/bin:$PATH

This fixed the issue and can be an acceptable solution. – SKPS Jan 19 '23 at 00:01 — SKPS, Jan 19 '23 at 00:01

score 0 · Answer 4 · edited Jul 06 '18 at 05:48

0

Add -isystem /usr/local/cuda-8.0/include to CXX_INCLUDES in flags.make file.

edited Jul 06 '18 at 05:48

geisterfurz007

5,292
5
33
54

answered Jul 06 '18 at 05:46

Vidhu

1
1

DomTomCat · Answer 5 · 2023-02-28T07:44:09.427

Using modern CMake https://cliutils.gitlab.io/modern-cmake/chapters/packages/CUDA.html - CUDA has become a CMake language just like CXX:

check_language(CUDA)
if(CMAKE_CUDA_COMPILER)
    enable_language(CUDA)

    # to tell your code it was found via #ifdef USE_CUDA:
    add_definitions(-DUSE_CUDA)    

    include_directories("${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES}")
endif()

and after definition of executable or library don't forget cuda arch version(s)

set_target_properties(${PROJECT_NAME} PROPERTIES CUDA_ARCHITECTURES "35;50;72")

error: cuda_runtime.h: No such file or directory

5 Answers5