Compilation Error When Building OpenCV 3.0-alpha on Linux (Bug #3894)

Added by Ryan Forbes over 10 years ago. Updated over 10 years ago.

Status:Cancelled Start date:2014-09-03
Priority:Normal Due date:
Assignee:Ryan Forbes % Done:

0%
Category:build/install
Target version:3.0-alpha
Affected version:branch 'master' (3.0-dev) Operating System:Linux
Difficulty: HW Platform:x64
Pull request:

Description

When trying to do "make" on a 64-bit SUSE Linux Enterprise Server 11.1 OS w/, I see the following compilation error.

@@[ 12%] Building NVCC (Device) object modules/core/CMakeFiles/cuda_compile.dir/src/cuda/./cuda_compile_generated_gpu_mat.cu.o
/recondev/vol3/forbes/opencv-master/modules/cudev/include/opencv2/cudev/ptr2d/texture.hpp(187): error: identifier "cudaTextureObject_t" is undefined

1 error detected in the compilation of "/tmp/tmpxft_00000911_00000000-12_gpu_mat.compute_30.cpp1.ii".
CMake Error at cuda_compile_generated_gpu_mat.cu.o.cmake:266 (message):
Error generating file
/recondev/vol3/forbes/opencv-master/release/modules/core/CMakeFiles/cuda_compile.dir/src/cuda/./cuda_compile_generated_gpu_mat.cu.o

make2: * [modules/core/CMakeFiles/cuda_compile.dir/src/cuda/./cuda_compile_generated_gpu_mat.cu.o] Error 1
make1: [modules/core/CMakeFiles/opencv_core.dir/all] Error 2
make: ** [all] Error 2@

It should be noted that this system has 6 nVidia M2075 GPUs. I saw on a post to add the -D CUDA_ARCH_BIN="3.0" to cmake. This is the cmake command being run:

cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D CUDA_ARCH_BIN="3.0" ..

However after adding that, the following error now shows up:

[ 12%] Building NVCC (Device) object modules/core/CMakeFiles/cuda_compile.dir/src/cuda/./cuda_compile_generated_gpu_mat.cu.o
/recondev/vol3/forbes/opencv-master/modules/cudev/include/opencv2/cudev/ptr2d/texture.hpp(187): error: identifier "cudaTextureObject_t" is undefined

1 error detected in the compilation of "/tmp/tmpxft_00000911_00000000-12_gpu_mat.compute_30.cpp1.ii".
CMake Error at cuda_compile_generated_gpu_mat.cu.o.cmake:266 (message):
Error generating file
/recondev/vol3/forbes/opencv-master/release/modules/core/CMakeFiles/cuda_compile.dir/src/cuda/./cuda_compile_generated_gpu_mat.cu.o

make2: * [modules/core/CMakeFiles/cuda_compile.dir/src/cuda/./cuda_compile_generated_gpu_mat.cu.o] Error 1
make1: [modules/core/CMakeFiles/opencv_core.dir/all] Error 2
make: ** [all] Error 2@

I was not able to find information on this error anywhwere. I should mention that I am able to build opencv2.4.9 on this same system.

History

#1
Updated by Steven Puttemans over 10 years ago

Hi Ryan, since you are building for that card, the CUDA GPU doc (https://developer.nvidia.com/cuda-gpus) clearly states you have compute capability 2.0. This means you are trying to build advanced cuda kernels for your system. That simply won't work.

Try changing your command to

cmake -D CMAKE_BUILD_TYPE=RELEASE -D CMAKE_INSTALL_PREFIX=/usr/local -D CUDA_ARCH_BIN="2.0" ..

Also are you aware you are not building the extra modules? This is a big change in openCV 3.0 alpha. I also suggest people to grab the latest 3.0 master branch from github. Me and some others have fixed several problems in the alpha build.

  • Status changed from New to Cancelled

#2
Updated by Steven Puttemans over 10 years ago

  • Assignee changed from Roman Donchenko to Ryan Forbes

Also available in: Atom PDF