nvcc fatal : Value 'sm_20' is not defined for option 'gpu-architecture'

Question

I've looked at many pages and either could not follow what they were saying because they were unclear and/or my knowledge is just not sufficient enough.

I am trying to run:

luarocks install https://raw.githubusercontent.com/qassemoquab/stnbhwd/master/stnbhwd-scm-1.rockspec

So that I may run DenseCap over some images using GPU Acceleration. When I run it, I get this error:

$ luarocks install https://raw.githubusercontent.com/qassemoquab/stnbhwd/master/stnbhwd-scm-1.rockspec
Using https://raw.githubusercontent.com/qassemoquab/stnbhwd/master/stnbhwd-scm-1.rockspec... switching to 'build' mode
Cloning into 'stnbhwd'...
remote: Counting objects: 24, done.
remote: Compressing objects: 100% (23/23), done.
remote: Total 24 (delta 0), reused 14 (delta 0), pack-reused 0
Receiving objects: 100% (24/24), 19.42 KiB | 0 bytes/s, done.
Checking connectivity... done.
cmake -E make_directory build && cd build && cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH="/home/tex/torch/install/bin/.." -DCMAKE_INSTALL_PREFIX="/home/tex/torch/install/lib/luarocks/rocks/stnbhwd/scm-1" && make

-- The C compiler identification is GNU 5.4.0
-- The CXX compiler identification is GNU 5.4.0
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Found Torch7 in /home/tex/torch/install
-- Try OpenMP C flag = [-fopenmp]
-- Performing Test OpenMP_FLAG_DETECTED
-- Performing Test OpenMP_FLAG_DETECTED - Success
-- Try OpenMP CXX flag = [-fopenmp]
-- Performing Test OpenMP_FLAG_DETECTED
-- Performing Test OpenMP_FLAG_DETECTED - Success
-- Found OpenMP: -fopenmp  
-- Compiling with OpenMP support
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Looking for pthread_create
-- Looking for pthread_create - not found
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - found
-- Found Threads: TRUE  
-- Found CUDA: /usr/local/cuda (found suitable version "9.0", minimum required is "5.5") 
-- Configuring done
-- Generating done
-- Build files have been written to: /tmp/luarocks_stnbhwd-scm-1-4197/stnbhwd/build
Scanning dependencies of target stn
[ 25%] Building C object CMakeFiles/stn.dir/init.c.o
[ 50%] Linking C shared module libstn.so
[ 50%] Built target stn
[ 75%] Building NVCC (Device) object CMakeFiles/custn.dir/custn_generated_init.cu.o
nvcc fatal   : Value 'sm_20' is not defined for option 'gpu-architecture'
CMake Error at custn_generated_init.cu.o.cmake:207 (message):
  Error generating
  /tmp/luarocks_stnbhwd-scm-1-4197/stnbhwd/build/CMakeFiles/custn.dir//./custn_generated_init.cu.o


CMakeFiles/custn.dir/build.make:63: recipe for target 'CMakeFiles/custn.dir/custn_generated_init.cu.o' failed
make[2]: *** [CMakeFiles/custn.dir/custn_generated_init.cu.o] Error 1
CMakeFiles/Makefile2:104: recipe for target 'CMakeFiles/custn.dir/all' failed
make[1]: *** [CMakeFiles/custn.dir/all] Error 2
Makefile:127: recipe for target 'all' failed
make: *** [all] Error 2

Error: Build error: Failed building.

I was able to luarocks install cutorch, luarocks install cunn, and luarocks install cudnn just fine.

I am running Ubuntu 16.04 on a GTX 1080ti.

$ nvidia-smi
Tue Dec  5 16:25:42 2017       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 384.90                 Driver Version: 384.90                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 108...  Off  | 00000000:29:00.0  On |                  N/A |
|  0%   47C    P8    16W / 250W |    716MiB / 11169MiB |      1%      Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes:                                                       GPU Memory |
|  GPU       PID   Type   Process name                             Usage      |
|=============================================================================|
|    0      1128      G   /usr/lib/xorg/Xorg                           479MiB |
|    0      1782      G   compiz                                       234MiB |
+-----------------------------------------------------------------------------+

For some reason, when I run $ nvcc -V, I get:

    The program 'nvcc' is currently not installed. You can install it by typing:
sudo apt install nvidia-cuda-toolkit

..and this is another big problem I am facing. When I install nvcc, it installs the toolkit for Cuda 7.5, but I have Cuda 9.0. I installed it using the .deb file that comes from Cuda's website.

$ sudo apt-get install cuda
Reading package lists... Done
Building dependency tree       
Reading state information... Done
cuda is already the newest version (9.0.176-1).
The following packages were automatically installed and are no longer required:
  libcublas7.5 libcudart7.5 libcufft7.5 libcufftw7.5 libcuinj64-7.5
  libcurand7.5 libcusolver7.5 libcusparse7.5 libnppc7.5 libnppi7.5 libnpps7.5
  libnvblas7.5 libnvrtc7.5 libnvtoolsext1 libnvvm3 libthrust-dev libvdpau-dev
  nvidia-cuda-dev nvidia-cuda-doc nvidia-cuda-gdb nvidia-opencl-dev
  nvidia-profiler nvidia-visual-profiler opencl-headers
Use 'sudo apt autoremove' to remove them.
0 upgraded, 0 newly installed, 0 to remove and 222 not upgraded.

I honestly expect that I need to install nvcc, but like I said, it installs for the wrong version, and I cannot for the life of me figure out how to install it for the right version, and Im really confused as to why I am getting this error now even though I already installed cutorch, cudnn, and cunn just fine. Any help is appreciated...

Thanks

Answer 1

Try to change the code architecture (such as sm_20) to some higher version in CMakeLists.txt of stnbhwd that you are trying to install.

From:

IF (CUDA_FOUND)
   LIST(APPEND CUDA_NVCC_FLAGS "-arch=sm_20")

to:

IF (CUDA_FOUND)
   LIST(APPEND CUDA_NVCC_FLAGS "-arch=sm_30")

Answer 2

As you mentioned:

but I have Cuda 9.0

Cuda 9.0 doesn't support SM_20 architecture (more info).

You have quite new gpu:

I am running Ubuntu 16.04 on a GTX 1080ti.

which is based on Pascal architecture (SM_60). So you probably don't need the support for SM_20 architecture.