nvcc fatal : Unsupported gpu architecture 'compute_30'

Question

I am using this GitHub repo https://github.com/davidstutz/mesh-fusion and I am following its instructions. The error I get after cmake .. is as follows. How should I fix it?

Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed

The error in CMakeError.log is:

/usr/bin/ld: cannot find -lpthreads

I followed this answer: https://stackoverflow.com/a/31948969/2414957

I do have the followings:

$ ls /usr/lib/x86_64-linux-gnu/libpthread.so
lrwxrwxrwx 1 root root 37 Aug 17 16:02 /usr/lib/x86_64-linux-gnu/libpthread.so -> /lib/x86_64-linux-gnu/libpthread.so.0

$ nm /lib/x86_64-linux-gnu/libpthread.so.0 | grep "pthread_create"
00000000000098d0 t __pthread_create_2_1
0000000000007af7 t __pthread_create_2_1.cold
00000000000098d0 T pthread_create@@GLIBC_2.2.5

I have:

[3645:3637 0:2047] 10:00:46 Tue Jan 12 [mona@goku:pts/0 +1] ~/mesh-fusion/libfusiongpu/build
$ cmake ..
CMake Warning (dev) in CMakeLists.txt:
  No project() command is present.  The top-level CMakeLists.txt file must
  contain a literal, direct call to the project() command.  Add a line of
  code such as

    project(ProjectName)

  near the top of the file, but after cmake_minimum_required().

  CMake is pretending there is a "project(Project)" command on the first
  line.
This warning is for project developers.  Use -Wno-dev to suppress it.

-- The C compiler identification is GNU 9.3.0
-- The CXX compiler identification is GNU 9.3.0
-- 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
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- 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-11.2 (found suitable version "11.2", minimum required is "6.5") 
-- Configuring done
-- Generating done
-- Build files have been written to: /home/mona/mesh-fusion/libfusiongpu/build

The CMakeError.log is:

[3645:3637 0:2051] 10:03:13 Tue Jan 12 [mona@goku:pts/0 +1] ~/mesh-fusion/libfusiongpu/build/CMakeFiles
$ cat CMakeError.log 
Performing C SOURCE FILE Test CMAKE_HAVE_LIBC_PTHREAD failed with the following output:
Change Dir: /home/mona/mesh-fusion/libfusiongpu/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/make cmTC_d8db1/fast && /usr/bin/make -f CMakeFiles/cmTC_d8db1.dir/build.make CMakeFiles/cmTC_d8db1.dir/build
make[1]: Entering directory '/home/mona/mesh-fusion/libfusiongpu/build/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_d8db1.dir/src.c.o
/usr/bin/cc   -DCMAKE_HAVE_LIBC_PTHREAD   -o CMakeFiles/cmTC_d8db1.dir/src.c.o   -c /home/mona/mesh-fusion/libfusiongpu/build/CMakeFiles/CMakeTmp/src.c
Linking C executable cmTC_d8db1
/usr/bin/cmake -E cmake_link_script CMakeFiles/cmTC_d8db1.dir/link.txt --verbose=1
/usr/bin/cc  -DCMAKE_HAVE_LIBC_PTHREAD    -rdynamic CMakeFiles/cmTC_d8db1.dir/src.c.o  -o cmTC_d8db1 
/usr/bin/ld: CMakeFiles/cmTC_d8db1.dir/src.c.o: in function `main':
src.c:(.text+0x46): undefined reference to `pthread_create'
/usr/bin/ld: src.c:(.text+0x52): undefined reference to `pthread_detach'
/usr/bin/ld: src.c:(.text+0x63): undefined reference to `pthread_join'
collect2: error: ld returned 1 exit status
make[1]: *** [CMakeFiles/cmTC_d8db1.dir/build.make:87: cmTC_d8db1] Error 1
make[1]: Leaving directory '/home/mona/mesh-fusion/libfusiongpu/build/CMakeFiles/CMakeTmp'
make: *** [Makefile:121: cmTC_d8db1/fast] Error 2


Source file was:
#include <pthread.h>

void* test_func(void* data)
{
  return data;
}

int main(void)
{
  pthread_t thread;
  pthread_create(&thread, NULL, test_func, NULL);
  pthread_detach(thread);
  pthread_join(thread, NULL);
  pthread_atfork(NULL, NULL, NULL);
  pthread_exit(NULL);

  return 0;
}

Determining if the function pthread_create exists in the pthreads failed with the following output:
Change Dir: /home/mona/mesh-fusion/libfusiongpu/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/make cmTC_4d856/fast && /usr/bin/make -f CMakeFiles/cmTC_4d856.dir/build.make CMakeFiles/cmTC_4d856.dir/build
make[1]: Entering directory '/home/mona/mesh-fusion/libfusiongpu/build/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_4d856.dir/CheckFunctionExists.c.o
/usr/bin/cc   -DCHECK_FUNCTION_EXISTS=pthread_create   -o CMakeFiles/cmTC_4d856.dir/CheckFunctionExists.c.o   -c /usr/share/cmake-3.16/Modules/CheckFunctionExists.c
Linking C executable cmTC_4d856
/usr/bin/cmake -E cmake_link_script CMakeFiles/cmTC_4d856.dir/link.txt --verbose=1
/usr/bin/cc  -DCHECK_FUNCTION_EXISTS=pthread_create    -rdynamic CMakeFiles/cmTC_4d856.dir/CheckFunctionExists.c.o  -o cmTC_4d856  -lpthreads 
/usr/bin/ld: cannot find -lpthreads
collect2: error: ld returned 1 exit status
make[1]: *** [CMakeFiles/cmTC_4d856.dir/build.make:87: cmTC_4d856] Error 1
make[1]: Leaving directory '/home/mona/mesh-fusion/libfusiongpu/build/CMakeFiles/CMakeTmp'
make: *** [Makefile:121: cmTC_4d856/fast] Error 2

I have:

$ lsb_release -a
LSB Version:    core-11.1.0ubuntu2-noarch:security-11.1.0ubuntu2-noarch
Distributor ID: Ubuntu
Description:    Ubuntu 20.04.1 LTS
Release:    20.04
Codename:   focal

Here's the CMakeLists.txt file (I added the first two lines):

cmake_minimum_required (VERSION 2.8.7) ## added by mona
find_package(Threads) ## added by mona

cmake_minimum_required(VERSION 2.8)
set(CMAKE_CXX_STANDARD 11)

# set(CMAKE_BUILD_TYPE Debug)
set(CMAKE_BUILD_TYPE Release)
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -msse -msse2 -msse3 -msse4.2 -fPIC")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -msse -msse2 -msse3 -msse4.2 -fPIC")

find_package(CUDA 6.5 REQUIRED)
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS};-std=c++11")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -D_FORCE_INLINES -Wall")

set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}; -gencode=arch=compute_30,code=sm_30")
set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}; -gencode=arch=compute_30,code=compute_30")

set(FUSION_GPU_SRC
  fusion.cu
  fusion_zach_tvl1.cu
)

cuda_add_library(fusion_gpu SHARED ${FUSION_GPU_SRC})
target_link_libraries(fusion_gpu ${CUDA_LIBRARIES})

Later, if I run the make command I'll get this error:

[3645:3637 0:2070] 10:33:37 Tue Jan 12 [mona@goku:pts/0 +1] ~/mesh-fusion/libfusiongpu/build
$ make
[ 33%] Building NVCC (Device) object CMakeFiles/fusion_gpu.dir/fusion_gpu_generated_fusion_zach_tvl1.cu.o
nvcc fatal   : Unsupported gpu architecture 'compute_30'
CMake Error at fusion_gpu_generated_fusion_zach_tvl1.cu.o.Release.cmake:220 (message):
  Error generating
  /home/mona/mesh-fusion/libfusiongpu/build/CMakeFiles/fusion_gpu.dir//./fusion_gpu_generated_fusion_zach_tvl1.cu.o


make[2]: *** [CMakeFiles/fusion_gpu.dir/build.make:72: CMakeFiles/fusion_gpu.dir/fusion_gpu_generated_fusion_zach_tvl1.cu.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:75: CMakeFiles/fusion_gpu.dir/all] Error 2
make: *** [Makefile:84: all] Error 2

As you have found, `CMAKE_HAVE_LIBC_PTHREAD` has nothing common with your problem. Could you edit your question post and its title for remove parts, unrelated to your problem? About the line `-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed` see that question: https://stackoverflow.com/q/64514666/3440745. — Tsyvarev, Jan 13 '21 at 08:26
Thank you for your suggestion. I went ahead and changed the title. — Mona Jalal, Jan 14 '21 at 01:14

score 3 · Accepted Answer · answered Jan 13 '21 at 03:53

I used lines 77-83 of this code: https://github.com/baidu-research/warp-ctc/blob/master/CMakeLists.txt

in my CMakeLists.txt file:

##set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}; -gencode=arch=compute_30,code=sm_30")
##set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS}; -gencode=arch=compute_30,code=compute_30")

IF ((CUDA_VERSION VERSION_GREATER "10.0") OR (CUDA_VERSION VERSION_EQUAL "10.0"))
    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_75,code=sm_75")
ENDIF()

IF ((CUDA_VERSION VERSION_GREATER "11.0") OR (CUDA_VERSION VERSION_EQUAL "11.0"))
    set(CUDA_NVCC_FLAGS "${CUDA_NVCC_FLAGS} -gencode arch=compute_80,code=sm_80")
ENDIF()

and the make was successful. Apparently, it's OK to ignore the error you see at cmake ..

The reason is the code was old and was using an older version of CUDA and my CUDA version is much newer.

score 0 · Answer 2 · answered Mar 12 '22 at 17:15

I got simialr error:

nvcc fatal : Unsupported gpu architecture 'compute_30'

I solved it by opening file: /path/to/openpose/3rdparty/caffe/cmake/Cuda.cmake

And updating:

set(AMPERE "80 86")

To:

#set(AMPERE "80 86")

This disables compute_80 and compute_86 at the caffe build stage.

To remove further errors, see: https://github.com/BVLC/caffe/issues/6381#issuecomment-386930727

nvcc fatal : Unsupported gpu architecture 'compute_30'

2 Answers2