Nvcc cannot compile the math_functions.h shipped along with it

Accelerated Computing CUDA CUDA Programming and Performance
1

I am using gcc 14.2.1 on my Arch linux machine, with CMake version 3.30. After I upgrading cuda toolkit to 12.8 I cannot use CMake with cuda anymore.

In a simple example, CMakeLists like this

cmake_minimum_required(VERSION 3.30)
project(demo)
enable_language(CXX CUDA)
add_executable(demo demo.cu)

and demo.cu

#include <cuda_runtime.h>
#include <cstdio>
int main() {
    printf("hello!\n");
}

the cmake config step failed with the following error message:

CMake Error at /usr/share/cmake/Modules/CMakeDetermineCompilerId.cmake:865 (message):
  Compiling the CUDA compiler identification source file
  "CMakeCUDACompilerId.cu" failed.

  Compiler: /opt/cuda/bin/nvcc

  Build flags:

  Id flags: --keep;--keep-dir;tmp -v



  The output was:

  1

  nvcc warning : Support for offline compilation for architectures prior to
  '<compute/sm/lto>_75' will be removed in a future release (Use
  -Wno-deprecated-gpu-targets to suppress warning).

  #$ _NVVM_BRANCH_=nvvm

  #$ _SPACE_=

  #$ _CUDART_=cudart

  #$ _HERE_=/opt/cuda/bin

  #$ _THERE_=/opt/cuda/bin

  #$ _TARGET_SIZE_=

  #$ _TARGET_DIR_=

  #$ _TARGET_DIR_=targets/x86_64-linux

  #$ TOP=/opt/cuda/bin/..

  #$ CICC_PATH=/opt/cuda/bin/../nvvm/bin

  #$ NVVMIR_LIBRARY_DIR=/opt/cuda/bin/../nvvm/libdevice

  #$ LD_LIBRARY_PATH=/opt/cuda/bin/../lib:

  #$
  PATH=/opt/cuda/bin/../nvvm/bin:/opt/cuda/bin:/opt/anaconda/bin:/opt/anaconda/condabin:/home/creeper/.cargo/bin:/home/creeper/.nix-profile/bin:/nix/var/nix/profiles/default/bin:/usr/local/sbin:/usr/local/bin:/usr/bin:/opt/anaconda/bin:/opt/cuda/bin:/opt/cuda/nsight_compute:/opt/cuda/nsight_systems/bin:/var/lib/flatpak/exports/bin:/usr/lib/jvm/default/bin:/usr/bin/site_perl:/usr/bin/vendor_perl:/usr/bin/core_perl:/home/creeper/.local/share/JetBrains/Toolbox/scripts:/home/creeper/nsight-systems-2024.6.1/bin:/usr/local/NVIDIA-Nsight-Compute-2024.3


  #$ INCLUDES="-I/opt/cuda/bin/../targets/x86_64-linux/include"

  #$ LIBRARIES= "-L/opt/cuda/bin/../targets/x86_64-linux/lib/stubs"
  "-L/opt/cuda/bin/../targets/x86_64-linux/lib"

  #$ CUDAFE_FLAGS=

  #$ PTXAS_FLAGS=

  #$ rm tmp/a_dlink.reg.c

  #$ "/usr/bin"/g++-13 -D__CUDA_ARCH_LIST__=520 -D__NV_LEGACY_LAUNCH -E -x
  c++ -D__CUDACC__ -D__NVCC__
  "-I/opt/cuda/bin/../targets/x86_64-linux/include" -D__CUDACC_VER_MAJOR__=12
  -D__CUDACC_VER_MINOR__=8 -D__CUDACC_VER_BUILD__=61
  -D__CUDA_API_VER_MAJOR__=12 -D__CUDA_API_VER_MINOR__=8
  -D__NVCC_DIAG_PRAGMA_SUPPORT__=1 -D__CUDACC_DEVICE_ATOMIC_BUILTINS__=1
  -include "cuda_runtime.h" -m64 "CMakeCUDACompilerId.cu" -o
  "tmp/CMakeCUDACompilerId.cpp4.ii"

  #$ cudafe++ --c++17 --gnu_version=130301 --display_error_number
  --orig_src_file_name "CMakeCUDACompilerId.cu" --orig_src_path_name
  "/home/creeper/demo/build/CMakeFiles/3.31.2/CompilerIdCUDA/CMakeCUDACompilerId.cu"
  --allow_managed --m64 --parse_templates --gen_c_file_name
  "tmp/CMakeCUDACompilerId.cudafe1.cpp" --stub_file_name
  "CMakeCUDACompilerId.cudafe1.stub.c" --gen_module_id_file
  --module_id_file_name "tmp/CMakeCUDACompilerId.module_id"
  "tmp/CMakeCUDACompilerId.cpp4.ii"

  #$ "/usr/bin"/g++-13 -D__CUDA_ARCH__=520 -D__CUDA_ARCH_LIST__=520
  -D__NV_LEGACY_LAUNCH -E -x c++ -DCUDA_DOUBLE_MATH_FUNCTIONS -D__CUDACC__
  -D__NVCC__ "-I/opt/cuda/bin/../targets/x86_64-linux/include"
  -D__CUDACC_VER_MAJOR__=12 -D__CUDACC_VER_MINOR__=8
  -D__CUDACC_VER_BUILD__=61 -D__CUDA_API_VER_MAJOR__=12
  -D__CUDA_API_VER_MINOR__=8 -D__NVCC_DIAG_PRAGMA_SUPPORT__=1
  -D__CUDACC_DEVICE_ATOMIC_BUILTINS__=1 -include "cuda_runtime.h" -m64
  "CMakeCUDACompilerId.cu" -o "tmp/CMakeCUDACompilerId.cpp1.ii"

  #$ "$CICC_PATH/cicc" --c++17 --gnu_version=130301 --display_error_number
  --orig_src_file_name "CMakeCUDACompilerId.cu" --orig_src_path_name
  "/home/creeper/demo/build/CMakeFiles/3.31.2/CompilerIdCUDA/CMakeCUDACompilerId.cu"
  --allow_managed -arch compute_52 -m64 --no-version-ident -ftz=0 -prec_div=1
  -prec_sqrt=1 -fmad=1 --include_file_name "CMakeCUDACompilerId.fatbin.c"
  -tused --module_id_file_name "tmp/CMakeCUDACompilerId.module_id"
  --gen_c_file_name "tmp/CMakeCUDACompilerId.cudafe1.c" --stub_file_name
  "tmp/CMakeCUDACompilerId.cudafe1.stub.c" --gen_device_file_name
  "tmp/CMakeCUDACompilerId.cudafe1.gpu" "tmp/CMakeCUDACompilerId.cpp1.ii" -o
  "tmp/CMakeCUDACompilerId.ptx"

  #$ ptxas -arch=sm_52 -m64 "tmp/CMakeCUDACompilerId.ptx" -o
  "tmp/CMakeCUDACompilerId.sm_52.cubin"

  #$ fatbinary --create="tmp/CMakeCUDACompilerId.fatbin" -64
  --cicc-cmdline="-ftz=0 -prec_div=1 -prec_sqrt=1 -fmad=1 "
  "--image3=kind=elf,sm=52,file=tmp/CMakeCUDACompilerId.sm_52.cubin"
  "--image3=kind=ptx,sm=52,file=tmp/CMakeCUDACompilerId.ptx"
  --embedded-fatbin="tmp/CMakeCUDACompilerId.fatbin.c"

  #$ "/usr/bin"/g++-13 -D__CUDA_ARCH__=520 -D__CUDA_ARCH_LIST__=520
  -D__NV_LEGACY_LAUNCH -c -x c++ -DCUDA_DOUBLE_MATH_FUNCTIONS -Wno-psabi
  "-I/opt/cuda/bin/../targets/x86_64-linux/include" -m64
  "tmp/CMakeCUDACompilerId.cudafe1.cpp" -o "tmp/CMakeCUDACompilerId.o"

  In file included from
  /opt/cuda/targets/x86_64-linux/include/crt/math_functions.h:5978,

                   from /opt/cuda/targets/x86_64-linux/include/crt/common_functions.h:303,
                   from /opt/cuda/bin/../targets/x86_64-linux/include/crt/host_runtime.h:285,
                   from tmp/CMakeCUDACompilerId.cudafe1.stub.c:8,
                   from CMakeCUDACompilerId.cudafe1.stub.c:1:

  /opt/cuda/targets/x86_64-linux/include/crt/math_functions.h:5990:1: error:
  declaration of ‘double sinpi(double)’ has a different exception
  specifier

   5990 | __func__(double sinpi(double a));
        | ^~~~~~~~

  
  /opt/cuda/bin/../targets/x86_64-linux/include/crt/math_functions.h:2556:15:
  note: from previous declaration ‘double sinpi(double) noexcept’

   2556 | extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double                 sinpi(double x) noexcept (true);
        |               ^~~~~

  /opt/cuda/targets/x86_64-linux/include/crt/math_functions.h:5992:1: error:
  declaration of ‘double cospi(double)’ has a different exception
  specifier

   5992 | __func__(double cospi(double a));
        | ^~~~~~~~

  
  /opt/cuda/bin/../targets/x86_64-linux/include/crt/math_functions.h:2601:15:
  note: from previous declaration ‘double cospi(double) noexcept’

   2601 | extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ double                 cospi(double x) noexcept (true);
        |               ^~~~~

  /opt/cuda/targets/x86_64-linux/include/crt/math_functions.h:6010:1: error:
  declaration of ‘float sinpif(float)’ has a different exception
  specifier

   6010 | __func__(float sinpif(float a));
        | ^~~~~~~~

  
  /opt/cuda/bin/../targets/x86_64-linux/include/crt/math_functions.h:2579:14:
  note: from previous declaration ‘float sinpif(float) noexcept’

   2579 | extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float                  sinpif(float x) noexcept (true);
        |              ^~~~~~

  /opt/cuda/targets/x86_64-linux/include/crt/math_functions.h:6012:1: error:
  declaration of ‘float cospif(float)’ has a different exception
  specifier

   6012 | __func__(float cospif(float a));
        | ^~~~~~~~

  
  /opt/cuda/bin/../targets/x86_64-linux/include/crt/math_functions.h:2623:14:
  note: from previous declaration ‘float cospif(float) noexcept’

   2623 | extern __DEVICE_FUNCTIONS_DECL__ __device_builtin__ float                  cospif(float x) noexcept (true);
        |              ^~~~~~

  # --error 0x1 --





Call Stack (most recent call first):
  /usr/share/cmake/Modules/CMakeDetermineCompilerId.cmake:8 (CMAKE_DETERMINE_COMPILER_ID_BUILD)
  /usr/share/cmake/Modules/CMakeDetermineCompilerId.cmake:53 (__determine_compiler_id_test)
  /usr/share/cmake/Modules/CMakeDetermineCUDACompiler.cmake:131 (CMAKE_DETERMINE_COMPILER_ID)
  CMakeLists.txt:3 (enable_language)

From the description, I guessed that it was because math_functions.h included by CMakeCUDACompilerId.cu failed to compile. So I reinstalled cuda 12.8 several times using sudo pacman -S cuda but this cannot solve the problem. What might be the cause and how to solve it?

2

this may be of interest

This is a toolchain problem. In general in such situations the recommendation I give is to adhere exactly to the support table in the CUDA version you are using. The first thing I note there is that there is no line for Arch Linux.