NVHPC compiler failure when building

[R0n12]

Currently we are testing the installation on TACC Vista. However, the NVHPC compiler doesn't seem to work well with compilation.

using CC=nvc will have the following errors:

-- The C compiler identification is NVHPC 24.7.0
-- The CXX compiler identification is NVHPC 24.7.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /home1/apps/nvidia/Linux_aarch64/24.7/compilers/bin/nvc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /home1/apps/nvidia/Linux_aarch64/24.7/compilers/bin/nvc++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Compiling with C standard: 99
-- Compiling with C++ standard: 98
-- Found OpenMP_C: -mp (found version "5.1")
-- Found OpenMP: TRUE (found version "5.1") found components: C
CMake Warning (dev) at CMakeLists.txt:222 (find_package):
  Policy CMP0146 is not set: The FindCUDA module is removed.  Run "cmake
  --help-policy CMP0146" for policy details.  Use the cmake_policy command to
  set the policy and suppress this warning.

This warning is for project developers.  Use -Wno-dev to suppress it.

-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
-- Found Threads: TRUE
-- Found CUDA: /home1/apps/nvidia/Linux_aarch64/24.7/cuda/12.5 (found version "12.5")
-- Performing Test HAVE_MATH
-- Performing Test HAVE_MATH - Success
-- Configuring done (2.5s)
-- Generating done (0.1s)
CMake Warning:
  Manually-specified variables were not used by the project:

    INSTALL_GTEST


-- Build files have been written to: /work/07936/tg872351/vista/project/zfp-opt/zfp_compression/build
[  2%] Building NVCC (Device) object src/CMakeFiles/zfp.dir/cuda/zfp_generated_interface.cu.o
nvcc fatal   : Unsupported NVHPC compiler found. nvc++ is the only NVHPC compiler that is supported.
CMake Error at zfp_generated_interface.cu.o.Release.cmake:220 (message):
  Error generating
  /work/07936/tg872351/vista/project/zfp-opt/zfp_compression/build/src/CMakeFiles/zfp.dir/cuda/./zfp_generated_interface.cu.o


make[2]: *** [src/CMakeFiles/zfp.dir/build.make:77: src/CMakeFiles/zfp.dir/cuda/zfp_generated_interface.cu.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:114: src/CMakeFiles/zfp.dir/all] Error 2

using CC=nvcc will have the following errors:

-- The C compiler identification is GNU 11.4.1
-- The CXX compiler identification is NVHPC 24.7.0
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - failed
-- Check for working C compiler: /home1/apps/nvidia/Linux_aarch64/24.7/compilers/bin/nvcc
-- Check for working C compiler: /home1/apps/nvidia/Linux_aarch64/24.7/compilers/bin/nvcc - works
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /home1/apps/nvidia/Linux_aarch64/24.7/compilers/bin/nvc++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Compiling with C standard: 99
-- Compiling with C++ standard: 98
-- Could NOT find OpenMP_C (missing: OpenMP_C_FLAGS OpenMP_C_LIB_NAMES)
-- Could NOT find OpenMP (missing: OpenMP_C_FOUND C)
CMake Warning (dev) at CMakeLists.txt:222 (find_package):
  Policy CMP0146 is not set: The FindCUDA module is removed.  Run "cmake
  --help-policy CMP0146" for policy details.  Use the cmake_policy command to
  set the policy and suppress this warning.

This warning is for project developers.  Use -Wno-dev to suppress it.

-- 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 - not found
-- Check if compiler accepts -pthread
-- Check if compiler accepts -pthread - no
CMake Error at /home1/apps/cmake/3.29.5/share/cmake-3.29/Modules/FindPackageHandleStandardArgs.cmake:230 (message):
  Could NOT find Threads (missing: Threads_FOUND)
Call Stack (most recent call first):
  /home1/apps/cmake/3.29.5/share/cmake-3.29/Modules/FindPackageHandleStandardArgs.cmake:600 (_FPHSA_FAILURE_MESSAGE)
  /home1/apps/cmake/3.29.5/share/cmake-3.29/Modules/FindThreads.cmake:226 (FIND_PACKAGE_HANDLE_STANDARD_ARGS)
  /home1/apps/cmake/3.29.5/share/cmake-3.29/Modules/FindCUDA.cmake:1071 (find_package)
  CMakeLists.txt:222 (find_package)

I noticed the ABI info failure for nvcc

[lindstro]

The CMake find_package issue has to do with the deprecation of FindCUDA in CMake as of version 3.27 (see #232). Any chance you can use an earlier version of CMake? I have partially updated the staging branch to use FindCUDAToolkit, which hopefully will address this. I will push these changes in the next few days but would like to do more testing first.

Any other issues with NVHPC may or may not be related to #112, #163, though as best I can tell this is purely a CMake issue. Or perhaps try CC=nvc++ per the error message nvc++ is the only NVHPC compiler that is supported. Sorry, I have no experience with NVHPC.

[R0n12]

The issue can be solved by adding following flags:

export NVCC_APPEND_FLAGS="-ccbin=/home1/apps/nvidia/Linux_aarch64/24.7/compilers/bin/nvc++ -arch=sm_90a -allow-unsupported-compiler"

ref:

https://forums.developer.nvidia.com/t/nvcc-fatal-unsupported-nvhpc-compiler-found-nvc-is-the-only-nvhpc-compiler-that-is-supported/291416/2?u=xulang7

NVIDIA/cuda-samples#179 (comment)

[lindstro]

@R0n12 Glad to hear you found a workaround. Since this appears to be a problem with NVHPC rather than zfp, I'm closing the issue.