Failure to build for AMD Ryzen + CUDA with 'error: identifier "__builtin_ia32_serialize" is undefined' #371
Description
Describe the bug
Fails to build for AMD Ryzen, CUDA (NVIDIA RTX), and recent gcc and nvcc versions with the following compile error:
[ 68%] Building CUDA object examples/cuda/CMakeFiles/cuda_device_property.dir/cuda_device_property.cu.o
cd /home/steinhagen/git/taskflow/build/examples/cuda && /usr/local/cuda-11.5/bin/nvcc -forward-unknown-to-host-compiler -I/home/steinhagen/git/taskflow -O3 -DNDEBUG --extended-lambda -Xcompiler=-Wall,-Wextra,-Wfatal-errors -MD -MT examples/cuda/CMakeFiles/cuda_device_property.dir/cuda_device_property.cu.o -MF CMakeFiles/cuda_device_property.dir/cuda_device_property.cu.o.d -x cu -c /home/steinhagen/git/taskflow/examples/cuda/cuda_device_property.cu -o CMakeFiles/cuda_device_property.dir/cuda_device_property.cu.o
/usr/lib64/gcc/x86_64-suse-linux/11/include/serializeintrin.h(41): error: identifier "__builtin_ia32_serialize" is undefined
The other CPU-only/non-CUDA examples seem to compile and run just fine, provided the unit tests have been disabled.
The plain nvcc SDK installation has been verified working fine using the NVIDIA_CUDA-11.5 examples.
To Reproduce
tested on the main git branch [ac1a3f8] with the following flags:
- 'TF_BUILD_CUDA==ON'
- 'TF_BUILD_TESTS==OFF' (possibly another bug)
Desktop:
- OS: Linux (OpenSUSE Tumbleweed)
- gcc: 11.2.1 20211124
- nvcc: release 11.5, V11.5.119
- cmake version 3.22.1
Any help would be much appreciated and keep up the good work! 👍