NSIMD is now available on the excellent #Godbolt. Thanks to their team NSIMD is now a library at #Godbolt. More formally known as Compiler Explorer, Godbolt allows one to write code and, at the same time, see the corresponding assembly. Code can be plain C, C++, Fortran, CUDA or many other. A lot of compilers are supported including GCC, Clang, MSVC, ICC and many other. The resulting assembly is the result of the compilation of the written code by the selected compiler. The interface is very versatile and it is easy to compare different compilers, optimization options...
NSIMD supports Intel CPUs, Arm CPUs and Nvidia GPUs. You can try and write code for the three architectures using NSIMD APIs and modules to see the resulting assembly. If you target Intel CPUs you will be able to run your code by selecting "Run the compiled output" from the "Output..." menu on the right panel. To test CUDA with NSIMD, select "CUDA" from the menu on the top right of the left panel and don't forget to add "-DCUDA" as a compiler option. You can get help from the documentation of NSIMD of from the unit tests of NSIMD in the tests directory.
Good hacking with #Godbolt!
 Godbolt: https://godbolt.org/
 NSIMD: https://github.com/agenium-scale/nsimd
 NSIMD documentation: https://agenium-scale.github.io/nsimd/index.html
 Intel SIMD technologies: https://www.intel.com/content/www/us/en/architecture-and-technology/avx-512-overview.html
 Arm SIMD technologies: https://developer.arm.com/architectures/instruction-sets/simd-isas
 Nvidia CUDA: https://developer.nvidia.com/cuda-zone