Intel's compilers produce code that checks if it is running on an Intel CPU, and runs an inferior version of
the code if it detects a non-Intel CPU.
Work around: Use another compiler if the compiled code may run on a non-Intel machine.
Documentation: https://www.agner.org/optimize/blog/read.php?i=63
The Intel C++ compiler version 19 cannot compile VCL version 2.00, apparently because of
problems with constexpr functions.
Fixed in 2021 version.
The __vectorcall calling mode on Intel C++ compiler version 19 is incompatible with other compilers. Example:
Vec4f __vectorcall add_25 (__m128 x) {
return _mm_add_ps (x, _mm_set1_ps(2.5));
}
The Intel compiler returns the result by a pointer, MS and Clang compilers return the result in xmm0 (__vectorcall is not currently supported on Gnu compiler )