Ambiguous overload on C++ structure returning _Float16

[thiagomacieira]

struct Float16
{
    _Float16 f; operator _Float16() const { return f; }
}; 
float f(float f, _Float16 f16) { return f + f16; }
float f(float f, Float16 f16) { return f + f16; }

Compile with either -march=sapphirerapids or -target aarch64-linux.

GCC 12 compiles fine. Clang fails with:

test.cpp:6:42: error: use of overloaded operator '+' is ambiguous (with operand types 'float' and 'Float16')
float f(float f, Float16 f16) { return f + f16; }
                                       ~ ^ ~~~
test.cpp:6:42: note: built-in candidate operator+(float, float)
test.cpp:6:42: note: built-in candidate operator+(float, double)
test.cpp:6:42: note: built-in candidate operator+(float, long double)
test.cpp:6:42: note: built-in candidate operator+(float, int)
test.cpp:6:42: note: built-in candidate operator+(float, long)
test.cpp:6:42: note: built-in candidate operator+(float, long long)
test.cpp:6:42: note: built-in candidate operator+(float, __int128)
test.cpp:6:42: note: built-in candidate operator+(float, unsigned int)
test.cpp:6:42: note: built-in candidate operator+(float, unsigned long)
test.cpp:6:42: note: built-in candidate operator+(float, unsigned long long)
test.cpp:6:42: note: built-in candidate operator+(float, unsigned __int128)
1 error generated.

Note that there's no offering for operator+(float, _Float16) in the listing, though that addition does work as seen in the previous line. The problem appears with the cast operator present only.

[llvmbot]

@llvm/issue-subscribers-clang-frontend