[HLSL] Integer conversion overload scoring

[llvm-beanz]

Because of C/C++ usual arithmetic conversions, C++ overload resolution doesn't handle some cases that HLSL needs to handle for 16-bit types.

The following code is ambiguous to resolve based on currently implemented overload rules:

void Fn4(int64_t L);
void Fn4(int I);

void Call4(int16_t H) { Fn4(H); }

A similar example is here

Per HLSL rules the conversion should prefer the next smallest integer type (int) when promoting 16-bit values.

Depends on microsoft/hlsl-specs#232.

Acceptance Criteria

• Tests verifying correct overload resolution of the case described above.

[llvmbot]

@llvm/issue-subscribers-clang-frontend

Author: Chris B (llvm-beanz)

Because of C/C++ usual arithmetic conversions, C++ overload resolution doesn't handle some cases that HLSL needs to handle for 16-bit types.

The following code is ambiguous to resolve based on currently implemented overload rules:

void Fn4(int64_t L);
void Fn4(int I);

void Call4(int16_t H) { Fn4(H); }

A similar example is here

Per HLSL rules the conversion should prefer the next smallest integer type (int) when promoting 16-bit values.

Depends on microsoft/hlsl-specs#232.

Acceptance Criteria

• Tests verifying correct overload resolution of the case described above.