| Issue |
128228
|
| Summary |
[HLSL] hlsl202x `double` compatability overloads
|
| Labels |
HLSL
|
| Assignees |
|
| Reporter |
llvm-beanz
|
Lots of HLSL builtin functions that operate on floating point elements do not have `double` overloads. In most cases DXC successfully resolves those overloads to `float` but does not emit conversion diagnostics.
In Clang we should provide wrapper overloads that call the `float` overloads and produce conversion diagnostics. We should use the C preprocessor to define these overloads and have them conditional on HLSL 202x.
Some macros I've been playing with:
```hlsl
#define DXC_COMPAT_UNARY_DOUBLE_OVERLOAD(fn) \
float fn(double V) { return fn((float)V); } \
float2 fn(double2 V) { return fn((float2)V); } \
float3 fn(double3 V) { return fn((float3)V); } \
float4 fn(double4 V) { return fn((float4)V); }
#define DXC_COMPAT_BINARY_DOUBLE_OVERLOAD(fn) \
float fn(double V1, double V2) { return fn((float)V1, (float)V2); } \
float2 fn(double2 V1, double2 V2) { return fn((float2)V1, (float2)V2); } \
float3 fn(double3 V1, double3 V2) { return fn((float3)V1, (float3)V2); } \
float4 fn(double4 V1, double4 V2) { return fn((float4)V1, (float4)V2); }
#define DXC_COMPAT_BINARY_TERNARY_OVERLOAD(fn) \
float fn(double V1, double V2, double V3) { \
return fn((float)V1, (float)V1, (float)V3); \
} \
float2 fn(double2 V1, double2 V2, double2 V3) { \
return fn((float2)V1, (float2)V2, (float2)V3); \
} \
float3 fn(double3 V1, double3 V2, double3 V3) { \
return fn((float3)V1, (float3)V2, (float3)V3); \
} \
float4 fn(double4 V1, double4 V2, double4 V3) { \
return fn((float4)V1, (float4)V2, (float4)V3); \
}
```
![]()
_______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
