Implement the `faceforward` HLSL Function

[farzonl]

• implement faceforward in the hlsl_intrinsics.h
• Implement faceforward spirv target builtin in clang/include/clang/Basic/BuiltinsSPIRV.td
• Add a spirv fast path in hlsl_detail.h in the form

#if (__has_builtin(__builtin_spirv_faceforward))
  return __builtin_spirv_faceforward(...);
#else
  return ...; // regular behavior
#endif

An alternative to doing a built-in here would be to do pattern matching. Say we have code like this

faceforward(p1, p2, p3) ->
    dResult = dot(p2, p3)
    if(dResult < 0)
        return 0 - dResult;
   return dResult;

We would expect the code gen to look something like this

faceforward(p1, p2, p3) ->
  dResult       = dot(p2, p3)
  cmpResult  = fcmp dResult, 0.0
  subResult   = fsub -0.0 p1
  select(cmpResult, p1, subResult)

that means we could pattern match in SPIRVCombine.td and SPIRVPreLegalizerCombiner.cpp to

select(
    fcmp(
        dot(p2, p3), 
        0),
    p1,
    0 - p1) 

• Add codegen for spirv faceforward builtin to EmitSPIRVBuiltinExpr in SPIR.cpp
• Add HLSL codegen tests to clang/test/CodeGenHLSL/builtins/faceforward.hlsl
• Add SPIRV builtin codegen tests to clang/test/CodeGenSPIRV/Builtins/faceforward.c
• Add sema tests to clang/test/SemaHLSL/BuiltIns/faceforward-errors.hlsl
• Add spirv sema tests to clang/test/SemaSPIRV/BuiltIns/faceforward-errors.c
• Create the int_spv_faceforward intrinsic in IntrinsicsSPIRV.td
• In SPIRVInstructionSelector.cpp create the faceforward lowering and map it to int_spv_faceforward in SPIRVInstructionSelector::selectIntrinsic.
• Create SPIR-V backend test case in llvm/test/CodeGen/SPIRV/hlsl-intrinsics/faceforward.ll
• Create SPIR-V backend test case in llvm/test/CodeGen/SPIRV/opencl/faceforward.ll

DirectX

No DXIL Opcodes were found.

SPIR-V

FaceForward:

Description:

FaceForward

If the dot product of Nref and I is negative, the result is N,
otherwise it is -N.

The operands must all be a scalar or vector whose component type is
floating-point.

Result Type and the type of all operands must be the same type.

Number	Operand 1	Operand 2	Operand 3	Operand 4
70	<id> N	<id> I	<id> Nref

Test Case(s)

Example 1

//dxc faceforward_test.hlsl -T lib_6_8 -enable-16bit-types -O0

export float4 fn(float4 p1, float4 p2, float4 p3) {
    return faceforward(p1, p2, p3);
}

HLSL:

Flips the surface-normal (if needed) to face in a direction opposite to i; returns the result in n.

ret faceforward(n, i, ng)

This function uses the following formula: -n sign(dot(i, ng)).

Parameters

Item	Description
n	[in] The resulting floating-point surface-normal vector.
i	[in] A floating-point, incident vector that points from the view position to the shading position.
ng	[in] A floating-point surface-normal vector.

Return Value

A floating-point, surface normal vector that is facing the view direction.

Type Description

Name	Template Type	Component Type	Size
n	vector	float	any
i	vector	float	same dimension(s) as input n
ng	vector	float	same dimensions as input n
ret	vector	float	same dimensions as input n

Minimum Shader Model

This function is supported in the following shader models.

Shader Model	Supported
Shader Model 2 (DirectX HLSL) and higher shader models	yes
Shader Model 1 (DirectX HLSL)	vs_1_1 and ps_1_4

See also

Intrinsic Functions (DirectX HLSL)

[davidcook-msft]

@farzonl to update this item outlining the difference between the DirectX impl and the SPIR-V impl

[damyanp]

@farzonl - update to mention pattern matching implementation option