Implement the QuadReadAcrossDiagonal HLSL Function

[farzonl]

• Implement QuadReadAcrossDiagonal clang builtin,
• Link QuadReadAcrossDiagonal clang builtin with hlsl_intrinsics.h
• Add sema checks for QuadReadAcrossDiagonal to CheckHLSLBuiltinFunctionCall in SemaChecking.cpp
• Add codegen for QuadReadAcrossDiagonal to EmitHLSLBuiltinExpr in CGBuiltin.cpp
• Add codegen tests to clang/test/CodeGenHLSL/builtins/QuadReadAcrossDiagonal.hlsl
• Add sema tests to clang/test/SemaHLSL/BuiltIns/QuadReadAcrossDiagonal-errors.hlsl
• Create the int_dx_QuadReadAcrossDiagonal intrinsic in IntrinsicsDirectX.td
• Create the DXILOpMapping of int_dx_QuadReadAcrossDiagonal to 123 in DXIL.td
• Create the QuadReadAcrossDiagonal.ll and QuadReadAcrossDiagonal_errors.ll tests in llvm/test/CodeGen/DirectX/
• Create the int_spv_QuadReadAcrossDiagonal intrinsic in IntrinsicsSPIRV.td
• In SPIRVInstructionSelector.cpp create the QuadReadAcrossDiagonal lowering and map it to int_spv_QuadReadAcrossDiagonal in SPIRVInstructionSelector::selectIntrinsic.
• Create SPIR-V backend test case in llvm/test/CodeGen/SPIRV/hlsl-intrinsics/QuadReadAcrossDiagonal.ll

DirectX

DXIL Opcode	DXIL OpName	Shader Model	Shader Stages
123	QuadOp	6.0	('library', 'compute', 'amplification', 'mesh', 'pixel', 'node')

SPIR-V

OpGroupNonUniformQuadSwap:

Description:

Swap the Value of the invocation within the
quad with another invocation in the quad using Direction.

Result Type must be a scalar or vector of floating-point
type, integer type, or Boolean
type.

Execution is a Scope, but has no effect on the
behavior of this instruction. It must be Subgroup.

The type of Value must be the same as Result Type.

Direction is the kind of swap to perform.

Direction must be a scalar of integer type, whose
Signedness operand is 0.

Direction must come from a constant
instruction.

The value returned in Result is the value provided to Value by
another invocation in the same quad scope instance. The invocation
providing this value is determined according to Direction.

A Direction of 0 indicates a horizontal swap;

• Invocations with quad indices of 0 and 1 swap values
• Invocations with quad indices of 2 and 3 swap values
  A Direction of 1 indicates a vertical swap;
• Invocations with quad indices of 0 and 2 swap values
• Invocations with quad indices of 1 and 3 swap values
  A Direction of 2 indicates a diagonal swap;
• Invocations with quad indices of 0 and 3 swap values
• Invocations with quad indices of 1 and 2 swap values

Direction must be one of the above values.

If an active invocation reads Value from an inactive invocation, the
resulting value is undefined.

Capability:
GroupNonUniformQuad

Missing before version 1.3.

Word Count	Opcode	Results	Operands
6	366	<id> Result Type	Result <id>	Scope <id> Execution	<id> Value	<id> Direction

Test Case(s)

Example 1

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

export float4 fn(float4 p1) {
    return QuadReadAcrossDiagonal(p1);
}

Example 2

//dxc QuadReadAcrossDiagonal_1_test.hlsl -T lib_6_8 -enable-16bit-types -O0

export uint4 fn(uint4 p1) {
    return QuadReadAcrossDiagonal(p1);
}

Example 3

//dxc QuadReadAcrossDiagonal_2_test.hlsl -T lib_6_8 -enable-16bit-types -O0

export int4 fn(int4 p1) {
    return QuadReadAcrossDiagonal(p1);
}

HLSL:

Returns the specified local value which is read from the diagonally opposite lane in this quad.

Syntax

<type> QuadReadAcrossDiagonal(
   <type> localValue
);

Parameters

localValue

The requested type.

Return value

The specified local value which is read from the diagonally opposite lane in this quad.

Remarks

For more information on quads, refer to Overview of Shader Model 6.

This function is supported from shader model 6.0 only in pixel and compute shaders.

 

See also

Shader Model 6