ue4 4.21 shader notes(1)

写在前面
经过大概三天的折磨，我希望从一个简单地切入点学习渲染的想法泡汤了。没有银弹，还是老老实实看ue4源码好了。

正文
4.21/Engine/Shaders/Public/FP16Math.ush

/** 定义使用16位浮点half，32位浮点float，还是32位浮点的低16位min16float */

/*=============================================================================
	FP16Math.usf: Defines for using FP16 or FP32
=============================================================================*/
#pragma once

#ifndef EXPERIMENTAL_FP16   // 默认开启16位 有两种选择：half 和 min16float
#define EXPERIMENTAL_FP16 1
#endif

#ifndef USE_MIN16FLOAT      // 默认不使用 min16float
#define USE_MIN16FLOAT 0
#endif

#if EXPERIMENTAL_FP16       
#if USE_MIN16FLOAT
	#define EXP_FLT min16float  // 开启USE_MIN16FLOAT宏，则使用 min16float
	#define EXP_FLT2 min16float2
	#define EXP_FLT3 min16float3
	#define EXP_FLT4 min16float4
	#define EXP_FLT3x3 min16float3x3
	#define EXP_FLT4x4 min16float4x4
	#define EXP_FLT4x3 min16float4x3
#else
	#define EXP_FLT half    // 否则使用 half
	#define EXP_FLT2 half2
	#define EXP_FLT3 half3
	#define EXP_FLT4 half4
	#define EXP_FLT3x3 half3x3
	#define EXP_FLT4x4 half4x4
	#define EXP_FLT4x3 half4x3
#endif
#else
	#define EXP_FLT float   // 关闭EXPERIMENTAL_FP16宏，则使用 float 
	#define EXP_FLT2 float2
	#define EXP_FLT3 float3
	#define EXP_FLT4 float4
	#define EXP_FLT3x3 float3x3
	#define EXP_FLT4x4 float4x4
	#define EXP_FLT4x3 float4x3
#endif

4.21/Engine/Shaders/Public/Platform.ush

/** 开启平台相关的宏，以及加入少量辅助方法 */

/*=============================================================================
	Prefix.usf: USF file automatically included by shader preprocessor.
=============================================================================*/

#pragma once

#include "FP16Math.ush"

// ---------------------------------------------------- Profile or compiler specific includes
// TODO: Have shader compiler including these platform specific USF files, that needs to work
// with ShaderCore.cpp's GetShaderIncludes().
//  引入平台对应的头文件
#if PS4_PROFILE  // ps4 平台
	// always #include PS4Common.usf so it can #define override anything in any other included file.
	#include "Platform/PS4/PS4Common.ush"
#endif
 
#if XBOXONE_PROFILE // xbox1 平台
	#include "Platform/XboxOne/XboxOneCommon.ush"
#endif

#if SWITCH_PROFILE || SWITCH_PROFILE_FORWARD // switch 平台
	#include "Platform/Switch/SwitchCommon.ush"
#endif

#if COMPILER_METAL  // metal 底层
	// Helps with iteration when changing Metal shader code generation backend.
	#include "Platform/Metal/MetalCommon.ush"
#endif

#if COMPILER_VULKAN // vulkan 底层
	// Helps with iteration when changing Vulkan shader code generation backend.
	#include "Platform/Vulkan/VulkanCommon.ush"
#endif


// ---------------------------------------------------- DDC invalidation
// to support the console command "r.InvalidateShaderCache"
#include "ShaderVersion.ush"


// ---------------------------------------------------- COMPILE_* and *_PROFILE defaults
// 定义shader语言宏
#ifndef COMPILER_HLSLCC // hlslcc 语言
#define COMPILER_HLSLCC 0
#endif

#ifndef COMPILER_HLSL   // hlsl 语言
#define COMPILER_HLSL 0
#endif

#ifndef COMPILER_GLSL   // glsl 语言
#define COMPILER_GLSL 0
#endif

#ifndef COMPILER_GLSL_ES2   //glsl es2 语言
#define COMPILER_GLSL_ES2 0
#endif

#ifndef COMPILER_GLSL_ES3_1     //glsl es3.1语言
#define COMPILER_GLSL_ES3_1 0
#endif

#ifndef COMPILER_GLSL_ES3_1_EXT //glsl es3.1 ext 语言
#define COMPILER_GLSL_ES3_1_EXT 0
#endif

#ifndef COMPILER_METAL  //metel 语言
#define COMPILER_METAL 0
#endif
// 定义语言特性宏
#ifndef COMPILER_SUPPORTS_ATTRIBUTES
#define COMPILER_SUPPORTS_ATTRIBUTES 0
#endif

#ifndef PLATFORM_SUPPORTS_SRV_UB
#define PLATFORM_SUPPORTS_SRV_UB 0
#endif

#ifndef SM5_PROFILE
#define SM5_PROFILE 0
#endif

#ifndef SM4_PROFILE
#define SM4_PROFILE 0
#endif

#ifndef OPENGL_PROFILE
#define OPENGL_PROFILE 0
#endif

#ifndef ES2_PROFILE
#define ES2_PROFILE 0
#endif

#ifndef ES3_1_PROFILE
#define ES3_1_PROFILE 0
#endif

#ifndef METAL_PROFILE
#define METAL_PROFILE 0
#endif

#ifndef METAL_MRT_PROFILE
#define METAL_MRT_PROFILE 0
#endif

#ifndef METAL_SM5_NOTESS_PROFILE
#define METAL_SM5_NOTESS_PROFILE 0
#endif

#ifndef METAL_SM5_PROFILE
#define METAL_SM5_PROFILE 0
#endif

#ifndef COMPILER_VULKAN
#define	COMPILER_VULKAN 0
#endif

#ifndef VULKAN_PROFILE
#define	VULKAN_PROFILE 0
#endif

#ifndef VULKAN_PROFILE_SM4
#define	VULKAN_PROFILE_SM4 0
#endif

#ifndef VULKAN_PROFILE_SM5
#define	VULKAN_PROFILE_SM5 0
#endif

#ifndef IOS
#define IOS 0
#endif

#ifndef MAC
#define MAC 0
#endif
// 依据上面定义的宏，开启对应平台特性
// 'static' asserts
#if COMPILER_GLSL || COMPILER_GLSL_ES2 || COMPILER_GLSL_ES3_1 || COMPILER_VULKAN || COMPILER_METAL
#if !COMPILER_HLSLCC    // 以上平台需要编译 hlslcc
#error "Missing COMPILER_HLSLCC define!"
#endif
#endif


#if PLATFORM_SUPPORTS_SRV_UB
#define PLATFORM_SUPPORTS_SRV_UB_MACRO(...) __VA_ARGS__ // __VA_ARGS__表示可变参数
#else
#define PLATFORM_SUPPORTS_SRV_UB_MACRO(...)
#endif

// ---------------------------------------------------- Alternative floating point types

#ifndef FORCE_FLOATS
#define FORCE_FLOATS 0
#endif
// 只对es2编译器开启 float
#if (!(COMPILER_GLSL_ES2 || COMPILER_GLSL_ES3_1 || METAL_PROFILE) || FORCE_FLOATS)
	// Always use floats when not using the ES2 compiler, because low precision modifiers are currently only tweaked for ES2, 
	// And we don't want potential side effects on other platforms
	#define half float
	#define half1 float1
	#define half2 float2
	#define half3 float3
	#define half4 float4
	#define half3x3 float3x3
	#define half4x4 float4x4 
	#define half4x3 float4x3 
	#define fixed float
	#define fixed1 float1
	#define fixed2 float2
	#define fixed3 float3
	#define fixed4 float4
	#define fixed3x3 float3x3
	#define fixed4x4 float4x4
	#define fixed4x3 float4x3
#endif

// ---------------------------------------------------- Profile config
// 定义shader model 等级
// Values of FEATURE_LEVEL.
#define FEATURE_LEVEL_ES2	1
#define FEATURE_LEVEL_ES3_1	2
#define FEATURE_LEVEL_SM3	3
#define FEATURE_LEVEL_SM4	4
#define FEATURE_LEVEL_SM5	5
#define FEATURE_LEVEL_MAX	6

#if PS4_PROFILE
	#define FEATURE_LEVEL FEATURE_LEVEL_SM5

#elif SM5_PROFILE
	// SM5 = full dx11 features (high end UE4 rendering)
	#define FEATURE_LEVEL FEATURE_LEVEL_SM5

#elif SM4_PROFILE
	#define FEATURE_LEVEL FEATURE_LEVEL_SM4

#elif SWITCH_PROFILE || SWITCH_PROFILE_FORWARD
	#undef ES3_1_PROFILE

	#if SWITCH_PROFILE
		#define FEATURE_LEVEL FEATURE_LEVEL_SM5
	#else
		#define FEATURE_LEVEL FEATURE_LEVEL_ES3_1
		// @todo switch: maybe all uses of this should check feature level not profile?
		#define ES3_1_PROFILE 1
	#endif

#elif VULKAN_PROFILE
	#define FEATURE_LEVEL FEATURE_LEVEL_ES3_1

#elif VULKAN_PROFILE_SM4
	#define FEATURE_LEVEL FEATURE_LEVEL_SM4

#elif VULKAN_PROFILE_SM5
	#define FEATURE_LEVEL FEATURE_LEVEL_SM5
	#define STENCIL_COMPONENT_SWIZZLE .x    // 这里定义stencil 返回x分量 
    // swizzle的用法举个例子:v.xxyz 
    

#elif GL3_PROFILE || GL4_PROFILE
	#if GL3_PROFILE
		#define FEATURE_LEVEL FEATURE_LEVEL_SM4
	#elif GL4_PROFILE
		#define FEATURE_LEVEL FEATURE_LEVEL_SM5
	#endif

	// A8 textures when sampled have their component in R.
	#define A8_SAMPLE_MASK .r   // 采样8bit a分量时返回的时r分量

	// hacks until the shader compiler supports those
	#if GL4_PROFILE
		#define class struct    // hack class 为 struct
	#endif

#elif METAL_PROFILE
	#define FEATURE_LEVEL  FEATURE_LEVEL_ES3_1
	// @todo metal: remove this and make sure all uses handle METAL_PROFILE
	#undef ES3_1_PROFILE
	#define ES3_1_PROFILE 1
	#define FCOLOR_COMPONENT_SWIZZLE .bgra
	#define FMANUALFETCH_COLOR_COMPONENT_SWIZZLE .bgra
	#define STENCIL_COMPONENT_SWIZZLE .x

#elif METAL_ES2_PROFILE
	#define FEATURE_LEVEL  FEATURE_LEVEL_ES2
	// @todo metal: remove this and make sure all uses handle METAL_ES2_PROFILE
	#undef ES2_PROFILE
	#define ES2_PROFILE 1
	#define FCOLOR_COMPONENT_SWIZZLE .bgra
	#define FMANUALFETCH_COLOR_COMPONENT_SWIZZLE .bgra
	#define STENCIL_COMPONENT_SWIZZLE .x

#elif METAL_MRT_PROFILE
	#define FEATURE_LEVEL FEATURE_LEVEL_SM5
	#define FCOLOR_COMPONENT_SWIZZLE .bgra
	#define FMANUALFETCH_COLOR_COMPONENT_SWIZZLE .bgra
	#define STENCIL_COMPONENT_SWIZZLE .x

#elif METAL_SM5_NOTESS_PROFILE
	#define FEATURE_LEVEL FEATURE_LEVEL_SM5
	#define FCOLOR_COMPONENT_SWIZZLE .bgra
	#define FMANUALFETCH_COLOR_COMPONENT_SWIZZLE .bgra
	#define STENCIL_COMPONENT_SWIZZLE .x

#elif METAL_SM5_PROFILE
	#define FEATURE_LEVEL FEATURE_LEVEL_SM5
	#define FCOLOR_COMPONENT_SWIZZLE .bgra
	#define FMANUALFETCH_COLOR_COMPONENT_SWIZZLE .bgra
	#define STENCIL_COMPONENT_SWIZZLE .x

#elif ES2_PROFILE || ES3_1_PROFILE
	#if ES3_1_PROFILE
		#define FEATURE_LEVEL FEATURE_LEVEL_ES3_1
	#else
		//@todo ES3_1 GL
		#define FEATURE_LEVEL FEATURE_LEVEL_ES2
	#endif

	#if COMPILER_GLSL_ES2 || COMPILER_GLSL_ES3_1
		// Swizzle as we only support GL_BGRA on non-ES2 platforms that have that extension
		#define FCOLOR_COMPONENT_SWIZZLE .bgra  
		#define FMANUALFETCH_COLOR_COMPONENT_SWIZZLE .bgra
	#else 
		#define FCOLOR_COMPONENT_SWIZZLE .rgba
		#define FMANUALFETCH_COLOR_COMPONENT_SWIZZLE .bgra
		#if COMPILER_GLSL
			// A8 textures when sampled have their component in R
			#define A8_SAMPLE_MASK .r
		#endif
	#endif

#else

	#error Add your platform here

	#define FEATURE_LEVEL FEATURE_LEVEL_MAX

#endif


// ---------------------------------------------------- Swizzle defaults

// If we didn't request color component swizzling, just make it empty
#ifndef FCOLOR_COMPONENT_SWIZZLE
#define FCOLOR_COMPONENT_SWIZZLE .rgba
#endif

#ifndef FMANUALFETCH_COLOR_COMPONENT_SWIZZLE
#define FMANUALFETCH_COLOR_COMPONENT_SWIZZLE .bgra
#endif

#ifndef STENCIL_COMPONENT_SWIZZLE
#define STENCIL_COMPONENT_SWIZZLE .g
#endif

#ifndef A8_SAMPLE_MASK
#define A8_SAMPLE_MASK .a
#endif

// ---------------------------------------------------- Platform dependent supports

// non-editor platforms generally never want development/editor features.
#ifndef PLATFORM_SUPPORTS_DEVELOPMENT_SHADERS
#define PLATFORM_SUPPORTS_DEVELOPMENT_SHADERS !ESDEFERRED_PROFILE
#endif

#ifndef MOBILE_EMULATION
#define MOBILE_EMULATION ((FEATURE_LEVEL == FEATURE_LEVEL_ES2 || FEATURE_LEVEL == FEATURE_LEVEL_ES3_1) && (!(COMPILER_GLSL_ES3_1 || COMPILER_GLSL_ES2) && USE_DEVELOPMENT_SHADERS && (!(METAL_PROFILE) || MAC) && !VULKAN_PROFILE))
#endif

// Whether the platform supports independent texture and samplers
// When enabled, different texture lookups can share samplers to allow more artist samplers in the base pass
// Ideally this would just be enabled for all SM4 and above feature level platforms
// @todo metal mrt: No reason this can't work with Metal, once cross compiler is fixed
#ifndef SUPPORTS_INDEPENDENT_SAMPLERS
#define SUPPORTS_INDEPENDENT_SAMPLERS (PS4_PROFILE || SM5_PROFILE || SM4_PROFILE || METAL_MRT_PROFILE || METAL_SM5_NOTESS_PROFILE || METAL_SM5_PROFILE || VULKAN_PROFILE_SM5 || VULKAN_PROFILE_SM4 || VULKAN_PROFILE)
#endif

// Whether the platform supports a global clip plane through SV_ClipDistance
// Ideally this would just be enabled for all SM4 and above feature level platforms, but not tested everywhere yet
#define PLATFORM_SUPPORTS_GLOBAL_CLIP_PLANE (PS4_PROFILE || SM5_PROFILE || SM4_PROFILE || METAL_PROFILE || METAL_MRT_PROFILE || METAL_SM5_NOTESS_PROFILE || METAL_SM5_PROFILE || GL4_PROFILE || GL3_PROFILE || VULKAN_PROFILE_SM4 || VULKAN_PROFILE_SM5)

// Whether the platform support pixel coverage on MSAA targets (SV_Coverage).
#define SUPPORTS_PIXEL_COVERAGE (FEATURE_LEVEL >= FEATURE_LEVEL_SM5 && !COMPILER_GLSL && !MOBILE_EMULATION)

// Must match C++ RHISupports4ComponentUAVReadWrite
// D3D11 does not support multi-component loads from a UAV: "error X3676: typed UAV loads are only allowed for single-component 32-bit element types"
#define PLATFORM_SUPPORTS_4COMPONENT_UAV_READ_WRITE (PS4_PROFILE || XBOXONE_PROFILE || COMPILER_METAL)


// ---------------------------------------------------- Compiler specific defaults and fallbacks

/** Defined only for Metal's combined Vertex + Hull shader */
#ifndef TESSELLATIONSHADER
#define TESSELLATIONSHADER 0
#endif

// Hlslcc platforms ignore the uniform keyword as it can't properly optimize flow
#if COMPILER_HLSLCC
#define uniform
#endif

// If compiler lane management in a wave.
//	WaveGetLaneCount()
//	WaveGetLaneIndex()
//	if (WaveOnce()) { ... }
#ifndef COMPILER_SUPPORTS_WAVE_ONCE
#define COMPILER_SUPPORTS_WAVE_ONCE 0
#endif

// Whether the compiler exposes voting on all lanes:
//	WaveAnyTrue(MyBool)
//	WaveAnyTrue(MyBool)
//	WaveAllEqual(MyBool)
#ifndef COMPILER_SUPPORTS_WAVE_VOTE
#define COMPILER_SUPPORTS_WAVE_VOTE 0
#endif

// Whether the compiler exposes min max instructions across all lane of the wave.
//	WaveAllMin(MyFloat)
//	WaveAllMin(MyInt)
//	WaveAllMin(MyUint)
//	WaveAllMax(MyFloat)
//	WaveAllMax(MyInt)
//	WaveAllMax(MyUint)
#ifndef COMPILER_SUPPORTS_WAVE_MINMAX
#define COMPILER_SUPPORTS_WAVE_MINMAX 0
#endif

// Whether the compiler exposes OR and AND bit operation all lanes:
//	WaveAllBitAnd(MyMask)
//	WaveAllBitOr(MyMask)
#ifndef COMPILER_SUPPORTS_WAVE_BIT_ORAND
#define COMPILER_SUPPORTS_WAVE_BIT_ORAND 0
#endif

// Whether the compiler exposes GCN's ds_swizzle_b32 instruction.
//	float WaveLaneSwizzleGCN(float x, const uint and_mask, const uint or_mask, const uint xor_mask)
#ifndef COMPILER_SUPPORTS_WAVE_SWIZZLE_GCN
#define COMPILER_SUPPORTS_WAVE_SWIZZLE_GCN 0
#endif

// Mirrors GRHISupportsRectTopology.
#ifndef PLATFORM_SUPPORTS_RECT_LIST
#define PLATFORM_SUPPORTS_RECT_LIST 0
#endif


// ---------------------------------------------------- Compiler attributes

#if SM5_PROFILE || COMPILER_SUPPORTS_ATTRIBUTES

/** Avoids flow control constructs. */
#define UNROLL [unroll]

/** Gives preference to flow control constructs. */
#define LOOP [loop]

/** Performs branching by using control flow instructions like jmp and label. */
#define BRANCH [branch]

/** Performs branching by using the cnd instructions. */
#define FLATTEN [flatten]

/** Allows a compute shader loop termination condition to be based off of a UAV read. The loop must not contain synchronization intrinsics. */
#define ALLOW_UAV_CONDITION [allow_uav_condition]

#endif // SM5_PROFILE || COMPILER_SUPPORTS_ATTRIBUTES

#if SM5_PROFILE || METAL_MRT_PROFILE || METAL_SM5_PROFILE || METAL_SM5_NOTESS_PROFILE
#define EARLYDEPTHSTENCIL [earlydepthstencil]
#endif


// ---------------------------------------------------- Compiler attribute fallbacks
// 对于不支持的特性，在这里抹去宏
#ifndef UNROLL
#define UNROLL
#endif

#ifndef LOOP
#define LOOP
#endif

#ifndef BRANCH
#define BRANCH
#endif

#ifndef FLATTEN
#define FLATTEN
#endif

#ifndef ALLOW_UAV_CONDITION	
#define ALLOW_UAV_CONDITION	
#endif

#ifndef INVARIANT
#define INVARIANT										
#endif

#ifndef ENABLE_RE_Z
#define ENABLE_RE_Z
#endif

#ifndef EARLYDEPTHSTENCIL
#define EARLYDEPTHSTENCIL
#endif

#ifndef STRONG_TYPE
#define STRONG_TYPE
#endif

#ifndef StrongTypedBuffer
#define StrongTypedBuffer Buffer
#endif

// ---------------------------------------------------- Interpolator attribute fallbacks

#ifndef COMPRESSED_16_FLOAT
#define COMPRESSED_16_FLOAT
#endif

#ifndef COMPRESSED_16_UNORM
#define COMPRESSED_16_UNORM
#endif

#ifndef COMPRESSED_16_SNORM
#define COMPRESSED_16_SNORM
#endif

#ifndef COMPRESSED_16_UINT
#define COMPRESSED_16_UINT
#endif

#ifndef COMPRESSED_16_INT
#define COMPRESSED_16_INT
#endif

#ifndef COMPRESSED_8_UNORM
#define COMPRESSED_8_UNORM
#endif

#ifndef COMPRESSED_8_SNORM
#define COMPRESSED_8_SNORM
#endif

#ifndef COMPRESSED_8_UINT
#define COMPRESSED_8_UINT
#endif


// ---------------------------------------------------- Global uses

#define USE_DEVELOPMENT_SHADERS (COMPILE_SHADERS_FOR_DEVELOPMENT && PLATFORM_SUPPORTS_DEVELOPMENT_SHADERS)


// ---------------------------------------------------- Standard sizes of the indirect parameter structs

// sizeof(FRHIDispatchIndirectParameters) / sizeof(uint)
#define DISPATCH_INDIRECT_UINT_COUNT 3

// sizeof(FRHIDrawIndirectParameters) / sizeof(uint)
#define DRAW_INDIRECT_UINT_COUNT 4

// sizeof(FRHIDrawIndexedIndirectParameters) / sizeof(uint)
#define DRAW_INDEXED_INDIRECT_UINT_COUNT 5


// ---------------------------------------------------- Compiler missing implementations
// 在特定平台补充辅助方法，求行列式
#if COMPILER_GLSL_ES2 || (COMPILER_METAL && MAX_SHADER_LANGUAGE_VERSION < 2) || COMPILER_SWITCH

	float determinant(float3x3 M)
	{
		return
			M[0][0] * (M[1][1] * M[2][2] - M[1][2] * M[2][1]) -
			M[1][0] * (M[0][1] * M[2][2] - M[0][2] * M[2][1]) +
			M[2][0] * (M[0][1] * M[1][2] - M[0][2] * M[1][1]);
	}

#endif

#if COMPILER_HLSLCC     // 补充 求log10(x)
#define log10(x)  log((x)) / log(10.0)
#endif


#if !COMPILER_SUPPORTS_MINMAX3  //补充 求min max 方法

float min3( float a, float b, float c )
{
	return min( a, min( b, c ) );
}

float max3( float a, float b, float c )
{
	return max( a, max( b, c ) );
}

float2 min3( float2 a, float2 b, float2 c )
{
	return float2(
		min3( a.x, b.x, c.x ),
		min3( a.y, b.y, c.y )
	);
}

float2 max3( float2 a, float2 b, float2 c )
{
	return float2(
		max3( a.x, b.x, c.x ),
		max3( a.y, b.y, c.y )
	);
}

float3 max3( float3 a, float3 b, float3 c )
{
	return float3(
		max3( a.x, b.x, c.x ),
		max3( a.y, b.y, c.y ),
		max3( a.z, b.z, c.z )
	);
}

float3 min3( float3 a, float3 b, float3 c )
{
	return float3(
		min3( a.x, b.x, c.x ),
		min3( a.y, b.y, c.y ),
		min3( a.z, b.z, c.z )
	);
}

float4 min3( float4 a, float4 b, float4 c )
{
	return float4(
		min3( a.x, b.x, c.x ),
		min3( a.y, b.y, c.y ),
		min3( a.z, b.z, c.z ),
		min3( a.w, b.w, c.w )
	);
}

float4 max3( float4 a, float4 b, float4 c )
{
	return float4(
		max3( a.x, b.x, c.x ),
		max3( a.y, b.y, c.y ),
		max3( a.z, b.z, c.z ),
		max3( a.w, b.w, c.w )
	);
}

#endif


#if COMPILER_HLSLCC //抹平求偏差方法
#define ddx_fine(x) ddx(x)
#define ddy_fine(y) ddy(y)
#endif

// Give hint to compiler to move one value to scalar unit.
#if !defined(ToScalarMemory) && !defined(COMPILER_SUPPORTS_TO_SCALAR_MEMORY)
	#define ToScalarMemory(x) (x)
#endif

#if FEATURE_LEVEL < FEATURE_LEVEL_ES3_1 && !COMPILER_METAL
	// DX11 (feature levels >= 10) feature sets natively supports uints in shaders; we just use floats on other platforms.
	#define uint4	int4    //抹平uint4支持
#endif

#if COMPILER_HLSLCC || PS4_PROFILE
	#define SNORM 
	#define UNORM
#else
	#define SNORM snorm
	#define UNORM unorm
#endif

4.21/Engine/Shaders/Public/ShaderVersion.ush


/** shader 版本号，用于版本校验 */
// This file is automatically generated by the console command r.InvalidateCachedShaders
// Each time the console command is executed it generates a new GUID. As this file is included
// in Platform.ush (which should be included in any shader) it allows to invalidate the shader DDC.
// 
// GUID = F7C9D61CA78A4854B1D6EE5D3F987298

4.21/Engine/Shaders/Public/WaveBroadcastIntrinsics.ush
Wave-Intrinsics

/** 不太清楚，msdn上说可以提升多线程中的性能。*/
/*=============================================================================
	WaveBroadcastIntrinsics.ush: Exposes intrisics to perform broadcasting
	within lanes of a same wave.
=============================================================================*/
  
#pragma once

#include "Platform.ush"


#define PLATFORM_SUPPORTS_WAVE_BROADCAST (COMPILER_SUPPORTS_WAVE_SWIZZLE_GCN)

#if PLATFORM_SUPPORTS_WAVE_BROADCAST

/** Swap left lane with righ lanes within lane group (size is power of two in [2; 64]). 
 *
 *  If a lane is not active, the VGPR value returned is 0.
 *
 *  LaneGroupSize = 8
 *  LaneId = 1
 *
 *         |  lane group (size=8)  |
 *     x = | 0  1  2  3| 4  5  6  7| 8  9 ...
 *
 *  return | 4  5  6  7| 0  1  2  3|12 13 ...
 */
float WaveSwapWithinLaneGroup(float x, const uint LaneGroupSize)
{
	const uint and_mask = 0x1F;
	const uint or_mask  = 0x00;
	const uint xor_mask = LaneGroupSize >> 1;
	return WaveLaneSwizzleGCN(x, and_mask, or_mask, xor_mask);
}

/**  Broadcast inner lane group over a lane group (size is power of two in [2; 64]).
 *
 *  If a lane is not active, the VGPR value returned is 0.
 *
 *  LaneGroupSize = 8
 *  InnerLaneGroupSize = 2
 *  InnerLaneGroupId = 1
 *
 *         |  lane group (size=8)  |
 *     x = | 0  1  2  3  4  5  6  7| 8  9 ...
 *
 *  return | 2  3  2  3  2  3  2  3|10 11 ...
 */
float WaveBroadcastLaneGroup(float x, const uint LaneGroupSize, const uint InnerLaneGroupSize, const uint InnerLaneGroupId)
{
	const uint InnerGroupCount = LaneGroupSize / InnerLaneGroupSize;

	const uint and_mask = ~((InnerGroupCount - 1) * InnerLaneGroupSize);
	const uint or_mask  = InnerLaneGroupId * InnerLaneGroupSize;
	const uint xor_mask = 0x00;
	return WaveLaneSwizzleGCN(x, and_mask, or_mask, xor_mask);
}


#define __WaveBroadcastOverrideType(Type) \
	Type WaveSwapWithinLaneGroup(Type x, const uint LaneGroupSize) \
	{ \
		return as##Type(WaveSwapWithinLaneGroup(asfloat(x), LaneGroupSize)); \
	} \
	Type WaveBroadcastLaneGroup(Type x, const uint LaneGroupSize, const uint InnerLaneGroupSize, const uint InnerLaneGroupId) \
	{ \
		return as##Type(WaveBroadcastLaneGroup(asfloat(x), LaneGroupSize, InnerLaneGroupSize, InnerLaneGroupId)); \
	} \


__WaveBroadcastOverrideType(int)
__WaveBroadcastOverrideType(uint)


#undef __WaveBroadcastOverrideType

#endif // PLATFORM_SUPPORTS_WAVE_BROADCAST