add asr
This commit is contained in:
parent
b4d543c222
commit
86f8a9e95b
@ -47546,7 +47546,7 @@ MonoBehaviour:
|
||||
m_PrefabInstance: {fileID: 0}
|
||||
m_PrefabAsset: {fileID: 0}
|
||||
m_GameObject: {fileID: 503266359}
|
||||
m_Enabled: 1
|
||||
m_Enabled: 0
|
||||
m_EditorHideFlags: 0
|
||||
m_Script: {fileID: 11500000, guid: 4d37d054e457113459accf4b905624c6, type: 3}
|
||||
m_Name:
|
||||
|
||||
@ -49,8 +49,8 @@ MonoBehaviour:
|
||||
v1setting:
|
||||
EasuCompute: {fileID: 7200000, guid: 787b0c165dad9074e9489817de945916, type: 3}
|
||||
RacsCompute: {fileID: 7200000, guid: a50f730ab549f794cbe91f005703e208, type: 3}
|
||||
Sharpness: 0.5
|
||||
MipMapBias: 0
|
||||
Sharpness: 0.439
|
||||
MipMapBias: -0
|
||||
v2setting:
|
||||
Shaders:
|
||||
computeLuminancePyramidPass: {fileID: 7200000, guid: 04c3480675e29a340808141e68d4cc8b, type: 3}
|
||||
@ -63,7 +63,7 @@ MonoBehaviour:
|
||||
tcrAutoGenPass: {fileID: 7200000, guid: f8b1c27fb6a544b43b38903592240500, type: 3}
|
||||
Sharpness: 0.5
|
||||
PreExposure: 1
|
||||
PerformSharpenPass: 0
|
||||
PerformSharpenPass: 1
|
||||
TransparencyAndCompositionMask: {fileID: 0}
|
||||
AutoGenerateTransparencyAndComposition: 0
|
||||
GenerateTransparencyAndCompositionParameters:
|
||||
@ -91,19 +91,19 @@ MonoBehaviour:
|
||||
autoGenReactivePass: {fileID: 7200000, guid: 3b0d05a92dc6af24fb6f30d6606b37cf, type: 3}
|
||||
tcrAutoGenPass: {fileID: 7200000, guid: 78865acef22baa24f9ce68fb50877fe8, type: 3}
|
||||
debugViewPass: {fileID: 7200000, guid: cb24a71d54164c54eb5e86839acd48c5, type: 3}
|
||||
AutoExposureSource: 0
|
||||
AutoExposureSource: 1
|
||||
TransparencyAndCompositionMask: {fileID: 0}
|
||||
PreExposure: 1
|
||||
PerformSharpenPass: 0
|
||||
Sharpness: 0
|
||||
PerformSharpenPass: 1
|
||||
Sharpness: 0.5
|
||||
VelocityFactor: 1
|
||||
AutoGenerateTransparencyAndComposition: 0
|
||||
AutoGenerateTransparencyAndComposition: 1
|
||||
GenerateTransparencyAndCompositionParameters:
|
||||
autoTcThreshold: 0.05
|
||||
autoTcScale: 1
|
||||
autoReactiveScale: 5
|
||||
autoReactiveMax: 0.9
|
||||
AutoGenerateReactiveMask: 0
|
||||
AutoGenerateReactiveMask: 1
|
||||
GenerateReactiveParameters:
|
||||
scale: 0.5
|
||||
cutoffThreshold: 0.2
|
||||
@ -138,7 +138,7 @@ MonoBehaviour:
|
||||
m_Name: GSR
|
||||
m_EditorClassIdentifier:
|
||||
m_Active: 0
|
||||
quality: 4
|
||||
quality: 1
|
||||
v1settings:
|
||||
EnableEdgeDirection: 1
|
||||
Mode: 1
|
||||
|
||||
80
Packages/com.unity.render-pipelines.universal@14.0.11/Runtime/SuperRendering/SR/Scripts/ASR.cs
vendored
Normal file
80
Packages/com.unity.render-pipelines.universal@14.0.11/Runtime/SuperRendering/SR/Scripts/ASR.cs
vendored
Normal file
@ -0,0 +1,80 @@
|
||||
using System;
|
||||
using UnityEngine;
|
||||
using UnityEngine.Rendering;
|
||||
using UnityEngine.Rendering.Universal;
|
||||
|
||||
namespace X.Rendering.Feature
|
||||
{
|
||||
internal class ASR : ScriptableRendererFeature, ISuperResolutionFeature
|
||||
{
|
||||
// TODO: wait for release. https://learn.arm.com/learning-paths/mobile-graphics-and-gaming/get-started-with-arm-asr/01-what-is-arm-asr/#:~:text=Arm%C2%AE%20Accuracy%20Super%20Resolution%E2%84%A2%20%28Arm%20ASR%29%20is%20a,from%20AMD%E2%80%99s%20Fidelity%20Super%20Resolution%202%20v2.2.2%20.
|
||||
[Serializable]
|
||||
internal class AsrShaders
|
||||
{
|
||||
public Material AccumulatePass;
|
||||
public ComputeShader ComputeLuminancePyramidPass;
|
||||
public Material ConvertVelocity;
|
||||
public ComputeShader CopyExposure;
|
||||
public Material CreateReactiveMask;
|
||||
public Material DepthClipPass;
|
||||
public ComputeShader LockPass;
|
||||
public Material RCASPass;
|
||||
public Material ReconstructPrevDepthPass;
|
||||
}
|
||||
|
||||
[SerializeField]
|
||||
private SrQuality quality;
|
||||
[SerializeField]
|
||||
private AsrShaders shaders;
|
||||
|
||||
|
||||
public override void AddRenderPasses(ScriptableRenderer renderer, ref RenderingData renderingData)
|
||||
{
|
||||
}
|
||||
|
||||
public override void Create()
|
||||
{
|
||||
}
|
||||
|
||||
public void DoSR(CommandBuffer cmd, RTHandle source, RTHandle destination, RTHandle motionVector, ref RenderingData renderingData)
|
||||
{
|
||||
}
|
||||
|
||||
public float GetRenderScale()
|
||||
{
|
||||
switch (quality)
|
||||
{
|
||||
case SrQuality.None:
|
||||
case SrQuality.NativeAA:
|
||||
return 1;
|
||||
case SrQuality.UltraQuality:
|
||||
return 1 / 1.5f;
|
||||
case SrQuality.Quality:
|
||||
return 1 / 1.7f;
|
||||
case SrQuality.Balanced:
|
||||
return 1 / 1.7f;
|
||||
case SrQuality.Performance:
|
||||
case SrQuality.UltraPerformance:
|
||||
return 1 / 2.0f;
|
||||
}
|
||||
return 1;
|
||||
}
|
||||
|
||||
public void SetSR(ESuperResolution resolution)
|
||||
{
|
||||
switch (resolution)
|
||||
{
|
||||
case ESuperResolution.ASR:
|
||||
{
|
||||
SetActive(true);
|
||||
break;
|
||||
}
|
||||
default:
|
||||
{
|
||||
SetActive(false);
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
@ -0,0 +1,11 @@
|
||||
fileFormatVersion: 2
|
||||
guid: d5133ad109868144fbe7ffd465805e59
|
||||
MonoImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
defaultReferences: []
|
||||
executionOrder: 0
|
||||
icon: {instanceID: 0}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -1,5 +1,5 @@
|
||||
fileFormatVersion: 2
|
||||
guid: ad00619c342280a40a4493d4353aa034
|
||||
guid: d275bab6995d15344b8827bf9edefcee
|
||||
folderAsset: yes
|
||||
DefaultImporter:
|
||||
externalObjects: {}
|
||||
@ -0,0 +1,17 @@
|
||||
// Each #kernel tells which function to compile; you can have many kernels
|
||||
#pragma kernel CSMain
|
||||
|
||||
// Create a RenderTexture with enableRandomWrite flag and set it
|
||||
// with cs.SetTexture
|
||||
|
||||
#include "fsr2/ffxm_fsr2_accumulate_pass_fs.hlsl"
|
||||
|
||||
RWTexture2D<float4> Result;
|
||||
|
||||
[numthreads(8,8,1)]
|
||||
void CSMain (uint3 id : SV_DispatchThreadID)
|
||||
{
|
||||
// TODO: insert actual code here!
|
||||
|
||||
Result[id.xy] = float4(id.x & id.y, (id.x & 15)/15.0, (id.y & 15)/15.0, 0.0);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 0ae100bd82a004e4c94d23a10b5e4cae
|
||||
ComputeShaderImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,14 @@
|
||||
// Each #kernel tells which function to compile; you can have many kernels
|
||||
#pragma kernel CSMain
|
||||
|
||||
// Create a RenderTexture with enableRandomWrite flag and set it
|
||||
// with cs.SetTexture
|
||||
RWTexture2D<float4> Result;
|
||||
|
||||
[numthreads(8,8,1)]
|
||||
void CSMain (uint3 id : SV_DispatchThreadID)
|
||||
{
|
||||
// TODO: insert actual code here!
|
||||
|
||||
Result[id.xy] = float4(id.x & id.y, (id.x & 15)/15.0, (id.y & 15)/15.0, 0.0);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 29ec3ac4b2760da40b3e2775f9cbf212
|
||||
ComputeShaderImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,14 @@
|
||||
// Each #kernel tells which function to compile; you can have many kernels
|
||||
#pragma kernel CSMain
|
||||
|
||||
// Create a RenderTexture with enableRandomWrite flag and set it
|
||||
// with cs.SetTexture
|
||||
RWTexture2D<float4> Result;
|
||||
|
||||
[numthreads(8,8,1)]
|
||||
void CSMain (uint3 id : SV_DispatchThreadID)
|
||||
{
|
||||
// TODO: insert actual code here!
|
||||
|
||||
Result[id.xy] = float4(id.x & id.y, (id.x & 15)/15.0, (id.y & 15)/15.0, 0.0);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: e805c8322764fc84fb292bf96eeac09e
|
||||
ComputeShaderImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,14 @@
|
||||
// Each #kernel tells which function to compile; you can have many kernels
|
||||
#pragma kernel CSMain
|
||||
|
||||
// Create a RenderTexture with enableRandomWrite flag and set it
|
||||
// with cs.SetTexture
|
||||
RWTexture2D<float4> Result;
|
||||
|
||||
[numthreads(8,8,1)]
|
||||
void CSMain (uint3 id : SV_DispatchThreadID)
|
||||
{
|
||||
// TODO: insert actual code here!
|
||||
|
||||
Result[id.xy] = float4(id.x & id.y, (id.x & 15)/15.0, (id.y & 15)/15.0, 0.0);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: b68196ffaf1ffe045a2c2f9836d67bcc
|
||||
ComputeShaderImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,14 @@
|
||||
// Each #kernel tells which function to compile; you can have many kernels
|
||||
#pragma kernel CSMain
|
||||
|
||||
// Create a RenderTexture with enableRandomWrite flag and set it
|
||||
// with cs.SetTexture
|
||||
RWTexture2D<float4> Result;
|
||||
|
||||
[numthreads(8,8,1)]
|
||||
void CSMain (uint3 id : SV_DispatchThreadID)
|
||||
{
|
||||
// TODO: insert actual code here!
|
||||
|
||||
Result[id.xy] = float4(id.x & id.y, (id.x & 15)/15.0, (id.y & 15)/15.0, 0.0);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: a3fac62a0b87bed4198c1dcd480c0673
|
||||
ComputeShaderImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,14 @@
|
||||
// Each #kernel tells which function to compile; you can have many kernels
|
||||
#pragma kernel CSMain
|
||||
|
||||
// Create a RenderTexture with enableRandomWrite flag and set it
|
||||
// with cs.SetTexture
|
||||
RWTexture2D<float4> Result;
|
||||
|
||||
[numthreads(8,8,1)]
|
||||
void CSMain (uint3 id : SV_DispatchThreadID)
|
||||
{
|
||||
// TODO: insert actual code here!
|
||||
|
||||
Result[id.xy] = float4(id.x & id.y, (id.x & 15)/15.0, (id.y & 15)/15.0, 0.0);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 7c2440e153528664ea4898beab51a1fc
|
||||
ComputeShaderImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,14 @@
|
||||
// Each #kernel tells which function to compile; you can have many kernels
|
||||
#pragma kernel CSMain
|
||||
|
||||
// Create a RenderTexture with enableRandomWrite flag and set it
|
||||
// with cs.SetTexture
|
||||
RWTexture2D<float4> Result;
|
||||
|
||||
[numthreads(8,8,1)]
|
||||
void CSMain (uint3 id : SV_DispatchThreadID)
|
||||
{
|
||||
// TODO: insert actual code here!
|
||||
|
||||
Result[id.xy] = float4(id.x & id.y, (id.x & 15)/15.0, (id.y & 15)/15.0, 0.0);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: c5addd720b87d2b428d0e0a4198376ed
|
||||
ComputeShaderImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,14 @@
|
||||
// Each #kernel tells which function to compile; you can have many kernels
|
||||
#pragma kernel CSMain
|
||||
|
||||
// Create a RenderTexture with enableRandomWrite flag and set it
|
||||
// with cs.SetTexture
|
||||
RWTexture2D<float4> Result;
|
||||
|
||||
[numthreads(8,8,1)]
|
||||
void CSMain (uint3 id : SV_DispatchThreadID)
|
||||
{
|
||||
// TODO: insert actual code here!
|
||||
|
||||
Result[id.xy] = float4(id.x & id.y, (id.x & 15)/15.0, (id.y & 15)/15.0, 0.0);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 6764c57ca8cde1e43b7601399dc98273
|
||||
ComputeShaderImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,14 @@
|
||||
// Each #kernel tells which function to compile; you can have many kernels
|
||||
#pragma kernel CSMain
|
||||
|
||||
// Create a RenderTexture with enableRandomWrite flag and set it
|
||||
// with cs.SetTexture
|
||||
RWTexture2D<float4> Result;
|
||||
|
||||
[numthreads(8,8,1)]
|
||||
void CSMain (uint3 id : SV_DispatchThreadID)
|
||||
{
|
||||
// TODO: insert actual code here!
|
||||
|
||||
Result[id.xy] = float4(id.x & id.y, (id.x & 15)/15.0, (id.y & 15)/15.0, 0.0);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 5b6ead1a89cc6c7429b3289b7be7dc35
|
||||
ComputeShaderImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,8 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 22dc99feef2190540b024441e72eb34a
|
||||
folderAsset: yes
|
||||
DefaultImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,522 @@
|
||||
// This file is part of the FidelityFX SDK.
|
||||
//
|
||||
// Copyright (C) 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024-2025 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files(the “Software”), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and /or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions :
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in
|
||||
// all copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
||||
// THE SOFTWARE.
|
||||
|
||||
#ifndef FFXM_COMMON_TYPES_H
|
||||
#define FFXM_COMMON_TYPES_H
|
||||
|
||||
#if defined(FFXM_CPU)
|
||||
#define FFXM_PARAMETER_IN
|
||||
#define FFXM_PARAMETER_OUT
|
||||
#define FFXM_PARAMETER_INOUT
|
||||
#define FFXM_PARAMETER_UNIFORM
|
||||
#elif defined(FFXM_HLSL)
|
||||
#define FFXM_PARAMETER_IN in
|
||||
#define FFXM_PARAMETER_OUT out
|
||||
#define FFXM_PARAMETER_INOUT inout
|
||||
#define FFXM_PARAMETER_UNIFORM uniform
|
||||
#elif defined(FFXM_GLSL)
|
||||
#define FFXM_PARAMETER_IN in
|
||||
#define FFXM_PARAMETER_OUT out
|
||||
#define FFXM_PARAMETER_INOUT inout
|
||||
#define FFXM_PARAMETER_UNIFORM const //[cacao_placeholder] until a better fit is found!
|
||||
#endif // #if defined(FFXM_CPU)
|
||||
|
||||
#if defined(FFXM_CPU)
|
||||
/// A typedef for a boolean value.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef bool FfxBoolean;
|
||||
|
||||
/// A typedef for a unsigned 8bit integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef uint8_t FfxUInt8;
|
||||
|
||||
/// A typedef for a unsigned 16bit integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef uint16_t FfxUInt16;
|
||||
|
||||
/// A typedef for a unsigned 32bit integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef uint32_t FfxUInt32;
|
||||
|
||||
/// A typedef for a unsigned 64bit integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef uint64_t FfxUInt64;
|
||||
|
||||
/// A typedef for a signed 8bit integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef int8_t FfxInt8;
|
||||
|
||||
/// A typedef for a signed 16bit integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef int16_t FfxInt16;
|
||||
|
||||
/// A typedef for a signed 32bit integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef int32_t FfxInt32;
|
||||
|
||||
/// A typedef for a signed 64bit integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef int64_t FfxInt64;
|
||||
|
||||
/// A typedef for a floating point value.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef float FfxFloat32;
|
||||
|
||||
/// A typedef for a 2-dimensional floating point value.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef float FfxFloat32x2[2];
|
||||
|
||||
/// A typedef for a 3-dimensional floating point value.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef float FfxFloat32x3[3];
|
||||
|
||||
/// A typedef for a 4-dimensional floating point value.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef float FfxFloat32x4[4];
|
||||
|
||||
/// A typedef for a 2-dimensional 32bit unsigned integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef uint32_t FfxUInt32x2[2];
|
||||
|
||||
/// A typedef for a 3-dimensional 32bit unsigned integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef uint32_t FfxUInt32x3[3];
|
||||
|
||||
/// A typedef for a 4-dimensional 32bit unsigned integer.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
typedef uint32_t FfxUInt32x4[4];
|
||||
#endif // #if defined(FFXM_CPU)
|
||||
|
||||
#if defined(FFXM_HLSL)
|
||||
|
||||
#define FfxFloat32Mat4 matrix <float, 4, 4>
|
||||
#define FfxFloat32Mat3 matrix <float, 3, 3>
|
||||
|
||||
/// A typedef for a boolean value.
|
||||
///
|
||||
/// @ingroup HLSLTypes
|
||||
typedef bool FfxBoolean;
|
||||
|
||||
#if FFXM_HLSL_6_2
|
||||
|
||||
/// @defgroup HLSL62Types HLSL 6.2 And Above Types
|
||||
/// HLSL 6.2 and above type defines for all commonly used variables
|
||||
///
|
||||
/// @ingroup HLSLTypes
|
||||
|
||||
/// A typedef for a floating point value.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef float32_t FfxFloat32;
|
||||
|
||||
/// A typedef for a 2-dimensional floating point value.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef float32_t2 FfxFloat32x2;
|
||||
|
||||
/// A typedef for a 3-dimensional floating point value.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef float32_t3 FfxFloat32x3;
|
||||
|
||||
/// A typedef for a 4-dimensional floating point value.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef float32_t4 FfxFloat32x4;
|
||||
|
||||
/// A [cacao_placeholder] typedef for matrix type until confirmed.
|
||||
typedef float4x4 FfxFloat32x4x4;
|
||||
typedef float3x3 FfxFloat32x3x3;
|
||||
typedef float2x2 FfxFloat32x2x2;
|
||||
|
||||
/// A typedef for a unsigned 32bit integer.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef uint32_t FfxUInt32;
|
||||
|
||||
/// A typedef for a 2-dimensional 32bit unsigned integer.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef uint32_t2 FfxUInt32x2;
|
||||
|
||||
/// A typedef for a 3-dimensional 32bit unsigned integer.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef uint32_t3 FfxUInt32x3;
|
||||
|
||||
/// A typedef for a 4-dimensional 32bit unsigned integer.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef uint32_t4 FfxUInt32x4;
|
||||
|
||||
/// A typedef for a signed 32bit integer.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef int32_t FfxInt32;
|
||||
|
||||
/// A typedef for a 2-dimensional signed 32bit integer.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef int32_t2 FfxInt32x2;
|
||||
|
||||
/// A typedef for a 3-dimensional signed 32bit integer.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef int32_t3 FfxInt32x3;
|
||||
|
||||
/// A typedef for a 4-dimensional signed 32bit integer.
|
||||
///
|
||||
/// @ingroup HLSL62Types
|
||||
typedef int32_t4 FfxInt32x4;
|
||||
|
||||
#else // #if defined(FFXM_HLSL_6_2)
|
||||
|
||||
/// @defgroup HLSLBaseTypes HLSL 6.1 And Below Types
|
||||
/// HLSL 6.1 and below type defines for all commonly used variables
|
||||
///
|
||||
/// @ingroup HLSLTypes
|
||||
|
||||
#define FfxFloat32 float
|
||||
#define FfxFloat32x2 float2
|
||||
#define FfxFloat32x3 float3
|
||||
#define FfxFloat32x4 float4
|
||||
|
||||
/// A [cacao_placeholder] typedef for matrix type until confirmed.
|
||||
#define FfxFloat32x4x4 float4x4
|
||||
#define FfxFloat32x3x3 float3x3
|
||||
#define FfxFloat32x2x2 float2x2
|
||||
|
||||
/// A typedef for a unsigned 32bit integer.
|
||||
///
|
||||
/// @ingroup GPU
|
||||
typedef uint FfxUInt32;
|
||||
typedef uint2 FfxUInt32x2;
|
||||
typedef uint3 FfxUInt32x3;
|
||||
typedef uint4 FfxUInt32x4;
|
||||
|
||||
typedef int FfxInt32;
|
||||
typedef int2 FfxInt32x2;
|
||||
typedef int3 FfxInt32x3;
|
||||
typedef int4 FfxInt32x4;
|
||||
|
||||
#endif // #if defined(FFXM_HLSL_6_2)
|
||||
|
||||
#if FFXM_HALF
|
||||
|
||||
#if FFXM_HLSL_6_2
|
||||
|
||||
typedef float16_t FfxFloat16;
|
||||
typedef float16_t2 FfxFloat16x2;
|
||||
typedef float16_t3 FfxFloat16x3;
|
||||
typedef float16_t4 FfxFloat16x4;
|
||||
|
||||
/// A typedef for an unsigned 16bit integer.
|
||||
///
|
||||
/// @ingroup HLSLTypes
|
||||
typedef uint16_t FfxUInt16;
|
||||
typedef uint16_t2 FfxUInt16x2;
|
||||
typedef uint16_t3 FfxUInt16x3;
|
||||
typedef uint16_t4 FfxUInt16x4;
|
||||
|
||||
/// A typedef for a signed 16bit integer.
|
||||
///
|
||||
/// @ingroup HLSLTypes
|
||||
typedef int16_t FfxInt16;
|
||||
typedef int16_t2 FfxInt16x2;
|
||||
typedef int16_t3 FfxInt16x3;
|
||||
typedef int16_t4 FfxInt16x4;
|
||||
#else // #if FFXM_HLSL_6_2
|
||||
typedef min16float FfxFloat16;
|
||||
typedef min16float2 FfxFloat16x2;
|
||||
typedef min16float3 FfxFloat16x3;
|
||||
typedef min16float4 FfxFloat16x4;
|
||||
|
||||
/// A typedef for an unsigned 16bit integer.
|
||||
///
|
||||
/// @ingroup HLSLTypes
|
||||
typedef min16uint FfxUInt16;
|
||||
typedef min16uint2 FfxUInt16x2;
|
||||
typedef min16uint3 FfxUInt16x3;
|
||||
typedef min16uint4 FfxUInt16x4;
|
||||
|
||||
/// A typedef for a signed 16bit integer.
|
||||
///
|
||||
/// @ingroup HLSLTypes
|
||||
typedef min16int FfxInt16;
|
||||
typedef min16int2 FfxInt16x2;
|
||||
typedef min16int3 FfxInt16x3;
|
||||
typedef min16int4 FfxInt16x4;
|
||||
#endif // #if FFXM_HLSL_6_2
|
||||
|
||||
#endif // FFXM_HALF
|
||||
|
||||
#endif // #if defined(FFXM_HLSL)
|
||||
|
||||
#if defined(FFXM_GLSL)
|
||||
|
||||
#define FfxFloat32Mat4 mat4
|
||||
#define FfxFloat32Mat3 mat3
|
||||
|
||||
/// A typedef for a boolean value.
|
||||
///
|
||||
/// @ingroup GLSLTypes
|
||||
#define FfxBoolean bool
|
||||
#define FfxFloat32 float
|
||||
#define FfxFloat32x2 vec2
|
||||
#define FfxFloat32x3 vec3
|
||||
#define FfxFloat32x4 vec4
|
||||
#define FfxUInt32 uint
|
||||
#define FfxUInt32x2 uvec2
|
||||
#define FfxUInt32x3 uvec3
|
||||
#define FfxUInt32x4 uvec4
|
||||
#define FfxInt32 int
|
||||
#define FfxInt32x2 ivec2
|
||||
#define FfxInt32x3 ivec3
|
||||
#define FfxInt32x4 ivec4
|
||||
|
||||
/// A [cacao_placeholder] typedef for matrix type until confirmed.
|
||||
#define FfxFloat32x4x4 mat4
|
||||
#define FfxFloat32x3x3 mat3
|
||||
#define FfxFloat32x2x2 mat2
|
||||
|
||||
#if FFXM_HALF
|
||||
#define FfxFloat16 float16_t
|
||||
#define FfxFloat16x2 f16vec2
|
||||
#define FfxFloat16x3 f16vec3
|
||||
#define FfxFloat16x4 f16vec4
|
||||
#define FfxUInt16 uint16_t
|
||||
#define FfxUInt16x2 u16vec2
|
||||
#define FfxUInt16x3 u16vec3
|
||||
#define FfxUInt16x4 u16vec4
|
||||
#define FfxInt16 int16_t
|
||||
#define FfxInt16x2 i16vec2
|
||||
#define FfxInt16x3 i16vec3
|
||||
#define FfxInt16x4 i16vec4
|
||||
#endif // FFXM_HALF
|
||||
#endif // #if defined(FFXM_GLSL)
|
||||
|
||||
// Global toggles:
|
||||
// #define FFXM_HALF (1)
|
||||
// #define FFXM_HLSL_6_2 (1)
|
||||
|
||||
#if FFXM_HALF
|
||||
|
||||
#if FFXM_HLSL_6_2
|
||||
|
||||
#define FFXM_MIN16_SCALAR( TypeName, BaseComponentType ) typedef BaseComponentType##16_t TypeName;
|
||||
#define FFXM_MIN16_VECTOR( TypeName, BaseComponentType, COL ) typedef vector<BaseComponentType##16_t, COL> TypeName;
|
||||
#define FFXM_MIN16_MATRIX( TypeName, BaseComponentType, ROW, COL ) typedef matrix<BaseComponentType##16_t, ROW, COL> TypeName;
|
||||
|
||||
#define FFXM_16BIT_SCALAR( TypeName, BaseComponentType ) typedef BaseComponentType##16_t TypeName;
|
||||
#define FFXM_16BIT_VECTOR( TypeName, BaseComponentType, COL ) typedef vector<BaseComponentType##16_t, COL> TypeName;
|
||||
#define FFXM_16BIT_MATRIX( TypeName, BaseComponentType, ROW, COL ) typedef matrix<BaseComponentType##16_t, ROW, COL> TypeName;
|
||||
|
||||
#else //FFXM_HLSL_6_2
|
||||
|
||||
#define FFXM_MIN16_SCALAR( TypeName, BaseComponentType ) typedef min16##BaseComponentType TypeName;
|
||||
#define FFXM_MIN16_VECTOR( TypeName, BaseComponentType, COL ) typedef vector<min16##BaseComponentType, COL> TypeName;
|
||||
#define FFXM_MIN16_MATRIX( TypeName, BaseComponentType, ROW, COL ) typedef matrix<min16##BaseComponentType, ROW, COL> TypeName;
|
||||
|
||||
#define FFXM_16BIT_SCALAR( TypeName, BaseComponentType ) FFXM_MIN16_SCALAR( TypeName, BaseComponentType );
|
||||
#define FFXM_16BIT_VECTOR( TypeName, BaseComponentType, COL ) FFXM_MIN16_VECTOR( TypeName, BaseComponentType, COL );
|
||||
#define FFXM_16BIT_MATRIX( TypeName, BaseComponentType, ROW, COL ) FFXM_MIN16_MATRIX( TypeName, BaseComponentType, ROW, COL );
|
||||
|
||||
#endif //FFXM_HLSL_6_2
|
||||
|
||||
#else //FFXM_HALF
|
||||
|
||||
#define FFXM_MIN16_SCALAR( TypeName, BaseComponentType ) typedef BaseComponentType TypeName;
|
||||
#define FFXM_MIN16_VECTOR( TypeName, BaseComponentType, COL ) typedef BaseComponentType##COL TypeName;
|
||||
#define FFXM_MIN16_MATRIX( TypeName, BaseComponentType, ROW, COL ) typedef matrix<BaseComponentType, ROW, COL> TypeName;
|
||||
|
||||
#define FFXM_16BIT_SCALAR( TypeName, BaseComponentType ) typedef BaseComponentType TypeName;
|
||||
#define FFXM_16BIT_VECTOR( TypeName, BaseComponentType, COL ) typedef BaseComponentType##COL TypeName;
|
||||
#define FFXM_16BIT_MATRIX( TypeName, BaseComponentType, ROW, COL ) typedef matrix<BaseComponentType, ROW, COL> TypeName;
|
||||
|
||||
#endif //FFXM_HALF
|
||||
|
||||
#if defined(FFXM_GPU)
|
||||
// Common typedefs:
|
||||
#if defined(FFXM_HLSL)
|
||||
FFXM_MIN16_SCALAR( FFXM_MIN16_F , float );
|
||||
FFXM_MIN16_VECTOR( FFXM_MIN16_F2, float, 2 );
|
||||
FFXM_MIN16_VECTOR( FFXM_MIN16_F3, float, 3 );
|
||||
FFXM_MIN16_VECTOR( FFXM_MIN16_F4, float, 4 );
|
||||
|
||||
FFXM_MIN16_SCALAR( FFXM_MIN16_I, int );
|
||||
FFXM_MIN16_VECTOR( FFXM_MIN16_I2, int, 2 );
|
||||
FFXM_MIN16_VECTOR( FFXM_MIN16_I3, int, 3 );
|
||||
FFXM_MIN16_VECTOR( FFXM_MIN16_I4, int, 4 );
|
||||
|
||||
FFXM_MIN16_SCALAR( FFXM_MIN16_U, uint );
|
||||
FFXM_MIN16_VECTOR( FFXM_MIN16_U2, uint, 2 );
|
||||
FFXM_MIN16_VECTOR( FFXM_MIN16_U3, uint, 3 );
|
||||
FFXM_MIN16_VECTOR( FFXM_MIN16_U4, uint, 4 );
|
||||
|
||||
FFXM_16BIT_SCALAR( FFXM_F16_t , float );
|
||||
FFXM_16BIT_VECTOR( FFXM_F16_t2, float, 2 );
|
||||
FFXM_16BIT_VECTOR( FFXM_F16_t3, float, 3 );
|
||||
FFXM_16BIT_VECTOR( FFXM_F16_t4, float, 4 );
|
||||
|
||||
FFXM_16BIT_SCALAR( FFXM_I16_t, int );
|
||||
FFXM_16BIT_VECTOR( FFXM_I16_t2, int, 2 );
|
||||
FFXM_16BIT_VECTOR( FFXM_I16_t3, int, 3 );
|
||||
FFXM_16BIT_VECTOR( FFXM_I16_t4, int, 4 );
|
||||
|
||||
FFXM_16BIT_SCALAR( FFXM_U16_t, uint );
|
||||
FFXM_16BIT_VECTOR( FFXM_U16_t2, uint, 2 );
|
||||
FFXM_16BIT_VECTOR( FFXM_U16_t3, uint, 3 );
|
||||
FFXM_16BIT_VECTOR( FFXM_U16_t4, uint, 4 );
|
||||
|
||||
#define TYPEDEF_MIN16_TYPES(Prefix) \
|
||||
typedef FFXM_MIN16_F Prefix##_F; \
|
||||
typedef FFXM_MIN16_F2 Prefix##_F2; \
|
||||
typedef FFXM_MIN16_F3 Prefix##_F3; \
|
||||
typedef FFXM_MIN16_F4 Prefix##_F4; \
|
||||
typedef FFXM_MIN16_I Prefix##_I; \
|
||||
typedef FFXM_MIN16_I2 Prefix##_I2; \
|
||||
typedef FFXM_MIN16_I3 Prefix##_I3; \
|
||||
typedef FFXM_MIN16_I4 Prefix##_I4; \
|
||||
typedef FFXM_MIN16_U Prefix##_U; \
|
||||
typedef FFXM_MIN16_U2 Prefix##_U2; \
|
||||
typedef FFXM_MIN16_U3 Prefix##_U3; \
|
||||
typedef FFXM_MIN16_U4 Prefix##_U4;
|
||||
|
||||
#define TYPEDEF_16BIT_TYPES(Prefix) \
|
||||
typedef FFXM_16BIT_F Prefix##_F; \
|
||||
typedef FFXM_16BIT_F2 Prefix##_F2; \
|
||||
typedef FFXM_16BIT_F3 Prefix##_F3; \
|
||||
typedef FFXM_16BIT_F4 Prefix##_F4; \
|
||||
typedef FFXM_16BIT_I Prefix##_I; \
|
||||
typedef FFXM_16BIT_I2 Prefix##_I2; \
|
||||
typedef FFXM_16BIT_I3 Prefix##_I3; \
|
||||
typedef FFXM_16BIT_I4 Prefix##_I4; \
|
||||
typedef FFXM_16BIT_U Prefix##_U; \
|
||||
typedef FFXM_16BIT_U2 Prefix##_U2; \
|
||||
typedef FFXM_16BIT_U3 Prefix##_U3; \
|
||||
typedef FFXM_16BIT_U4 Prefix##_U4;
|
||||
|
||||
#define TYPEDEF_FULL_PRECISION_TYPES(Prefix) \
|
||||
typedef FfxFloat32 Prefix##_F; \
|
||||
typedef FfxFloat32x2 Prefix##_F2; \
|
||||
typedef FfxFloat32x3 Prefix##_F3; \
|
||||
typedef FfxFloat32x4 Prefix##_F4; \
|
||||
typedef FfxInt32 Prefix##_I; \
|
||||
typedef FfxInt32x2 Prefix##_I2; \
|
||||
typedef FfxInt32x3 Prefix##_I3; \
|
||||
typedef FfxInt32x4 Prefix##_I4; \
|
||||
typedef FfxUInt32 Prefix##_U; \
|
||||
typedef FfxUInt32x2 Prefix##_U2; \
|
||||
typedef FfxUInt32x3 Prefix##_U3; \
|
||||
typedef FfxUInt32x4 Prefix##_U4;
|
||||
#endif // #if defined(FFXM_HLSL)
|
||||
|
||||
#if defined(FFXM_GLSL)
|
||||
|
||||
#if FFXM_HALF
|
||||
|
||||
#define FFXM_MIN16_F float16_t
|
||||
#define FFXM_MIN16_F2 f16vec2
|
||||
#define FFXM_MIN16_F3 f16vec3
|
||||
#define FFXM_MIN16_F4 f16vec4
|
||||
|
||||
#define FFXM_MIN16_I int16_t
|
||||
#define FFXM_MIN16_I2 i16vec2
|
||||
#define FFXM_MIN16_I3 i16vec3
|
||||
#define FFXM_MIN16_I4 i16vec4
|
||||
|
||||
#define FFXM_MIN16_U uint16_t
|
||||
#define FFXM_MIN16_U2 u16vec2
|
||||
#define FFXM_MIN16_U3 u16vec3
|
||||
#define FFXM_MIN16_U4 u16vec4
|
||||
|
||||
#define FFXM_16BIT_F float16_t
|
||||
#define FFXM_16BIT_F2 f16vec2
|
||||
#define FFXM_16BIT_F3 f16vec3
|
||||
#define FFXM_16BIT_F4 f16vec4
|
||||
|
||||
#define FFXM_16BIT_I int16_t
|
||||
#define FFXM_16BIT_I2 i16vec2
|
||||
#define FFXM_16BIT_I3 i16vec3
|
||||
#define FFXM_16BIT_I4 i16vec4
|
||||
|
||||
#define FFXM_16BIT_U uint16_t
|
||||
#define FFXM_16BIT_U2 u16vec2
|
||||
#define FFXM_16BIT_U3 u16vec3
|
||||
#define FFXM_16BIT_U4 u16vec4
|
||||
|
||||
#else // FFXM_HALF
|
||||
|
||||
#define FFXM_MIN16_F float
|
||||
#define FFXM_MIN16_F2 vec2
|
||||
#define FFXM_MIN16_F3 vec3
|
||||
#define FFXM_MIN16_F4 vec4
|
||||
|
||||
#define FFXM_MIN16_I int
|
||||
#define FFXM_MIN16_I2 ivec2
|
||||
#define FFXM_MIN16_I3 ivec3
|
||||
#define FFXM_MIN16_I4 ivec4
|
||||
|
||||
#define FFXM_MIN16_U uint
|
||||
#define FFXM_MIN16_U2 uvec2
|
||||
#define FFXM_MIN16_U3 uvec3
|
||||
#define FFXM_MIN16_U4 uvec4
|
||||
|
||||
#define FFXM_16BIT_F float
|
||||
#define FFXM_16BIT_F2 vec2
|
||||
#define FFXM_16BIT_F3 vec3
|
||||
#define FFXM_16BIT_F4 vec4
|
||||
|
||||
#define FFXM_16BIT_I int
|
||||
#define FFXM_16BIT_I2 ivec2
|
||||
#define FFXM_16BIT_I3 ivec3
|
||||
#define FFXM_16BIT_I4 ivec4
|
||||
|
||||
#define FFXM_16BIT_U uint
|
||||
#define FFXM_16BIT_U2 uvec2
|
||||
#define FFXM_16BIT_U3 uvec3
|
||||
#define FFXM_16BIT_U4 uvec4
|
||||
|
||||
#endif // FFXM_HALF
|
||||
|
||||
#endif // #if defined(FFXM_GLSL)
|
||||
|
||||
#endif // #if defined(FFXM_GPU)
|
||||
#endif // #ifndef FFXM_COMMON_TYPES_H
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 7735112b854e65849b3f98a63d29fca6
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,72 @@
|
||||
// This file is part of the FidelityFX SDK.
|
||||
//
|
||||
// Copyright (C) 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files(the “Software”), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and /or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions :
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in
|
||||
// all copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
||||
// THE SOFTWARE.
|
||||
|
||||
/// @defgroup FfxGPU GPU
|
||||
/// The FidelityFX SDK GPU References
|
||||
///
|
||||
/// @ingroup ffxSDK
|
||||
|
||||
/// @defgroup FfxHLSL HLSL References
|
||||
/// FidelityFX SDK HLSL GPU References
|
||||
///
|
||||
/// @ingroup FfxGPU
|
||||
|
||||
/// @defgroup FfxGLSL GLSL References
|
||||
/// FidelityFX SDK GLSL GPU References
|
||||
///
|
||||
/// @ingroup FfxGPU
|
||||
|
||||
/// @defgroup FfxGPUEffects FidelityFX GPU References
|
||||
/// FidelityFX Effect GPU Reference Documentation
|
||||
///
|
||||
/// @ingroup FfxGPU
|
||||
|
||||
/// @defgroup GPUCore GPU Core
|
||||
/// GPU defines and functions
|
||||
///
|
||||
/// @ingroup FfxGPU
|
||||
|
||||
#if !defined(FFXM_CORE_H)
|
||||
#define FFXM_CORE_H
|
||||
|
||||
#include "ffxm_common_types.h"
|
||||
|
||||
#if defined(FFXM_CPU)
|
||||
#include "ffxm_core_cpu.h"
|
||||
#endif // #if defined(FFXM_CPU)
|
||||
|
||||
#if defined(FFXM_GLSL) && defined(FFXM_GPU)
|
||||
#include "ffxm_core_glsl.h"
|
||||
#endif // #if defined(FFXM_GLSL) && defined(FFXM_GPU)
|
||||
|
||||
#if defined(FFXM_HLSL) && defined(FFXM_GPU)
|
||||
#include "ffxm_core_hlsl.h"
|
||||
#endif // #if defined(FFXM_HLSL) && defined(FFXM_GPU)
|
||||
|
||||
#if defined(FFXM_GPU)
|
||||
#include "ffxm_core_gpu_common.h"
|
||||
#include "ffxm_core_gpu_common_half.h"
|
||||
#include "ffxm_core_portability.h"
|
||||
#endif // #if defined(FFXM_GPU)
|
||||
#endif // #if !defined(FFXM_CORE_H)
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 8c61a8ebcd9482c43ad8a2942265d70d
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,338 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
/// A define for a true value in a boolean expression.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
#define FFXM_TRUE (1)
|
||||
|
||||
/// A define for a false value in a boolean expression.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
#define FFXM_FALSE (0)
|
||||
|
||||
#if !defined(FFXM_STATIC)
|
||||
/// A define to abstract declaration of static variables and functions.
|
||||
///
|
||||
/// @ingroup CPUTypes
|
||||
#define FFXM_STATIC static
|
||||
#endif // #if !defined(FFXM_STATIC)
|
||||
|
||||
/// @defgroup CPUCore CPU Core
|
||||
/// Core CPU-side defines and functions
|
||||
///
|
||||
/// @ingroup ffxHost
|
||||
|
||||
#ifdef __clang__
|
||||
#pragma clang diagnostic ignored "-Wunused-variable"
|
||||
#endif
|
||||
|
||||
/// Interpret the bit layout of an IEEE-754 floating point value as an unsigned integer.
|
||||
///
|
||||
/// @param [in] x A 32bit floating value.
|
||||
///
|
||||
/// @returns
|
||||
/// An unsigned 32bit integer value containing the bit pattern of <c><i>x</i></c>.
|
||||
///
|
||||
/// @ingroup CPUCore
|
||||
FFXM_STATIC FfxUInt32 ffxAsUInt32(FfxFloat32 x)
|
||||
{
|
||||
union
|
||||
{
|
||||
FfxFloat32 f;
|
||||
FfxUInt32 u;
|
||||
} bits;
|
||||
|
||||
bits.f = x;
|
||||
return bits.u;
|
||||
}
|
||||
|
||||
FFXM_STATIC FfxFloat32 ffxDot2(FfxFloat32x2 a, FfxFloat32x2 b)
|
||||
{
|
||||
return a[0] * b[0] + a[1] * b[1];
|
||||
}
|
||||
|
||||
FFXM_STATIC FfxFloat32 ffxDot3(FfxFloat32x3 a, FfxFloat32x3 b)
|
||||
{
|
||||
return a[0] * b[0] + a[1] * b[1] + a[2] * b[2];
|
||||
}
|
||||
|
||||
FFXM_STATIC FfxFloat32 ffxDot4(FfxFloat32x4 a, FfxFloat32x4 b)
|
||||
{
|
||||
return a[0] * b[0] + a[1] * b[1] + a[2] * b[2] + a[3] * b[3];
|
||||
}
|
||||
|
||||
/// Compute the linear interopation between two values.
|
||||
///
|
||||
/// Implemented by calling the GLSL <c><i>mix</i></c> instrinsic function. Implements the
|
||||
/// following math:
|
||||
///
|
||||
/// (1 - t) * x + t * y
|
||||
///
|
||||
/// @param [in] x The first value to lerp between.
|
||||
/// @param [in] y The second value to lerp between.
|
||||
/// @param [in] t The value to determine how much of <c><i>x</i></c> and how much of <c><i>y</i></c>.
|
||||
///
|
||||
/// @returns
|
||||
/// A linearly interpolated value between <c><i>x</i></c> and <c><i>y</i></c> according to <c><i>t</i></c>.
|
||||
///
|
||||
/// @ingroup CPUCore
|
||||
FFXM_STATIC FfxFloat32 ffxLerp(FfxFloat32 x, FfxFloat32 y, FfxFloat32 t)
|
||||
{
|
||||
return y * t + (-x * t + x);
|
||||
}
|
||||
|
||||
/// Compute the reciprocal of a value.
|
||||
///
|
||||
/// @param [in] x The value to compute the reciprocal for.
|
||||
///
|
||||
/// @returns
|
||||
/// The reciprocal value of <c><i>x</i></c>.
|
||||
///
|
||||
/// @ingroup CPUCore
|
||||
FFXM_STATIC FfxFloat32 ffxReciprocal(FfxFloat32 x)
|
||||
{
|
||||
return 1.0f / x;
|
||||
}
|
||||
|
||||
/// Compute the square root of a value.
|
||||
///
|
||||
/// @param [in] x The first value to compute the min of.
|
||||
///
|
||||
/// @returns
|
||||
/// The the square root of <c><i>x</i></c>.
|
||||
///
|
||||
/// @ingroup CPUCore
|
||||
FFXM_STATIC FfxFloat32 ffxSqrt(FfxFloat32 x)
|
||||
{
|
||||
return sqrt(x);
|
||||
}
|
||||
|
||||
FFXM_STATIC FfxUInt32 AShrSU1(FfxUInt32 a, FfxUInt32 b)
|
||||
{
|
||||
return FfxUInt32(FfxInt32(a) >> FfxInt32(b));
|
||||
}
|
||||
|
||||
/// Compute the factional part of a decimal value.
|
||||
///
|
||||
/// This function calculates <c><i>x - floor(x)</i></c>.
|
||||
///
|
||||
/// @param [in] x The value to compute the fractional part from.
|
||||
///
|
||||
/// @returns
|
||||
/// The fractional part of <c><i>x</i></c>.
|
||||
///
|
||||
/// @ingroup CPUCore
|
||||
FFXM_STATIC FfxFloat32 ffxFract(FfxFloat32 x)
|
||||
{
|
||||
return x - floor(x);
|
||||
}
|
||||
|
||||
/// Compute the reciprocal square root of a value.
|
||||
///
|
||||
/// @param [in] x The value to compute the reciprocal for.
|
||||
///
|
||||
/// @returns
|
||||
/// The reciprocal square root value of <c><i>x</i></c>.
|
||||
///
|
||||
/// @ingroup CPUCore
|
||||
FFXM_STATIC FfxFloat32 rsqrt(FfxFloat32 x)
|
||||
{
|
||||
return ffxReciprocal(ffxSqrt(x));
|
||||
}
|
||||
|
||||
FFXM_STATIC FfxFloat32 ffxMin(FfxFloat32 x, FfxFloat32 y)
|
||||
{
|
||||
return x < y ? x : y;
|
||||
}
|
||||
|
||||
FFXM_STATIC FfxUInt32 ffxMin(FfxUInt32 x, FfxUInt32 y)
|
||||
{
|
||||
return x < y ? x : y;
|
||||
}
|
||||
|
||||
FFXM_STATIC FfxFloat32 ffxMax(FfxFloat32 x, FfxFloat32 y)
|
||||
{
|
||||
return x > y ? x : y;
|
||||
}
|
||||
|
||||
FFXM_STATIC FfxUInt32 ffxMax(FfxUInt32 x, FfxUInt32 y)
|
||||
{
|
||||
return x > y ? x : y;
|
||||
}
|
||||
|
||||
/// Clamp a value to a [0..1] range.
|
||||
///
|
||||
/// @param [in] x The value to clamp to [0..1] range.
|
||||
///
|
||||
/// @returns
|
||||
/// The clamped version of <c><i>x</i></c>.
|
||||
///
|
||||
/// @ingroup CPUCore
|
||||
FFXM_STATIC FfxFloat32 ffxSaturate(FfxFloat32 x)
|
||||
{
|
||||
return ffxMin(1.0f, ffxMax(0.0f, x));
|
||||
}
|
||||
|
||||
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
|
||||
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
|
||||
|
||||
FFXM_STATIC void opAAddOneF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32 b)
|
||||
{
|
||||
d[0] = a[0] + b;
|
||||
d[1] = a[1] + b;
|
||||
d[2] = a[2] + b;
|
||||
return;
|
||||
}
|
||||
|
||||
FFXM_STATIC void opACpyF3(FfxFloat32x3 d, FfxFloat32x3 a)
|
||||
{
|
||||
d[0] = a[0];
|
||||
d[1] = a[1];
|
||||
d[2] = a[2];
|
||||
return;
|
||||
}
|
||||
|
||||
FFXM_STATIC void opAMulF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32x3 b)
|
||||
{
|
||||
d[0] = a[0] * b[0];
|
||||
d[1] = a[1] * b[1];
|
||||
d[2] = a[2] * b[2];
|
||||
return;
|
||||
}
|
||||
|
||||
FFXM_STATIC void opAMulOneF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32 b)
|
||||
{
|
||||
d[0] = a[0] * b;
|
||||
d[1] = a[1] * b;
|
||||
d[2] = a[2] * b;
|
||||
return;
|
||||
}
|
||||
|
||||
FFXM_STATIC void opARcpF3(FfxFloat32x3 d, FfxFloat32x3 a)
|
||||
{
|
||||
d[0] = ffxReciprocal(a[0]);
|
||||
d[1] = ffxReciprocal(a[1]);
|
||||
d[2] = ffxReciprocal(a[2]);
|
||||
return;
|
||||
}
|
||||
|
||||
/// Convert FfxFloat32 to half (in lower 16-bits of output).
|
||||
///
|
||||
/// This function implements the same fast technique that is documented here: ftp://ftp.fox-toolkit.org/pub/fasthalffloatconversion.pdf
|
||||
///
|
||||
/// The function supports denormals.
|
||||
///
|
||||
/// Some conversion rules are to make computations possibly "safer" on the GPU,
|
||||
/// -INF & -NaN -> -65504
|
||||
/// +INF & +NaN -> +65504
|
||||
///
|
||||
/// @param [in] f The 32bit floating point value to convert.
|
||||
///
|
||||
/// @returns
|
||||
/// The closest 16bit floating point value to <c><i>f</i></c>.
|
||||
///
|
||||
/// @ingroup CPUCore
|
||||
FFXM_STATIC FfxUInt32 f32tof16(FfxFloat32 f)
|
||||
{
|
||||
static FfxUInt16 base[512] = {
|
||||
0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000,
|
||||
0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000,
|
||||
0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000,
|
||||
0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000,
|
||||
0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000,
|
||||
0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0001, 0x0002, 0x0004, 0x0008, 0x0010, 0x0020, 0x0040, 0x0080, 0x0100, 0x0200, 0x0400,
|
||||
0x0800, 0x0c00, 0x1000, 0x1400, 0x1800, 0x1c00, 0x2000, 0x2400, 0x2800, 0x2c00, 0x3000, 0x3400, 0x3800, 0x3c00, 0x4000, 0x4400, 0x4800, 0x4c00, 0x5000,
|
||||
0x5400, 0x5800, 0x5c00, 0x6000, 0x6400, 0x6800, 0x6c00, 0x7000, 0x7400, 0x7800, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff,
|
||||
0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff,
|
||||
0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff,
|
||||
0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff,
|
||||
0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff,
|
||||
0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff,
|
||||
0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x7bff, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000,
|
||||
0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000,
|
||||
0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000,
|
||||
0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000,
|
||||
0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000,
|
||||
0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8000, 0x8001, 0x8002,
|
||||
0x8004, 0x8008, 0x8010, 0x8020, 0x8040, 0x8080, 0x8100, 0x8200, 0x8400, 0x8800, 0x8c00, 0x9000, 0x9400, 0x9800, 0x9c00, 0xa000, 0xa400, 0xa800, 0xac00,
|
||||
0xb000, 0xb400, 0xb800, 0xbc00, 0xc000, 0xc400, 0xc800, 0xcc00, 0xd000, 0xd400, 0xd800, 0xdc00, 0xe000, 0xe400, 0xe800, 0xec00, 0xf000, 0xf400, 0xf800,
|
||||
0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff,
|
||||
0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff,
|
||||
0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff,
|
||||
0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff,
|
||||
0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff,
|
||||
0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff, 0xfbff
|
||||
};
|
||||
|
||||
static FfxUInt8 shift[512] = {
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0f, 0x0e, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d,
|
||||
0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x17, 0x16, 0x15, 0x14, 0x13, 0x12, 0x11, 0x10, 0x0f, 0x0e, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d,
|
||||
0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x0d, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18,
|
||||
0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18, 0x18
|
||||
};
|
||||
|
||||
union
|
||||
{
|
||||
FfxFloat32 f;
|
||||
FfxUInt32 u;
|
||||
} bits;
|
||||
|
||||
bits.f = f;
|
||||
FfxUInt32 u = bits.u;
|
||||
FfxUInt32 i = u >> 23;
|
||||
return (FfxUInt32)(base[i]) + ((u & 0x7fffff) >> shift[i]);
|
||||
}
|
||||
|
||||
/// Pack 2x32-bit floating point values in a single 32bit value.
|
||||
///
|
||||
/// This function first converts each component of <c><i>value</i></c> into their nearest 16-bit floating
|
||||
/// point representation, and then stores the X and Y components in the lower and upper 16 bits of the
|
||||
/// 32bit unsigned integer respectively.
|
||||
///
|
||||
/// @param [in] x A 2-dimensional floating point value to convert and pack.
|
||||
///
|
||||
/// @returns
|
||||
/// A packed 32bit value containing 2 16bit floating point values.
|
||||
///
|
||||
/// @ingroup CPUCore
|
||||
FFXM_STATIC FfxUInt32 packHalf2x16(FfxFloat32x2 x)
|
||||
{
|
||||
return f32tof16(x[0]) + (f32tof16(x[1]) << 16);
|
||||
}
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: c8712c6d1271c464eaad4f12f51bb9df
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
File diff suppressed because it is too large
Load Diff
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 4d827c0232dfcc042baaae76e1338571
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
File diff suppressed because it is too large
Load Diff
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: e0499458c2ba17940866481f316cd05e
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
File diff suppressed because it is too large
Load Diff
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: aa5024eb5a51476429a97cd28e5f38c0
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,52 @@
|
||||
// This file is part of the FidelityFX SDK.
|
||||
//
|
||||
// Copyright (C) 2023 Advanced Micro Devices, Inc.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files(the “Software”), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and /or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions :
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in
|
||||
// all copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
|
||||
// THE SOFTWARE.
|
||||
|
||||
FfxFloat32x3 opAAddOneF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32 b)
|
||||
{
|
||||
d = a + ffxBroadcast3(b);
|
||||
return d;
|
||||
}
|
||||
|
||||
FfxFloat32x3 opACpyF3(FfxFloat32x3 d, FfxFloat32x3 a)
|
||||
{
|
||||
d = a;
|
||||
return d;
|
||||
}
|
||||
|
||||
FfxFloat32x3 opAMulF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32x3 b)
|
||||
{
|
||||
d = a * b;
|
||||
return d;
|
||||
}
|
||||
|
||||
FfxFloat32x3 opAMulOneF3(FfxFloat32x3 d, FfxFloat32x3 a, FfxFloat32 b)
|
||||
{
|
||||
d = a * ffxBroadcast3(b);
|
||||
return d;
|
||||
}
|
||||
|
||||
FfxFloat32x3 opARcpF3(FfxFloat32x3 d, FfxFloat32x3 a)
|
||||
{
|
||||
d = rcp(a);
|
||||
return d;
|
||||
}
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 05dc457a61251e040b3f95a705eff4b1
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
File diff suppressed because it is too large
Load Diff
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: ce0215c747039664eabc4b52d3d18cc4
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,380 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#ifndef FFXM_FSR2_ACCUMULATE_H
|
||||
#define FFXM_FSR2_ACCUMULATE_H
|
||||
|
||||
struct AccumulateOutputs
|
||||
{
|
||||
#if !FFXM_SHADER_QUALITY_OPT_SEPARATE_TEMPORAL_REACTIVE
|
||||
FfxFloat32x4 fColorAndWeight;
|
||||
#else
|
||||
FfxFloat32x3 fUpscaledColor;
|
||||
FfxFloat32 fTemporalReactive;
|
||||
#endif
|
||||
FfxFloat32x2 fLockStatus;
|
||||
FfxFloat32x4 fLumaHistory;
|
||||
#if (FFXM_FSR2_OPTION_APPLY_SHARPENING == 0)
|
||||
FfxFloat32x3 fColor;
|
||||
#endif
|
||||
};
|
||||
|
||||
FfxFloat32 GetPxHrVelocity(FfxFloat32x2 fMotionVector)
|
||||
{
|
||||
return length(fMotionVector * DisplaySize());
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F GetPxHrVelocity(FFXM_MIN16_F2 fMotionVector)
|
||||
{
|
||||
return length(fMotionVector * FFXM_MIN16_F2(DisplaySize()));
|
||||
}
|
||||
#endif
|
||||
|
||||
void Accumulate(const AccumulationPassCommonParams params, FFXM_PARAMETER_INOUT FfxFloat32x3 fHistoryColor, FfxFloat32x3 fAccumulation, FFXM_PARAMETER_IN FfxFloat32x4 fUpsampledColorAndWeight)
|
||||
{
|
||||
// Aviod invalid values when accumulation and upsampled weight is 0
|
||||
fAccumulation = ffxMax(FSR2_EPSILON.xxx, fAccumulation + fUpsampledColorAndWeight.www);
|
||||
|
||||
#if FFXM_FSR2_OPTION_HDR_COLOR_INPUT
|
||||
#if FFXM_SHADER_QUALITY_OPT_TONEMAPPED_RGB_PREPARED_INPUT_COLOR
|
||||
fHistoryColor = Tonemap(fHistoryColor);
|
||||
#else
|
||||
//YCoCg -> RGB -> Tonemap -> YCoCg (Use RGB tonemapper to avoid color desaturation)
|
||||
fUpsampledColorAndWeight.xyz = RGBToYCoCg(Tonemap(YCoCgToRGB(fUpsampledColorAndWeight.xyz)));
|
||||
fHistoryColor = RGBToYCoCg(Tonemap(YCoCgToRGB(fHistoryColor)));
|
||||
#endif
|
||||
#endif
|
||||
|
||||
const FfxFloat32x3 fAlpha = fUpsampledColorAndWeight.www / fAccumulation;
|
||||
fHistoryColor = ffxLerp(fHistoryColor, fUpsampledColorAndWeight.xyz, fAlpha);
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_TONEMAPPED_RGB_PREPARED_INPUT_COLOR
|
||||
fHistoryColor = YCoCgToRGB(fHistoryColor);
|
||||
#endif
|
||||
|
||||
#if FFXM_FSR2_OPTION_HDR_COLOR_INPUT
|
||||
fHistoryColor = InverseTonemap(fHistoryColor);
|
||||
#endif
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
void RectifyHistory(
|
||||
const AccumulationPassCommonParams params,
|
||||
RectificationBoxMin16 clippingBox,
|
||||
FFXM_PARAMETER_INOUT FfxFloat32x3 fHistoryColor,
|
||||
FFXM_PARAMETER_INOUT FfxFloat32x3 fAccumulation,
|
||||
FfxFloat32 fLockContributionThisFrame,
|
||||
FfxFloat32 fTemporalReactiveFactor,
|
||||
FfxFloat32 fLumaInstabilityFactor)
|
||||
#else
|
||||
void RectifyHistory(
|
||||
const AccumulationPassCommonParams params,
|
||||
RectificationBox clippingBox,
|
||||
FFXM_PARAMETER_INOUT FfxFloat32x3 fHistoryColor,
|
||||
FFXM_PARAMETER_INOUT FfxFloat32x3 fAccumulation,
|
||||
FfxFloat32 fLockContributionThisFrame,
|
||||
FfxFloat32 fTemporalReactiveFactor,
|
||||
FfxFloat32 fLumaInstabilityFactor)
|
||||
#endif
|
||||
{
|
||||
FfxFloat32 fScaleFactorInfluence = ffxMin(20.0f, ffxPow(FfxFloat32(1.0f / length(DownscaleFactor().x * DownscaleFactor().y)), 3.0f));
|
||||
|
||||
const FfxFloat32 fVecolityFactor = ffxSaturate(params.fHrVelocity / 20.0f);
|
||||
const FfxFloat32 fBoxScaleT = ffxMax(params.fDepthClipFactor, ffxMax(params.fAccumulationMask, fVecolityFactor));
|
||||
FfxFloat32 fBoxScale = ffxLerp(fScaleFactorInfluence, 1.0f, fBoxScaleT);
|
||||
|
||||
FfxFloat32x3 fScaledBoxVec = clippingBox.boxVec * fBoxScale;
|
||||
FfxFloat32x3 boxMin = clippingBox.boxCenter - fScaledBoxVec;
|
||||
FfxFloat32x3 boxMax = clippingBox.boxCenter + fScaledBoxVec;
|
||||
FfxFloat32x3 boxCenter = clippingBox.boxCenter;
|
||||
FfxFloat32 boxVecSize = length(clippingBox.boxVec);
|
||||
|
||||
boxMin = ffxMax(clippingBox.aabbMin, boxMin);
|
||||
boxMax = ffxMin(clippingBox.aabbMax, boxMax);
|
||||
#if FFXM_SHADER_QUALITY_OPT_TONEMAPPED_RGB_PREPARED_INPUT_COLOR
|
||||
boxMin = InverseTonemap(boxMin);
|
||||
boxMax = InverseTonemap(boxMax);
|
||||
#endif
|
||||
|
||||
if (any(FFXM_GREATER_THAN(boxMin, fHistoryColor)) || any(FFXM_GREATER_THAN(fHistoryColor, boxMax))) {
|
||||
|
||||
const FfxFloat32x3 fClampedHistoryColor = clamp(fHistoryColor, boxMin, boxMax);
|
||||
|
||||
FfxFloat32x3 fHistoryContribution = ffxMax(fLumaInstabilityFactor, fLockContributionThisFrame).xxx;
|
||||
|
||||
const FfxFloat32 fReactiveFactor = params.fDilatedReactiveFactor;
|
||||
const FfxFloat32 fReactiveContribution = 1.0f - ffxPow(fReactiveFactor, 1.0f / 2.0f);
|
||||
fHistoryContribution *= fReactiveContribution;
|
||||
|
||||
// Scale history color using rectification info, also using accumulation mask to avoid potential invalid color protection
|
||||
fHistoryColor = ffxLerp(fClampedHistoryColor, fHistoryColor, ffxSaturate(fHistoryContribution));
|
||||
|
||||
// Scale accumulation using rectification info
|
||||
const FfxFloat32x3 fAccumulationMin = ffxMin(fAccumulation, FFXM_BROADCAST_FLOAT32X3(0.1f));
|
||||
fAccumulation = ffxLerp(fAccumulationMin, fAccumulation, ffxSaturate(fHistoryContribution));
|
||||
}
|
||||
}
|
||||
|
||||
void FinalizeLockStatus(const AccumulationPassCommonParams params, FfxFloat32x2 fLockStatus, FfxFloat32 fUpsampledWeight, FFXM_PARAMETER_INOUT AccumulateOutputs result)
|
||||
{
|
||||
// we expect similar motion for next frame
|
||||
// kill lock if that location is outside screen, avoid locks to be clamped to screen borders
|
||||
FfxFloat32x2 fEstimatedUvNextFrame = params.fHrUv - params.fMotionVector;
|
||||
if (IsUvInside(fEstimatedUvNextFrame) == false) {
|
||||
KillLock(fLockStatus);
|
||||
}
|
||||
else {
|
||||
// Decrease lock lifetime
|
||||
const FfxFloat32 fLifetimeDecreaseLanczosMax = FfxFloat32(JitterSequenceLength()) * FfxFloat32(fAverageLanczosWeightPerFrame);
|
||||
const FfxFloat32 fLifetimeDecrease = FfxFloat32(fUpsampledWeight / fLifetimeDecreaseLanczosMax);
|
||||
fLockStatus[LOCK_LIFETIME_REMAINING] = ffxMax(FfxFloat32(0), fLockStatus[LOCK_LIFETIME_REMAINING] - fLifetimeDecrease);
|
||||
}
|
||||
|
||||
result.fLockStatus = fLockStatus;
|
||||
}
|
||||
|
||||
|
||||
FfxFloat32x3 ComputeBaseAccumulationWeight(const AccumulationPassCommonParams params, FfxFloat32 fThisFrameReactiveFactor, FfxBoolean bInMotionLastFrame, FfxFloat32 fUpsampledWeight, LockState lockState)
|
||||
{
|
||||
// Always assume max accumulation was reached
|
||||
FfxFloat32 fBaseAccumulation = fMaxAccumulationLanczosWeight * FfxFloat32(params.bIsExistingSample) * (1.0f - fThisFrameReactiveFactor) * (1.0f - params.fDepthClipFactor);
|
||||
|
||||
fBaseAccumulation = ffxMin(fBaseAccumulation, ffxLerp(fBaseAccumulation, fUpsampledWeight * 10.0f, ffxMax(FfxFloat32(bInMotionLastFrame), ffxSaturate(params.fHrVelocity * FfxFloat32(10)))));
|
||||
|
||||
fBaseAccumulation = ffxMin(fBaseAccumulation, ffxLerp(fBaseAccumulation, fUpsampledWeight, ffxSaturate(params.fHrVelocity / FfxFloat32(20))));
|
||||
|
||||
return fBaseAccumulation.xxx;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FfxFloat32 ComputeLumaInstabilityFactor(const AccumulationPassCommonParams params, RectificationBoxMin16 clippingBox, FfxFloat32 fThisFrameReactiveFactor, FfxFloat32 fLuminanceDiff, FFXM_PARAMETER_INOUT AccumulateOutputs result)
|
||||
#else
|
||||
FfxFloat32 ComputeLumaInstabilityFactor(const AccumulationPassCommonParams params, RectificationBox clippingBox, FfxFloat32 fThisFrameReactiveFactor, FfxFloat32 fLuminanceDiff, FFXM_PARAMETER_INOUT AccumulateOutputs result)
|
||||
#endif
|
||||
{
|
||||
const FfxFloat32 fUnormThreshold = 1.0f / 255.0f;
|
||||
const FfxInt32 N_MINUS_1 = 0;
|
||||
const FfxInt32 N_MINUS_2 = 1;
|
||||
const FfxInt32 N_MINUS_3 = 2;
|
||||
const FfxInt32 N_MINUS_4 = 3;
|
||||
|
||||
FfxFloat32 fCurrentFrameLuma = clippingBox.boxCenter.x;
|
||||
|
||||
#if FFXM_FSR2_OPTION_HDR_COLOR_INPUT
|
||||
fCurrentFrameLuma = fCurrentFrameLuma / (1.0f + ffxMax(0.0f, fCurrentFrameLuma));
|
||||
#endif
|
||||
|
||||
fCurrentFrameLuma = round(fCurrentFrameLuma * 255.0f) / 255.0f;
|
||||
|
||||
const FfxBoolean bSampleLumaHistory = (ffxMax(ffxMax(params.fDepthClipFactor, params.fAccumulationMask), fLuminanceDiff) < 0.1f) && (params.bIsNewSample == false);
|
||||
FfxFloat32x4 fCurrentFrameLumaHistory = bSampleLumaHistory ? SampleLumaHistory(params.fReprojectedHrUv) : FFXM_BROADCAST_FLOAT32X4(0.0f);
|
||||
|
||||
FfxFloat32 fLumaInstability = 0.0f;
|
||||
FfxFloat32 fDiffs0 = (fCurrentFrameLuma - fCurrentFrameLumaHistory[N_MINUS_1]);
|
||||
|
||||
FfxFloat32 fMin = abs(fDiffs0);
|
||||
|
||||
if (fMin >= fUnormThreshold) {
|
||||
for (int i = N_MINUS_2; i <= N_MINUS_4; i++) {
|
||||
FfxFloat32 fDiffs1 = (fCurrentFrameLuma - fCurrentFrameLumaHistory[i]);
|
||||
|
||||
if (sign(fDiffs0) == sign(fDiffs1)) {
|
||||
|
||||
// Scale difference to protect historically similar values
|
||||
const FfxFloat32 fMinBias = 1.0f;
|
||||
fMin = ffxMin(fMin, abs(fDiffs1) * fMinBias);
|
||||
}
|
||||
}
|
||||
|
||||
const FfxFloat32 fBoxSize = clippingBox.boxVec.x;
|
||||
const FfxFloat32 fBoxSizeFactor = ffxPow(ffxSaturate(fBoxSize / 0.1f), 6.0f);
|
||||
|
||||
fLumaInstability = FfxFloat32(fMin != abs(fDiffs0)) * fBoxSizeFactor;
|
||||
fLumaInstability = FfxFloat32(fLumaInstability > fUnormThreshold);
|
||||
|
||||
fLumaInstability *= 1.0f - ffxMax(params.fAccumulationMask, ffxPow(fThisFrameReactiveFactor, 1.0f / 6.0f));
|
||||
}
|
||||
|
||||
//shift history
|
||||
fCurrentFrameLumaHistory[N_MINUS_4] = fCurrentFrameLumaHistory[N_MINUS_3];
|
||||
fCurrentFrameLumaHistory[N_MINUS_3] = fCurrentFrameLumaHistory[N_MINUS_2];
|
||||
fCurrentFrameLumaHistory[N_MINUS_2] = fCurrentFrameLumaHistory[N_MINUS_1];
|
||||
fCurrentFrameLumaHistory[N_MINUS_1] = fCurrentFrameLuma;
|
||||
|
||||
result.fLumaHistory = fCurrentFrameLumaHistory;
|
||||
|
||||
return fLumaInstability * FfxFloat32(fCurrentFrameLumaHistory[N_MINUS_4] != 0);
|
||||
}
|
||||
|
||||
FfxFloat32 ComputeTemporalReactiveFactor(const AccumulationPassCommonParams params, FfxFloat32 fTemporalReactiveFactor)
|
||||
{
|
||||
FfxFloat32 fNewFactor = ffxMin(0.99f, fTemporalReactiveFactor);
|
||||
|
||||
fNewFactor = ffxMax(fNewFactor, ffxLerp(fNewFactor, 0.4f, ffxSaturate(params.fHrVelocity)));
|
||||
|
||||
fNewFactor = ffxMax(fNewFactor * fNewFactor, ffxMax(params.fDepthClipFactor * 0.1f, params.fDilatedReactiveFactor));
|
||||
|
||||
// Force reactive factor for new samples
|
||||
fNewFactor = params.bIsNewSample ? 1.0f : fNewFactor;
|
||||
|
||||
if (ffxSaturate(params.fHrVelocity * 10.0f) >= 1.0f) {
|
||||
fNewFactor = ffxMax(FSR2_EPSILON, fNewFactor) * -1.0f;
|
||||
}
|
||||
|
||||
return fNewFactor;
|
||||
}
|
||||
|
||||
void initReactiveMaskFactors(FFXM_PARAMETER_INOUT AccumulationPassCommonParams params)
|
||||
{
|
||||
const FFXM_MIN16_F2 fDilatedReactiveMasks = FFXM_MIN16_F2(SampleDilatedReactiveMasks(params.fLrUv_HwSampler));
|
||||
params.fDilatedReactiveFactor = fDilatedReactiveMasks.x;
|
||||
params.fAccumulationMask = fDilatedReactiveMasks.y;
|
||||
}
|
||||
|
||||
void initDepthClipFactors(FFXM_PARAMETER_INOUT AccumulationPassCommonParams params)
|
||||
{
|
||||
params.fDepthClipFactor = FFXM_MIN16_F(ffxSaturate(SampleDepthClip(params.fLrUv_HwSampler)));
|
||||
}
|
||||
|
||||
void initIsNewSample(FFXM_PARAMETER_INOUT AccumulationPassCommonParams params)
|
||||
{
|
||||
const FfxBoolean bIsResetFrame = (0 == FrameIndex());
|
||||
params.bIsNewSample = (params.bIsExistingSample == false || bIsResetFrame);
|
||||
}
|
||||
|
||||
|
||||
AccumulationPassCommonParams InitParams(FfxInt32x2 iPxHrPos)
|
||||
{
|
||||
AccumulationPassCommonParams params;
|
||||
|
||||
params.iPxHrPos = iPxHrPos;
|
||||
const FfxFloat32x2 fHrUv = (iPxHrPos + 0.5f) / DisplaySize();
|
||||
params.fHrUv = fHrUv;
|
||||
|
||||
const FfxFloat32x2 fLrUvJittered = fHrUv + Jitter() / RenderSize();
|
||||
params.fLrUv_HwSampler = ClampUv(fLrUvJittered, RenderSize(), MaxRenderSize());
|
||||
|
||||
params.fMotionVector = GetMotionVector(iPxHrPos, fHrUv);
|
||||
params.fHrVelocity = GetPxHrVelocity(params.fMotionVector);
|
||||
|
||||
ComputeReprojectedUVs(params, params.fReprojectedHrUv, params.bIsExistingSample);
|
||||
|
||||
return params;
|
||||
}
|
||||
|
||||
AccumulateOutputs Accumulate(FfxInt32x2 iPxHrPos)
|
||||
{
|
||||
AccumulationPassCommonParams params = InitParams(iPxHrPos);
|
||||
|
||||
FfxFloat32x3 fHistoryColor = FfxFloat32x3(0, 0, 0);
|
||||
FFXM_MIN16_F2 fLockStatus;
|
||||
InitializeNewLockSample(fLockStatus);
|
||||
|
||||
FFXM_MIN16_F fTemporalReactiveFactor = FFXM_MIN16_F(0.0f);
|
||||
FfxBoolean bInMotionLastFrame = FFXM_FALSE;
|
||||
LockState lockState = { FFXM_FALSE , FFXM_FALSE };
|
||||
const FfxBoolean bIsResetFrame = (0 == FrameIndex());
|
||||
if (params.bIsExistingSample && !bIsResetFrame) {
|
||||
ReprojectHistoryColor(params, fHistoryColor, fTemporalReactiveFactor, bInMotionLastFrame);
|
||||
lockState = ReprojectHistoryLockStatus(params, fLockStatus);
|
||||
}
|
||||
|
||||
initReactiveMaskFactors(params);
|
||||
initDepthClipFactors(params);
|
||||
|
||||
FfxFloat32 fThisFrameReactiveFactor = ffxMax(params.fDilatedReactiveFactor, fTemporalReactiveFactor);
|
||||
|
||||
FfxFloat32 fLuminanceDiff = 0.0f;
|
||||
FfxFloat32 fLockContributionThisFrame = 0.0f;
|
||||
FfxFloat32x2 fLockStatus32 = {fLockStatus.x, fLockStatus.y};
|
||||
UpdateLockStatus(params, fThisFrameReactiveFactor, lockState, fLockStatus32, fLockContributionThisFrame, fLuminanceDiff);
|
||||
fLockStatus = FFXM_MIN16_F2(fLockStatus32);
|
||||
|
||||
#ifdef FFXM_HLSL
|
||||
AccumulateOutputs results = (AccumulateOutputs)0;
|
||||
#else
|
||||
AccumulateOutputs results;
|
||||
#endif
|
||||
|
||||
// Load upsampled input color
|
||||
#if FFXM_HALF
|
||||
#ifdef FFXM_HLSL
|
||||
RectificationBoxMin16 clippingBox = (RectificationBoxMin16)0;
|
||||
#else
|
||||
RectificationBoxMin16 clippingBox;
|
||||
#endif
|
||||
#else
|
||||
#ifdef FFXM_HLSL
|
||||
RectificationBox clippingBox = (RectificationBox)0;
|
||||
#else
|
||||
RectificationBox clippingBox;
|
||||
#endif
|
||||
#endif
|
||||
|
||||
initIsNewSample(params);
|
||||
|
||||
FfxFloat32x4 fUpsampledColorAndWeight = ComputeUpsampledColorAndWeight(params, clippingBox, fThisFrameReactiveFactor);
|
||||
|
||||
FinalizeLockStatus(params, fLockStatus, fUpsampledColorAndWeight.w, results);
|
||||
|
||||
#if FFXM_SHADER_QUALITY_OPT_DISABLE_LUMA_INSTABILITY
|
||||
const FfxFloat32 fLumaInstabilityFactor = 0.0f;
|
||||
#else
|
||||
const FfxFloat32 fLumaInstabilityFactor = ComputeLumaInstabilityFactor(params, clippingBox, fThisFrameReactiveFactor, fLuminanceDiff, results);
|
||||
#endif
|
||||
|
||||
FfxFloat32x3 fAccumulation = ComputeBaseAccumulationWeight(params, fThisFrameReactiveFactor, bInMotionLastFrame, fUpsampledColorAndWeight.w, lockState);
|
||||
|
||||
if (params.bIsNewSample) {
|
||||
#if FFXM_SHADER_QUALITY_OPT_TONEMAPPED_RGB_PREPARED_INPUT_COLOR
|
||||
fHistoryColor = InverseTonemap(fUpsampledColorAndWeight.xyz);
|
||||
#else
|
||||
fHistoryColor = YCoCgToRGB(fUpsampledColorAndWeight.xyz);
|
||||
#endif
|
||||
}
|
||||
else {
|
||||
RectifyHistory(params, clippingBox, fHistoryColor, fAccumulation, fLockContributionThisFrame, fThisFrameReactiveFactor, fLumaInstabilityFactor);
|
||||
|
||||
Accumulate(params, fHistoryColor, fAccumulation, fUpsampledColorAndWeight);
|
||||
}
|
||||
|
||||
fHistoryColor = UnprepareRgb(fHistoryColor, Exposure());
|
||||
|
||||
// Get new temporal reactive factor
|
||||
fTemporalReactiveFactor = FFXM_MIN16_F(ComputeTemporalReactiveFactor(params, fThisFrameReactiveFactor));
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_SEPARATE_TEMPORAL_REACTIVE
|
||||
results.fColorAndWeight = FfxFloat32x4(fHistoryColor, fTemporalReactiveFactor);
|
||||
#else
|
||||
// Output the upscaled color and the temporal reactive factor if these are contained in separate textures
|
||||
results.fUpscaledColor = fHistoryColor;
|
||||
results.fTemporalReactive = fTemporalReactiveFactor;
|
||||
#endif
|
||||
// Output final color when RCAS is disabled
|
||||
#if FFXM_FSR2_OPTION_APPLY_SHARPENING == 0
|
||||
results.fColor = fHistoryColor;
|
||||
#endif
|
||||
|
||||
return results;
|
||||
}
|
||||
|
||||
#endif // FFXM_FSR2_ACCUMULATE_H
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 5a36cab1c2463cb4d8643abb8be995e7
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,103 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#define FSR2_BIND_SRV_INPUT_EXPOSURE 0
|
||||
#define FSR2_BIND_SRV_DILATED_REACTIVE_MASKS 1
|
||||
#if FFXM_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS
|
||||
#define FSR2_BIND_SRV_DILATED_MOTION_VECTORS 2
|
||||
#else
|
||||
#define FSR2_BIND_SRV_INPUT_MOTION_VECTORS 2
|
||||
#endif
|
||||
#define FSR2_BIND_SRV_INTERNAL_UPSCALED 3
|
||||
#define FSR2_BIND_SRV_LOCK_STATUS 4
|
||||
#define FSR2_BIND_SRV_PREPARED_INPUT_COLOR 5
|
||||
#define FSR2_BIND_SRV_LANCZOS_LUT 6
|
||||
#define FSR2_BIND_SRV_UPSCALE_MAXIMUM_BIAS_LUT 7
|
||||
#define FSR2_BIND_SRV_SCENE_LUMINANCE_MIPS 8
|
||||
#define FSR2_BIND_SRV_AUTO_EXPOSURE 9
|
||||
#define FSR2_BIND_SRV_LUMA_HISTORY 10
|
||||
#define FSR2_BIND_SRV_TEMPORAL_REACTIVE 11
|
||||
#define FSR2_BIND_SRV_NEW_LOCKS 12
|
||||
|
||||
#define FSR2_BIND_CB_FSR2 0
|
||||
|
||||
// Global mandatory defines
|
||||
#if !defined(FFXM_HALF)
|
||||
#define FFXM_HALF 1
|
||||
#endif
|
||||
#if !defined(FFXM_GPU)
|
||||
#define FFXM_GPU 1
|
||||
#endif
|
||||
#if !defined(FFXM_HLSL)
|
||||
#define FFXM_HLSL 1
|
||||
#endif
|
||||
|
||||
#include "ffxm_fsr2_callbacks_hlsl.h"
|
||||
#include "ffxm_fsr2_common.h"
|
||||
#include "ffxm_fsr2_sample.h"
|
||||
#include "ffxm_fsr2_upsample.h"
|
||||
#include "ffxm_fsr2_postprocess_lock_status.h"
|
||||
#include "ffxm_fsr2_reproject.h"
|
||||
#include "ffxm_fsr2_accumulate.h"
|
||||
|
||||
struct VertexOut
|
||||
{
|
||||
float4 position : SV_POSITION;
|
||||
};
|
||||
|
||||
struct AccumulateOutputsFS
|
||||
{
|
||||
#if !FFXM_SHADER_QUALITY_BALANCED_OR_PERFORMANCE
|
||||
FfxFloat32x4 fColorAndWeight : SV_TARGET0;
|
||||
FfxFloat32x2 fLockStatus : SV_TARGET1;
|
||||
FfxFloat32x4 fLumaHistory : SV_TARGET2;
|
||||
#if FFXM_FSR2_OPTION_APPLY_SHARPENING == 0
|
||||
FfxFloat32x3 fColor : SV_TARGET3;
|
||||
#endif
|
||||
#else // FFXM_SHADER_QUALITY_BALANCED_OR_PERFORMANCE
|
||||
FfxFloat32x3 fUpscaledColor : SV_TARGET0;
|
||||
FfxFloat32 fTemporalReactive : SV_TARGET1;
|
||||
FfxFloat32x2 fLockStatus : SV_TARGET2;
|
||||
#if FFXM_FSR2_OPTION_APPLY_SHARPENING == 0
|
||||
FfxFloat32x3 fColor : SV_TARGET3;
|
||||
#endif
|
||||
#endif
|
||||
};
|
||||
|
||||
AccumulateOutputsFS main(float4 SvPosition : SV_POSITION)
|
||||
{
|
||||
uint2 uPixelCoord = uint2(SvPosition.xy);
|
||||
AccumulateOutputs result = Accumulate(uPixelCoord);
|
||||
AccumulateOutputsFS output = (AccumulateOutputsFS)0;
|
||||
#if !FFXM_SHADER_QUALITY_BALANCED_OR_PERFORMANCE
|
||||
output.fColorAndWeight = result.fColorAndWeight;
|
||||
output.fLumaHistory = result.fLumaHistory;
|
||||
#else
|
||||
output.fUpscaledColor = result.fUpscaledColor;
|
||||
output.fTemporalReactive = result.fTemporalReactive;
|
||||
#endif
|
||||
output.fLockStatus = result.fLockStatus;
|
||||
#if FFXM_FSR2_OPTION_APPLY_SHARPENING == 0
|
||||
output.fColor = result.fColor;
|
||||
#endif
|
||||
return output;
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: fe41a21263058324f963a49e20618f53
|
||||
ShaderIncludeImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,84 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#define FSR2_BIND_SRV_INPUT_OPAQUE_ONLY 0
|
||||
#define FSR2_BIND_SRV_INPUT_COLOR 1
|
||||
|
||||
#define FSR2_BIND_CB_FSR2 0
|
||||
#define FSR2_BIND_CB_REACTIVE 1
|
||||
|
||||
// Global mandatory defines
|
||||
#if !defined(FFXM_HALF)
|
||||
#define FFXM_HALF 1
|
||||
#endif
|
||||
#if !defined(FFXM_GPU)
|
||||
#define FFXM_GPU 1
|
||||
#endif
|
||||
#if !defined(FFXM_HLSL)
|
||||
#define FFXM_HLSL 1
|
||||
#endif
|
||||
|
||||
#include "ffxm_fsr2_callbacks_hlsl.h"
|
||||
#include "ffxm_fsr2_common.h"
|
||||
|
||||
struct GenReactiveMaskOutputs
|
||||
{
|
||||
FfxFloat32 fReactiveMask : SV_TARGET0;
|
||||
};
|
||||
|
||||
struct VertexOut
|
||||
{
|
||||
float4 position : SV_POSITION;
|
||||
};
|
||||
|
||||
GenReactiveMaskOutputs main(float4 SvPosition : SV_POSITION)
|
||||
{
|
||||
uint2 uPixelCoord = uint2(SvPosition.xy);
|
||||
|
||||
float3 ColorPreAlpha = LoadOpaqueOnly( FFXM_MIN16_I2(uPixelCoord) ).rgb;
|
||||
float3 ColorPostAlpha = LoadInputColor(uPixelCoord).rgb;
|
||||
|
||||
if (GenReactiveFlags() & FFXM_FSR2_AUTOREACTIVEFLAGS_APPLY_TONEMAP)
|
||||
{
|
||||
ColorPreAlpha = Tonemap(ColorPreAlpha);
|
||||
ColorPostAlpha = Tonemap(ColorPostAlpha);
|
||||
}
|
||||
|
||||
if (GenReactiveFlags() & FFXM_FSR2_AUTOREACTIVEFLAGS_APPLY_INVERSETONEMAP)
|
||||
{
|
||||
ColorPreAlpha = InverseTonemap(ColorPreAlpha);
|
||||
ColorPostAlpha = InverseTonemap(ColorPostAlpha);
|
||||
}
|
||||
|
||||
float out_reactive_value = 0.f;
|
||||
float3 delta = abs(ColorPostAlpha - ColorPreAlpha);
|
||||
|
||||
out_reactive_value = (GenReactiveFlags() & FFXM_FSR2_AUTOREACTIVEFLAGS_USE_COMPONENTS_MAX) ? max(delta.x, max(delta.y, delta.z)) : length(delta);
|
||||
out_reactive_value *= GenReactiveScale();
|
||||
|
||||
out_reactive_value = (GenReactiveFlags() & FFXM_FSR2_AUTOREACTIVEFLAGS_APPLY_THRESHOLD) ? (out_reactive_value < GenReactiveThreshold() ? 0 : GenReactiveBinaryValue()) : out_reactive_value;
|
||||
|
||||
GenReactiveMaskOutputs results = (GenReactiveMaskOutputs)0;
|
||||
results.fReactiveMask = out_reactive_value;
|
||||
|
||||
return results;
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: fd9d546b114abca4cad6326ff4f6769c
|
||||
ShaderIncludeImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
File diff suppressed because it is too large
Load Diff
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: c905e516c694e204cb184632584a74e6
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,596 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#if !defined(FFXM_FSR2_COMMON_H)
|
||||
#define FFXM_FSR2_COMMON_H
|
||||
|
||||
#if defined(FFXM_CPU) || defined(FFXM_GPU)
|
||||
//Locks
|
||||
#define LOCK_LIFETIME_REMAINING 0
|
||||
#define LOCK_TEMPORAL_LUMA 1
|
||||
#endif // #if defined(FFXM_CPU) || defined(FFXM_GPU)
|
||||
|
||||
#if defined(FFXM_GPU)
|
||||
FFXM_STATIC const FfxFloat32 FSR2_FP16_MIN = 6.10e-05f;
|
||||
FFXM_STATIC const FfxFloat32 FSR2_FP16_MAX = 65504.0f;
|
||||
FFXM_STATIC const FfxFloat32 FSR2_EPSILON = 1e-03f;
|
||||
FFXM_STATIC const FfxFloat32 FSR2_TONEMAP_EPSILON = 1.0f / FSR2_FP16_MAX;
|
||||
FFXM_STATIC const FfxFloat32 FSR2_FLT_MAX = 3.402823466e+38f;
|
||||
FFXM_STATIC const FfxFloat32 FSR2_FLT_MIN = 1.175494351e-38f;
|
||||
|
||||
// treat vector truncation warnings as errors
|
||||
#pragma warning(error: 3206)
|
||||
|
||||
// suppress warnings
|
||||
#pragma warning(disable: 3205) // conversion from larger type to smaller
|
||||
#pragma warning(disable: 3571) // in ffxPow(f, e), f could be negative
|
||||
|
||||
// Reconstructed depth usage
|
||||
FFXM_STATIC const FfxFloat32 fReconstructedDepthBilinearWeightThreshold = 0.01f;
|
||||
|
||||
// Accumulation
|
||||
#if !FFXM_SHADER_QUALITY_OPT_UPSCALING_LANCZOS_5TAP
|
||||
FFXM_STATIC const FfxFloat32 fUpsampleLanczosWeightScale = 1.0f / 9.0f;
|
||||
#else
|
||||
FFXM_STATIC const FfxFloat32 fUpsampleLanczosWeightScale = 1.0f / 5.0f;
|
||||
#endif
|
||||
FFXM_STATIC const FfxFloat32 fMaxAccumulationLanczosWeight = 1.0f;
|
||||
FFXM_STATIC const FfxFloat32 fAverageLanczosWeightPerFrame = 0.74f * fUpsampleLanczosWeightScale; // Average lanczos weight for jitter accumulated samples
|
||||
FFXM_STATIC const FfxFloat32 fAccumulationMaxOnMotion = 3.0f * fUpsampleLanczosWeightScale;
|
||||
|
||||
// Auto exposure
|
||||
FFXM_STATIC const FfxFloat32 resetAutoExposureAverageSmoothing = 1e8f;
|
||||
|
||||
// Optimizations defines
|
||||
#ifndef FFXM_OPT_USE_GATHER_OPS
|
||||
#define FFXM_OPT_USE_GATHER_OPS 0
|
||||
#endif
|
||||
|
||||
struct AccumulationPassCommonParams
|
||||
{
|
||||
FfxInt32x2 iPxHrPos;
|
||||
FfxFloat32x2 fHrUv;
|
||||
FfxFloat32x2 fLrUv_HwSampler;
|
||||
FfxFloat32x2 fMotionVector;
|
||||
FfxFloat32x2 fReprojectedHrUv;
|
||||
FfxFloat32 fHrVelocity;
|
||||
FFXM_MIN16_F fDepthClipFactor;
|
||||
FFXM_MIN16_F fDilatedReactiveFactor;
|
||||
FFXM_MIN16_F fAccumulationMask;
|
||||
|
||||
//FfxBoolean bIsResetFrame;
|
||||
FfxBoolean bIsExistingSample;
|
||||
FfxBoolean bIsNewSample;
|
||||
};
|
||||
|
||||
struct LockState
|
||||
{
|
||||
FfxBoolean NewLock; //Set for both unique new and re-locked new
|
||||
FfxBoolean WasLockedPrevFrame; //Set to identify if the pixel was already locked (relock)
|
||||
};
|
||||
|
||||
void InitializeNewLockSample(FFXM_PARAMETER_OUT FfxFloat32x2 fLockStatus)
|
||||
{
|
||||
fLockStatus = FfxFloat32x2(0, 0);
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
void InitializeNewLockSample(FFXM_PARAMETER_OUT FFXM_MIN16_F2 fLockStatus)
|
||||
{
|
||||
fLockStatus = FFXM_MIN16_F2(0, 0);
|
||||
}
|
||||
#endif
|
||||
|
||||
|
||||
void KillLock(FFXM_PARAMETER_INOUT FfxFloat32x2 fLockStatus)
|
||||
{
|
||||
fLockStatus[LOCK_LIFETIME_REMAINING] = 0;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
void KillLock(FFXM_PARAMETER_INOUT FFXM_MIN16_F2 fLockStatus)
|
||||
{
|
||||
fLockStatus[LOCK_LIFETIME_REMAINING] = FFXM_MIN16_F(0);
|
||||
}
|
||||
#endif
|
||||
|
||||
struct RectificationBox
|
||||
{
|
||||
FfxFloat32x3 boxCenter;
|
||||
FfxFloat32x3 boxVec;
|
||||
FfxFloat32x3 aabbMin;
|
||||
FfxFloat32x3 aabbMax;
|
||||
FfxFloat32 fBoxCenterWeight;
|
||||
};
|
||||
#if FFXM_HALF
|
||||
struct RectificationBoxMin16
|
||||
{
|
||||
FFXM_MIN16_F3 boxCenter;
|
||||
FFXM_MIN16_F3 boxVec;
|
||||
FFXM_MIN16_F3 aabbMin;
|
||||
FFXM_MIN16_F3 aabbMax;
|
||||
FFXM_MIN16_F fBoxCenterWeight;
|
||||
};
|
||||
#endif
|
||||
|
||||
void RectificationBoxReset(FFXM_PARAMETER_INOUT RectificationBox rectificationBox)
|
||||
{
|
||||
rectificationBox.fBoxCenterWeight = FfxFloat32(0);
|
||||
|
||||
rectificationBox.boxCenter = FfxFloat32x3(0, 0, 0);
|
||||
rectificationBox.boxVec = FfxFloat32x3(0, 0, 0);
|
||||
rectificationBox.aabbMin = FfxFloat32x3(FSR2_FLT_MAX, FSR2_FLT_MAX, FSR2_FLT_MAX);
|
||||
rectificationBox.aabbMax = -FfxFloat32x3(FSR2_FLT_MAX, FSR2_FLT_MAX, FSR2_FLT_MAX);
|
||||
}
|
||||
#if FFXM_HALF
|
||||
void RectificationBoxReset(FFXM_PARAMETER_INOUT RectificationBoxMin16 rectificationBox)
|
||||
{
|
||||
rectificationBox.fBoxCenterWeight = FFXM_MIN16_F(0);
|
||||
|
||||
rectificationBox.boxCenter = FFXM_MIN16_F3(0, 0, 0);
|
||||
rectificationBox.boxVec = FFXM_MIN16_F3(0, 0, 0);
|
||||
rectificationBox.aabbMin = FFXM_MIN16_F3(FSR2_FP16_MAX, FSR2_FP16_MAX, FSR2_FP16_MAX);
|
||||
rectificationBox.aabbMax = -FFXM_MIN16_F3(FSR2_FP16_MAX, FSR2_FP16_MAX, FSR2_FP16_MAX);
|
||||
}
|
||||
#endif
|
||||
|
||||
void RectificationBoxAddInitialSample(FFXM_PARAMETER_INOUT RectificationBox rectificationBox, const FfxFloat32x3 colorSample, const FfxFloat32 fSampleWeight)
|
||||
{
|
||||
rectificationBox.aabbMin = colorSample;
|
||||
rectificationBox.aabbMax = colorSample;
|
||||
|
||||
FfxFloat32x3 weightedSample = colorSample * fSampleWeight;
|
||||
rectificationBox.boxCenter = weightedSample;
|
||||
rectificationBox.boxVec = colorSample * weightedSample;
|
||||
rectificationBox.fBoxCenterWeight = fSampleWeight;
|
||||
}
|
||||
|
||||
void RectificationBoxAddSample(FfxBoolean bInitialSample, FFXM_PARAMETER_INOUT RectificationBox rectificationBox, const FfxFloat32x3 colorSample, const FfxFloat32 fSampleWeight)
|
||||
{
|
||||
if (bInitialSample) {
|
||||
RectificationBoxAddInitialSample(rectificationBox, colorSample, fSampleWeight);
|
||||
} else {
|
||||
rectificationBox.aabbMin = ffxMin(rectificationBox.aabbMin, colorSample);
|
||||
rectificationBox.aabbMax = ffxMax(rectificationBox.aabbMax, colorSample);
|
||||
|
||||
FfxFloat32x3 weightedSample = colorSample * fSampleWeight;
|
||||
rectificationBox.boxCenter += weightedSample;
|
||||
rectificationBox.boxVec += colorSample * weightedSample;
|
||||
rectificationBox.fBoxCenterWeight += fSampleWeight;
|
||||
}
|
||||
}
|
||||
#if FFXM_HALF
|
||||
void RectificationBoxAddInitialSample(FFXM_PARAMETER_INOUT RectificationBoxMin16 rectificationBox, const FFXM_MIN16_F3 colorSample, const FFXM_MIN16_F fSampleWeight)
|
||||
{
|
||||
rectificationBox.aabbMin = colorSample;
|
||||
rectificationBox.aabbMax = colorSample;
|
||||
|
||||
FFXM_MIN16_F3 weightedSample = colorSample * fSampleWeight;
|
||||
rectificationBox.boxCenter = weightedSample;
|
||||
rectificationBox.boxVec = colorSample * weightedSample;
|
||||
rectificationBox.fBoxCenterWeight = fSampleWeight;
|
||||
}
|
||||
|
||||
void RectificationBoxAddSample(FfxBoolean bInitialSample, FFXM_PARAMETER_INOUT RectificationBoxMin16 rectificationBox, const FFXM_MIN16_F3 colorSample, const FFXM_MIN16_F fSampleWeight)
|
||||
{
|
||||
if (bInitialSample) {
|
||||
RectificationBoxAddInitialSample(rectificationBox, colorSample, fSampleWeight);
|
||||
} else {
|
||||
rectificationBox.aabbMin = ffxMin(rectificationBox.aabbMin, colorSample);
|
||||
rectificationBox.aabbMax = ffxMax(rectificationBox.aabbMax, colorSample);
|
||||
|
||||
FFXM_MIN16_F3 weightedSample = colorSample * fSampleWeight;
|
||||
rectificationBox.boxCenter += weightedSample;
|
||||
rectificationBox.boxVec += colorSample * weightedSample;
|
||||
rectificationBox.fBoxCenterWeight += fSampleWeight;
|
||||
}
|
||||
}
|
||||
#endif
|
||||
|
||||
void RectificationBoxComputeVarianceBoxData(FFXM_PARAMETER_INOUT RectificationBox rectificationBox)
|
||||
{
|
||||
rectificationBox.fBoxCenterWeight = (abs(rectificationBox.fBoxCenterWeight) > FfxFloat32(FSR2_EPSILON) ? rectificationBox.fBoxCenterWeight : FfxFloat32(1.f));
|
||||
rectificationBox.boxCenter /= rectificationBox.fBoxCenterWeight;
|
||||
rectificationBox.boxVec /= rectificationBox.fBoxCenterWeight;
|
||||
FfxFloat32x3 stdDev = sqrt(abs(rectificationBox.boxVec - rectificationBox.boxCenter * rectificationBox.boxCenter));
|
||||
rectificationBox.boxVec = stdDev;
|
||||
}
|
||||
#if FFXM_HALF
|
||||
void RectificationBoxComputeVarianceBoxData(FFXM_PARAMETER_INOUT RectificationBoxMin16 rectificationBox)
|
||||
{
|
||||
rectificationBox.fBoxCenterWeight = (abs(rectificationBox.fBoxCenterWeight) > FFXM_MIN16_F(FSR2_EPSILON) ? rectificationBox.fBoxCenterWeight : FFXM_MIN16_F(1.f));
|
||||
rectificationBox.boxCenter /= rectificationBox.fBoxCenterWeight;
|
||||
rectificationBox.boxVec /= rectificationBox.fBoxCenterWeight;
|
||||
FFXM_MIN16_F3 stdDev = sqrt(abs(rectificationBox.boxVec - rectificationBox.boxCenter * rectificationBox.boxCenter));
|
||||
rectificationBox.boxVec = stdDev;
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32x3 SafeRcp3(FfxFloat32x3 v)
|
||||
{
|
||||
return (all(FFXM_NOT_EQUAL(v, FfxFloat32x3(0, 0, 0)))) ? (FfxFloat32x3(1, 1, 1) / v) : FfxFloat32x3(0, 0, 0);
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F3 SafeRcp3(FFXM_MIN16_F3 v)
|
||||
{
|
||||
return (all(FFXM_NOT_EQUAL(v, FFXM_MIN16_F3(0, 0, 0)))) ? (FFXM_MIN16_F3(1, 1, 1) / v) : FFXM_MIN16_F3(0, 0, 0);
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32 MinDividedByMax(const FfxFloat32 v0, const FfxFloat32 v1)
|
||||
{
|
||||
const FfxFloat32 m = ffxMax(v0, v1);
|
||||
return m != 0 ? ffxMin(v0, v1) / m : 0;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F MinDividedByMax(const FFXM_MIN16_F v0, const FFXM_MIN16_F v1)
|
||||
{
|
||||
const FFXM_MIN16_F m = ffxMax(v0, v1);
|
||||
return m != FFXM_MIN16_F(0) ? ffxMin(v0, v1) / m : FFXM_MIN16_F(0);
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32x3 YCoCgToRGB(FfxFloat32x3 fYCoCg)
|
||||
{
|
||||
FfxFloat32x3 fRgb;
|
||||
|
||||
fRgb = FfxFloat32x3(
|
||||
fYCoCg.x + fYCoCg.y - fYCoCg.z,
|
||||
fYCoCg.x + fYCoCg.z,
|
||||
fYCoCg.x - fYCoCg.y - fYCoCg.z);
|
||||
|
||||
return fRgb;
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F3 YCoCgToRGB(FFXM_MIN16_F3 fYCoCg)
|
||||
{
|
||||
FFXM_MIN16_F3 fRgb;
|
||||
|
||||
fRgb = FFXM_MIN16_F3(
|
||||
fYCoCg.x + fYCoCg.y - fYCoCg.z,
|
||||
fYCoCg.x + fYCoCg.z,
|
||||
fYCoCg.x - fYCoCg.y - fYCoCg.z);
|
||||
|
||||
return fRgb;
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32x3 RGBToYCoCg(FfxFloat32x3 fRgb)
|
||||
{
|
||||
FfxFloat32x3 fYCoCg;
|
||||
|
||||
fYCoCg = FfxFloat32x3(
|
||||
0.25f * fRgb.r + 0.5f * fRgb.g + 0.25f * fRgb.b,
|
||||
0.5f * fRgb.r - 0.5f * fRgb.b,
|
||||
-0.25f * fRgb.r + 0.5f * fRgb.g - 0.25f * fRgb.b);
|
||||
|
||||
return fYCoCg;
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F3 RGBToYCoCg(FFXM_MIN16_F3 fRgb)
|
||||
{
|
||||
FFXM_MIN16_F3 fYCoCg;
|
||||
|
||||
fYCoCg = FFXM_MIN16_F3(
|
||||
0.25 * fRgb.r + 0.5 * fRgb.g + 0.25 * fRgb.b,
|
||||
0.5 * fRgb.r - 0.5 * fRgb.b,
|
||||
-0.25 * fRgb.r + 0.5 * fRgb.g - 0.25 * fRgb.b);
|
||||
|
||||
return fYCoCg;
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32 RGBToLuma(FfxFloat32x3 fLinearRgb)
|
||||
{
|
||||
return dot(fLinearRgb, FfxFloat32x3(0.2126f, 0.7152f, 0.0722f));
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F RGBToLuma(FFXM_MIN16_F3 fLinearRgb)
|
||||
{
|
||||
return dot(fLinearRgb, FFXM_MIN16_F3(0.2126f, 0.7152f, 0.0722f));
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32 RGBToPerceivedLuma(FfxFloat32x3 fLinearRgb)
|
||||
{
|
||||
FfxFloat32 fLuminance = RGBToLuma(fLinearRgb);
|
||||
|
||||
FfxFloat32 fPercievedLuminance = 0;
|
||||
if (fLuminance <= 216.0f / 24389.0f) {
|
||||
fPercievedLuminance = fLuminance * (24389.0f / 27.0f);
|
||||
}
|
||||
else {
|
||||
fPercievedLuminance = ffxPow(fLuminance, 1.0f / 3.0f) * 116.0f - 16.0f;
|
||||
}
|
||||
|
||||
return fPercievedLuminance * 0.01f;
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F RGBToPerceivedLuma(FFXM_MIN16_F3 fLinearRgb)
|
||||
{
|
||||
FFXM_MIN16_F fLuminance = RGBToLuma(fLinearRgb);
|
||||
|
||||
FFXM_MIN16_F fPercievedLuminance = FFXM_MIN16_F(0);
|
||||
if (fLuminance <= FFXM_MIN16_F(216.0f / 24389.0f)) {
|
||||
fPercievedLuminance = fLuminance * FFXM_MIN16_F(24389.0f / 27.0f);
|
||||
}
|
||||
else {
|
||||
fPercievedLuminance = ffxPow(fLuminance, FFXM_MIN16_F(1.0f / 3.0f)) * FFXM_MIN16_F(116.0f) - FFXM_MIN16_F(16.0f);
|
||||
}
|
||||
|
||||
return fPercievedLuminance * FFXM_MIN16_F(0.01f);
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32x3 Tonemap(FfxFloat32x3 fRgb)
|
||||
{
|
||||
return fRgb / (ffxMax(ffxMax(0.f, fRgb.r), ffxMax(fRgb.g, fRgb.b)) + 1.f).xxx;
|
||||
}
|
||||
|
||||
FfxFloat32x3 InverseTonemap(FfxFloat32x3 fRgb)
|
||||
{
|
||||
return fRgb / ffxMax(FSR2_TONEMAP_EPSILON, 1.f - ffxMax(fRgb.r, ffxMax(fRgb.g, fRgb.b))).xxx;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F3 Tonemap(FFXM_MIN16_F3 fRgb)
|
||||
{
|
||||
return fRgb / (ffxMax(ffxMax(FFXM_MIN16_F(0.f), fRgb.r), ffxMax(fRgb.g, fRgb.b)) + FFXM_MIN16_F(1.f)).xxx;
|
||||
}
|
||||
|
||||
FFXM_MIN16_F3 InverseTonemap(FFXM_MIN16_F3 fRgb)
|
||||
{
|
||||
return fRgb / ffxMax(FFXM_MIN16_F(FSR2_TONEMAP_EPSILON), FFXM_MIN16_F(1.f) - ffxMax(fRgb.r, ffxMax(fRgb.g, fRgb.b))).xxx;
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxInt32x2 ClampLoad(FfxInt32x2 iPxSample, FfxInt32x2 iPxOffset, FfxInt32x2 iTextureSize)
|
||||
{
|
||||
FfxInt32x2 result = iPxSample + iPxOffset;
|
||||
result.x = (iPxOffset.x < 0) ? ffxMax(result.x, 0) : result.x;
|
||||
result.x = (iPxOffset.x > 0) ? ffxMin(result.x, iTextureSize.x - 1) : result.x;
|
||||
result.y = (iPxOffset.y < 0) ? ffxMax(result.y, 0) : result.y;
|
||||
result.y = (iPxOffset.y > 0) ? ffxMin(result.y, iTextureSize.y - 1) : result.y;
|
||||
return result;
|
||||
|
||||
// return ffxMed3(iPxSample + iPxOffset, FfxInt32x2(0, 0), iTextureSize - FfxInt32x2(1, 1));
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_I2 ClampLoad(FFXM_MIN16_I2 iPxSample, FFXM_MIN16_I2 iPxOffset, FFXM_MIN16_I2 iTextureSize)
|
||||
{
|
||||
FFXM_MIN16_I2 result = iPxSample + iPxOffset;
|
||||
result.x = (iPxOffset.x < 0) ? ffxMax(result.x, FFXM_MIN16_I(0)) : result.x;
|
||||
result.x = (iPxOffset.x > 0) ? ffxMin(result.x, iTextureSize.x - FFXM_MIN16_I(1)) : result.x;
|
||||
result.y = (iPxOffset.y < 0) ? ffxMax(result.y, FFXM_MIN16_I(0)) : result.y;
|
||||
result.y = (iPxOffset.y > 0) ? ffxMin(result.y, iTextureSize.y - FFXM_MIN16_I(1)) : result.y;
|
||||
return result;
|
||||
|
||||
// return ffxMed3Half(iPxSample + iPxOffset, FFXM_MIN16_I2(0, 0), iTextureSize - FFXM_MIN16_I2(1, 1));
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32x2 ClampUv(FfxFloat32x2 fUv, FfxInt32x2 iTextureSize, FfxInt32x2 iResourceSize)
|
||||
{
|
||||
const FfxFloat32x2 fSampleLocation = fUv * iTextureSize;
|
||||
const FfxFloat32x2 fClampedLocation = ffxMax(FfxFloat32x2(0.5f, 0.5f), ffxMin(fSampleLocation, FfxFloat32x2(iTextureSize) - FfxFloat32x2(0.5f, 0.5f)));
|
||||
const FfxFloat32x2 fClampedUv = fClampedLocation / FfxFloat32x2(iResourceSize);
|
||||
|
||||
return fClampedUv;
|
||||
}
|
||||
|
||||
FfxBoolean IsOnScreen(FfxInt32x2 pos, FfxInt32x2 size)
|
||||
{
|
||||
return all(FFXM_LESS_THAN(FfxUInt32x2(pos), FfxUInt32x2(size)));
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FfxBoolean IsOnScreen(FFXM_MIN16_I2 pos, FFXM_MIN16_I2 size)
|
||||
{
|
||||
return all(FFXM_LESS_THAN(FFXM_MIN16_U2(pos), FFXM_MIN16_U2(size)));
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32 ComputeAutoExposureFromLavg(FfxFloat32 Lavg)
|
||||
{
|
||||
Lavg = exp(Lavg);
|
||||
|
||||
const FfxFloat32 S = 100.0f; //ISO arithmetic speed
|
||||
const FfxFloat32 K = 12.5f;
|
||||
FfxFloat32 ExposureISO100 = log2((Lavg * S) / K);
|
||||
|
||||
const FfxFloat32 q = 0.65f;
|
||||
FfxFloat32 Lmax = (78.0f / (q * S)) * ffxPow(2.0f, ExposureISO100);
|
||||
|
||||
return 1 / Lmax;
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F ComputeAutoExposureFromLavg(FFXM_MIN16_F Lavg)
|
||||
{
|
||||
Lavg = exp(Lavg);
|
||||
|
||||
const FFXM_MIN16_F S = FFXM_MIN16_F(100.0f); //ISO arithmetic speed
|
||||
const FFXM_MIN16_F K = FFXM_MIN16_F(12.5f);
|
||||
const FFXM_MIN16_F ExposureISO100 = log2((Lavg * S) / K);
|
||||
|
||||
const FFXM_MIN16_F q = FFXM_MIN16_F(0.65f);
|
||||
const FFXM_MIN16_F Lmax = (FFXM_MIN16_F(78.0f) / (q * S)) * ffxPow(FFXM_MIN16_F(2.0f), ExposureISO100);
|
||||
|
||||
return FFXM_MIN16_F(1) / Lmax;
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxInt32x2 ComputeHrPosFromLrPos(FfxInt32x2 iPxLrPos)
|
||||
{
|
||||
FfxFloat32x2 fSrcJitteredPos = FfxFloat32x2(iPxLrPos) + 0.5f - Jitter();
|
||||
FfxFloat32x2 fLrPosInHr = (fSrcJitteredPos / RenderSize()) * DisplaySize();
|
||||
FfxInt32x2 iPxHrPos = FfxInt32x2(floor(fLrPosInHr));
|
||||
return iPxHrPos;
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_I2 ComputeHrPosFromLrPos(FFXM_MIN16_I2 iPxLrPos)
|
||||
{
|
||||
FFXM_MIN16_F2 fSrcJitteredPos = FFXM_MIN16_F2(iPxLrPos) + FFXM_MIN16_F(0.5f) - FFXM_MIN16_F2(Jitter());
|
||||
FFXM_MIN16_F2 fLrPosInHr = (fSrcJitteredPos / FFXM_MIN16_F2(RenderSize())) * FFXM_MIN16_F2(DisplaySize());
|
||||
FFXM_MIN16_I2 iPxHrPos = FFXM_MIN16_I2(floor(fLrPosInHr));
|
||||
return iPxHrPos;
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32x2 ComputeNdc(FfxFloat32x2 fPxPos, FfxInt32x2 iSize)
|
||||
{
|
||||
return fPxPos / FfxFloat32x2(iSize) * FfxFloat32x2(2.0f, -2.0f) + FfxFloat32x2(-1.0f, 1.0f);
|
||||
}
|
||||
|
||||
FfxFloat32 GetViewSpaceDepth(FfxFloat32 fDeviceDepth)
|
||||
{
|
||||
const FfxFloat32x4 fDeviceToViewDepth = DeviceToViewSpaceTransformFactors();
|
||||
|
||||
// fDeviceToViewDepth details found in ffx_fsr2.cpp
|
||||
return (fDeviceToViewDepth[1] / (fDeviceDepth - fDeviceToViewDepth[0]));
|
||||
}
|
||||
|
||||
FfxFloat32 GetViewSpaceDepthInMeters(FfxFloat32 fDeviceDepth)
|
||||
{
|
||||
return GetViewSpaceDepth(fDeviceDepth) * ViewSpaceToMetersFactor();
|
||||
}
|
||||
|
||||
FfxFloat32x3 GetViewSpacePosition(FfxInt32x2 iViewportPos, FfxInt32x2 iViewportSize, FfxFloat32 fDeviceDepth)
|
||||
{
|
||||
const FfxFloat32x4 fDeviceToViewDepth = DeviceToViewSpaceTransformFactors();
|
||||
|
||||
const FfxFloat32 Z = GetViewSpaceDepth(fDeviceDepth);
|
||||
|
||||
const FfxFloat32x2 fNdcPos = ComputeNdc(iViewportPos, iViewportSize);
|
||||
const FfxFloat32 X = fDeviceToViewDepth[2] * fNdcPos.x * Z;
|
||||
const FfxFloat32 Y = fDeviceToViewDepth[3] * fNdcPos.y * Z;
|
||||
|
||||
return FfxFloat32x3(X, Y, Z);
|
||||
}
|
||||
|
||||
FfxFloat32x3 GetViewSpacePositionInMeters(FfxInt32x2 iViewportPos, FfxInt32x2 iViewportSize, FfxFloat32 fDeviceDepth)
|
||||
{
|
||||
return GetViewSpacePosition(iViewportPos, iViewportSize, fDeviceDepth) * ViewSpaceToMetersFactor();
|
||||
}
|
||||
|
||||
FfxFloat32 GetMaxDistanceInMeters()
|
||||
{
|
||||
#if FFXM_FSR2_OPTION_INVERTED_DEPTH
|
||||
return GetViewSpaceDepth(0.0f) * ViewSpaceToMetersFactor();
|
||||
#else
|
||||
return GetViewSpaceDepth(1.0f) * ViewSpaceToMetersFactor();
|
||||
#endif
|
||||
}
|
||||
|
||||
FfxFloat32x3 PrepareRgb(FfxFloat32x3 fRgb, FfxFloat32 fExposure, FfxFloat32 fPreExposure)
|
||||
{
|
||||
fRgb /= fPreExposure;
|
||||
fRgb *= fExposure;
|
||||
|
||||
fRgb = clamp(fRgb, 0.0f, FSR2_FP16_MAX);
|
||||
|
||||
return fRgb;
|
||||
}
|
||||
|
||||
FfxFloat32x3 UnprepareRgb(FfxFloat32x3 fRgb, FfxFloat32 fExposure)
|
||||
{
|
||||
fRgb /= fExposure;
|
||||
fRgb *= PreExposure();
|
||||
|
||||
return fRgb;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FfxFloat16x3 PrepareRgb(FfxFloat16x3 fRgb, FfxFloat16 fExposure, FfxFloat16 fPreExposure)
|
||||
{
|
||||
fRgb /= fPreExposure;
|
||||
fRgb *= fExposure;
|
||||
|
||||
fRgb = clamp(fRgb, FfxFloat16(0.0f), FfxFloat16(FSR2_FP16_MAX));
|
||||
|
||||
return fRgb;
|
||||
}
|
||||
|
||||
FfxFloat16x3 UnprepareRgb(FfxFloat16x3 fRgb, FfxFloat16 fExposure)
|
||||
{
|
||||
fRgb /= fExposure;
|
||||
fRgb *= FfxFloat16(PreExposure());
|
||||
|
||||
return fRgb;
|
||||
}
|
||||
#endif
|
||||
|
||||
struct BilinearSamplingData
|
||||
{
|
||||
FfxInt32x2 iOffsets[4];
|
||||
FfxFloat32 fWeights[4];
|
||||
FfxInt32x2 iBasePos;
|
||||
FfxFloat32x2 fQuadCenterUv;
|
||||
};
|
||||
|
||||
BilinearSamplingData GetBilinearSamplingData(FfxFloat32x2 fUv, FfxInt32x2 iSize)
|
||||
{
|
||||
BilinearSamplingData data;
|
||||
|
||||
FfxFloat32x2 fPxSample = (fUv * iSize) - FfxFloat32x2(0.5f, 0.5f);
|
||||
data.iBasePos = FfxInt32x2(floor(fPxSample));
|
||||
data.fQuadCenterUv = fPxSample / FfxFloat32x2(iSize);
|
||||
FfxFloat32x2 fPxFrac = ffxFract(fPxSample);
|
||||
|
||||
data.iOffsets[0] = FfxInt32x2(0, 0);
|
||||
data.iOffsets[1] = FfxInt32x2(1, 0);
|
||||
data.iOffsets[2] = FfxInt32x2(0, 1);
|
||||
data.iOffsets[3] = FfxInt32x2(1, 1);
|
||||
|
||||
data.fWeights[0] = (1 - fPxFrac.x) * (1 - fPxFrac.y);
|
||||
data.fWeights[1] = (fPxFrac.x) * (1 - fPxFrac.y);
|
||||
data.fWeights[2] = (1 - fPxFrac.x) * (fPxFrac.y);
|
||||
data.fWeights[3] = (fPxFrac.x) * (fPxFrac.y);
|
||||
|
||||
return data;
|
||||
}
|
||||
|
||||
struct PlaneData
|
||||
{
|
||||
FfxFloat32x3 fNormal;
|
||||
FfxFloat32 fDistanceFromOrigin;
|
||||
};
|
||||
|
||||
PlaneData GetPlaneFromPoints(FfxFloat32x3 fP0, FfxFloat32x3 fP1, FfxFloat32x3 fP2)
|
||||
{
|
||||
PlaneData plane;
|
||||
|
||||
FfxFloat32x3 v0 = fP0 - fP1;
|
||||
FfxFloat32x3 v1 = fP0 - fP2;
|
||||
plane.fNormal = normalize(cross(v0, v1));
|
||||
plane.fDistanceFromOrigin = -dot(fP0, plane.fNormal);
|
||||
|
||||
return plane;
|
||||
}
|
||||
|
||||
FfxFloat32 PointToPlaneDistance(PlaneData plane, FfxFloat32x3 fPoint)
|
||||
{
|
||||
return abs(dot(plane.fNormal, fPoint) + plane.fDistanceFromOrigin);
|
||||
}
|
||||
|
||||
#endif // #if defined(FFXM_GPU)
|
||||
|
||||
#endif //!defined(FFXM_FSR2_COMMON_H)
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 04d4500acc8f72748b3c536a1af0637a
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,212 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
FFXM_GROUPSHARED FfxUInt32 spdCounter;
|
||||
|
||||
void SpdIncreaseAtomicCounter(FfxUInt32 slice)
|
||||
{
|
||||
SPD_IncreaseAtomicCounter(spdCounter);
|
||||
}
|
||||
|
||||
FfxUInt32 SpdGetAtomicCounter()
|
||||
{
|
||||
return spdCounter;
|
||||
}
|
||||
|
||||
void SpdResetAtomicCounter(FfxUInt32 slice)
|
||||
{
|
||||
SPD_ResetAtomicCounter();
|
||||
}
|
||||
|
||||
#ifndef SPD_PACKED_ONLY
|
||||
FFXM_GROUPSHARED FfxFloat32 spdIntermediateR[16][16];
|
||||
FFXM_GROUPSHARED FfxFloat32 spdIntermediateG[16][16];
|
||||
FFXM_GROUPSHARED FfxFloat32 spdIntermediateB[16][16];
|
||||
FFXM_GROUPSHARED FfxFloat32 spdIntermediateA[16][16];
|
||||
|
||||
FfxFloat32x4 SpdLoadSourceImage(FfxFloat32x2 tex, FfxUInt32 slice)
|
||||
{
|
||||
FfxFloat32x2 fUv = (tex + 0.5f + Jitter()) / RenderSize();
|
||||
fUv = ClampUv(fUv, RenderSize(), InputColorResourceDimensions());
|
||||
FfxFloat32x3 fRgb = SampleInputColor(fUv);
|
||||
|
||||
fRgb /= PreExposure();
|
||||
|
||||
//compute log luma
|
||||
const FfxFloat32 fLogLuma = log(ffxMax(FSR2_EPSILON, RGBToLuma(fRgb)));
|
||||
|
||||
// Make sure out of screen pixels contribute no value to the end result
|
||||
const FfxFloat32 result = all(FFXM_LESS_THAN(tex, RenderSize())) ? fLogLuma : 0.0f;
|
||||
|
||||
return FfxFloat32x4(result, 0, 0, 0);
|
||||
}
|
||||
|
||||
FfxFloat32x4 SpdLoad(FfxInt32x2 tex, FfxUInt32 slice)
|
||||
{
|
||||
return SPD_LoadMipmap5(tex);
|
||||
}
|
||||
|
||||
void SpdStore(FfxInt32x2 pix, FfxFloat32x4 outValue, FfxUInt32 index, FfxUInt32 slice)
|
||||
{
|
||||
if (index == LumaMipLevelToUse() || index == 5)
|
||||
{
|
||||
SPD_SetMipmap(pix, index, outValue.r);
|
||||
}
|
||||
|
||||
if (index == MipCount() - 1) { //accumulate on 1x1 level
|
||||
|
||||
if (all(FFXM_EQUAL(pix, FfxInt32x2(0, 0))))
|
||||
{
|
||||
FfxFloat32 prev = SPD_LoadExposureBuffer().y;
|
||||
FfxFloat32 result = outValue.r;
|
||||
|
||||
if (prev < resetAutoExposureAverageSmoothing) // Compare Lavg, so small or negative values
|
||||
{
|
||||
FfxFloat32 rate = 1.0f;
|
||||
result = prev + (result - prev) * (1 - exp(-DeltaTime() * rate));
|
||||
}
|
||||
FfxFloat32x2 spdOutput = FfxFloat32x2(ComputeAutoExposureFromLavg(result), result);
|
||||
SPD_SetExposureBuffer(spdOutput);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
FfxFloat32x4 SpdLoadIntermediate(FfxUInt32 x, FfxUInt32 y)
|
||||
{
|
||||
return FfxFloat32x4(
|
||||
spdIntermediateR[x][y],
|
||||
spdIntermediateG[x][y],
|
||||
spdIntermediateB[x][y],
|
||||
spdIntermediateA[x][y]);
|
||||
}
|
||||
void SpdStoreIntermediate(FfxUInt32 x, FfxUInt32 y, FfxFloat32x4 value)
|
||||
{
|
||||
spdIntermediateR[x][y] = value.x;
|
||||
spdIntermediateG[x][y] = value.y;
|
||||
spdIntermediateB[x][y] = value.z;
|
||||
spdIntermediateA[x][y] = value.w;
|
||||
}
|
||||
FfxFloat32x4 SpdReduce4(FfxFloat32x4 v0, FfxFloat32x4 v1, FfxFloat32x4 v2, FfxFloat32x4 v3)
|
||||
{
|
||||
return (v0 + v1 + v2 + v3) * 0.25f;
|
||||
}
|
||||
#endif
|
||||
|
||||
// define fetch and store functions Packed
|
||||
#if FFXM_HALF
|
||||
|
||||
FFXM_GROUPSHARED FfxFloat16x2 spdIntermediateRG[16][16];
|
||||
FFXM_GROUPSHARED FfxFloat16x2 spdIntermediateBA[16][16];
|
||||
|
||||
FfxFloat16x4 SpdLoadSourceImageH(FfxFloat32x2 tex, FfxUInt32 slice)
|
||||
{
|
||||
FfxFloat16x2 fUv = FfxFloat16x2((tex + 0.5f + Jitter()) / RenderSize());
|
||||
fUv = FfxFloat16x2(ClampUv(fUv, RenderSize(), InputColorResourceDimensions()));
|
||||
FfxFloat16x3 fRgb = FfxFloat16x3(SampleInputColor(fUv));
|
||||
|
||||
fRgb /= FfxFloat16(PreExposure());
|
||||
|
||||
//compute log luma
|
||||
const FfxFloat16 fLogLuma = FfxFloat16(log(ffxMax(FSR2_EPSILON, RGBToLuma(fRgb))));
|
||||
|
||||
// Make sure out of screen pixels contribute no value to the end result
|
||||
const FfxFloat16 result = all(FFXM_LESS_THAN(tex, RenderSize())) ? fLogLuma : FfxFloat16(0.0f);
|
||||
|
||||
return FfxFloat16x4(result, 0, 0, 0);
|
||||
}
|
||||
|
||||
FfxFloat16x4 SpdLoadH(FfxInt32x2 p, FfxUInt32 slice)
|
||||
{
|
||||
return FfxFloat16x4(SPD_LoadMipmap5(p));
|
||||
}
|
||||
|
||||
void SpdStoreH(FfxInt32x2 pix, FfxFloat16x4 outValue, FfxUInt32 index, FfxUInt32 slice)
|
||||
{
|
||||
if (index == LumaMipLevelToUse() || index == 5)
|
||||
{
|
||||
SPD_SetMipmap(pix, index, outValue.r);
|
||||
}
|
||||
|
||||
if (index == MipCount() - 1) { //accumulate on 1x1 level
|
||||
|
||||
if (all(FFXM_EQUAL(pix, FfxInt16x2(0, 0))))
|
||||
{
|
||||
FfxFloat16 result = outValue.r;
|
||||
|
||||
// Remove the smooth exposure transision with GLES 3.2
|
||||
#if !FFXM_SHADER_PLATFORM_GLES_3_2
|
||||
FfxFloat16 prev = FfxFloat16(SPD_LoadExposureBuffer().y);
|
||||
if (prev < resetAutoExposureAverageSmoothing) // Compare Lavg, so small or negative values
|
||||
{
|
||||
FfxFloat16 rate = FfxFloat16(1.0f);
|
||||
result = FfxFloat16(prev + (result - prev) * (1 - exp(-DeltaTime() * rate)));
|
||||
}
|
||||
#endif
|
||||
FfxFloat16x2 spdOutput = FfxFloat16x2(ComputeAutoExposureFromLavg(result), result);
|
||||
SPD_SetExposureBuffer(spdOutput);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
FfxFloat16x4 SpdLoadIntermediateH(FfxUInt32 x, FfxUInt32 y)
|
||||
{
|
||||
return FfxFloat16x4(
|
||||
spdIntermediateRG[x][y].x,
|
||||
spdIntermediateRG[x][y].y,
|
||||
spdIntermediateBA[x][y].x,
|
||||
spdIntermediateBA[x][y].y);
|
||||
}
|
||||
|
||||
void SpdStoreIntermediateH(FfxUInt32 x, FfxUInt32 y, FfxFloat16x4 value)
|
||||
{
|
||||
spdIntermediateRG[x][y] = value.xy;
|
||||
spdIntermediateBA[x][y] = value.zw;
|
||||
}
|
||||
|
||||
FfxFloat16x4 SpdReduce4H(FfxFloat16x4 v0, FfxFloat16x4 v1, FfxFloat16x4 v2, FfxFloat16x4 v3)
|
||||
{
|
||||
return (v0 + v1 + v2 + v3) * FfxFloat16(0.25);
|
||||
}
|
||||
#endif
|
||||
|
||||
#include "ffxm_spd.h"
|
||||
|
||||
void ComputeAutoExposure(FfxUInt32x3 WorkGroupId, FfxUInt32 LocalThreadIndex)
|
||||
{
|
||||
#if FFXM_HALF
|
||||
SpdDownsampleH(
|
||||
FfxUInt32x2(WorkGroupId.xy),
|
||||
FfxUInt32(LocalThreadIndex),
|
||||
FfxUInt32(MipCount()),
|
||||
FfxUInt32(NumWorkGroups()),
|
||||
FfxUInt32(WorkGroupId.z),
|
||||
FfxUInt32x2(WorkGroupOffset()));
|
||||
#else
|
||||
SpdDownsample(
|
||||
FfxUInt32x2(WorkGroupId.xy),
|
||||
FfxUInt32(LocalThreadIndex),
|
||||
FfxUInt32(MipCount()),
|
||||
FfxUInt32(NumWorkGroups()),
|
||||
FfxUInt32(WorkGroupId.z),
|
||||
FfxUInt32x2(WorkGroupOffset()));
|
||||
#endif
|
||||
}
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 527e868f91bb5d44e9fa1a9dd92d118e
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,67 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#define FSR2_BIND_SRV_INPUT_COLOR 0
|
||||
|
||||
#define FSR2_BIND_UAV_SPD_GLOBAL_ATOMIC 1
|
||||
#define FSR2_BIND_UAV_EXPOSURE_MIP_LUMA_CHANGE 2
|
||||
#define FSR2_BIND_UAV_EXPOSURE_MIP_5 3
|
||||
#define FSR2_BIND_UAV_AUTO_EXPOSURE 4
|
||||
|
||||
#define FSR2_BIND_CB_FSR2 0
|
||||
#define FSR2_BIND_CB_SPD 1
|
||||
|
||||
// Global mandatory defines
|
||||
#if !defined(FFXM_HALF)
|
||||
#define FFXM_HALF 1
|
||||
#endif
|
||||
#if !defined(FFXM_GPU)
|
||||
#define FFXM_GPU 1
|
||||
#endif
|
||||
#if !defined(FFXM_HLSL)
|
||||
#define FFXM_HLSL 1
|
||||
#endif
|
||||
|
||||
#include "ffxm_fsr2_callbacks_hlsl.h"
|
||||
#include "ffxm_fsr2_common.h"
|
||||
#include "ffxm_fsr2_compute_luminance_pyramid.h"
|
||||
|
||||
#ifndef FFXM_FSR2_THREAD_GROUP_WIDTH
|
||||
#define FFXM_FSR2_THREAD_GROUP_WIDTH 256
|
||||
#endif // #ifndef FFXM_FSR2_THREAD_GROUP_WIDTH
|
||||
#ifndef FFXM_FSR2_THREAD_GROUP_HEIGHT
|
||||
#define FFXM_FSR2_THREAD_GROUP_HEIGHT 1
|
||||
#endif // #ifndef FFXM_FSR2_THREAD_GROUP_HEIGHT
|
||||
#ifndef FFXM_FSR2_THREAD_GROUP_DEPTH
|
||||
#define FFXM_FSR2_THREAD_GROUP_DEPTH 1
|
||||
#endif // #ifndef FFXM_FSR2_THREAD_GROUP_DEPTH
|
||||
#ifndef FFXM_FSR2_NUM_THREADS
|
||||
#define FFXM_FSR2_NUM_THREADS [numthreads(FFXM_FSR2_THREAD_GROUP_WIDTH, FFXM_FSR2_THREAD_GROUP_HEIGHT, FFXM_FSR2_THREAD_GROUP_DEPTH)]
|
||||
#endif // #ifndef FFXM_FSR2_NUM_THREADS
|
||||
|
||||
FFXM_PREFER_WAVE64
|
||||
FFXM_FSR2_NUM_THREADS
|
||||
FFXM_FSR2_EMBED_CB2_ROOTSIG_CONTENT
|
||||
void main(uint3 WorkGroupId : SV_GroupID, uint LocalThreadIndex : SV_GroupIndex)
|
||||
{
|
||||
ComputeAutoExposure(WorkGroupId, LocalThreadIndex);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: c5f334cff899f8547b79b95577007742
|
||||
ShaderIncludeImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,350 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#ifndef FFXM_FSR2_DEPTH_CLIP_H
|
||||
#define FFXM_FSR2_DEPTH_CLIP_H
|
||||
|
||||
// Can casue some temporal instability
|
||||
#define OPT_PREFETCH_PREVDEPTH_WITH_GATHER 0
|
||||
|
||||
struct DepthClipOutputs
|
||||
{
|
||||
FfxFloat32x2 fDilatedReactiveMasks;
|
||||
FfxFloat32x4 fTonemapped;
|
||||
};
|
||||
|
||||
FFXM_STATIC const FfxFloat32 DepthClipBaseScale = 4.0f;
|
||||
|
||||
FfxFloat32 ComputeDepthClip(FfxFloat32x2 fUvSample, FfxFloat32 fCurrentDepthSample)
|
||||
{
|
||||
FfxFloat32 fCurrentDepthViewSpace = GetViewSpaceDepth(fCurrentDepthSample);
|
||||
BilinearSamplingData bilinearInfo = GetBilinearSamplingData(fUvSample, RenderSize());
|
||||
|
||||
FfxFloat32 fDilatedSum = 0.0f;
|
||||
FfxFloat32 fDepth = 0.0f;
|
||||
FfxFloat32 fWeightSum = 0.0f;
|
||||
|
||||
|
||||
#if OPT_PREFETCH_PREVDEPTH_WITH_GATHER
|
||||
FfxFloat32 fDepthSamples[4];
|
||||
GatherReconstructedPreviousDepthRQuad(bilinearInfo.fQuadCenterUv,
|
||||
fDepthSamples[0], fDepthSamples[1], fDepthSamples[2], fDepthSamples[3]);
|
||||
#endif
|
||||
|
||||
for (FfxInt32 iSampleIndex = 0; iSampleIndex < 4; iSampleIndex++)
|
||||
{
|
||||
const FfxInt32x2 iOffset = bilinearInfo.iOffsets[iSampleIndex];
|
||||
const FfxInt32x2 iSamplePos = bilinearInfo.iBasePos + iOffset;
|
||||
|
||||
if (IsOnScreen(iSamplePos, RenderSize()))
|
||||
{
|
||||
const FfxFloat32 fWeight = bilinearInfo.fWeights[iSampleIndex];
|
||||
if (fWeight > fReconstructedDepthBilinearWeightThreshold)
|
||||
{
|
||||
#if OPT_PREFETCH_PREVDEPTH_WITH_GATHER
|
||||
const FfxFloat32 fPrevDepthSample = fDepthSamples[iSampleIndex];
|
||||
#else
|
||||
const FfxFloat32 fPrevDepthSample = LoadReconstructedPrevDepth(iSamplePos);
|
||||
#endif
|
||||
const FfxFloat32 fPrevNearestDepthViewSpace = GetViewSpaceDepth(fPrevDepthSample);
|
||||
const FfxFloat32 fDepthDiff = fCurrentDepthViewSpace - fPrevNearestDepthViewSpace;
|
||||
|
||||
if (fDepthDiff > 0.0f) {
|
||||
|
||||
#if FFXM_FSR2_OPTION_INVERTED_DEPTH
|
||||
const FfxFloat32 fPlaneDepth = ffxMin(fPrevDepthSample, fCurrentDepthSample);
|
||||
#else
|
||||
const FfxFloat32 fPlaneDepth = ffxMax(fPrevDepthSample, fCurrentDepthSample);
|
||||
#endif
|
||||
|
||||
const FfxFloat32x3 fCenter = GetViewSpacePosition(FfxInt32x2(RenderSize() * 0.5f), RenderSize(), fPlaneDepth);
|
||||
const FfxFloat32x3 fCorner = GetViewSpacePosition(FfxInt32x2(0, 0), RenderSize(), fPlaneDepth);
|
||||
|
||||
const FfxFloat32 fHalfViewportWidth = length(FfxFloat32x2(RenderSize()));
|
||||
const FfxFloat32 fDepthThreshold = ffxMax(fCurrentDepthViewSpace, fPrevNearestDepthViewSpace);
|
||||
|
||||
const FfxFloat32 Ksep = 1.37e-05f;
|
||||
const FfxFloat32 Kfov = length(fCorner) / length(fCenter);
|
||||
const FfxFloat32 fRequiredDepthSeparation = Ksep * Kfov * fHalfViewportWidth * fDepthThreshold;
|
||||
|
||||
const FfxFloat32 fResolutionFactor = ffxSaturate(length(FfxFloat32x2(RenderSize())) / length(FfxFloat32x2(1920.0f, 1080.0f)));
|
||||
const FfxFloat32 fPower = ffxLerp(1.0f, 3.0f, fResolutionFactor);
|
||||
fDepth += ffxPow(ffxSaturate(FfxFloat32(fRequiredDepthSeparation / fDepthDiff)), fPower) * fWeight;
|
||||
fWeightSum += fWeight;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return (fWeightSum > 0) ? ffxSaturate(1.0f - fDepth / fWeightSum) : 0.0f;
|
||||
}
|
||||
|
||||
FfxFloat32 ComputeMotionDivergence(FfxInt32x2 iPxPos, FfxInt32x2 iPxInputMotionVectorSize)
|
||||
{
|
||||
FfxFloat32 minconvergence = 1.0f;
|
||||
|
||||
FfxFloat32x2 fMotionVectorNucleus = LoadInputMotionVector(iPxPos);
|
||||
FfxFloat32 fNucleusVelocityLr = length(fMotionVectorNucleus * RenderSize());
|
||||
FfxFloat32 fMaxVelocityUv = length(fMotionVectorNucleus);
|
||||
|
||||
const FfxFloat32 MotionVectorVelocityEpsilon = 1e-02f;
|
||||
|
||||
|
||||
const FfxFloat32x2 fMVSize = FfxFloat32x2(iPxInputMotionVectorSize);
|
||||
FfxFloat32x2 fPxBaseUv = FfxFloat32x2(iPxPos) / fMVSize;
|
||||
FfxFloat32x2 fUnitUv = FfxFloat32x2(1.0f, 1.0f) / fMVSize;
|
||||
|
||||
FFXM_MIN16_F2 fMotionVectorSamples[9];
|
||||
FFXM_MIN16_F2 fTmpDummy = FFXM_MIN16_F2(0.0f, 0.0f);
|
||||
GatherInputMotionVectorRGQuad(fPxBaseUv,
|
||||
fMotionVectorSamples[0], fMotionVectorSamples[1],
|
||||
fMotionVectorSamples[3], fMotionVectorSamples[4]);
|
||||
GatherInputMotionVectorRGQuad(fUnitUv + fPxBaseUv,
|
||||
fTmpDummy, fMotionVectorSamples[5],
|
||||
fMotionVectorSamples[7], fMotionVectorSamples[8]);
|
||||
fMotionVectorSamples[2] = LoadInputMotionVector(iPxPos + FfxInt32x2(1, -1));
|
||||
fMotionVectorSamples[6] = LoadInputMotionVector(iPxPos + FfxInt32x2(-1, 1));
|
||||
|
||||
if (fNucleusVelocityLr > MotionVectorVelocityEpsilon) {
|
||||
for (FfxInt32 y = -1; y <= 1; ++y)
|
||||
{
|
||||
for (FfxInt32 x = -1; x <= 1; ++x)
|
||||
{
|
||||
FfxInt32 sampleIdx = (y + 1) * 3 + x + 1;
|
||||
|
||||
FfxFloat32x2 fMotionVector = fMotionVectorSamples[sampleIdx]; //LoadInputMotionVector(sp);
|
||||
FfxFloat32 fVelocityUv = length(fMotionVector);
|
||||
|
||||
fMaxVelocityUv = ffxMax(fVelocityUv, fMaxVelocityUv);
|
||||
fVelocityUv = ffxMax(fVelocityUv, fMaxVelocityUv);
|
||||
minconvergence = ffxMin(minconvergence, dot(fMotionVector / fVelocityUv, fMotionVectorNucleus / fVelocityUv));
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return ffxSaturate(1.0f - minconvergence) * ffxSaturate(fMaxVelocityUv / 0.01f);
|
||||
}
|
||||
|
||||
FfxFloat32 ComputeDepthDivergence(FfxInt32x2 iPxPos)
|
||||
{
|
||||
const FfxFloat32 fMaxDistInMeters = GetMaxDistanceInMeters();
|
||||
FfxFloat32 fDepthMax = 0.0f;
|
||||
FfxFloat32 fDepthMin = fMaxDistInMeters;
|
||||
|
||||
FfxInt32 iMaxDistFound = 0;
|
||||
|
||||
FfxInt32x2 iRenderSize = RenderSize();
|
||||
const FfxFloat32x2 fRenderSize = FfxFloat32x2(iRenderSize);
|
||||
FfxFloat32x2 fPxPosBase = FfxFloat32x2(iPxPos) / fRenderSize;
|
||||
FfxFloat32x2 fUnitUv = FfxFloat32x2(1.0f, 1.0f) / fRenderSize;
|
||||
|
||||
FfxFloat32 fDilatedDepthSamples[9];
|
||||
FfxFloat32 fTmpDummy = 0.0f;
|
||||
GatherDilatedDepthRQuad(fPxPosBase,
|
||||
fDilatedDepthSamples[0], fDilatedDepthSamples[1],
|
||||
fDilatedDepthSamples[3], fDilatedDepthSamples[4]);
|
||||
GatherDilatedDepthRQuad(fUnitUv + fPxPosBase,
|
||||
fTmpDummy, fDilatedDepthSamples[5],
|
||||
fDilatedDepthSamples[7], fDilatedDepthSamples[8]);
|
||||
fDilatedDepthSamples[2] = LoadDilatedDepth(iPxPos + FfxInt32x2(1, -1));
|
||||
fDilatedDepthSamples[6] = LoadDilatedDepth(iPxPos + FfxInt32x2(-1, 1));
|
||||
|
||||
for (FfxInt32 y = -1; y < 2; y++)
|
||||
{
|
||||
for (FfxInt32 x = -1; x < 2; x++)
|
||||
{
|
||||
FfxInt32 sampleIdx = (y + 1) * 3 + x + 1;
|
||||
const FfxInt32x2 iOffset = FfxInt32x2(x, y);
|
||||
const FfxInt32x2 iSamplePos = iPxPos + iOffset;
|
||||
|
||||
const FfxFloat32 fOnScreenFactor = IsOnScreen(iSamplePos, iRenderSize) ? 1.0f : 0.0f;
|
||||
// FfxFloat32 fDepth = GetViewSpaceDepthInMeters(LoadDilatedDepth(iSamplePos)) * fOnScreenFactor;
|
||||
FfxFloat32 fDepth = GetViewSpaceDepthInMeters(fDilatedDepthSamples[sampleIdx]) * fOnScreenFactor;
|
||||
|
||||
iMaxDistFound |= FfxInt32(fMaxDistInMeters == fDepth);
|
||||
|
||||
fDepthMin = ffxMin(fDepthMin, fDepth);
|
||||
fDepthMax = ffxMax(fDepthMax, fDepth);
|
||||
}
|
||||
}
|
||||
|
||||
return (1.0f - fDepthMin / fDepthMax) * (FfxBoolean(iMaxDistFound) ? 0.0f : 1.0f);
|
||||
}
|
||||
|
||||
FfxFloat32 ComputeTemporalMotionDivergence(FfxInt32x2 iPxPos)
|
||||
{
|
||||
const FfxFloat32x2 fUv = FfxFloat32x2(iPxPos + 0.5f) / RenderSize();
|
||||
|
||||
FfxFloat32x2 fMotionVector = LoadDilatedMotionVector(iPxPos);
|
||||
FfxFloat32x2 fReprojectedUv = fUv + fMotionVector;
|
||||
fReprojectedUv = ClampUv(fReprojectedUv, RenderSize(), MaxRenderSize());
|
||||
FfxFloat32x2 fPrevMotionVector = SamplePreviousDilatedMotionVector(fReprojectedUv);
|
||||
|
||||
float fPxDistance = length(fMotionVector * DisplaySize());
|
||||
return fPxDistance > 1.0f ? ffxLerp(0.0f, 1.0f - ffxSaturate(length(fPrevMotionVector) / length(fMotionVector)), ffxSaturate(ffxPow(fPxDistance / 20.0f, 3.0f))) : 0;
|
||||
}
|
||||
|
||||
void PreProcessReactiveMasks(FfxInt32x2 iPxLrPos, FfxFloat32 fMotionDivergence, FFXM_PARAMETER_INOUT DepthClipOutputs results)
|
||||
{
|
||||
// Compensate for bilinear sampling in accumulation pass
|
||||
|
||||
const FfxInt32x2 iRenderSize = RenderSize();
|
||||
const FfxFloat32x2 fRenderSize = FfxFloat32x2(iRenderSize);
|
||||
FfxFloat32x2 fPxPosBase = FfxFloat32x2(iPxLrPos) / fRenderSize;
|
||||
FfxFloat32x2 fUnitUv = FfxFloat32x2(1.0f, 1.0f) / fRenderSize;
|
||||
|
||||
FFXM_MIN16_F2 fReactiveFactor = FFXM_MIN16_F2(0.0f, fMotionDivergence);
|
||||
FFXM_MIN16_F fMasksSum = FFXM_MIN16_F(0.0f);
|
||||
|
||||
FFXM_MIN16_F fTmpDummy = FFXM_MIN16_F(0.0f);
|
||||
// Reactive samples
|
||||
FFXM_MIN16_F fReactiveSamples[9];
|
||||
GatherReactiveRQuad(fPxPosBase,
|
||||
fReactiveSamples[0], fReactiveSamples[1],
|
||||
fReactiveSamples[3], fReactiveSamples[4]);
|
||||
GatherReactiveRQuad(fUnitUv + fPxPosBase,
|
||||
fTmpDummy, fReactiveSamples[5],
|
||||
fReactiveSamples[7], fReactiveSamples[8]);
|
||||
fReactiveSamples[2] = FFXM_MIN16_F(LoadReactiveMask(iPxLrPos + FfxInt32x2(1, -1)));
|
||||
fReactiveSamples[6] = FFXM_MIN16_F(LoadReactiveMask(iPxLrPos + FfxInt32x2(-1, 1)));
|
||||
|
||||
// Transparency and composition mask samples
|
||||
FFXM_MIN16_F fTransparencyAndCompositionSamples[9];
|
||||
GatherTransparencyAndCompositionMaskRQuad(fPxPosBase,
|
||||
fTransparencyAndCompositionSamples[0], fTransparencyAndCompositionSamples[1],
|
||||
fTransparencyAndCompositionSamples[3], fTransparencyAndCompositionSamples[4]);
|
||||
GatherTransparencyAndCompositionMaskRQuad(fUnitUv + fPxPosBase,
|
||||
fTmpDummy, fTransparencyAndCompositionSamples[5],
|
||||
fTransparencyAndCompositionSamples[7], fTransparencyAndCompositionSamples[8]);
|
||||
fTransparencyAndCompositionSamples[2] = FFXM_MIN16_F(LoadTransparencyAndCompositionMask(iPxLrPos + FfxInt32x2(1, -1)));
|
||||
fTransparencyAndCompositionSamples[6] = FFXM_MIN16_F(LoadTransparencyAndCompositionMask(iPxLrPos + FfxInt32x2(-1, 1)));
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 y = -1; y < 2; y++)
|
||||
{
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 x = -1; x < 2; x++)
|
||||
{
|
||||
FfxInt32 sampleIdx = (y + 1) * 3 + x + 1;
|
||||
fMasksSum += (fReactiveSamples[sampleIdx] + fTransparencyAndCompositionSamples[sampleIdx]);
|
||||
}
|
||||
}
|
||||
|
||||
if (fMasksSum > FFXM_MIN16_F(0))
|
||||
{
|
||||
const FfxFloat32x2 InputColorSize = FfxFloat32x2(InputColorResourceDimensions());
|
||||
FfxFloat32x2 Base = FfxFloat32x2(iPxLrPos) / InputColorSize;
|
||||
FFXM_MIN16_F3 fInputColorSamples[9];
|
||||
// Input color samples
|
||||
GatherInputColorRGBQuad(Base,
|
||||
fInputColorSamples[0], fInputColorSamples[1], fInputColorSamples[3], fInputColorSamples[4]);
|
||||
fInputColorSamples[2] = LoadInputColor(iPxLrPos + FfxInt32x2(1, -1));
|
||||
fInputColorSamples[5] = LoadInputColor(iPxLrPos + FfxInt32x2(1, 0) );
|
||||
fInputColorSamples[6] = LoadInputColor(iPxLrPos + FfxInt32x2(-1, 1));
|
||||
fInputColorSamples[7] = LoadInputColor(iPxLrPos + FfxInt32x2(0, 1) );
|
||||
fInputColorSamples[8] = LoadInputColor(iPxLrPos + FfxInt32x2(1, 1) );
|
||||
|
||||
FFXM_MIN16_F3 fReferenceColor = fInputColorSamples[4];
|
||||
|
||||
for (FfxInt32 sampleIdx = 0; sampleIdx < 9; sampleIdx++)
|
||||
{
|
||||
FFXM_MIN16_F3 fColorSample = fInputColorSamples[sampleIdx];
|
||||
FFXM_MIN16_F fReactiveSample = fReactiveSamples[sampleIdx];
|
||||
FFXM_MIN16_F fTransparencyAndCompositionSample = fTransparencyAndCompositionSamples[sampleIdx];
|
||||
|
||||
const FfxFloat32 fMaxLenSq = ffxMax(dot(fReferenceColor, fReferenceColor), dot(fColorSample, fColorSample));
|
||||
const FFXM_MIN16_F fSimilarity = dot(fReferenceColor, fColorSample) / fMaxLenSq;
|
||||
|
||||
// Increase power for non-similar samples
|
||||
const FFXM_MIN16_F fPowerBiasMax = FFXM_MIN16_F(6.0f);
|
||||
const FFXM_MIN16_F fSimilarityPower = FFXM_MIN16_F(1.0f + (fPowerBiasMax - fSimilarity * fPowerBiasMax));
|
||||
const FFXM_MIN16_F fWeightedReactiveSample = ffxPow(fReactiveSample, fSimilarityPower);
|
||||
const FFXM_MIN16_F fWeightedTransparencyAndCompositionSample = ffxPow(fTransparencyAndCompositionSample, fSimilarityPower);
|
||||
|
||||
fReactiveFactor = ffxMax(fReactiveFactor, FFXM_MIN16_F2(fWeightedReactiveSample, fWeightedTransparencyAndCompositionSample));
|
||||
}
|
||||
}
|
||||
|
||||
results.fDilatedReactiveMasks = fReactiveFactor;
|
||||
}
|
||||
|
||||
FfxFloat32x3 ComputePreparedInputColor(FfxInt32x2 iPxLrPos)
|
||||
{
|
||||
//We assume linear data. if non-linear input (sRGB, ...),
|
||||
//then we should convert to linear first and back to sRGB on output.
|
||||
FfxFloat32x3 fRgb = ffxMax(FfxFloat32x3(0, 0, 0), LoadInputColor(iPxLrPos));
|
||||
|
||||
fRgb = PrepareRgb(fRgb, Exposure(), PreExposure());
|
||||
|
||||
#if FFXM_SHADER_QUALITY_OPT_TONEMAPPED_RGB_PREPARED_INPUT_COLOR
|
||||
const FfxFloat32x3 fPreparedYCoCg = Tonemap(fRgb);
|
||||
#else
|
||||
const FfxFloat32x3 fPreparedYCoCg = RGBToYCoCg(fRgb);
|
||||
#endif
|
||||
|
||||
return fPreparedYCoCg;
|
||||
}
|
||||
|
||||
FfxFloat32 EvaluateSurface(FfxInt32x2 iPxPos, FfxFloat32x2 fMotionVector)
|
||||
{
|
||||
FfxFloat32 d0 = GetViewSpaceDepth(LoadReconstructedPrevDepth(iPxPos + FfxInt32x2(0, -1)));
|
||||
FfxFloat32 d1 = GetViewSpaceDepth(LoadReconstructedPrevDepth(iPxPos + FfxInt32x2(0, 0)));
|
||||
FfxFloat32 d2 = GetViewSpaceDepth(LoadReconstructedPrevDepth(iPxPos + FfxInt32x2(0, 1)));
|
||||
|
||||
return 1.0f - FfxFloat32(((d0 - d1) > (d1 * 0.01f)) && ((d1 - d2) > (d2 * 0.01f)));
|
||||
}
|
||||
|
||||
DepthClipOutputs DepthClip(FfxInt32x2 iPxPos)
|
||||
{
|
||||
FfxFloat32x2 fDepthUv = (iPxPos + 0.5f) / RenderSize();
|
||||
FfxFloat32x2 fMotionVector = LoadDilatedMotionVector(iPxPos);
|
||||
|
||||
// Discard tiny mvs
|
||||
fMotionVector *= FfxFloat32(length(fMotionVector * DisplaySize()) > 0.01f);
|
||||
|
||||
const FfxFloat32x2 fDilatedUv = fDepthUv + fMotionVector;
|
||||
const FfxFloat32 fDilatedDepth = LoadDilatedDepth(iPxPos);
|
||||
const FfxFloat32 fCurrentDepthViewSpace = GetViewSpaceDepth(LoadInputDepth(iPxPos));
|
||||
|
||||
DepthClipOutputs results;
|
||||
|
||||
// Compute prepared input color and depth clip
|
||||
FfxFloat32 fDepthClip = ComputeDepthClip(fDilatedUv, fDilatedDepth) * EvaluateSurface(iPxPos, fMotionVector);
|
||||
FfxFloat32x3 fPreparedYCoCg = ComputePreparedInputColor(iPxPos);
|
||||
results.fTonemapped = FfxFloat32x4(fPreparedYCoCg, fDepthClip);
|
||||
|
||||
// Compute dilated reactive mask
|
||||
#if FFXM_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS
|
||||
FfxInt32x2 iSamplePos = iPxPos;
|
||||
#else
|
||||
FfxInt32x2 iSamplePos = ComputeHrPosFromLrPos(iPxPos);
|
||||
#endif
|
||||
|
||||
FfxFloat32 fMotionDivergence = ComputeMotionDivergence(iSamplePos, RenderSize());
|
||||
FfxFloat32 fTemporalMotionDifference = ffxSaturate(ComputeTemporalMotionDivergence(iPxPos) - ComputeDepthDivergence(iPxPos));
|
||||
|
||||
PreProcessReactiveMasks(iPxPos, ffxMax(fTemporalMotionDifference, fMotionDivergence), results);
|
||||
|
||||
return results;
|
||||
}
|
||||
|
||||
#endif //!defined( FFXM_FSR2_DEPTH_CLIPH )
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 69553130c6fc19f44830c8f8c088a239
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,72 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#define FSR2_BIND_SRV_RECONSTRUCTED_PREV_NEAREST_DEPTH 0
|
||||
#define FSR2_BIND_SRV_DILATED_MOTION_VECTORS 1
|
||||
#define FSR2_BIND_SRV_DILATED_DEPTH 2
|
||||
#define FSR2_BIND_SRV_REACTIVE_MASK 3
|
||||
#define FSR2_BIND_SRV_TRANSPARENCY_AND_COMPOSITION_MASK 4
|
||||
#define FSR2_BIND_SRV_PREVIOUS_DILATED_MOTION_VECTORS 5
|
||||
#define FSR2_BIND_SRV_INPUT_MOTION_VECTORS 6
|
||||
#define FSR2_BIND_SRV_INPUT_COLOR 7
|
||||
#define FSR2_BIND_SRV_INPUT_DEPTH 8
|
||||
#define FSR2_BIND_SRV_INPUT_EXPOSURE 9
|
||||
|
||||
#define FSR2_BIND_CB_FSR2 0
|
||||
|
||||
// Global mandatory defines
|
||||
#if !defined(FFXM_HALF)
|
||||
#define FFXM_HALF 1
|
||||
#endif
|
||||
#if !defined(FFXM_GPU)
|
||||
#define FFXM_GPU 1
|
||||
#endif
|
||||
#if !defined(FFXM_HLSL)
|
||||
#define FFXM_HLSL 1
|
||||
#endif
|
||||
|
||||
#include "ffxm_fsr2_callbacks_hlsl.h"
|
||||
#include "ffxm_fsr2_common.h"
|
||||
#include "ffxm_fsr2_sample.h"
|
||||
#include "ffxm_fsr2_depth_clip.h"
|
||||
|
||||
struct VertexOut
|
||||
{
|
||||
float4 position : SV_POSITION;
|
||||
};
|
||||
|
||||
|
||||
struct DepthClipOutputsFS
|
||||
{
|
||||
FfxFloat32x2 fDilatedReactiveMasks : SV_TARGET0;
|
||||
FfxFloat32x4 fTonemapped : SV_TARGET1;
|
||||
};
|
||||
|
||||
DepthClipOutputsFS main(float4 SvPosition : SV_POSITION)
|
||||
{
|
||||
uint2 uPixelCoord = uint2(SvPosition.xy);
|
||||
DepthClipOutputs result = DepthClip(uPixelCoord);
|
||||
DepthClipOutputsFS output = (DepthClipOutputsFS)0;
|
||||
output.fDilatedReactiveMasks = result.fDilatedReactiveMasks;
|
||||
output.fTonemapped = result.fTonemapped;
|
||||
return output;
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 995ceded3b171f54fa7fd3570f9c467b
|
||||
ShaderIncludeImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,131 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#ifndef FFXM_FSR2_LOCK_H
|
||||
#define FFXM_FSR2_LOCK_H
|
||||
|
||||
void ClearResourcesForNextFrame(in FfxInt32x2 iPxHrPos)
|
||||
{
|
||||
if (all(FFXM_LESS_THAN(iPxHrPos, FfxInt32x2(RenderSize()))))
|
||||
{
|
||||
#if FFXM_FSR2_OPTION_INVERTED_DEPTH
|
||||
const FfxUInt32 farZ = 0x0;
|
||||
#else
|
||||
const FfxUInt32 farZ = 0x3f800000;
|
||||
#endif
|
||||
SetReconstructedDepth(iPxHrPos, farZ);
|
||||
}
|
||||
}
|
||||
|
||||
FfxBoolean ComputeThinFeatureConfidence(FfxInt32x2 pos)
|
||||
{
|
||||
const FfxInt32 RADIUS = 1;
|
||||
|
||||
FFXM_MIN16_F fNucleus = LoadLockInputLuma(pos);
|
||||
|
||||
FFXM_MIN16_F similar_threshold = FFXM_MIN16_F(1.05f);
|
||||
FFXM_MIN16_F dissimilarLumaMin = FFXM_MIN16_F(FSR2_FP16_MAX);
|
||||
FFXM_MIN16_F dissimilarLumaMax = FFXM_MIN16_F(0);
|
||||
|
||||
/*
|
||||
0 1 2
|
||||
3 4 5
|
||||
6 7 8
|
||||
*/
|
||||
|
||||
#define SETBIT(x) (1U << x)
|
||||
|
||||
FfxUInt32 mask = SETBIT(4); //flag fNucleus as similar
|
||||
|
||||
const FfxUInt32 uNumRejectionMasks = 4;
|
||||
const FfxUInt32 uRejectionMasks[uNumRejectionMasks] = {
|
||||
SETBIT(0) | SETBIT(1) | SETBIT(3) | SETBIT(4), //Upper left
|
||||
SETBIT(1) | SETBIT(2) | SETBIT(4) | SETBIT(5), //Upper right
|
||||
SETBIT(3) | SETBIT(4) | SETBIT(6) | SETBIT(7), //Lower left
|
||||
SETBIT(4) | SETBIT(5) | SETBIT(7) | SETBIT(8), //Lower right
|
||||
};
|
||||
|
||||
FFXM_MIN16_F lumaSamples [9];
|
||||
FFXM_MIN16_F fTmpDummy = FFXM_MIN16_F(0.0f);
|
||||
const FfxFloat32x2 fInputLumaSize = FfxFloat32x2(RenderSize());
|
||||
const FfxFloat32x2 fPxBaseUv = FfxFloat32x2(pos) / fInputLumaSize;
|
||||
const FfxFloat32x2 fUnitUv = FfxFloat32x2(1.0f, 1.0f) / fInputLumaSize;
|
||||
|
||||
// Gather samples
|
||||
GatherLockInputLumaRQuad(fPxBaseUv,
|
||||
lumaSamples[0], lumaSamples[1],
|
||||
lumaSamples[3], lumaSamples[4]);
|
||||
GatherLockInputLumaRQuad(fUnitUv + fPxBaseUv,
|
||||
fTmpDummy, lumaSamples[5],
|
||||
lumaSamples[7], lumaSamples[8]);
|
||||
lumaSamples[2] = LoadLockInputLuma(pos + FfxInt32x2(1, -1));
|
||||
lumaSamples[6] = LoadLockInputLuma(pos + FfxInt32x2(-1, 1));
|
||||
|
||||
FfxInt32 idx = 0;
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 y = -RADIUS; y <= RADIUS; y++) {
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 x = -RADIUS; x <= RADIUS; x++, idx++) {
|
||||
if (x == 0 && y == 0) continue;
|
||||
|
||||
FfxInt32 sampleIdx = (y + 1) * 3 + x + 1;
|
||||
FFXM_MIN16_F sampleLuma = lumaSamples[sampleIdx];
|
||||
|
||||
FFXM_MIN16_F difference = ffxMax(sampleLuma, fNucleus) / ffxMin(sampleLuma, fNucleus);
|
||||
|
||||
if (difference > FFXM_MIN16_F(0) && (difference < similar_threshold)) {
|
||||
mask |= SETBIT(idx);
|
||||
} else {
|
||||
dissimilarLumaMin = ffxMin(dissimilarLumaMin, sampleLuma);
|
||||
dissimilarLumaMax = ffxMax(dissimilarLumaMax, sampleLuma);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
FfxBoolean isRidge = fNucleus > dissimilarLumaMax || fNucleus < dissimilarLumaMin;
|
||||
|
||||
if (FFXM_FALSE == isRidge) {
|
||||
|
||||
return false;
|
||||
}
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 i = 0; i < 4; i++) {
|
||||
|
||||
if ((mask & uRejectionMasks[i]) == uRejectionMasks[i]) {
|
||||
return false;
|
||||
}
|
||||
}
|
||||
|
||||
return true;
|
||||
}
|
||||
|
||||
void ComputeLock(FfxInt32x2 iPxLrPos)
|
||||
{
|
||||
if (ComputeThinFeatureConfidence(iPxLrPos))
|
||||
{
|
||||
StoreNewLocks(ComputeHrPosFromLrPos(iPxLrPos), 1.f);
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
#endif // FFXM_FSR2_LOCK_H
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 59e5fd93b44cc1a41a8018ee57143e14
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,66 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#define FSR2_BIND_SRV_LOCK_INPUT_LUMA 0
|
||||
|
||||
#define FSR2_BIND_UAV_NEW_LOCKS 1
|
||||
|
||||
#define FSR2_BIND_CB_FSR2 0
|
||||
|
||||
// Global mandatory defines
|
||||
#if !defined(FFXM_HALF)
|
||||
#define FFXM_HALF 1
|
||||
#endif
|
||||
#if !defined(FFXM_GPU)
|
||||
#define FFXM_GPU 1
|
||||
#endif
|
||||
#if !defined(FFXM_HLSL)
|
||||
#define FFXM_HLSL 1
|
||||
#endif
|
||||
|
||||
#include "ffxm_fsr2_callbacks_hlsl.h"
|
||||
#include "ffxm_fsr2_common.h"
|
||||
#include "ffxm_fsr2_sample.h"
|
||||
#include "ffxm_fsr2_lock.h"
|
||||
|
||||
#ifndef FFXM_FSR2_THREAD_GROUP_WIDTH
|
||||
#define FFXM_FSR2_THREAD_GROUP_WIDTH 8
|
||||
#endif // #ifndef FFXM_FSR2_THREAD_GROUP_WIDTH
|
||||
#ifndef FFXM_FSR2_THREAD_GROUP_HEIGHT
|
||||
#define FFXM_FSR2_THREAD_GROUP_HEIGHT 8
|
||||
#endif // #ifndef FFXM_FSR2_THREAD_GROUP_HEIGHT
|
||||
#ifndef FFXM_FSR2_THREAD_GROUP_DEPTH
|
||||
#define FFXM_FSR2_THREAD_GROUP_DEPTH 1
|
||||
#endif // #ifndef FFXM_FSR2_THREAD_GROUP_DEPTH
|
||||
#ifndef FFXM_FSR2_NUM_THREADS
|
||||
#define FFXM_FSR2_NUM_THREADS [numthreads(FFXM_FSR2_THREAD_GROUP_WIDTH, FFXM_FSR2_THREAD_GROUP_HEIGHT, FFXM_FSR2_THREAD_GROUP_DEPTH)]
|
||||
#endif // #ifndef FFXM_FSR2_NUM_THREADS
|
||||
|
||||
FFXM_PREFER_WAVE64
|
||||
FFXM_FSR2_NUM_THREADS
|
||||
FFXM_FSR2_EMBED_ROOTSIG_CONTENT
|
||||
void main(uint2 uGroupId : SV_GroupID, uint2 uGroupThreadId : SV_GroupThreadID)
|
||||
{
|
||||
uint2 uDispatchThreadId = uGroupId * uint2(FFXM_FSR2_THREAD_GROUP_WIDTH, FFXM_FSR2_THREAD_GROUP_HEIGHT) + uGroupThreadId;
|
||||
|
||||
ComputeLock(uDispatchThreadId);
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: b593df7c75a95fa44a6e39bf993f3f3a
|
||||
ShaderIncludeImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,102 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#ifndef FFXM_FSR2_POSTPROCESS_LOCK_STATUS_H
|
||||
#define FFXM_FSR2_POSTPROCESS_LOCK_STATUS_H
|
||||
|
||||
FfxFloat32x4 WrapShadingChangeLuma(FfxInt32x2 iPxSample)
|
||||
{
|
||||
return FfxFloat32x4(LoadMipLuma(iPxSample, LumaMipLevelToUse()), 0, 0, 0);
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F4 WrapShadingChangeLuma(FFXM_MIN16_I2 iPxSample)
|
||||
{
|
||||
return FFXM_MIN16_F4(LoadMipLuma(iPxSample, LumaMipLevelToUse()), 0, 0, 0);
|
||||
}
|
||||
#endif
|
||||
|
||||
#if FFXM_HALF
|
||||
DeclareCustomFetchBilinearSamplesMin16(FetchShadingChangeLumaSamples, WrapShadingChangeLuma)
|
||||
DeclareCustomTextureSampleMin16(ShadingChangeLumaSample, Bilinear, FetchShadingChangeLumaSamples)
|
||||
#else
|
||||
DeclareCustomFetchBicubicSamples(FetchShadingChangeLumaSamples, WrapShadingChangeLuma)
|
||||
DeclareCustomTextureSample(ShadingChangeLumaSample, Lanczos2, FetchShadingChangeLumaSamples)
|
||||
#endif
|
||||
|
||||
FfxFloat32 GetShadingChangeLuma(FfxInt32x2 iPxHrPos, FfxFloat32x2 fUvCoord)
|
||||
{
|
||||
FfxFloat32 fShadingChangeLuma = 0;
|
||||
const FfxFloat32 fDiv = FfxFloat32(FfxInt32(2) << LumaMipLevelToUse());
|
||||
FfxInt32x2 iMipRenderSize = FfxInt32x2(RenderSize() / fDiv);
|
||||
|
||||
fUvCoord = ClampUv(fUvCoord, iMipRenderSize, LumaMipDimensions());
|
||||
fShadingChangeLuma = Exposure() * exp(FfxFloat32(SampleMipLuma(fUvCoord, LumaMipLevelToUse())));
|
||||
|
||||
fShadingChangeLuma = ffxPow(fShadingChangeLuma, 1.0f / 6.0f);
|
||||
|
||||
return fShadingChangeLuma;
|
||||
}
|
||||
|
||||
void UpdateLockStatus(AccumulationPassCommonParams params,
|
||||
FFXM_PARAMETER_INOUT FfxFloat32 fReactiveFactor, LockState state,
|
||||
FFXM_PARAMETER_INOUT FfxFloat32x2 fLockStatus,
|
||||
FFXM_PARAMETER_OUT FfxFloat32 fLockContributionThisFrame,
|
||||
FFXM_PARAMETER_OUT FfxFloat32 fLuminanceDiff) {
|
||||
|
||||
const FfxFloat32 fShadingChangeLuma = GetShadingChangeLuma(params.iPxHrPos, params.fHrUv);
|
||||
|
||||
//init temporal shading change factor, init to -1 or so in reproject to know if "true new"?
|
||||
fLockStatus[LOCK_TEMPORAL_LUMA] = (fLockStatus[LOCK_TEMPORAL_LUMA] == FfxFloat32(0.0f)) ? fShadingChangeLuma : fLockStatus[LOCK_TEMPORAL_LUMA];
|
||||
|
||||
FfxFloat32 fPreviousShadingChangeLuma = fLockStatus[LOCK_TEMPORAL_LUMA];
|
||||
|
||||
fLuminanceDiff = 1.0f - MinDividedByMax(fPreviousShadingChangeLuma, fShadingChangeLuma);
|
||||
|
||||
if (state.NewLock) {
|
||||
fLockStatus[LOCK_TEMPORAL_LUMA] = fShadingChangeLuma;
|
||||
|
||||
fLockStatus[LOCK_LIFETIME_REMAINING] = (fLockStatus[LOCK_LIFETIME_REMAINING] != 0.0f) ? 2.0f : 1.0f;
|
||||
}
|
||||
else if(fLockStatus[LOCK_LIFETIME_REMAINING] <= 1.0f) {
|
||||
fLockStatus[LOCK_TEMPORAL_LUMA] = ffxLerp(fLockStatus[LOCK_TEMPORAL_LUMA], FfxFloat32(fShadingChangeLuma), 0.5f);
|
||||
}
|
||||
else {
|
||||
if (fLuminanceDiff > 0.1f) {
|
||||
KillLock(fLockStatus);
|
||||
}
|
||||
}
|
||||
|
||||
fReactiveFactor = ffxMax(fReactiveFactor, ffxSaturate((fLuminanceDiff - 0.1f) * 10.0f));
|
||||
fLockStatus[LOCK_LIFETIME_REMAINING] *= (1.0f - fReactiveFactor);
|
||||
|
||||
fLockStatus[LOCK_LIFETIME_REMAINING] *= ffxSaturate(1.0f - params.fAccumulationMask);
|
||||
fLockStatus[LOCK_LIFETIME_REMAINING] *= FfxFloat32(params.fDepthClipFactor < 0.1f);
|
||||
|
||||
// Compute this frame lock contribution
|
||||
const FfxFloat32 fLifetimeContribution = ffxSaturate(fLockStatus[LOCK_LIFETIME_REMAINING] - 1.0f);
|
||||
const FfxFloat32 fShadingChangeContribution = ffxSaturate(MinDividedByMax(fLockStatus[LOCK_TEMPORAL_LUMA], fShadingChangeLuma));
|
||||
|
||||
fLockContributionThisFrame = ffxSaturate(ffxSaturate(fLifetimeContribution * 4.0f) * fShadingChangeContribution);
|
||||
}
|
||||
|
||||
#endif //!defined( FFXM_FSR2_POSTPROCESS_LOCK_STATUS_H )
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: ac8f3985631853c458b022f379c55fff
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,92 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#define GROUP_SIZE 8
|
||||
#define FSR_RCAS_DENOISE 1
|
||||
|
||||
#include "ffxm_core.h"
|
||||
|
||||
struct RCASOutputs
|
||||
{
|
||||
FfxFloat32x3 fUpscaledColor;
|
||||
};
|
||||
|
||||
#if FFXM_HALF
|
||||
#define USE_FSR_RCASH 1
|
||||
#else
|
||||
#define USE_FSR_RCASH 0
|
||||
#endif
|
||||
|
||||
#if USE_FSR_RCASH
|
||||
#define FSR_RCAS_H 1
|
||||
FfxFloat16x4 FsrRcasLoadH(FfxInt16x2 p)
|
||||
{
|
||||
FfxFloat16x4 fColor = LoadRCAS_Input(p);
|
||||
fColor.rgb = FfxFloat16x3(PrepareRgb(fColor.rgb, Exposure(), PreExposure()));
|
||||
return fColor;
|
||||
}
|
||||
void FsrRcasInputH(inout FfxFloat16 r,inout FfxFloat16 g,inout FfxFloat16 b)
|
||||
{
|
||||
|
||||
}
|
||||
|
||||
#else
|
||||
#define FSR_RCAS_F 1
|
||||
FfxFloat32x4 FsrRcasLoadF(FfxInt32x2 p)
|
||||
{
|
||||
FfxFloat32x4 fColor = LoadRCAS_Input(p);
|
||||
|
||||
fColor.rgb = PrepareRgb(fColor.rgb, Exposure(), PreExposure());
|
||||
|
||||
return fColor;
|
||||
}
|
||||
void FsrRcasInputF(inout FfxFloat32 r, inout FfxFloat32 g, inout FfxFloat32 b) {}
|
||||
#endif
|
||||
|
||||
#include "ffxm_fsr1.h"
|
||||
|
||||
void CurrFilter(FFXM_MIN16_U2 pos, FFXM_PARAMETER_INOUT RCASOutputs results)
|
||||
{
|
||||
#if USE_FSR_RCASH
|
||||
FfxFloat16x3 c;
|
||||
FsrRcasH(c.r, c.g, c.b, pos, RCASConfig());
|
||||
|
||||
c = UnprepareRgb(c, FfxFloat16(Exposure()));
|
||||
#else
|
||||
FfxFloat32x3 c;
|
||||
FsrRcasF(c.r, c.g, c.b, pos, RCASConfig());
|
||||
|
||||
c = UnprepareRgb(c, Exposure());
|
||||
#endif
|
||||
results.fUpscaledColor = c;
|
||||
}
|
||||
|
||||
RCASOutputs RCAS(FfxUInt32x2 gxy)
|
||||
{
|
||||
#ifdef FFXM_HLSL
|
||||
RCASOutputs results = (RCASOutputs)0;
|
||||
#else
|
||||
RCASOutputs results;
|
||||
#endif
|
||||
CurrFilter(FFXM_MIN16_U2(gxy), results);
|
||||
return results;
|
||||
}
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 3ad625febda45744da4de219101f1515
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,61 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#define FSR2_BIND_SRV_INPUT_EXPOSURE 0
|
||||
#define FSR2_BIND_SRV_RCAS_INPUT 1
|
||||
|
||||
#define FSR2_BIND_CB_FSR2 0
|
||||
#define FSR2_BIND_CB_RCAS 1
|
||||
|
||||
// Global mandatory defines
|
||||
#if !defined(FFXM_HALF)
|
||||
#define FFXM_HALF 1
|
||||
#endif
|
||||
#if !defined(FFXM_GPU)
|
||||
#define FFXM_GPU 1
|
||||
#endif
|
||||
#if !defined(FFXM_HLSL)
|
||||
#define FFXM_HLSL 1
|
||||
#endif
|
||||
|
||||
#include "ffxm_fsr2_callbacks_hlsl.h"
|
||||
#include "ffxm_fsr2_common.h"
|
||||
#include "ffxm_fsr2_rcas.h"
|
||||
|
||||
struct VertexOut
|
||||
{
|
||||
float4 position : SV_POSITION;
|
||||
};
|
||||
|
||||
struct RCASOutputsFS
|
||||
{
|
||||
FfxFloat32x3 fUpscaledColor : SV_TARGET0;
|
||||
};
|
||||
|
||||
RCASOutputsFS main(float4 SvPosition : SV_POSITION)
|
||||
{
|
||||
uint2 uPixelCoord = uint2(SvPosition.xy);
|
||||
RCASOutputs result = RCAS(uPixelCoord);
|
||||
RCASOutputsFS output = (RCASOutputsFS)0;
|
||||
output.fUpscaledColor = result.fUpscaledColor;
|
||||
return output;
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 3f2db786e458b8944bb05f46f33e00e2
|
||||
ShaderIncludeImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,156 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#ifndef FFXM_FSR2_RECONSTRUCT_DILATED_VELOCITY_AND_PREVIOUS_DEPTH_H
|
||||
#define FFXM_FSR2_RECONSTRUCT_DILATED_VELOCITY_AND_PREVIOUS_DEPTH_H
|
||||
|
||||
struct ReconstructPrevDepthOutputs
|
||||
{
|
||||
FfxFloat32 fDepth;
|
||||
FfxFloat32x2 fMotionVector;
|
||||
FfxFloat32 fLuma;
|
||||
};
|
||||
|
||||
|
||||
void ReconstructPrevDepth(FfxInt32x2 iPxPos, FfxFloat32 fDepth, FfxFloat32x2 fMotionVector, FfxInt32x2 iPxDepthSize)
|
||||
{
|
||||
fMotionVector *= FfxFloat32(length(fMotionVector * DisplaySize()) > 0.1f);
|
||||
|
||||
FfxFloat32x2 fUv = (iPxPos + FfxFloat32(0.5)) / iPxDepthSize;
|
||||
FfxFloat32x2 fReprojectedUv = fUv + fMotionVector;
|
||||
|
||||
BilinearSamplingData bilinearInfo = GetBilinearSamplingData(fReprojectedUv, RenderSize());
|
||||
|
||||
// Project current depth into previous frame locations.
|
||||
// Push to all pixels having some contribution if reprojection is using bilinear logic.
|
||||
for (FfxInt32 iSampleIndex = 0; iSampleIndex < 4; iSampleIndex++) {
|
||||
|
||||
const FfxInt32x2 iOffset = bilinearInfo.iOffsets[iSampleIndex];
|
||||
FfxFloat32 fWeight = bilinearInfo.fWeights[iSampleIndex];
|
||||
|
||||
if (fWeight > fReconstructedDepthBilinearWeightThreshold) {
|
||||
|
||||
FfxInt32x2 iStorePos = bilinearInfo.iBasePos + iOffset;
|
||||
if (IsOnScreen(iStorePos, iPxDepthSize)) {
|
||||
StoreReconstructedDepth(iStorePos, fDepth);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
void FindNearestDepth(FFXM_PARAMETER_IN FfxInt32x2 iPxPos, FFXM_PARAMETER_IN FfxInt32x2 iPxSize, FFXM_PARAMETER_OUT FfxFloat32 fNearestDepth, FFXM_PARAMETER_OUT FfxInt32x2 fNearestDepthCoord)
|
||||
{
|
||||
const FfxInt32 iSampleCount = 9;
|
||||
const FfxInt32x2 iSampleOffsets[iSampleCount] = {
|
||||
FfxInt32x2(+0, +0),
|
||||
FfxInt32x2(+1, +0),
|
||||
FfxInt32x2(+0, +1),
|
||||
FfxInt32x2(+0, -1),
|
||||
FfxInt32x2(-1, +0),
|
||||
FfxInt32x2(-1, +1),
|
||||
FfxInt32x2(+1, +1),
|
||||
FfxInt32x2(-1, -1),
|
||||
FfxInt32x2(+1, -1),
|
||||
};
|
||||
|
||||
// pull out the depth loads to allow SC to batch them
|
||||
FfxFloat32 depth[9];
|
||||
FfxInt32 iSampleIndex = 0;
|
||||
FFXM_UNROLL
|
||||
for (iSampleIndex = 0; iSampleIndex < iSampleCount; ++iSampleIndex) {
|
||||
|
||||
FfxInt32x2 iPos = iPxPos + iSampleOffsets[iSampleIndex];
|
||||
depth[iSampleIndex] = LoadInputDepth(iPos);
|
||||
}
|
||||
|
||||
// find closest depth
|
||||
fNearestDepthCoord = iPxPos;
|
||||
fNearestDepth = depth[0];
|
||||
FFXM_UNROLL
|
||||
for (iSampleIndex = 1; iSampleIndex < iSampleCount; ++iSampleIndex) {
|
||||
|
||||
FfxInt32x2 iPos = iPxPos + iSampleOffsets[iSampleIndex];
|
||||
if (IsOnScreen(iPos, iPxSize)) {
|
||||
|
||||
FfxFloat32 fNdDepth = depth[iSampleIndex];
|
||||
#if FFXM_FSR2_OPTION_INVERTED_DEPTH
|
||||
if (fNdDepth > fNearestDepth) {
|
||||
#else
|
||||
if (fNdDepth < fNearestDepth) {
|
||||
#endif
|
||||
fNearestDepthCoord = iPos;
|
||||
fNearestDepth = fNdDepth;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
FfxFloat32 ComputeLockInputLuma(FfxInt32x2 iPxLrPos)
|
||||
{
|
||||
//We assume linear data. if non-linear input (sRGB, ...),
|
||||
//then we should convert to linear first and back to sRGB on output.
|
||||
FfxFloat32x3 fRgb = ffxMax(FfxFloat32x3(0, 0, 0), LoadInputColor(iPxLrPos));
|
||||
|
||||
// Use internal auto exposure for locking logic
|
||||
fRgb /= PreExposure();
|
||||
fRgb *= Exposure();
|
||||
|
||||
#if FFXM_FSR2_OPTION_HDR_COLOR_INPUT
|
||||
fRgb = Tonemap(fRgb);
|
||||
#endif
|
||||
|
||||
//compute luma used to lock pixels, if used elsewhere the ffxPow must be moved!
|
||||
const FfxFloat32 fLockInputLuma = ffxPow(RGBToPerceivedLuma(fRgb), FfxFloat32(1.0 / 6.0));
|
||||
|
||||
return fLockInputLuma;
|
||||
}
|
||||
|
||||
ReconstructPrevDepthOutputs ReconstructAndDilate(FfxInt32x2 iPxLrPos)
|
||||
{
|
||||
FfxFloat32 fDilatedDepth;
|
||||
FfxInt32x2 iNearestDepthCoord;
|
||||
|
||||
FindNearestDepth(iPxLrPos, RenderSize(), fDilatedDepth, iNearestDepthCoord);
|
||||
|
||||
#if FFXM_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS
|
||||
FfxInt32x2 iSamplePos = iPxLrPos;
|
||||
FfxInt32x2 iMotionVectorPos = iNearestDepthCoord;
|
||||
#else
|
||||
FfxInt32x2 iSamplePos = ComputeHrPosFromLrPos(iPxLrPos);
|
||||
FfxInt32x2 iMotionVectorPos = ComputeHrPosFromLrPos(iNearestDepthCoord);
|
||||
#endif
|
||||
|
||||
FfxFloat32x2 fDilatedMotionVector = LoadInputMotionVector(iMotionVectorPos);
|
||||
|
||||
ReconstructPrevDepthOutputs results;
|
||||
|
||||
results.fDepth = fDilatedDepth;
|
||||
results.fMotionVector = fDilatedMotionVector;
|
||||
ReconstructPrevDepth(iPxLrPos, fDilatedDepth, fDilatedMotionVector, RenderSize());
|
||||
FfxFloat32 fLockInputLuma = ComputeLockInputLuma(iPxLrPos);
|
||||
results.fLuma = fLockInputLuma;
|
||||
|
||||
return results;
|
||||
}
|
||||
|
||||
|
||||
#endif //!defined( FFXM_FSR2_RECONSTRUCT_DILATED_VELOCITY_AND_PREVIOUS_DEPTH_H )
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: bfecdfb9980efa94a9f94cb96d959870
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,70 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#define FSR2_BIND_SRV_INPUT_MOTION_VECTORS 0
|
||||
#define FSR2_BIND_SRV_INPUT_DEPTH 1
|
||||
#define FSR2_BIND_SRV_INPUT_COLOR 2
|
||||
#define FSR2_BIND_SRV_INPUT_EXPOSURE 3
|
||||
|
||||
#define FSR2_BIND_UAV_RECONSTRUCTED_PREV_NEAREST_DEPTH 4
|
||||
|
||||
#define FSR2_BIND_CB_FSR2 0
|
||||
|
||||
// Global mandatory defines
|
||||
#if !defined(FFXM_HALF)
|
||||
#define FFXM_HALF 1
|
||||
#endif
|
||||
#if !defined(FFXM_GPU)
|
||||
#define FFXM_GPU 1
|
||||
#endif
|
||||
#if !defined(FFXM_HLSL)
|
||||
#define FFXM_HLSL 1
|
||||
#endif
|
||||
|
||||
#include "ffxm_fsr2_callbacks_hlsl.h"
|
||||
#include "ffxm_fsr2_common.h"
|
||||
#include "ffxm_fsr2_sample.h"
|
||||
#include "ffxm_fsr2_reconstruct_dilated_velocity_and_previous_depth.h"
|
||||
|
||||
struct VertexOut
|
||||
{
|
||||
float4 position : SV_POSITION;
|
||||
};
|
||||
|
||||
struct ReconstructPrevDepthOutputsFS
|
||||
{
|
||||
FfxFloat32 fDepth : SV_TARGET0;
|
||||
FfxFloat32x2 fMotionVector : SV_TARGET1;
|
||||
FfxFloat32 fLuma : SV_TARGET2;
|
||||
};
|
||||
|
||||
|
||||
ReconstructPrevDepthOutputsFS main(float4 SvPosition : SV_POSITION)
|
||||
{
|
||||
uint2 uPixelCoord = uint2(SvPosition.xy);
|
||||
ReconstructPrevDepthOutputs result = ReconstructAndDilate(uPixelCoord);
|
||||
ReconstructPrevDepthOutputsFS output = (ReconstructPrevDepthOutputsFS)0;
|
||||
output.fDepth = result.fDepth;
|
||||
output.fMotionVector = result.fMotionVector;
|
||||
output.fLuma = result.fLuma;
|
||||
return output;
|
||||
}
|
||||
@ -0,0 +1,7 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 940757c9e1206ee4297ed09063f780bd
|
||||
ShaderIncludeImporter:
|
||||
externalObjects: {}
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,387 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#ifndef FFXM_FSR2_REPROJECT_H
|
||||
#define FFXM_FSR2_REPROJECT_H
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F4 WrapHistory(FFXM_MIN16_I2 iPxSample)
|
||||
{
|
||||
return FFXM_MIN16_F4(LoadHistory(iPxSample));
|
||||
}
|
||||
FFXM_MIN16_F4 SampleHistory(FfxFloat32x2 fUV)
|
||||
{
|
||||
return SampleUpscaledHistory(fUV);
|
||||
}
|
||||
#else
|
||||
FfxFloat32x4 WrapHistory(FfxInt32x2 iPxSample)
|
||||
{
|
||||
return LoadHistory(iPxSample);
|
||||
}
|
||||
FfxFloat32x4 SampleHistory(FfxFloat32x2 fUV)
|
||||
{
|
||||
return SampleUpscaledHistory(fUV);
|
||||
}
|
||||
#endif
|
||||
|
||||
|
||||
#if FFXM_HALF
|
||||
|
||||
#define FFXM_FSR2_REPROJECT_CATMULL_9TAP 0
|
||||
#define FFXM_FSR2_REPROJECT_LANCZOS_APPROX_9TAP 1
|
||||
#define FFXM_FSR2_REPROJECT_CATMULL_5TAP 2
|
||||
|
||||
#if FFXM_SHADER_QUALITY_OPT_REPROJECT_CATMULL_5TAP
|
||||
#define FFXM_FSR2_REPROJECT_MODE FFXM_FSR2_REPROJECT_CATMULL_5TAP
|
||||
#elif FFXM_SHADER_QUALITY_OPT_REPROJECT_CATMULL_9TAP
|
||||
#define FFXM_FSR2_REPROJECT_MODE FFXM_FSR2_REPROJECT_CATMULL_9TAP
|
||||
#else // QUALITY
|
||||
#define FFXM_FSR2_REPROJECT_MODE FFXM_FSR2_REPROJECT_CATMULL_9TAP
|
||||
#endif
|
||||
|
||||
#if (FFXM_FSR2_REPROJECT_MODE == FFXM_FSR2_REPROJECT_CATMULL_9TAP)
|
||||
struct CatmullRomSamples9Tap
|
||||
{
|
||||
// bilinear sampling UV coordinates of the samples
|
||||
FfxFloat32x2 UV[3];
|
||||
|
||||
// weights of the samples
|
||||
FFXM_MIN16_F2 Weight[3];
|
||||
|
||||
// final multiplier (it is faster to multiply 3 RGB values than reweights the 5 weights)
|
||||
FFXM_MIN16_F FinalMultiplier;
|
||||
};
|
||||
|
||||
CatmullRomSamples9Tap Get2DCatmullRom9Kernel(FfxFloat32x2 uv, FfxFloat32x2 size, in FfxFloat32x2 invSize)
|
||||
{
|
||||
CatmullRomSamples9Tap catmullSamples;
|
||||
FfxFloat32x2 samplePos = uv * size;
|
||||
FfxFloat32x2 texPos1 = floor(samplePos - 0.5f) + 0.5f;
|
||||
FfxFloat32x2 f = samplePos - texPos1;
|
||||
|
||||
FfxFloat32x2 w0 = f * (-0.5f + f * (1.0f - 0.5f * f));
|
||||
FfxFloat32x2 w1 = 1.0f + f * f * (-2.5f + 1.5f * f);
|
||||
FfxFloat32x2 w2 = f * (0.5f + f * (2.0f - 1.5f * f));
|
||||
FfxFloat32x2 w3 = f * f * (-0.5f + 0.5f * f);
|
||||
|
||||
catmullSamples.Weight[0] = FFXM_MIN16_F2(w0);
|
||||
catmullSamples.Weight[1] = FFXM_MIN16_F2(w1 + w2);
|
||||
catmullSamples.Weight[2] = FFXM_MIN16_F2(w3);
|
||||
|
||||
FfxFloat32x2 offset12 = w2 / (w1 + w2);
|
||||
|
||||
// Compute the final UV coordinates we'll use for sampling the texture
|
||||
catmullSamples.UV[0] = FfxFloat32x2(texPos1 - 1);
|
||||
catmullSamples.UV[1] = FfxFloat32x2(texPos1 + 2);
|
||||
catmullSamples.UV[2] = FfxFloat32x2(texPos1 + offset12);
|
||||
|
||||
catmullSamples.UV[0] = FfxFloat32x2(catmullSamples.UV[0]*invSize);
|
||||
catmullSamples.UV[1] = FfxFloat32x2(catmullSamples.UV[1]*invSize);
|
||||
catmullSamples.UV[2] = FfxFloat32x2(catmullSamples.UV[2]*invSize);
|
||||
return catmullSamples;
|
||||
}
|
||||
|
||||
FFXM_MIN16_F4 HistorySample(FfxFloat32x2 fUvSample, FfxInt32x2 iTextureSize)
|
||||
{
|
||||
FfxFloat32x2 fPxSample = (fUvSample * FfxFloat32x2(iTextureSize)) - FfxFloat32x2(0.5f, 0.5f);
|
||||
FfxFloat32x2 fTextureSize = FfxFloat32x2(iTextureSize);
|
||||
FfxFloat32x2 fInvTextureSize = FfxFloat32x2(1.0f, 1.0f) / fTextureSize;
|
||||
CatmullRomSamples9Tap samples = Get2DCatmullRom9Kernel(fUvSample, fTextureSize, fInvTextureSize);
|
||||
|
||||
FFXM_MIN16_F4 fColor = FFXM_MIN16_F4(0.0f, 0.0f, 0.0f, 0.0f);
|
||||
|
||||
FFXM_MIN16_F4 fColor00 = SampleHistory(FfxFloat32x2(samples.UV[0]));
|
||||
fColor += fColor00 * samples.Weight[0].x * samples.Weight[0].y;
|
||||
FFXM_MIN16_F4 fColor20 = SampleHistory(FfxFloat32x2(samples.UV[2].x, samples.UV[0].y));
|
||||
fColor += fColor20 * samples.Weight[1].x * samples.Weight[0].y;
|
||||
fColor += SampleHistory(FfxFloat32x2(samples.UV[1].x, samples.UV[0].y)) * samples.Weight[2].x * samples.Weight[0].y;
|
||||
|
||||
FFXM_MIN16_F4 fColor02 = SampleHistory(FfxFloat32x2(samples.UV[0].x, samples.UV[2].y));
|
||||
fColor += SampleHistory(FfxFloat32x2(samples.UV[0].x, samples.UV[2].y)) * samples.Weight[0].x * samples.Weight[1].y;
|
||||
FFXM_MIN16_F4 fColor22 = SampleHistory(FfxFloat32x2(samples.UV[2]));
|
||||
fColor += fColor22 * samples.Weight[1].x * samples.Weight[1].y;
|
||||
fColor += SampleHistory(FfxFloat32x2(samples.UV[1].x, samples.UV[2].y)) * samples.Weight[2].x * samples.Weight[1].y;
|
||||
|
||||
fColor += SampleHistory(FfxFloat32x2(samples.UV[0].x, samples.UV[1].y)) * samples.Weight[0].x * samples.Weight[2].y;
|
||||
fColor += SampleHistory(FfxFloat32x2(samples.UV[2].x, samples.UV[1].y)) * samples.Weight[1].x * samples.Weight[2].y;
|
||||
fColor += SampleHistory(FfxFloat32x2(samples.UV[1])) * samples.Weight[2].x * samples.Weight[2].y;
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
const FFXM_MIN16_F4 fDeringingSamples[4] = {fColor00, fColor20, fColor02, fColor22};
|
||||
|
||||
FFXM_MIN16_F4 fDeringingMin = fDeringingSamples[0];
|
||||
FFXM_MIN16_F4 fDeringingMax = fDeringingSamples[0];
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 iSampleIndex = 1; iSampleIndex < 4; ++iSampleIndex)
|
||||
{
|
||||
fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]);
|
||||
fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]);
|
||||
}
|
||||
fColor = clamp(fColor, fDeringingMin, fDeringingMax);
|
||||
#endif
|
||||
return fColor;
|
||||
}
|
||||
#elif (FFXM_FSR2_REPROJECT_MODE == FFXM_FSR2_REPROJECT_CATMULL_5TAP)
|
||||
#define ARM_CATMULL_5TAP_SAMPLE_COUNT 5
|
||||
struct CatmullRomSamples
|
||||
{
|
||||
// bilinear sampling UV coordinates of the samples
|
||||
FfxFloat32x2 UV[ARM_CATMULL_5TAP_SAMPLE_COUNT];
|
||||
// weights of the samples
|
||||
FFXM_MIN16_F Weight[ARM_CATMULL_5TAP_SAMPLE_COUNT];
|
||||
// final multiplier (it is faster to multiply 3 RGB values than reweights the 5 weights)
|
||||
FFXM_MIN16_F FinalMultiplier;
|
||||
};
|
||||
|
||||
void Bicubic2DCatmullRom(in FfxFloat32x2 uv, in FfxFloat32x2 size, in FfxFloat32x2 invSize, FFXM_PARAMETER_OUT FfxFloat32x2 samples[3], FFXM_PARAMETER_OUT FfxFloat32x2 weights[3])
|
||||
{
|
||||
uv *= size;
|
||||
FfxFloat32x2 tc = floor(uv - 0.5) + 0.5;
|
||||
FfxFloat32x2 f = uv - tc;
|
||||
FfxFloat32x2 f2 = f * f;
|
||||
FfxFloat32x2 f3 = f2 * f;
|
||||
FfxFloat32x2 w0 = f2 - 0.5 * (f3 + f);
|
||||
FfxFloat32x2 w1 = 1.5 * f3 - 2.5 * f2 + 1.f;
|
||||
FfxFloat32x2 w3 = 0.5 * (f3 - f2);
|
||||
FfxFloat32x2 w2 = 1.f - w0 - w1 - w3;
|
||||
|
||||
samples[0] = tc - 1.f;
|
||||
samples[1] = tc + w2 / (w1 + w2);
|
||||
samples[2] = tc + 2.f;
|
||||
|
||||
samples[0] *= invSize;
|
||||
samples[1] *= invSize;
|
||||
samples[2] *= invSize;
|
||||
weights[0] = w0;
|
||||
weights[1] = w1 + w2;
|
||||
weights[2] = w3;
|
||||
}
|
||||
|
||||
CatmullRomSamples GetBicubic2DCatmullRomSamples(FfxFloat32x2 uv, FfxFloat32x2 size, in FfxFloat32x2 invSize)
|
||||
{
|
||||
FfxFloat32x2 weights[3];
|
||||
FfxFloat32x2 samples[3];
|
||||
Bicubic2DCatmullRom(uv, size, invSize, samples, weights);
|
||||
|
||||
CatmullRomSamples crSamples;
|
||||
// optimized by removing corner samples
|
||||
crSamples.UV[0] = FfxFloat32x2(samples[1].x, samples[0].y);
|
||||
crSamples.UV[1] = FfxFloat32x2(samples[0].x, samples[1].y);
|
||||
crSamples.UV[2] = FfxFloat32x2(samples[1].x, samples[1].y);
|
||||
crSamples.UV[3] = FfxFloat32x2(samples[2].x, samples[1].y);
|
||||
crSamples.UV[4] = FfxFloat32x2(samples[1].x, samples[2].y);
|
||||
|
||||
crSamples.Weight[0] = FFXM_MIN16_F(weights[1].x * weights[0].y);
|
||||
crSamples.Weight[1] = FFXM_MIN16_F(weights[0].x * weights[1].y);
|
||||
crSamples.Weight[2] = FFXM_MIN16_F(weights[1].x * weights[1].y);
|
||||
crSamples.Weight[3] = FFXM_MIN16_F(weights[2].x * weights[1].y);
|
||||
crSamples.Weight[4] = FFXM_MIN16_F(weights[1].x * weights[2].y);
|
||||
|
||||
// reweight after removing the corners
|
||||
FFXM_MIN16_F cornerWeights;
|
||||
cornerWeights = crSamples.Weight[0];
|
||||
cornerWeights += crSamples.Weight[1];
|
||||
cornerWeights += crSamples.Weight[2];
|
||||
cornerWeights += crSamples.Weight[3];
|
||||
cornerWeights += crSamples.Weight[4];
|
||||
crSamples.FinalMultiplier = FFXM_MIN16_F(1.f / cornerWeights);
|
||||
return crSamples;
|
||||
}
|
||||
|
||||
FFXM_MIN16_F4 HistorySample(FfxFloat32x2 fUvSample, FfxInt32x2 iTextureSize)
|
||||
{
|
||||
FfxFloat32x2 fTextureSize = FfxFloat32x2(iTextureSize);
|
||||
FfxFloat32x2 fInvTextureSize = FfxFloat32x2(1.0f, 1.0f) / fTextureSize;
|
||||
CatmullRomSamples samples = GetBicubic2DCatmullRomSamples(fUvSample, fTextureSize, fInvTextureSize);
|
||||
|
||||
FFXM_MIN16_F4 fColor = FFXM_MIN16_F4(0.0f, 0.0f, 0.0f, 0.0f);
|
||||
fColor = SampleHistory(FfxFloat32x2(samples.UV[0])) * samples.Weight[0];
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
FFXM_MIN16_F4 fDeringingMin = fColor;
|
||||
FFXM_MIN16_F4 fDeringingMax = fColor;
|
||||
#endif
|
||||
for(FfxInt32 iSampleIndex = 1; iSampleIndex < ARM_CATMULL_5TAP_SAMPLE_COUNT; iSampleIndex++)
|
||||
{
|
||||
FFXM_MIN16_F4 fSample = SampleHistory(FfxFloat32x2(samples.UV[iSampleIndex])) * samples.Weight[iSampleIndex];
|
||||
fColor += fSample;
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
fDeringingMin = ffxMin(fDeringingMin, fSample);
|
||||
fDeringingMax = ffxMax(fDeringingMax, fSample);
|
||||
#endif
|
||||
}
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
fColor = clamp(fColor, fDeringingMin, fDeringingMax);
|
||||
#endif
|
||||
return fColor;
|
||||
}
|
||||
#elif (FFXM_FSR2_REPROJECT_MODE == FFXM_FSR2_REPROJECT_LANCZOS_APPROX_9TAP)
|
||||
|
||||
Fetched9TapSamplesMin16 FetchHistorySamples(FfxInt32x2 iPxSample, FfxInt32x2 iTextureSize)
|
||||
{
|
||||
Fetched9TapSamplesMin16 Samples;
|
||||
FfxFloat32x2 iSrcInputUv = FfxFloat32x2(iPxSample) / FfxFloat32x2(iTextureSize);
|
||||
FfxFloat32x2 unitOffsetUv = FfxFloat32x2(1.0f, 1.0f) / FfxFloat32x2(iTextureSize);
|
||||
|
||||
// Collect samples
|
||||
GatherHistoryColorRGBQuad(FfxFloat32x2(-0.5, -0.5) * unitOffsetUv + iSrcInputUv,
|
||||
Samples.fColor00, Samples.fColor10, Samples.fColor01, Samples.fColor11);
|
||||
Samples.fColor20 = WrapHistory(FfxFloat32x2(1, -1) + iPxSample);
|
||||
Samples.fColor21 = WrapHistory(FfxFloat32x2(1, 0) + iPxSample);
|
||||
Samples.fColor02 = WrapHistory(FfxFloat32x2(-1, 1) + iPxSample);
|
||||
Samples.fColor12 = WrapHistory(FfxFloat32x2(0, 1) + iPxSample);
|
||||
Samples.fColor22 = WrapHistory(FfxFloat32x2(1, 1) + iPxSample);
|
||||
|
||||
return Samples;
|
||||
}
|
||||
//DeclareCustomFetch9TapSamplesMin16(FetchHistorySamples, WrapHistory)
|
||||
DeclareCustomTextureSampleMin16(HistorySample, Lanczos2Approx, FetchHistorySamples)
|
||||
#endif // FFXM_FSR2_REPROJECT_MODE
|
||||
|
||||
#else // !FFXM_HALF
|
||||
|
||||
#ifndef FFXM_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE
|
||||
#define FFXM_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE 0 // Reference
|
||||
#endif
|
||||
DeclareCustomFetchBicubicSamples(FetchHistorySamples, WrapHistory)
|
||||
DeclareCustomTextureSample(HistorySample, FFXM_FSR2_GET_LANCZOS_SAMPLER1D(FFXM_FSR2_OPTION_REPROJECT_USE_LANCZOS_TYPE), FetchHistorySamples)
|
||||
#endif
|
||||
|
||||
FfxFloat32x4 WrapLockStatus(FfxInt32x2 iPxSample)
|
||||
{
|
||||
FfxFloat32x4 fSample = FfxFloat32x4(LoadLockStatus(iPxSample), 0.0f, 0.0f);
|
||||
return fSample;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F4 WrapLockStatus(FFXM_MIN16_I2 iPxSample)
|
||||
{
|
||||
FFXM_MIN16_F4 fSample = FFXM_MIN16_F4(LoadLockStatus(iPxSample), 0.0, 0.0);
|
||||
|
||||
return fSample;
|
||||
}
|
||||
#endif
|
||||
|
||||
#if FFXM_HALF
|
||||
DeclareCustomFetchBilinearSamplesMin16(FetchLockStatusSamples, WrapLockStatus)
|
||||
DeclareCustomTextureSampleMin16(LockStatusSample, Bilinear, FetchLockStatusSamples)
|
||||
#else
|
||||
DeclareCustomFetchBilinearSamples(FetchLockStatusSamples, WrapLockStatus)
|
||||
DeclareCustomTextureSample(LockStatusSample, Bilinear, FetchLockStatusSamples)
|
||||
#endif
|
||||
|
||||
FfxFloat32x2 GetMotionVector(FfxInt32x2 iPxHrPos, FfxFloat32x2 fHrUv)
|
||||
{
|
||||
#if FFXM_FSR2_OPTION_LOW_RESOLUTION_MOTION_VECTORS
|
||||
FfxFloat32x2 fDilatedMotionVector = LoadDilatedMotionVector(FfxInt32x2(fHrUv * RenderSize()));
|
||||
#else
|
||||
FfxFloat32x2 fDilatedMotionVector = LoadInputMotionVector(iPxHrPos);
|
||||
#endif
|
||||
|
||||
return fDilatedMotionVector;
|
||||
}
|
||||
|
||||
FfxBoolean IsUvInside(FfxFloat32x2 fUv)
|
||||
{
|
||||
return (fUv.x >= 0.0f && fUv.x <= 1.0f) && (fUv.y >= 0.0f && fUv.y <= 1.0f);
|
||||
}
|
||||
|
||||
void ComputeReprojectedUVs(const AccumulationPassCommonParams params, FFXM_PARAMETER_OUT FfxFloat32x2 fReprojectedHrUv, FFXM_PARAMETER_OUT FfxBoolean bIsExistingSample)
|
||||
{
|
||||
fReprojectedHrUv = params.fHrUv + params.fMotionVector;
|
||||
|
||||
bIsExistingSample = IsUvInside(fReprojectedHrUv);
|
||||
}
|
||||
|
||||
#if !FFXM_HALF
|
||||
void ReprojectHistoryColor(const AccumulationPassCommonParams params, FFXM_PARAMETER_OUT FfxFloat32x3 fHistoryColor, FFXM_PARAMETER_OUT FfxFloat32 fTemporalReactiveFactor, FFXM_PARAMETER_OUT FfxBoolean bInMotionLastFrame)
|
||||
{
|
||||
FfxFloat32x4 fHistory = HistorySample(params.fReprojectedHrUv, DisplaySize());
|
||||
|
||||
fHistoryColor = PrepareRgb(fHistory.rgb, Exposure(), PreviousFramePreExposure());
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_TONEMAPPED_RGB_PREPARED_INPUT_COLOR
|
||||
fHistoryColor = RGBToYCoCg(fHistoryColor);
|
||||
#endif
|
||||
|
||||
//Compute temporal reactivity info
|
||||
fTemporalReactiveFactor = ffxSaturate(abs(fHistory.w));
|
||||
bInMotionLastFrame = (fHistory.w < 0.0f);
|
||||
}
|
||||
|
||||
LockState ReprojectHistoryLockStatus(const AccumulationPassCommonParams params, FFXM_PARAMETER_OUT FfxFloat32x2 fReprojectedLockStatus)
|
||||
{
|
||||
LockState state = { FFXM_FALSE, FFXM_FALSE };
|
||||
const FfxFloat32 fNewLockIntensity = LoadNewLocks(params.iPxHrPos);
|
||||
state.NewLock = fNewLockIntensity > (127.0f / 255.0f);
|
||||
|
||||
FfxFloat32 fInPlaceLockLifetime = state.NewLock ? fNewLockIntensity : 0;
|
||||
|
||||
fReprojectedLockStatus = SampleLockStatus(params.fReprojectedHrUv);
|
||||
|
||||
if (fReprojectedLockStatus[LOCK_LIFETIME_REMAINING] != FfxFloat32(0.0f)) {
|
||||
state.WasLockedPrevFrame = true;
|
||||
}
|
||||
|
||||
return state;
|
||||
}
|
||||
#else //FFXM_HALF
|
||||
|
||||
void ReprojectHistoryColor(const AccumulationPassCommonParams params, FFXM_PARAMETER_OUT FfxFloat16x3 fHistoryColor, FFXM_PARAMETER_OUT FfxFloat16 fTemporalReactiveFactor, FFXM_PARAMETER_OUT FfxBoolean bInMotionLastFrame)
|
||||
{
|
||||
FfxFloat16x4 fHistory = HistorySample(params.fReprojectedHrUv, DisplaySize());
|
||||
|
||||
fHistoryColor = FfxFloat16x3(PrepareRgb(fHistory.rgb, Exposure(), PreviousFramePreExposure()));
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_TONEMAPPED_RGB_PREPARED_INPUT_COLOR
|
||||
fHistoryColor = RGBToYCoCg(fHistoryColor);
|
||||
#endif
|
||||
|
||||
//Compute temporal reactivity info
|
||||
#if FFXM_FSR2_OPTION_SHADER_OPT_BALANCED
|
||||
fTemporalReactiveFactor = FfxFloat16(ffxSaturate(abs(SampleTemporalReactive(params.fReprojectedHrUv))));
|
||||
#else
|
||||
fTemporalReactiveFactor = FfxFloat16(ffxSaturate(abs(fHistory.w)));
|
||||
#endif
|
||||
bInMotionLastFrame = (fHistory.w < 0.0f);
|
||||
}
|
||||
|
||||
LockState ReprojectHistoryLockStatus(const AccumulationPassCommonParams params, FFXM_PARAMETER_OUT FfxFloat16x2 fReprojectedLockStatus)
|
||||
{
|
||||
LockState state = { FFXM_FALSE, FFXM_FALSE };
|
||||
const FfxFloat16 fNewLockIntensity = FfxFloat16(LoadNewLocks(params.iPxHrPos));
|
||||
state.NewLock = fNewLockIntensity > (127.0f / 255.0f);
|
||||
|
||||
FfxFloat16 fInPlaceLockLifetime = state.NewLock ? fNewLockIntensity : FfxFloat16(0);
|
||||
|
||||
fReprojectedLockStatus = FfxFloat16x2(SampleLockStatus(params.fReprojectedHrUv));
|
||||
|
||||
if (fReprojectedLockStatus[LOCK_LIFETIME_REMAINING] != FfxFloat16(0.0f)) {
|
||||
state.WasLockedPrevFrame = true;
|
||||
}
|
||||
return state;
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
#endif //!defined( FFXM_FSR2_REPROJECT_H )
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 37bdf71c574e81547b136545f7807249
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,101 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#ifndef FFXM_FSR2_RESOURCES_H
|
||||
#define FFXM_FSR2_RESOURCES_H
|
||||
|
||||
#if defined(FFXM_CPU) || defined(FFXM_GPU)
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_NULL 0
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INPUT_OPAQUE_ONLY 1
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INPUT_COLOR 2
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INPUT_MOTION_VECTORS 3
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INPUT_DEPTH 4
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INPUT_EXPOSURE 5
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INPUT_REACTIVE_MASK 6
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INPUT_TRANSPARENCY_AND_COMPOSITION_MASK 7
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_RECONSTRUCTED_PREVIOUS_NEAREST_DEPTH 8
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_DILATED_MOTION_VECTORS 9
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_DILATED_DEPTH 10
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_UPSCALED_COLOR 11
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_LOCK_STATUS 12
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_NEW_LOCKS 13
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_PREPARED_INPUT_COLOR 14
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_LUMA_HISTORY 15
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_DEBUG_OUTPUT 16
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_LANCZOS_LUT 17
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SPD_ATOMIC_COUNT 18
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_UPSCALED_OUTPUT 19
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_RCAS_INPUT 20
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_LOCK_STATUS_1 21
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_LOCK_STATUS_2 22
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_UPSCALED_COLOR_1 23
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_UPSCALED_COLOR_2 24
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_DEFAULT_REACTIVITY 25
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_DEFAULT_TRANSPARENCY_AND_COMPOSITION 26
|
||||
#define FFXM_FSR2_RESOURCE_IDENTITIER_UPSAMPLE_MAXIMUM_BIAS_LUT 27
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_DILATED_REACTIVE_MASKS 28
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE 29 // same as FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_0
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_0 29
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_1 30
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_2 31
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_3 32
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_4 33
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_5 34
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_6 35
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_7 36
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_8 37
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_9 38
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_10 39
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_11 40
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_12 41
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_DEFAULT_EXPOSURE 42
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_AUTO_EXPOSURE 43
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_AUTOREACTIVE 44
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_PREVIOUS_DILATED_MOTION_VECTORS 45
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_DILATED_MOTION_VECTORS_1 46
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_DILATED_MOTION_VECTORS_2 47
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_LUMA_HISTORY_1 48
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_LUMA_HISTORY_2 49
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_LOCK_INPUT_LUMA 50
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_TEMPORAL_REACTIVE 51
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_TEMPORAL_REACTIVE_1 52
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_INTERNAL_TEMPORAL_REACTIVE_2 53
|
||||
|
||||
// Shading change detection mip level setting, value must be in the range [FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_0, FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_12]
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_SHADING_CHANGE FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_4
|
||||
#define FFXM_FSR2_SHADING_CHANGE_MIP_LEVEL (FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE_MIPMAP_SHADING_CHANGE - FFXM_FSR2_RESOURCE_IDENTIFIER_SCENE_LUMINANCE)
|
||||
|
||||
#define FFXM_FSR2_RESOURCE_IDENTIFIER_COUNT 54
|
||||
|
||||
#define FFXM_FSR2_CONSTANTBUFFER_IDENTIFIER_FSR2 0
|
||||
#define FFXM_FSR2_CONSTANTBUFFER_IDENTIFIER_SPD 1
|
||||
#define FFXM_FSR2_CONSTANTBUFFER_IDENTIFIER_RCAS 2
|
||||
#define FFXM_FSR2_CONSTANTBUFFER_IDENTIFIER_GENREACTIVE 3
|
||||
|
||||
#define FFXM_FSR2_AUTOREACTIVEFLAGS_APPLY_TONEMAP 1
|
||||
#define FFXM_FSR2_AUTOREACTIVEFLAGS_APPLY_INVERSETONEMAP 2
|
||||
#define FFXM_FSR2_AUTOREACTIVEFLAGS_APPLY_THRESHOLD 4
|
||||
#define FFXM_FSR2_AUTOREACTIVEFLAGS_USE_COMPONENTS_MAX 8
|
||||
|
||||
#endif // #if defined(FFXM_CPU) || defined(FFXM_GPU)
|
||||
|
||||
#endif //!defined( FFXM_FSR2_RESOURCES_H )
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 403379c4f891efb4697f34a3ca44c6ac
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,700 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#ifndef FFXM_FSR2_SAMPLE_H
|
||||
#define FFXM_FSR2_SAMPLE_H
|
||||
|
||||
// suppress warnings
|
||||
#ifdef FFXM_HLSL
|
||||
#pragma warning(disable: 4008) // potentially divide by zero
|
||||
#endif //FFXM_HLSL
|
||||
|
||||
struct FetchedBilinearSamples {
|
||||
|
||||
FfxFloat32x4 fColor00;
|
||||
FfxFloat32x4 fColor10;
|
||||
|
||||
FfxFloat32x4 fColor01;
|
||||
FfxFloat32x4 fColor11;
|
||||
};
|
||||
|
||||
struct FetchedBicubicSamples {
|
||||
|
||||
FfxFloat32x4 fColor00;
|
||||
FfxFloat32x4 fColor10;
|
||||
FfxFloat32x4 fColor20;
|
||||
FfxFloat32x4 fColor30;
|
||||
|
||||
FfxFloat32x4 fColor01;
|
||||
FfxFloat32x4 fColor11;
|
||||
FfxFloat32x4 fColor21;
|
||||
FfxFloat32x4 fColor31;
|
||||
|
||||
FfxFloat32x4 fColor02;
|
||||
FfxFloat32x4 fColor12;
|
||||
FfxFloat32x4 fColor22;
|
||||
FfxFloat32x4 fColor32;
|
||||
|
||||
FfxFloat32x4 fColor03;
|
||||
FfxFloat32x4 fColor13;
|
||||
FfxFloat32x4 fColor23;
|
||||
FfxFloat32x4 fColor33;
|
||||
};
|
||||
|
||||
#if FFXM_HALF
|
||||
struct FetchedBilinearSamplesMin16 {
|
||||
|
||||
FFXM_MIN16_F4 fColor00;
|
||||
FFXM_MIN16_F4 fColor10;
|
||||
|
||||
FFXM_MIN16_F4 fColor01;
|
||||
FFXM_MIN16_F4 fColor11;
|
||||
};
|
||||
|
||||
struct FetchedBicubicSamplesMin16 {
|
||||
|
||||
FFXM_MIN16_F4 fColor00;
|
||||
FFXM_MIN16_F4 fColor10;
|
||||
FFXM_MIN16_F4 fColor20;
|
||||
FFXM_MIN16_F4 fColor30;
|
||||
|
||||
FFXM_MIN16_F4 fColor01;
|
||||
FFXM_MIN16_F4 fColor11;
|
||||
FFXM_MIN16_F4 fColor21;
|
||||
FFXM_MIN16_F4 fColor31;
|
||||
|
||||
FFXM_MIN16_F4 fColor02;
|
||||
FFXM_MIN16_F4 fColor12;
|
||||
FFXM_MIN16_F4 fColor22;
|
||||
FFXM_MIN16_F4 fColor32;
|
||||
|
||||
FFXM_MIN16_F4 fColor03;
|
||||
FFXM_MIN16_F4 fColor13;
|
||||
FFXM_MIN16_F4 fColor23;
|
||||
FFXM_MIN16_F4 fColor33;
|
||||
};
|
||||
|
||||
struct Fetched9TapSamplesMin16 {
|
||||
|
||||
FFXM_MIN16_F4 fColor00;
|
||||
FFXM_MIN16_F4 fColor10;
|
||||
FFXM_MIN16_F4 fColor20;
|
||||
|
||||
FFXM_MIN16_F4 fColor01;
|
||||
FFXM_MIN16_F4 fColor11;
|
||||
FFXM_MIN16_F4 fColor21;
|
||||
|
||||
FFXM_MIN16_F4 fColor02;
|
||||
FFXM_MIN16_F4 fColor12;
|
||||
FFXM_MIN16_F4 fColor22;
|
||||
};
|
||||
|
||||
#else //FFXM_HALF
|
||||
#define FetchedBicubicSamplesMin16 FetchedBicubicSamples
|
||||
#define FetchedBilinearSamplesMin16 FetchedBilinearSamples
|
||||
#endif //FFXM_HALF
|
||||
|
||||
FfxFloat32x4 Linear(FfxFloat32x4 A, FfxFloat32x4 B, FfxFloat32 t)
|
||||
{
|
||||
return A + (B - A) * t;
|
||||
}
|
||||
|
||||
FfxFloat32x4 Bilinear(FetchedBilinearSamples BilinearSamples, FfxFloat32x2 fPxFrac)
|
||||
{
|
||||
FfxFloat32x4 fColorX0 = Linear(BilinearSamples.fColor00, BilinearSamples.fColor10, fPxFrac.x);
|
||||
FfxFloat32x4 fColorX1 = Linear(BilinearSamples.fColor01, BilinearSamples.fColor11, fPxFrac.x);
|
||||
FfxFloat32x4 fColorXY = Linear(fColorX0, fColorX1, fPxFrac.y);
|
||||
return fColorXY;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F4 Linear(FFXM_MIN16_F4 A, FFXM_MIN16_F4 B, FFXM_MIN16_F t)
|
||||
{
|
||||
return A + (B - A) * t;
|
||||
}
|
||||
|
||||
FFXM_MIN16_F4 Bilinear(FetchedBilinearSamplesMin16 BilinearSamples, FFXM_MIN16_F2 fPxFrac)
|
||||
{
|
||||
FFXM_MIN16_F4 fColorX0 = Linear(BilinearSamples.fColor00, BilinearSamples.fColor10, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX1 = Linear(BilinearSamples.fColor01, BilinearSamples.fColor11, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorXY = Linear(fColorX0, fColorX1, fPxFrac.y);
|
||||
return fColorXY;
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32 Lanczos2NoClamp(FfxFloat32 x)
|
||||
{
|
||||
const FfxFloat32 PI = 3.141592653589793f; // TODO: share SDK constants
|
||||
return abs(x) < FSR2_EPSILON ? 1.f : (sin(PI * x) / (PI * x)) * (sin(0.5f * PI * x) / (0.5f * PI * x));
|
||||
}
|
||||
|
||||
FfxFloat32 Lanczos2(FfxFloat32 x)
|
||||
{
|
||||
x = ffxMin(abs(x), 2.0f);
|
||||
return Lanczos2NoClamp(x);
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
|
||||
#if 1
|
||||
FFXM_MIN16_F Lanczos2NoClamp(FFXM_MIN16_F x)
|
||||
{
|
||||
const FFXM_MIN16_F PI = FFXM_MIN16_F(3.141592653589793f); // TODO: share SDK constants
|
||||
return abs(x) < FFXM_MIN16_F(FSR2_EPSILON) ? FFXM_MIN16_F(1.f) : (sin(PI * x) / (PI * x)) * (sin(FFXM_MIN16_F(0.5f) * PI * x) / (FFXM_MIN16_F(0.5f) * PI * x));
|
||||
}
|
||||
#endif
|
||||
|
||||
FFXM_MIN16_F Lanczos2(FFXM_MIN16_F x)
|
||||
{
|
||||
x = ffxMin(abs(x), FFXM_MIN16_F(2.0f));
|
||||
return FFXM_MIN16_F(Lanczos2NoClamp(x));
|
||||
}
|
||||
#endif //FFXM_HALF
|
||||
|
||||
// FSR1 lanczos approximation. Input is x*x and must be <= 4.
|
||||
FfxFloat32 Lanczos2ApproxSqNoClamp(FfxFloat32 x2)
|
||||
{
|
||||
FfxFloat32 a = (2.0f / 5.0f) * x2 - 1;
|
||||
FfxFloat32 b = (1.0f / 4.0f) * x2 - 1;
|
||||
return ((25.0f / 16.0f) * a * a - (25.0f / 16.0f - 1)) * (b * b);
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F Lanczos2ApproxSqNoClamp(FFXM_MIN16_F x2)
|
||||
{
|
||||
FFXM_MIN16_F a = FFXM_MIN16_F(2.0f / 5.0f) * x2 - FFXM_MIN16_F(1);
|
||||
FFXM_MIN16_F b = FFXM_MIN16_F(1.0f / 4.0f) * x2 - FFXM_MIN16_F(1);
|
||||
return (FFXM_MIN16_F(25.0f / 16.0f) * a * a - FFXM_MIN16_F(25.0f / 16.0f - 1)) * (b * b);
|
||||
}
|
||||
#endif //FFXM_HALF
|
||||
|
||||
FfxFloat32 Lanczos2ApproxSq(FfxFloat32 x2)
|
||||
{
|
||||
x2 = ffxMin(x2, 4.0f);
|
||||
return Lanczos2ApproxSqNoClamp(x2);
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F Lanczos2ApproxSq(FFXM_MIN16_F x2)
|
||||
{
|
||||
x2 = ffxMin(x2, FFXM_MIN16_F(4.0f));
|
||||
return Lanczos2ApproxSqNoClamp(x2);
|
||||
}
|
||||
#endif //FFXM_HALF
|
||||
|
||||
FfxFloat32 Lanczos2ApproxNoClamp(FfxFloat32 x)
|
||||
{
|
||||
return Lanczos2ApproxSqNoClamp(x * x);
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F Lanczos2ApproxNoClamp(FFXM_MIN16_F x)
|
||||
{
|
||||
return Lanczos2ApproxSqNoClamp(x * x);
|
||||
}
|
||||
#endif //FFXM_HALF
|
||||
|
||||
FfxFloat32 Lanczos2Approx(FfxFloat32 x)
|
||||
{
|
||||
return Lanczos2ApproxSq(x * x);
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F Lanczos2Approx(FFXM_MIN16_F x)
|
||||
{
|
||||
return Lanczos2ApproxSq(x * x);
|
||||
}
|
||||
#endif //FFXM_HALF
|
||||
|
||||
FfxFloat32 Lanczos2_UseLUT(FfxFloat32 x)
|
||||
{
|
||||
return SampleLanczos2Weight(abs(x));
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F Lanczos2_UseLUT(FFXM_MIN16_F x)
|
||||
{
|
||||
return FFXM_MIN16_F(SampleLanczos2Weight(abs(x)));
|
||||
}
|
||||
#endif //FFXM_HALF
|
||||
|
||||
FfxFloat32x4 Lanczos2_UseLUT(FfxFloat32x4 fColor0, FfxFloat32x4 fColor1, FfxFloat32x4 fColor2, FfxFloat32x4 fColor3, FfxFloat32 t)
|
||||
{
|
||||
FfxFloat32 fWeight0 = Lanczos2_UseLUT(-1.f - t);
|
||||
FfxFloat32 fWeight1 = Lanczos2_UseLUT(-0.f - t);
|
||||
FfxFloat32 fWeight2 = Lanczos2_UseLUT(+1.f - t);
|
||||
FfxFloat32 fWeight3 = Lanczos2_UseLUT(+2.f - t);
|
||||
return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3);
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F4 Lanczos2_UseLUT(FFXM_MIN16_F4 fColor0, FFXM_MIN16_F4 fColor1, FFXM_MIN16_F4 fColor2, FFXM_MIN16_F4 fColor3, FFXM_MIN16_F t)
|
||||
{
|
||||
FFXM_MIN16_F fWeight0 = Lanczos2_UseLUT(FFXM_MIN16_F(-1.f) - t);
|
||||
FFXM_MIN16_F fWeight1 = Lanczos2_UseLUT(FFXM_MIN16_F(-0.f) - t);
|
||||
FFXM_MIN16_F fWeight2 = Lanczos2_UseLUT(FFXM_MIN16_F(+1.f) - t);
|
||||
FFXM_MIN16_F fWeight3 = Lanczos2_UseLUT(FFXM_MIN16_F(+2.f) - t);
|
||||
return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3);
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32x4 Lanczos2(FfxFloat32x4 fColor0, FfxFloat32x4 fColor1, FfxFloat32x4 fColor2, FfxFloat32x4 fColor3, FfxFloat32 t)
|
||||
{
|
||||
FfxFloat32 fWeight0 = Lanczos2(-1.f - t);
|
||||
FfxFloat32 fWeight1 = Lanczos2(-0.f - t);
|
||||
FfxFloat32 fWeight2 = Lanczos2(+1.f - t);
|
||||
FfxFloat32 fWeight3 = Lanczos2(+2.f - t);
|
||||
return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3);
|
||||
}
|
||||
|
||||
FfxFloat32x4 Lanczos2(FetchedBicubicSamples Samples, FfxFloat32x2 fPxFrac)
|
||||
{
|
||||
FfxFloat32x4 fColorX0 = Lanczos2(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x);
|
||||
FfxFloat32x4 fColorX1 = Lanczos2(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x);
|
||||
FfxFloat32x4 fColorX2 = Lanczos2(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x);
|
||||
FfxFloat32x4 fColorX3 = Lanczos2(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x);
|
||||
FfxFloat32x4 fColorXY = Lanczos2(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y);
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
// Deringing
|
||||
|
||||
// TODO: only use 4 by checking jitter
|
||||
const FfxInt32 iDeringingSampleCount = 4;
|
||||
const FfxFloat32x4 fDeringingSamples[4] = {
|
||||
Samples.fColor11,
|
||||
Samples.fColor21,
|
||||
Samples.fColor12,
|
||||
Samples.fColor22,
|
||||
};
|
||||
|
||||
FfxFloat32x4 fDeringingMin = fDeringingSamples[0];
|
||||
FfxFloat32x4 fDeringingMax = fDeringingSamples[0];
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex) {
|
||||
|
||||
fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]);
|
||||
fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]);
|
||||
}
|
||||
|
||||
fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax);
|
||||
#endif
|
||||
return fColorXY;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F4 Lanczos2(FFXM_MIN16_F4 fColor0, FFXM_MIN16_F4 fColor1, FFXM_MIN16_F4 fColor2, FFXM_MIN16_F4 fColor3, FFXM_MIN16_F t)
|
||||
{
|
||||
FFXM_MIN16_F fWeight0 = Lanczos2(FFXM_MIN16_F(-1.f) - t);
|
||||
FFXM_MIN16_F fWeight1 = Lanczos2(FFXM_MIN16_F(-0.f) - t);
|
||||
FFXM_MIN16_F fWeight2 = Lanczos2(FFXM_MIN16_F(+1.f) - t);
|
||||
FFXM_MIN16_F fWeight3 = Lanczos2(FFXM_MIN16_F(+2.f) - t);
|
||||
return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3);
|
||||
}
|
||||
|
||||
FFXM_MIN16_F4 Lanczos2(FetchedBicubicSamplesMin16 Samples, FFXM_MIN16_F2 fPxFrac)
|
||||
{
|
||||
FFXM_MIN16_F4 fColorX0 = Lanczos2(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX1 = Lanczos2(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX2 = Lanczos2(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX3 = Lanczos2(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorXY = Lanczos2(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y);
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
// Deringing
|
||||
|
||||
// TODO: only use 4 by checking jitter
|
||||
const FfxInt32 iDeringingSampleCount = 4;
|
||||
const FFXM_MIN16_F4 fDeringingSamples[4] = {
|
||||
Samples.fColor11,
|
||||
Samples.fColor21,
|
||||
Samples.fColor12,
|
||||
Samples.fColor22,
|
||||
};
|
||||
|
||||
FFXM_MIN16_F4 fDeringingMin = fDeringingSamples[0];
|
||||
FFXM_MIN16_F4 fDeringingMax = fDeringingSamples[0];
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex)
|
||||
{
|
||||
fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]);
|
||||
fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]);
|
||||
}
|
||||
|
||||
fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax);
|
||||
#endif
|
||||
return fColorXY;
|
||||
}
|
||||
|
||||
FFXM_MIN16_F4 Lanczos2(FFXM_MIN16_F4 fColor0, FFXM_MIN16_F4 fColor1, FFXM_MIN16_F4 fColor2, FFXM_MIN16_F t)
|
||||
{
|
||||
FFXM_MIN16_F fWeight0 = Lanczos2(FFXM_MIN16_F(-1.f) - t);
|
||||
FFXM_MIN16_F fWeight1 = Lanczos2(FFXM_MIN16_F(-0.f) - t);
|
||||
FFXM_MIN16_F fWeight2 = Lanczos2(FFXM_MIN16_F(+1.f) - t);
|
||||
return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2) / (fWeight0 + fWeight1 + fWeight2);
|
||||
}
|
||||
|
||||
FFXM_MIN16_F4 Lanczos2Approx(FFXM_MIN16_F4 fColor0, FFXM_MIN16_F4 fColor1, FFXM_MIN16_F4 fColor2, FFXM_MIN16_F t)
|
||||
{
|
||||
FFXM_MIN16_F fWeight0 = Lanczos2ApproxNoClamp(FFXM_MIN16_F(-1.f) - t);
|
||||
FFXM_MIN16_F fWeight1 = Lanczos2ApproxNoClamp(FFXM_MIN16_F(-0.f) - t);
|
||||
FFXM_MIN16_F fWeight2 = Lanczos2ApproxNoClamp(FFXM_MIN16_F(+1.f) - t);
|
||||
return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2) / (fWeight0 + fWeight1 + fWeight2);
|
||||
}
|
||||
|
||||
FFXM_MIN16_F4 Lanczos2Approx(Fetched9TapSamplesMin16 Samples, FFXM_MIN16_F2 fPxFrac)
|
||||
{
|
||||
FFXM_MIN16_F4 fColorX0 = Lanczos2Approx(Samples.fColor00, Samples.fColor10, Samples.fColor20, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX1 = Lanczos2Approx(Samples.fColor01, Samples.fColor11, Samples.fColor21, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX2 = Lanczos2Approx(Samples.fColor02, Samples.fColor12, Samples.fColor22, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorXY = Lanczos2Approx(fColorX0, fColorX1, fColorX2, fPxFrac.y);
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
// Deringing
|
||||
const FfxInt32 iDeringingSampleCount = 4;
|
||||
const FFXM_MIN16_F4 fDeringingSamples[4] = {
|
||||
Samples.fColor11,
|
||||
Samples.fColor21,
|
||||
Samples.fColor12,
|
||||
Samples.fColor22,
|
||||
};
|
||||
|
||||
FFXM_MIN16_F4 fDeringingMin = fDeringingSamples[0];
|
||||
FFXM_MIN16_F4 fDeringingMax = fDeringingSamples[0];
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex)
|
||||
{
|
||||
fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]);
|
||||
fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]);
|
||||
}
|
||||
|
||||
fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax);
|
||||
#endif
|
||||
return fColorXY;
|
||||
}
|
||||
|
||||
#endif //FFXM_HALF
|
||||
|
||||
|
||||
FfxFloat32x4 Lanczos2LUT(FetchedBicubicSamples Samples, FfxFloat32x2 fPxFrac)
|
||||
{
|
||||
FfxFloat32x4 fColorX0 = Lanczos2_UseLUT(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x);
|
||||
FfxFloat32x4 fColorX1 = Lanczos2_UseLUT(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x);
|
||||
FfxFloat32x4 fColorX2 = Lanczos2_UseLUT(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x);
|
||||
FfxFloat32x4 fColorX3 = Lanczos2_UseLUT(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x);
|
||||
FfxFloat32x4 fColorXY = Lanczos2_UseLUT(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y);
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
// Deringing
|
||||
|
||||
// TODO: only use 4 by checking jitter
|
||||
const FfxInt32 iDeringingSampleCount = 4;
|
||||
const FfxFloat32x4 fDeringingSamples[4] = {
|
||||
Samples.fColor11,
|
||||
Samples.fColor21,
|
||||
Samples.fColor12,
|
||||
Samples.fColor22,
|
||||
};
|
||||
|
||||
FfxFloat32x4 fDeringingMin = fDeringingSamples[0];
|
||||
FfxFloat32x4 fDeringingMax = fDeringingSamples[0];
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex) {
|
||||
|
||||
fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]);
|
||||
fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]);
|
||||
}
|
||||
|
||||
fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax);
|
||||
#endif
|
||||
return fColorXY;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F4 Lanczos2LUT(FetchedBicubicSamplesMin16 Samples, FFXM_MIN16_F2 fPxFrac)
|
||||
{
|
||||
FFXM_MIN16_F4 fColorX0 = Lanczos2_UseLUT(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX1 = Lanczos2_UseLUT(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX2 = Lanczos2_UseLUT(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX3 = Lanczos2_UseLUT(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorXY = Lanczos2_UseLUT(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y);
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
// Deringing
|
||||
|
||||
// TODO: only use 4 by checking jitter
|
||||
const FfxInt32 iDeringingSampleCount = 4;
|
||||
const FFXM_MIN16_F4 fDeringingSamples[4] = {
|
||||
Samples.fColor11,
|
||||
Samples.fColor21,
|
||||
Samples.fColor12,
|
||||
Samples.fColor22,
|
||||
};
|
||||
|
||||
FFXM_MIN16_F4 fDeringingMin = fDeringingSamples[0];
|
||||
FFXM_MIN16_F4 fDeringingMax = fDeringingSamples[0];
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex)
|
||||
{
|
||||
fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]);
|
||||
fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]);
|
||||
}
|
||||
|
||||
fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax);
|
||||
#endif
|
||||
return fColorXY;
|
||||
}
|
||||
#endif //FFXM_HALF
|
||||
|
||||
|
||||
|
||||
FfxFloat32x4 Lanczos2Approx(FfxFloat32x4 fColor0, FfxFloat32x4 fColor1, FfxFloat32x4 fColor2, FfxFloat32x4 fColor3, FfxFloat32 t)
|
||||
{
|
||||
FfxFloat32 fWeight0 = Lanczos2ApproxNoClamp(-1.f - t);
|
||||
FfxFloat32 fWeight1 = Lanczos2ApproxNoClamp(-0.f - t);
|
||||
FfxFloat32 fWeight2 = Lanczos2ApproxNoClamp(+1.f - t);
|
||||
FfxFloat32 fWeight3 = Lanczos2ApproxNoClamp(+2.f - t);
|
||||
return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3);
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F4 Lanczos2Approx(FFXM_MIN16_F4 fColor0, FFXM_MIN16_F4 fColor1, FFXM_MIN16_F4 fColor2, FFXM_MIN16_F4 fColor3, FFXM_MIN16_F t)
|
||||
{
|
||||
FFXM_MIN16_F fWeight0 = Lanczos2ApproxNoClamp(FFXM_MIN16_F(-1.f) - t);
|
||||
FFXM_MIN16_F fWeight1 = Lanczos2ApproxNoClamp(FFXM_MIN16_F(-0.f) - t);
|
||||
FFXM_MIN16_F fWeight2 = Lanczos2ApproxNoClamp(FFXM_MIN16_F(+1.f) - t);
|
||||
FFXM_MIN16_F fWeight3 = Lanczos2ApproxNoClamp(FFXM_MIN16_F(+2.f) - t);
|
||||
return (fWeight0 * fColor0 + fWeight1 * fColor1 + fWeight2 * fColor2 + fWeight3 * fColor3) / (fWeight0 + fWeight1 + fWeight2 + fWeight3);
|
||||
}
|
||||
#endif //FFXM_HALF
|
||||
|
||||
FfxFloat32x4 Lanczos2Approx(FetchedBicubicSamples Samples, FfxFloat32x2 fPxFrac)
|
||||
{
|
||||
FfxFloat32x4 fColorX0 = Lanczos2Approx(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x);
|
||||
FfxFloat32x4 fColorX1 = Lanczos2Approx(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x);
|
||||
FfxFloat32x4 fColorX2 = Lanczos2Approx(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x);
|
||||
FfxFloat32x4 fColorX3 = Lanczos2Approx(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x);
|
||||
FfxFloat32x4 fColorXY = Lanczos2Approx(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y);
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
// Deringing
|
||||
|
||||
// TODO: only use 4 by checking jitter
|
||||
const FfxInt32 iDeringingSampleCount = 4;
|
||||
const FfxFloat32x4 fDeringingSamples[4] = {
|
||||
Samples.fColor11,
|
||||
Samples.fColor21,
|
||||
Samples.fColor12,
|
||||
Samples.fColor22,
|
||||
};
|
||||
|
||||
FfxFloat32x4 fDeringingMin = fDeringingSamples[0];
|
||||
FfxFloat32x4 fDeringingMax = fDeringingSamples[0];
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex)
|
||||
{
|
||||
fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]);
|
||||
fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]);
|
||||
}
|
||||
|
||||
fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax);
|
||||
#endif
|
||||
return fColorXY;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F4 Lanczos2Approx(FetchedBicubicSamplesMin16 Samples, FFXM_MIN16_F2 fPxFrac)
|
||||
{
|
||||
FFXM_MIN16_F4 fColorX0 = Lanczos2Approx(Samples.fColor00, Samples.fColor10, Samples.fColor20, Samples.fColor30, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX1 = Lanczos2Approx(Samples.fColor01, Samples.fColor11, Samples.fColor21, Samples.fColor31, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX2 = Lanczos2Approx(Samples.fColor02, Samples.fColor12, Samples.fColor22, Samples.fColor32, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorX3 = Lanczos2Approx(Samples.fColor03, Samples.fColor13, Samples.fColor23, Samples.fColor33, fPxFrac.x);
|
||||
FFXM_MIN16_F4 fColorXY = Lanczos2Approx(fColorX0, fColorX1, fColorX2, fColorX3, fPxFrac.y);
|
||||
|
||||
#if !FFXM_SHADER_QUALITY_OPT_DISABLE_DERINGING
|
||||
// Deringing
|
||||
|
||||
// TODO: only use 4 by checking jitter
|
||||
const FfxInt32 iDeringingSampleCount = 4;
|
||||
const FFXM_MIN16_F4 fDeringingSamples[4] = {
|
||||
Samples.fColor11,
|
||||
Samples.fColor21,
|
||||
Samples.fColor12,
|
||||
Samples.fColor22,
|
||||
};
|
||||
|
||||
FFXM_MIN16_F4 fDeringingMin = fDeringingSamples[0];
|
||||
FFXM_MIN16_F4 fDeringingMax = fDeringingSamples[0];
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 iSampleIndex = 1; iSampleIndex < iDeringingSampleCount; ++iSampleIndex)
|
||||
{
|
||||
fDeringingMin = ffxMin(fDeringingMin, fDeringingSamples[iSampleIndex]);
|
||||
fDeringingMax = ffxMax(fDeringingMax, fDeringingSamples[iSampleIndex]);
|
||||
}
|
||||
|
||||
fColorXY = clamp(fColorXY, fDeringingMin, fDeringingMax);
|
||||
#endif
|
||||
return fColorXY;
|
||||
}
|
||||
#endif
|
||||
|
||||
// Clamp by offset direction. Assuming iPxSample is already in range and iPxOffset is compile time constant.
|
||||
FfxInt32x2 ClampCoord(FfxInt32x2 iPxSample, FfxInt32x2 iPxOffset, FfxInt32x2 iTextureSize)
|
||||
{
|
||||
FfxInt32x2 result = iPxSample + iPxOffset;
|
||||
result.x = (iPxOffset.x < 0) ? ffxMax(result.x, 0) : result.x;
|
||||
result.x = (iPxOffset.x > 0) ? ffxMin(result.x, iTextureSize.x - 1) : result.x;
|
||||
result.y = (iPxOffset.y < 0) ? ffxMax(result.y, 0) : result.y;
|
||||
result.y = (iPxOffset.y > 0) ? ffxMin(result.y, iTextureSize.y - 1) : result.y;
|
||||
return result;
|
||||
}
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_I2 ClampCoord(FFXM_MIN16_I2 iPxSample, FFXM_MIN16_I2 iPxOffset, FFXM_MIN16_I2 iTextureSize)
|
||||
{
|
||||
FFXM_MIN16_I2 result = iPxSample + iPxOffset;
|
||||
result.x = (iPxOffset.x < FFXM_MIN16_I(0)) ? ffxMax(result.x, FFXM_MIN16_I(0)) : result.x;
|
||||
result.x = (iPxOffset.x > FFXM_MIN16_I(0)) ? ffxMin(result.x, iTextureSize.x - FFXM_MIN16_I(1)) : result.x;
|
||||
result.y = (iPxOffset.y < FFXM_MIN16_I(0)) ? ffxMax(result.y, FFXM_MIN16_I(0)) : result.y;
|
||||
result.y = (iPxOffset.y > FFXM_MIN16_I(0)) ? ffxMin(result.y, iTextureSize.y - FFXM_MIN16_I(1)) : result.y;
|
||||
return result;
|
||||
}
|
||||
#endif //FFXM_HALF
|
||||
|
||||
|
||||
#define DeclareCustomFetchBicubicSamplesWithType(SampleType, TextureType, AddrType, Name, LoadTexture) \
|
||||
SampleType Name(AddrType iPxSample, AddrType iTextureSize) \
|
||||
{ \
|
||||
SampleType Samples; \
|
||||
\
|
||||
Samples.fColor00 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, -1), iTextureSize))); \
|
||||
Samples.fColor10 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, -1), iTextureSize))); \
|
||||
Samples.fColor20 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, -1), iTextureSize))); \
|
||||
Samples.fColor30 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+2, -1), iTextureSize))); \
|
||||
\
|
||||
Samples.fColor01 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, +0), iTextureSize))); \
|
||||
Samples.fColor11 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +0), iTextureSize))); \
|
||||
Samples.fColor21 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +0), iTextureSize))); \
|
||||
Samples.fColor31 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+2, +0), iTextureSize))); \
|
||||
\
|
||||
Samples.fColor02 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, +1), iTextureSize))); \
|
||||
Samples.fColor12 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +1), iTextureSize))); \
|
||||
Samples.fColor22 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +1), iTextureSize))); \
|
||||
Samples.fColor32 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+2, +1), iTextureSize))); \
|
||||
\
|
||||
Samples.fColor03 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, +2), iTextureSize))); \
|
||||
Samples.fColor13 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +2), iTextureSize))); \
|
||||
Samples.fColor23 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +2), iTextureSize))); \
|
||||
Samples.fColor33 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+2, +2), iTextureSize))); \
|
||||
\
|
||||
return Samples; \
|
||||
}
|
||||
|
||||
#define DeclareCustomFetch9TapSamplesWithType(SampleType, TextureType, AddrType, Name, LoadTexture) \
|
||||
SampleType Name(AddrType iPxSample, AddrType iTextureSize) \
|
||||
{ \
|
||||
SampleType Samples; \
|
||||
\
|
||||
Samples.fColor00 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, -1), iTextureSize))); \
|
||||
Samples.fColor10 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, -1), iTextureSize))); \
|
||||
Samples.fColor20 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, -1), iTextureSize))); \
|
||||
\
|
||||
Samples.fColor01 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, +0), iTextureSize))); \
|
||||
Samples.fColor11 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +0), iTextureSize))); \
|
||||
Samples.fColor21 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +0), iTextureSize))); \
|
||||
\
|
||||
Samples.fColor02 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(-1, +1), iTextureSize))); \
|
||||
Samples.fColor12 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +1), iTextureSize))); \
|
||||
Samples.fColor22 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +1), iTextureSize))); \
|
||||
\
|
||||
return Samples; \
|
||||
}
|
||||
|
||||
#define DeclareCustomFetchBicubicSamples(Name, LoadTexture) \
|
||||
DeclareCustomFetchBicubicSamplesWithType(FetchedBicubicSamples, FfxFloat32x4, FfxInt32x2, Name, LoadTexture)
|
||||
|
||||
#define DeclareCustomFetchBicubicSamplesMin16(Name, LoadTexture) \
|
||||
DeclareCustomFetchBicubicSamplesWithType(FetchedBicubicSamplesMin16, FFXM_MIN16_F4, FfxInt32x2, Name, LoadTexture)
|
||||
|
||||
#define DeclareCustomFetch9TapSamplesMin16(Name, LoadTexture) \
|
||||
DeclareCustomFetch9TapSamplesWithType(Fetched9TapSamplesMin16, FFXM_MIN16_F4, FfxInt32x2, Name, LoadTexture)
|
||||
|
||||
#define DeclareCustomFetchBilinearSamplesWithType(SampleType, TextureType,AddrType, Name, LoadTexture) \
|
||||
SampleType Name(AddrType iPxSample, AddrType iTextureSize) \
|
||||
{ \
|
||||
SampleType Samples; \
|
||||
Samples.fColor00 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +0), iTextureSize))); \
|
||||
Samples.fColor10 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +0), iTextureSize))); \
|
||||
Samples.fColor01 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+0, +1), iTextureSize))); \
|
||||
Samples.fColor11 = TextureType(LoadTexture(ClampCoord(iPxSample, AddrType(+1, +1), iTextureSize))); \
|
||||
return Samples; \
|
||||
}
|
||||
|
||||
#define DeclareCustomFetchBilinearSamples(Name, LoadTexture) \
|
||||
DeclareCustomFetchBilinearSamplesWithType(FetchedBilinearSamples, FfxFloat32x4, FfxInt32x2, Name, LoadTexture)
|
||||
|
||||
#define DeclareCustomFetchBilinearSamplesMin16(Name, LoadTexture) \
|
||||
DeclareCustomFetchBilinearSamplesWithType(FetchedBilinearSamplesMin16, FFXM_MIN16_F4, FfxInt32x2, Name, LoadTexture)
|
||||
|
||||
// BE CAREFUL: there is some precision issues and (3253, 125) leading to (3252.9989778, 125.001102)
|
||||
// is common, so iPxSample can "jitter"
|
||||
#define DeclareCustomTextureSample(Name, InterpolateSamples, FetchSamples) \
|
||||
FfxFloat32x4 Name(FfxFloat32x2 fUvSample, FfxInt32x2 iTextureSize) \
|
||||
{ \
|
||||
FfxFloat32x2 fPxSample = (fUvSample * FfxFloat32x2(iTextureSize)) - FfxFloat32x2(0.5f, 0.5f); \
|
||||
/* Clamp base coords */ \
|
||||
fPxSample.x = ffxMax(0.0f, ffxMin(FfxFloat32(iTextureSize.x), fPxSample.x)); \
|
||||
fPxSample.y = ffxMax(0.0f, ffxMin(FfxFloat32(iTextureSize.y), fPxSample.y)); \
|
||||
/* */ \
|
||||
FfxInt32x2 iPxSample = FfxInt32x2(floor(fPxSample)); \
|
||||
FfxFloat32x2 fPxFrac = ffxFract(fPxSample); \
|
||||
FfxFloat32x4 fColorXY = FfxFloat32x4(InterpolateSamples(FetchSamples(iPxSample, iTextureSize), fPxFrac)); \
|
||||
return fColorXY; \
|
||||
}
|
||||
|
||||
#define DeclareCustomTextureSampleMin16(Name, InterpolateSamples, FetchSamples) \
|
||||
FFXM_MIN16_F4 Name(FfxFloat32x2 fUvSample, FfxInt32x2 iTextureSize) \
|
||||
{ \
|
||||
FfxFloat32x2 fPxSample = (fUvSample * FfxFloat32x2(iTextureSize)) - FfxFloat32x2(0.5f, 0.5f); \
|
||||
/* Clamp base coords */ \
|
||||
fPxSample.x = ffxMax(0.0f, ffxMin(FfxFloat32(iTextureSize.x), fPxSample.x)); \
|
||||
fPxSample.y = ffxMax(0.0f, ffxMin(FfxFloat32(iTextureSize.y), fPxSample.y)); \
|
||||
/* */ \
|
||||
FfxInt32x2 iPxSample = FfxInt32x2(floor(fPxSample)); \
|
||||
FFXM_MIN16_F2 fPxFrac = FFXM_MIN16_F2(ffxFract(fPxSample)); \
|
||||
FFXM_MIN16_F4 fColorXY = FFXM_MIN16_F4(InterpolateSamples(FetchSamples(iPxSample, iTextureSize), fPxFrac)); \
|
||||
return fColorXY; \
|
||||
}
|
||||
|
||||
#define FFXM_FSR2_CONCAT_ID(x, y) x ## y
|
||||
#define FFXM_FSR2_CONCAT(x, y) FFXM_FSR2_CONCAT_ID(x, y)
|
||||
#define FFXM_FSR2_SAMPLER_1D_0 Lanczos2
|
||||
#define FFXM_FSR2_SAMPLER_1D_1 Lanczos2LUT
|
||||
#define FFXM_FSR2_SAMPLER_1D_2 Lanczos2Approx
|
||||
|
||||
#define FFXM_FSR2_GET_LANCZOS_SAMPLER1D(x) FFXM_FSR2_CONCAT(FFXM_FSR2_SAMPLER_1D_, x)
|
||||
|
||||
#endif //!defined( FFXM_FSR2_SAMPLE_H )
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 3f7412f41446e2447a7abb0d7cf287bc
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
@ -0,0 +1,196 @@
|
||||
// Copyright © 2023 Advanced Micro Devices, Inc.
|
||||
// Copyright © 2024 Arm Limited.
|
||||
// SPDX-License-Identifier: MIT
|
||||
//
|
||||
// Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
// of this software and associated documentation files (the "Software"), to deal
|
||||
// in the Software without restriction, including without limitation the rights
|
||||
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
// copies of the Software, and to permit persons to whom the Software is
|
||||
// furnished to do so, subject to the following conditions:
|
||||
//
|
||||
// The above copyright notice and this permission notice shall be included in all
|
||||
// copies or substantial portions of the Software.
|
||||
//
|
||||
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
// SOFTWARE.
|
||||
|
||||
#ifndef FFXM_FSR2_UPSAMPLE_H
|
||||
#define FFXM_FSR2_UPSAMPLE_H
|
||||
|
||||
#define FFXM_FSR2_UPSAMPLE_USE_LANCZOS_9_TAP 0
|
||||
#define FFXM_FSR2_UPSAMPLE_USE_LANCZOS_5_TAP 1
|
||||
|
||||
#if FFXM_SHADER_QUALITY_OPT_UPSCALING_LANCZOS_5TAP
|
||||
#define FFXM_FSR2_UPSAMPLE_KERNEL FFXM_FSR2_UPSAMPLE_USE_LANCZOS_5_TAP
|
||||
FFXM_STATIC const FfxInt32 iLanczos2SampleCount = 5;
|
||||
#else
|
||||
#define FFXM_FSR2_UPSAMPLE_KERNEL FFXM_FSR2_UPSAMPLE_USE_LANCZOS_9_TAP
|
||||
FFXM_STATIC const FfxUInt32 iLanczos2SampleCount = 16;
|
||||
#endif
|
||||
|
||||
|
||||
void Deringing(RectificationBox clippingBox, FFXM_PARAMETER_INOUT FfxFloat32x3 fColor)
|
||||
{
|
||||
fColor = clamp(fColor, clippingBox.aabbMin, clippingBox.aabbMax);
|
||||
}
|
||||
#if FFXM_HALF
|
||||
void Deringing(RectificationBoxMin16 clippingBox, FFXM_PARAMETER_INOUT FFXM_MIN16_F3 fColor)
|
||||
{
|
||||
fColor = clamp(fColor, clippingBox.aabbMin, clippingBox.aabbMax);
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32 GetUpsampleLanczosWeight(FfxFloat32x2 fSrcSampleOffset, FfxFloat32 fKernelWeight)
|
||||
{
|
||||
FfxFloat32x2 fSrcSampleOffsetBiased = fSrcSampleOffset * fKernelWeight.xx;
|
||||
FfxFloat32 fSampleWeight = Lanczos2ApproxSq(dot(fSrcSampleOffsetBiased, fSrcSampleOffsetBiased));
|
||||
return fSampleWeight;
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FFXM_MIN16_F GetUpsampleLanczosWeight(FFXM_MIN16_F2 fSrcSampleOffset, FFXM_MIN16_F fKernelWeight)
|
||||
{
|
||||
FFXM_MIN16_F2 fSrcSampleOffsetBiased = fSrcSampleOffset * fKernelWeight.xx;
|
||||
FFXM_MIN16_F fSampleWeight = Lanczos2ApproxSq(dot(fSrcSampleOffsetBiased, fSrcSampleOffsetBiased));
|
||||
return fSampleWeight;
|
||||
}
|
||||
#endif
|
||||
|
||||
FfxFloat32 ComputeMaxKernelWeight() {
|
||||
const FfxFloat32 fKernelSizeBias = 1.0f;
|
||||
|
||||
FfxFloat32 fKernelWeight = FfxFloat32(1) + (FfxFloat32(1.0f) / FfxFloat32x2(DownscaleFactor()) - FfxFloat32(1)).x * FfxFloat32(fKernelSizeBias);
|
||||
|
||||
return ffxMin(FfxFloat32(1.99f), fKernelWeight);
|
||||
}
|
||||
|
||||
#if FFXM_HALF
|
||||
FfxFloat32x4 ComputeUpsampledColorAndWeight(const AccumulationPassCommonParams params,
|
||||
FFXM_PARAMETER_INOUT RectificationBoxMin16 clippingBox, FfxFloat32 fReactiveFactor)
|
||||
#else
|
||||
FfxFloat32x4 ComputeUpsampledColorAndWeight(const AccumulationPassCommonParams params,
|
||||
FFXM_PARAMETER_INOUT RectificationBox clippingBox, FfxFloat32 fReactiveFactor)
|
||||
#endif
|
||||
{
|
||||
// We compute a sliced lanczos filter with 2 lobes (other slices are accumulated temporaly)
|
||||
FfxFloat32x2 fDstOutputPos = FfxFloat32x2(params.iPxHrPos) + FFXM_BROADCAST_FLOAT32X2(0.5f); // Destination resolution output pixel center position
|
||||
FfxFloat32x2 fSrcOutputPos = fDstOutputPos * DownscaleFactor(); // Source resolution output pixel center position
|
||||
FfxInt32x2 iSrcInputPos = FfxInt32x2(floor(fSrcOutputPos)); // TODO: what about weird upscale factors...
|
||||
|
||||
FfxFloat32x2 fSrcUnjitteredPos = (FfxFloat32x2(iSrcInputPos) + FfxFloat32x2(0.5f, 0.5f)) - Jitter(); // This is the un-jittered position of the sample at offset 0,0
|
||||
|
||||
FfxFloat32x2 iSrcInputUv = FfxFloat32x2(fSrcOutputPos) / FfxFloat32x2(RenderSize());
|
||||
FfxFloat32x2 unitOffsetUv = FfxFloat32x2(1.0f, 1.0f) / FfxFloat32x2(RenderSize());
|
||||
|
||||
FFXM_MIN16_F4 fColorAndWeight = FFXM_MIN16_F4(0.0f, 0.0f, 0.0f, 0.0f);
|
||||
|
||||
FFXM_MIN16_F2 fBaseSampleOffset = FFXM_MIN16_F2(fSrcUnjitteredPos - fSrcOutputPos);
|
||||
|
||||
// Identify how much of each upsampled color to be used for this frame
|
||||
const FFXM_MIN16_F fKernelReactiveFactor = FFXM_MIN16_F(ffxMax(fReactiveFactor, FfxFloat32(params.bIsNewSample)));
|
||||
const FFXM_MIN16_F fKernelBiasMax = FFXM_MIN16_F(ComputeMaxKernelWeight() * (1.0f - fKernelReactiveFactor));
|
||||
|
||||
const FFXM_MIN16_F fKernelBiasMin = FFXM_MIN16_F(ffxMax(1.0f, ((1.0f + fKernelBiasMax) * 0.3f)));
|
||||
const FFXM_MIN16_F fKernelBiasFactor = FFXM_MIN16_F(ffxMax(0.0f, ffxMax(0.25f * params.fDepthClipFactor, fKernelReactiveFactor)));
|
||||
const FFXM_MIN16_F fKernelBias = ffxLerp(fKernelBiasMax, fKernelBiasMin, fKernelBiasFactor);
|
||||
|
||||
const FFXM_MIN16_F fRectificationCurveBias = FFXM_MIN16_F(ffxLerp(-2.0f, -3.0f, ffxSaturate(params.fHrVelocity / 50.0f)));
|
||||
|
||||
FFXM_MIN16_F2 offsetTL;
|
||||
offsetTL.x = FFXM_MIN16_F(-1);
|
||||
offsetTL.y = FFXM_MIN16_F(-1);
|
||||
|
||||
FFXM_MIN16_F2 fOffsetTL = offsetTL;
|
||||
|
||||
#if FFXM_FSR2_UPSAMPLE_KERNEL == FFXM_FSR2_UPSAMPLE_USE_LANCZOS_9_TAP
|
||||
FFXM_MIN16_F3 fSamples[iLanczos2SampleCount];
|
||||
// Collect samples
|
||||
GatherPreparedInputColorRGBQuad(FfxFloat32x2(-0.5, -0.5) * unitOffsetUv + iSrcInputUv,
|
||||
fSamples[0], fSamples[1], fSamples[4], fSamples[5]);
|
||||
fSamples[2] = LoadPreparedInputColor(FfxInt32x2(1, -1) + iSrcInputPos);
|
||||
fSamples[6] = LoadPreparedInputColor(FfxInt32x2(1, 0) + iSrcInputPos);
|
||||
fSamples[8] = LoadPreparedInputColor(FfxInt32x2(-1, 1) + iSrcInputPos);
|
||||
fSamples[9] = LoadPreparedInputColor(FfxInt32x2(0, 1) + iSrcInputPos);
|
||||
fSamples[10] = LoadPreparedInputColor(FfxInt32x2(1, 1) + iSrcInputPos);
|
||||
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 row = 0; row < 3; row++)
|
||||
{
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 col = 0; col < 3; col++)
|
||||
{
|
||||
FfxInt32 iSampleIndex = col + (row << 2);
|
||||
const FfxInt32x2 sampleColRow = FfxInt32x2(col, row);
|
||||
const FFXM_MIN16_F2 fOffset = fOffsetTL + FFXM_MIN16_F2(sampleColRow);
|
||||
FFXM_MIN16_F2 fSrcSampleOffset = fBaseSampleOffset + fOffset;
|
||||
|
||||
FfxInt32x2 iSrcSamplePos = FfxInt32x2(iSrcInputPos) + FfxInt32x2(offsetTL) + sampleColRow;
|
||||
FFXM_MIN16_F fSampleWeight = FFXM_MIN16_F(GetUpsampleLanczosWeight(fSrcSampleOffset, fKernelBias));
|
||||
|
||||
fColorAndWeight += FFXM_MIN16_F4(fSamples[iSampleIndex] * fSampleWeight, fSampleWeight);
|
||||
|
||||
// Update rectification box
|
||||
{
|
||||
const FFXM_MIN16_F fSrcSampleOffsetSq = dot(fSrcSampleOffset, fSrcSampleOffset);
|
||||
const FFXM_MIN16_F fBoxSampleWeight = exp(fRectificationCurveBias * fSrcSampleOffsetSq);
|
||||
|
||||
const FfxBoolean bInitialSample = (row == 0) && (col == 0);
|
||||
RectificationBoxAddSample(bInitialSample, clippingBox, fSamples[iSampleIndex], fBoxSampleWeight);
|
||||
}
|
||||
}
|
||||
}
|
||||
#elif FFXM_FSR2_UPSAMPLE_KERNEL == FFXM_FSR2_UPSAMPLE_USE_LANCZOS_5_TAP
|
||||
|
||||
FFXM_MIN16_F3 fSamples[iLanczos2SampleCount];
|
||||
// Collect samples
|
||||
FfxInt32x2 rowCol [iLanczos2SampleCount] = {FfxInt32x2(0, -1), FfxInt32x2(-1, 0), FfxInt32x2(0, 0), FfxInt32x2(1, 0), FfxInt32x2(0, 1)};
|
||||
fSamples[0] = LoadPreparedInputColor(rowCol[0] + iSrcInputPos);
|
||||
fSamples[1] = LoadPreparedInputColor(rowCol[1] + iSrcInputPos);
|
||||
fSamples[2] = LoadPreparedInputColor(rowCol[2] + iSrcInputPos);
|
||||
fSamples[3] = LoadPreparedInputColor(rowCol[3] + iSrcInputPos);
|
||||
fSamples[4] = LoadPreparedInputColor(rowCol[4] + iSrcInputPos);
|
||||
FFXM_UNROLL
|
||||
for (FfxInt32 idx = 0; idx < iLanczos2SampleCount; idx++)
|
||||
{
|
||||
const FfxInt32x2 sampleColRow = rowCol[idx];
|
||||
const FFXM_MIN16_F2 fOffset = FFXM_MIN16_F2(sampleColRow);
|
||||
FFXM_MIN16_F2 fSrcSampleOffset = fBaseSampleOffset + fOffset;
|
||||
|
||||
FfxInt32x2 iSrcSamplePos = FfxInt32x2(iSrcInputPos) + FfxInt32x2(offsetTL) + sampleColRow;
|
||||
FFXM_MIN16_F fSampleWeight = FFXM_MIN16_F(GetUpsampleLanczosWeight(fSrcSampleOffset, fKernelBias));
|
||||
|
||||
fColorAndWeight += FFXM_MIN16_F4(fSamples[idx] * fSampleWeight, fSampleWeight);
|
||||
|
||||
// Update rectification box
|
||||
{
|
||||
const FFXM_MIN16_F fSrcSampleOffsetSq = dot(fSrcSampleOffset, fSrcSampleOffset);
|
||||
const FFXM_MIN16_F fBoxSampleWeight = exp(fRectificationCurveBias * fSrcSampleOffsetSq);
|
||||
|
||||
const FfxBoolean bInitialSample = (idx == 0);
|
||||
RectificationBoxAddSample(bInitialSample, clippingBox, fSamples[idx], fBoxSampleWeight);
|
||||
}
|
||||
}
|
||||
|
||||
#endif
|
||||
|
||||
RectificationBoxComputeVarianceBoxData(clippingBox);
|
||||
|
||||
fColorAndWeight.w *= FFXM_MIN16_F(fColorAndWeight.w > FSR2_EPSILON);
|
||||
|
||||
if (fColorAndWeight.w > FSR2_EPSILON) {
|
||||
// Normalize for deringing (we need to compare colors)
|
||||
fColorAndWeight.xyz = fColorAndWeight.xyz / fColorAndWeight.w;
|
||||
fColorAndWeight.w = FFXM_MIN16_F(fColorAndWeight.w*fUpsampleLanczosWeightScale);
|
||||
|
||||
Deringing(clippingBox, fColorAndWeight.xyz);
|
||||
}
|
||||
return fColorAndWeight;
|
||||
}
|
||||
|
||||
#endif //!defined( FFXM_FSR2_UPSAMPLE_H )
|
||||
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 6464cf76ffee2754a8a41d09263e3d5a
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
File diff suppressed because it is too large
Load Diff
@ -0,0 +1,27 @@
|
||||
fileFormatVersion: 2
|
||||
guid: 651248762c096be40bb4913af51e11b7
|
||||
PluginImporter:
|
||||
externalObjects: {}
|
||||
serializedVersion: 2
|
||||
iconMap: {}
|
||||
executionOrder: {}
|
||||
defineConstraints: []
|
||||
isPreloaded: 0
|
||||
isOverridable: 1
|
||||
isExplicitlyReferenced: 0
|
||||
validateReferences: 1
|
||||
platformData:
|
||||
- first:
|
||||
Any:
|
||||
second:
|
||||
enabled: 1
|
||||
settings: {}
|
||||
- first:
|
||||
Editor: Editor
|
||||
second:
|
||||
enabled: 0
|
||||
settings:
|
||||
DefaultValueInitialized: true
|
||||
userData:
|
||||
assetBundleName:
|
||||
assetBundleVariant:
|
||||
Loading…
x
Reference in New Issue
Block a user