v4.11.4:

HIGHLIGHTS: - improved robustness DETAILS: - NRD: GLSL-related replacements come into play if not already defined - NRD: improved robustness - RELAX: fixed touching of pixels outside of denoising range (can be garbage) in "HistoryClamping" pass - RELAX/REBLUR: fixed touching of wrong data (can be garbage) in "checkerboard" mode on the left side of the screen - REBLUR: pedantic IQ changes in "TemporalStabilization" pass if "split screen" mode is active - Integration: NRI updated to v1.161 - updated docs
NVIDIA-RTX · Jan 21, 2025 · c559760 · c559760
1 parent 33e10fa
commit c559760
Show file tree

Hide file tree

Showing 24 changed files with 222 additions and 140 deletions.
diff --git a/External/MathLib b/External/MathLib
diff --git a/Include/NRD.h b/Include/NRD.h
@@ -29,8 +29,8 @@ license agreement from NVIDIA CORPORATION is strictly prohibited.
 
 #define NRD_VERSION_MAJOR 4
 #define NRD_VERSION_MINOR 11
-#define NRD_VERSION_BUILD 3
-#define NRD_VERSION_DATE "3 January 2024"
+#define NRD_VERSION_BUILD 4
+#define NRD_VERSION_DATE "20 January 2025"
 
 #if defined(_WIN32)
     #define NRD_CALL __stdcall

diff --git a/Include/NRDSettings.h b/Include/NRDSettings.h
@@ -134,6 +134,7 @@ namespace nrd
         float timeDeltaBetweenFrames = 0.0f;
 
         // (units > 0) - use TLAS or tracing range
+        // It's highly recommended to use "viewZ > denoisingRange" for INF (sky) pixels
         float denoisingRange = 500000.0f;
 
         // [0.01; 0.02] - two samples considered occluded if relative distance difference is greater than this slope-scaled threshold

diff --git a/Integration/NRDIntegration.hpp b/Integration/NRDIntegration.hpp
@@ -17,8 +17,8 @@ license agreement from NVIDIA CORPORATION is strictly prohibited.
     #include <alloca.h>
 #endif
 
-static_assert(NRD_VERSION_MAJOR >= 4 && NRD_VERSION_MINOR >= 10, "Unsupported NRD version!");
-static_assert(NRI_VERSION_MAJOR >= 1 && NRI_VERSION_MINOR >= 158, "Unsupported NRI version!");
+static_assert(NRD_VERSION_MAJOR >= 4 && NRD_VERSION_MINOR >= 11, "Unsupported NRD version!");
+static_assert(NRI_VERSION_MAJOR >= 1 && NRI_VERSION_MINOR >= 161, "Unsupported NRI version!");
 
 namespace nrd
 {
@@ -341,15 +341,15 @@ void Integration::CreateResources(uint16_t resourceWidth, uint16_t resourceHeigh
             snprintf(name, sizeof(name), "%s::P(%u)", m_Name, i);
         else
             snprintf(name, sizeof(name), "%s::T(%u)", m_Name, i - instanceDesc.permanentPoolSize);
-        m_NRI->SetTextureDebugName(*texture, name);
+        m_NRI->SetDebugName(texture, name);
 
         // Construct NRD texture
         nri::TextureBarrierDesc& nrdTexture = m_TexturePool[i];
         nrdTexture = nri::TextureBarrierFromUnknown(texture, {nri::AccessBits::UNKNOWN, nri::Layout::UNKNOWN}, 0, 1);
 
         // Adjust memory usage
         nri::MemoryDesc memoryDesc = {};
-        m_NRI->GetTextureMemoryDesc(*m_Device, textureDesc, nri::MemoryLocation::DEVICE, memoryDesc);
+        m_NRI->GetTextureMemoryDesc(*texture, nri::MemoryLocation::DEVICE, memoryDesc);
 
         if (i < instanceDesc.permanentPoolSize)
             m_PermanentPoolSize += memoryDesc.size;

diff --git a/README.md b/README.md
@@ -1,4 +1,4 @@
-# NVIDIA REAL-TIME DENOISERS v4.11.3 (NRD)
+# NVIDIA REAL-TIME DENOISERS v4.11.4 (NRD)
 
 [![Build NRD SDK](https://github.com/NVIDIAGameWorks/RayTracingDenoiser/actions/workflows/build.yml/badge.svg)](https://github.com/NVIDIAGameWorks/RayTracingDenoiser/actions/workflows/build.yml)
 
@@ -98,8 +98,6 @@ NRD sample has *TESTS* section in the bottom of the UI, a new test can be added
 - If nothing helps
   - describe the issue, attach a video and steps to reproduce
 
-Additionally, for any information, suggestions or general requests please feel free to contact us at NRD-SDK-Support@nvidia.com
-
 # API
 
 Terminology:
@@ -154,8 +152,6 @@ Commons inputs for primary hits (if *PSR* is not used, common use case) or for s
   - project on screen using matrices passed to NRD
   - `.w` component is positive view Z (or just transform world-space position to main view space and take `.z` component)
 
-All textures should be *NaN* free at each pixel, even at pixels outside of denoising range.
-
 The illustration below shows expected inputs for primary hits:
 
 ![Input without PSR](Images/InputsWithoutPsr.png)
@@ -208,6 +204,17 @@ Distance to occluder (*SIGMA*):
 
 See `NRDDescs.h` and `NRD.hlsli` for more details and descriptions of other inputs and outputs.
 
+# NOISY & NON-NOISY DATA REQUIREMENTS
+
+Noisy inputs:
+ - garbage values are allowed outside of active viewport, i.e. `>= CommonSettings::rectSize`
+ - garbage values are allowed outside of denoising range, i.e. `>= CommonSettings::denoisingRange`
+
+Non-noisy inputs (guides):
+ - must not contain garbage
+
+Where "garbage" is `NAN/INF` or undesired value.
+
 # IMPROVING OUTPUT QUALITY
 
 The temporal part of *NRD* naturally suppresses jitter, which is essential for upscaling techniques. If an *SH* denoiser is in use, a high quality resolve can be applied to the final output to regain back macro details, micro details and per-pixel jittering. As an example, the image below demonstrates the results *after* and *before* resolve with active *DLSS* (quality mode).
@@ -665,13 +672,6 @@ IN_NORMAL_ROUGHNESS = GetVirtualSpaceNormalAndRoughnessAt( B );
 IN_MV = GetMotionAt( B );
 ```
 
-## INTERACTION WITH `INFs` AND `NANs`
-
-- *NRD* doesn't touch pixels outside of viewport: `INFs / NANs` are allowed
-- *NRD* doesn't touch pixels outside of denoising range: `INFs / NANs` are allowed
-- `INFs / NANs` are not allowed for pixels inside the viewport and denoising range
-  - `INFs` can be used in `IN_VIEWZ`, but not recommended
-
 ## INTERACTION WITH FRAME GENERATION TECHNIQUES
 
 Frame generation (FG) techniques boost FPS by interpolating between 2 last available frames. *NRD* works better when frame rate increases, because it gets more data per second. It's not the case for FG, because all rendering pipeline underlying passes (like, denoising) continue to work on the original non-boosted framerate. `GetMaxAccumulatedFrameNum` helper should get a real FPS, not a fake one.
@@ -708,7 +708,7 @@ Hair strands tangent vectors *can't* be used as "normals guide" for *NRD* due to
 
 **[NRD]** When upgrading to the latest version keep an eye on `ResourceType` enumeration. The order of the input slots can be changed or something can be added, you need to adjust the inputs accordingly to match the mapping. Or use *NRD integration* to simplify the process.
 
-**[NRD]** All pixels in floating point textures should be INF / NAN free to avoid propagation, because such values are used in weight calculations and accumulation of a weighted sum. Functions `XXX_FrontEnd_PackRadianceAndHitDist` perform optional NAN / INF clearing of the input signal. There is a boolean to skip these checks.
+**[NRD]** Functions `XXX_FrontEnd_PackRadianceAndHitDist` perform optional `NAN/INF` clearing of the input signal. There is a boolean to skip these checks.
 
 **[NRD]** All denoisers work with positive RGB inputs (some denoisers can change color space in *front end* functions). For better image quality, HDR color inputs need to be in a sane range [0; 250], because the internal pipeline uses FP16 and *RELAX* tracks second moments of the input signal, i.e. `x^2` must fit into FP16 range. If the color input is in a wider range, any form of non-aggressive color compression can be applied (linear scaling, pow-based or log-based methods). *REBLUR* supports wider HDR ranges, because it doesn't track second moments. Passing pre-exposured colors (i.e. `color * exposure`) is not recommended, because a significant momentary change in exposure is hard to react to in this case.
 

diff --git a/Resources/Version.h b/Resources/Version.h
@@ -23,6 +23,6 @@ Versioning rules:
 
 #define VERSION_MAJOR                   4
 #define VERSION_MINOR                   11
-#define VERSION_BUILD                   3
+#define VERSION_BUILD                   4
 
 #define VERSION_STRING STR(VERSION_MAJOR.VERSION_MINOR.VERSION_BUILD encoding=NRD_NORMAL_ENCODING.NRD_ROUGHNESS_ENCODING)
diff --git a/Shaders/Include/Common.hlsli b/Shaders/Include/Common.hlsli
@@ -278,7 +278,7 @@ float4 GetBlurKernelRotation( compiletime const uint mode, uint2 pixelPos, float
 
 float IsInScreenNearest( float2 uv )
 {
-    return float( all( uv >= 0.0 ) && all( uv < 1.0 ) );
+    return float( all( uv > 0.0 ) && all( uv < 1.0 ) );
 }
 
 // x y
@@ -296,6 +296,7 @@ float4 IsInScreenBilinear( float2 footprintOrigin, float2 rectSize )
 float2 ApplyCheckerboardShift( float2 pos, uint mode, uint counter, uint frameIndex )
 {
     // IMPORTANT: "pos" must be snapped to the pixel center
+    pos += 16384.0; // apply pattern-neutral bias, to make "pos" non-negative
     uint checkerboard = Sequence::CheckerBoard( uint2( pos ), frameIndex );
     float shift = ( ( counter & 0x1 ) == 0 ) ? -1.0 : 1.0;
 

diff --git a/Shaders/Include/NRD.hlsli b/Shaders/Include/NRD.hlsli
@@ -236,16 +236,36 @@ NOISY INPUTS:
 //=================================================================================================================================
 
 #ifdef NRD_GLSL
-    #define float4                                                                      vec4
-    #define float3                                                                      vec3
-    #define float2                                                                      vec2
-    #define float2x2                                                                    mat2x2
-    #define float2x3                                                                    mat2x3
-    #define float3x3                                                                    mat3x3
-    #define rsqrt                                                                       inversesqrt
-    #define saturate( x )                                                               clamp( x, 0.0, 1.0 )
-    #define lerp                                                                        mix
-    #define mul( x, y )                                                                 ( x * y )
+    #ifndef float4
+        #define float4                                                                  vec4
+    #endif
+    #ifndef float3
+        #define float3                                                                  vec3
+    #endif
+    #ifndef float2
+        #define float2                                                                  vec2
+    #endif
+    #ifndef float2x2
+        #define float2x2                                                                mat2x2
+    #endif
+    #ifndef float2x3
+        #define float2x3                                                                mat2x3
+    #endif
+    #ifndef float3x3
+        #define float3x3                                                                mat3x3
+    #endif
+    #ifndef rsqrt
+        #define rsqrt                                                                   inversesqrt
+    #endif
+    #ifndef saturate
+        #define saturate( x )                                                           clamp( x, 0.0, 1.0 )
+    #endif
+    #ifndef lerp
+        #define lerp                                                                    mix
+    #endif
+    #ifndef mul
+        #define mul( x, y )                                                             ( x * y )
+    #endif
 #endif
 
 //=================================================================================================================================

diff --git a/Shaders/Include/REBLUR_Common_DiffuseSpatialFilter.hlsli b/Shaders/Include/REBLUR_Common_DiffuseSpatialFilter.hlsli
@@ -106,14 +106,16 @@ license agreement from NVIDIA CORPORATION is strictly prohibited.
         #endif
 
             // Snap to the pixel center!
-            uv = ( floor( uv * gRectSize ) + 0.5 ) * gRectSizeInv;
+            uv = floor( uv * gRectSize ) + 0.5;
 
             // Apply checkerboard shift
         #if( REBLUR_SPATIAL_MODE == REBLUR_PRE_BLUR )
-            uv = ApplyCheckerboardShift( uv * gRectSize, gDiffCheckerboard, n, gFrameIndex ) * gRectSizeInv;
+            uv = ApplyCheckerboardShift( uv, gDiffCheckerboard, n, gFrameIndex );
         #endif
 
             // Texture coordinates
+            uv *= gRectSizeInv;
+
             float2 uvScaled = ClampUvToViewport( uv );
             float2 checkerboardUvScaled = uvScaled;
         #if( REBLUR_SPATIAL_MODE == REBLUR_PRE_BLUR )

diff --git a/Shaders/Include/REBLUR_Common_SpecularSpatialFilter.hlsli b/Shaders/Include/REBLUR_Common_SpecularSpatialFilter.hlsli
@@ -130,14 +130,16 @@ license agreement from NVIDIA CORPORATION is strictly prohibited.
         #endif
 
             // Snap to the pixel center!
-            uv = ( floor( uv * gRectSize ) + 0.5 ) * gRectSizeInv;
+            uv = floor( uv * gRectSize ) + 0.5;
 
             // Apply checkerboard shift
         #if( REBLUR_SPATIAL_MODE == REBLUR_PRE_BLUR )
-            uv = ApplyCheckerboardShift( uv * gRectSize, gSpecCheckerboard, n, gFrameIndex ) * gRectSizeInv;
+            uv = ApplyCheckerboardShift( uv, gSpecCheckerboard, n, gFrameIndex );
         #endif
 
             // Texture coordinates
+            uv *= gRectSizeInv;
+
             float2 uvScaled = ClampUvToViewport( uv );
             float2 checkerboardUvScaled = uvScaled;
         #if( REBLUR_SPATIAL_MODE == REBLUR_PRE_BLUR )

diff --git a/Shaders/Include/REBLUR_Config.hlsli b/Shaders/Include/REBLUR_Config.hlsli
@@ -170,6 +170,7 @@ license agreement from NVIDIA CORPORATION is strictly prohibited.
     NRD_CONSTANT( float, gMinRectDimMulUnproject ) \
     NRD_CONSTANT( float, gUsePrepassNotOnlyForSpecularMotionEstimation ) \
     NRD_CONSTANT( float, gSplitScreen ) \
+    NRD_CONSTANT( float, gSplitScreenPrev ) \
     NRD_CONSTANT( float, gCheckerboardResolveAccumSpeed ) \
     NRD_CONSTANT( float, gViewZScale ) \
     NRD_CONSTANT( float, gFireflySuppressorMinRelativeScale ) \

diff --git a/Shaders/Include/REBLUR_TemporalStabilization.hlsli b/Shaders/Include/REBLUR_TemporalStabilization.hlsli
@@ -262,7 +262,7 @@ NRD_EXPORT void NRD_CS_MAIN( NRD_CS_MAIN_ARGS )
         float diffHistoryWeight = diffTemporalAccumulationParams.x;
         diffHistoryWeight *= diffAntilag; // this is important
         diffHistoryWeight *= float( pixelUv.x >= gSplitScreen );
-        diffHistoryWeight *= float( smbPixelUv.x >= gSplitScreen );
+        diffHistoryWeight *= float( smbPixelUv.x >= gSplitScreenPrev );
 
         smbDiffHistory = Color::Clamp( diffM1, diffSigma * diffTemporalAccumulationParams.y, smbDiffHistory );
 
@@ -407,8 +407,8 @@ NRD_EXPORT void NRD_CS_MAIN( NRD_CS_MAIN_ARGS )
         float specHistoryWeight = specTemporalAccumulationParams.x;
         specHistoryWeight *= specAntilag; // this is important
         specHistoryWeight *= float( pixelUv.x >= gSplitScreen );
-        specHistoryWeight *= virtualHistoryAmount != 1.0 ? float( smbPixelUv.x >= gSplitScreen ) : 1.0;
-        specHistoryWeight *= virtualHistoryAmount != 0.0 ? float( vmbPixelUv.x >= gSplitScreen ) : 1.0;
+        specHistoryWeight *= virtualHistoryAmount != 1.0 ? float( smbPixelUv.x >= gSplitScreenPrev ) : 1.0;
+        specHistoryWeight *= virtualHistoryAmount != 0.0 ? float( vmbPixelUv.x >= gSplitScreenPrev ) : 1.0;
 
         float responsiveFactor = RemapRoughnessToResponsiveFactor( roughness );
         float smc = GetSpecMagicCurve( roughness );