v1.2.07 release

.gitmodules

@@ -14,3 +14,11 @@
 	path = thirdparty/Vulkan-Headers
 	url = https://github.com/KhronosGroup/Vulkan-Headers.git
 	branch = sdk-1.2.170
+[submodule "thirdparty/libpng"]
+	path = thirdparty/libpng
+	url = https://github.com/glennrp/libpng.git
+	branch = libpng16
+[submodule "thirdparty/zlib"]
+	path = thirdparty/zlib
+	url = https://github.com/madler/zlib.git
+	branch = master

CMakeLists.txt

@@ -19,7 +19,7 @@ set(CMAKE_RUNTIME_OUTPUT_DIRECTORY "bin")
 
 set_property(GLOBAL PROPERTY USE_FOLDERS ON)
 
-# Add Vulkan headers on ARM since there is no official Vulkan SDK
+# Add Vulkan headers on ARM since there is no official ARM Vulkan SDK
 if(${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64")
     add_subdirectory(thirdparty/Vulkan-Headers)
     set(RTXGI_API_VULKAN_SDK "1")
@@ -47,6 +47,26 @@ if(RTXGI_BUILD_SAMPLES)
     target_compile_definitions(tinygltf PUBLIC _CRT_SECURE_NO_WARNINGS) # suppress the sprintf CRT warnings
     set_target_properties(tinygltf PROPERTIES FOLDER "Thirdparty/")
 
+    # zlib
+    set(CMAKE_WARN_DEPRECATED OFF CACHE BOOL "" FORCE)
+    set(CMAKE_DEBUG_POSTFIX "" CACHE STRING "override debug postfix")
+    option(ASM686 "" OFF)
+    option(AMD64 "" OFF)
+    add_subdirectory(thirdparty/zlib)
+    target_compile_options(zlib BEFORE PUBLIC /wd4267) # suppress the implicit conversion warning
+    set_target_properties(zlib zlibstatic example minigzip PROPERTIES FOLDER "Thirdparty/zlib" DEBUG_POSTFIX "")
+
+    # libpng
+    option(PNG_BUILD_ZLIB "" ON)
+    option(PNG_STATIC "" ON)
+    option(PNG_SHARED "" OFF)
+    option(PNG_EXECUTABLES "" OFF)
+    option(PNG_FRAMEWORK "" OFF)
+    option(PNG_TESTS "" OFF)
+    set(ZLIB_INCLUDE_DIRS "${CMAKE_SOURCE_DIR}/thirdparty/zlib" "${CMAKE_BINARY_DIR}/thirdparty/zlib")
+    add_subdirectory(thirdparty/libpng)
+    set_target_properties(png_static genfiles PROPERTIES FOLDER "Thirdparty/libpng" DEBUG_POSTFIX "")
+
     # Samples
     add_subdirectory(samples)
 endif()

ChangeLog.md

@@ -1,5 +1,34 @@
 # RTXGI SDK Change Log
 
+## 1.2.07
+
+### SDK
+
+Features and Improvements:
+- Shader improvements for probe data reads/writes
+  - The probe data texture now uses the same layout as the irradiance and distance textures
+  - This change makes probe data visualizations easier to understand
+- Adds a RWTexture2D variant of ```DDGILoadProbeState()```
+- Uses the ```DDGILoadProbeState()``` convenience function where appropriate
+- Moved the ```data``` member of the ```DDGIVolumeDescGPU``` struct inside the ```GetPackedData()``` function so the struct size is the same on the CPU and GPU
+- Updated stale code comments (non-functional)
+- Bumped SDK revision number and version string
+
+Bug Fixes:
+- ```DDGIGetVolumeBlendWeight```: fixed regression related to volume edge fade not respecting rotations
+
+### Test Harness
+
+Features and Improvements:
+- ```ProbeTraceRGS.hlsl```: no longer performs lighting for fixed rays when probe relocation is enabled (performance optimization)
+- Adds a performance benchmark mode that collects CPU and GPU performance data and outputs the results to ```*.csv``` files
+  - Press the ```F2``` key to run a benchmark
+- Adds the ability to store intermediate textures (back buffer, GBuffer, RTAO, and DDGIVolume textures) to disk
+  - Uses the zlib and libpng libraries for cross-platform compatibility
+  - Press the ```F1``` key to save the current image data to disk
+- ```VolumeTexturesCS.hlsl```: minor updates to the layout order of visualized textures
+- Config file and documentation updates
+
 ## 1.2.04
 
 ### SDK

docs/DDGIVolume.md

@@ -369,11 +369,11 @@ This texture stores world-space offsets and classification states for all probes
   - ```ProbeDataCommon.hlsl``` contains helper functions for reading and writing world-space offset data.
 - Probe classification state is stored in the W channel.
 
-The texture's layout is the same as the irradiance and distance texture altases, except each probe uses just *one texel*. Below is a visualization of the probe data texture's world-space offsets (top) and probe states (bottom).
+The texture's layout is the same as the irradiance and distance texture altases, except here each probe is represented by just *one texel*. Below is a visualization of the probe data texture's world-space offsets (top) and probe states (bottom).
 
 <figure>
 <img src="images/ddgivolume-textures-probedata.jpg" width=800px></img>
-<figcaption><b>Figure 6: The Probe Data texture (zoomed) from the Crytek Sponza scene</b></figcaption>
+<figcaption><b>Figure 6: A visualization of the Probe Data texture (zoomed) for the Crytek Sponza scene</b></figcaption>
 </figure>
 
 

docs/ShaderAPI.md

@@ -52,9 +52,14 @@ The Test Harness sample application provides [an example ray generation shader](
 
 Computes the 3D grid-space coordinates for the probe at a given probe index in the range [0, numProbes]. Provide these coordinates to ```DDGIGetProbeWorldPosition()``` and ```DDGIGetScrollingProbeIndex()```.
 
-#### ```DDGIGetProbeState(...)```
+#### ```DDGILoadProbeState(...)```
 
-    float DDGIGetProbeState(
+    float DDGILoadProbeState(
+        int probeIndex,
+        RWTexture2D<float4> probeData,
+        DDGIVolumeDescGPU volume)
+
+    float DDGILoadProbeState(
         int probeIndex,
         Texture2D<float4> probeData,
         DDGIVolumeDescGPU volume)

rtxgi-sdk/include/rtxgi/Common.h

@@ -20,11 +20,11 @@ namespace rtxgi
     {
         static const int major = 1;
         static const int minor = 2;
-        static const int revision = 4;
+        static const int revision = 7;
 
         inline static const char* getVersionString()
         {
-            return "1.2.04";
+            return "1.2.07";
         }
     };
 

rtxgi-sdk/include/rtxgi/ddgi/DDGIVolumeDescGPU.h

@@ -95,9 +95,9 @@ struct DDGIVolumeDescGPU
     bool     probeScrollClear[3];                // whether probes of a plane need to be cleared due to scrolling movement
 
 #ifndef HLSL
-    rtxgi::DDGIVolumeDescGPUPacked data;
     rtxgi::DDGIVolumeDescGPUPacked GetPackedData()
     {
+        rtxgi::DDGIVolumeDescGPUPacked data;
         data.origin = origin;
         data.probeHysteresis = probeHysteresis;
         data.rotation = rotation;

rtxgi-sdk/include/rtxgi/ddgi/gfx/DDGIVolume_D3D12.h

@@ -128,7 +128,7 @@ namespace rtxgi
 
             ID3D12Resource*                   constantsBuffer = nullptr;                    // Constants structured buffer resource pointer (device)
 
-            // Provide these resources if you use DDGIVolume::Upload() to transfer constants to the GPU
+            // Provide these resources if you use UploadDDGIVolumeConstants() to transfer constants to the GPU
             ID3D12Resource*                   constantsBufferUpload = nullptr;              // [Optional] Constants structured buffer resource pointer (upload)
             UINT64                            constantsBufferSizeInBytes = 0;               // [Optional] Size (in bytes) of the constants structured buffer
         };
@@ -261,7 +261,7 @@ namespace rtxgi
             ID3D12Device*               m_device = nullptr;                                 // D3D12 device pointer
         #endif
 
-            // Constants (if you use the DDGIVolume::Upload() to transfer constants to the GPU)
+            // Constants (if you use UploadDDGIVolumeConstants() to transfer constants to the GPU)
             ID3D12Resource*             m_constantsBuffer = nullptr;                        // Structured buffer that stores the volume's constants (device)
             ID3D12Resource*             m_constantsBufferUpload = nullptr;                  // Structured buffer that stores the volume's constants (upload)
             UINT64                      m_constantsBufferSizeInBytes = 0;                   // Size (in bytes) of the structured buffer that stores *all* volumes constants

rtxgi-sdk/include/rtxgi/ddgi/gfx/DDGIVolume_VK.h

@@ -142,7 +142,7 @@ namespace rtxgi
 
             VkBuffer                constantsBuffer = nullptr;                              // Constants structured buffer (device)
 
-            // Provide these resources if you use DDGIVolume::Upload() to transfer constants to the GPU
+            // Provide these resources if you use UploadDDGIVolumeConstants() to transfer constants to the GPU
             VkBuffer                constantsBufferUpload = nullptr;                        // [Optional] Constants structured buffer (upload)
             VkDeviceMemory          constantsBufferUploadMemory = nullptr;                  // [Optional] Constants structured buffer memory (upload)
             uint64_t                constantsBufferSizeInBytes = 0;                         // [Optional] Size (in bytes) of the constants structured buffer
@@ -307,7 +307,7 @@ namespace rtxgi
             VkDescriptorPool            m_descriptorPool = nullptr;                            // Vulkan descriptor pool handle
         #endif
 
-            // Constants (if you use the DDGIVolume::Upload() to transfer constants to the GPU)
+            // Constants (if you use UploadDDGIVolumeConstants() to transfer constants to the GPU)
             VkBuffer                    m_constantsBuffer = nullptr;                           // Structured buffer that stores the volume's constants (device)
             VkBuffer                    m_constantsBufferUpload = nullptr;                     // Structured buffer that stores the volume's constants (upload)
             VkDeviceMemory              m_constantsBufferUploadMemory = nullptr;               // Memory for the volume's constants upload structured buffer

rtxgi-sdk/shaders/ddgi/Irradiance.hlsl

@@ -38,12 +38,14 @@ float3 DDGIGetSurfaceBias(float3 surfaceNormal, float3 cameraDirection, DDGIVolu
  */ 
 float DDGIGetVolumeBlendWeight(float3 worldPosition, DDGIVolumeDescGPU volume)
 {
-    // Get the volume's origin
+    // Get the volume's origin and extent
     float3 origin = volume.origin + (volume.probeScrollOffsets * volume.probeSpacing);
-
-    // Get spatial delta between the world position and the volume
     float3 extent = (volume.probeSpacing * (volume.probeCounts - 1)) * 0.5f;
-    float3 position = abs(worldPosition - origin);
+
+    // Get the delta between the (rotated volume) and the world-space position
+    float3 position = (worldPosition - origin);
+    position = abs(RTXGIQuaternionRotate(position, RTXGIQuaternionConjugate(volume.rotation)));
+
     float3 delta = position - extent;
     if(all(delta < 0)) return 1.f;
 
@@ -99,13 +101,8 @@ float3 DDGIGetVolumeIrradiance(
         int adjacentProbeIndex = DDGIGetScrollingProbeIndex(adjacentProbeCoords, volume);
 
         // Early Out: don't allow inactive probes to contribute to irradiance
-        if (volume.probeClassificationEnabled)
-        {
-            // Get the probe state
-            int2 probeDataCoords = DDGIGetProbeDataTexelCoords(adjacentProbeIndex, volume);
-            int  probeState = resources.probeData[probeDataCoords].w;
-            if (probeState == RTXGI_DDGI_PROBE_STATE_INACTIVE) continue;
-        }
+        int probeState = DDGILoadProbeState(adjacentProbeIndex, resources.probeData, volume);
+        if (probeState == RTXGI_DDGI_PROBE_STATE_INACTIVE) continue;
 
         // Get the adjacent probe's world position
         float3 adjacentProbeWorldPosition = DDGIGetProbeWorldPosition(adjacentProbeCoords, volume, resources.probeData);

rtxgi-sdk/shaders/ddgi/ProbeBlendingCS.hlsl

@@ -406,13 +406,8 @@ void DDGIProbeBlendingCS(uint3 DispatchThreadID : SV_DispatchThreadID, uint Grou
     }
 
     // Early Out: don't blend rays for probes that are inactive
-    if(volume.probeClassificationEnabled)
-    {
-        // Get the probe state
-        int2 probeDataCoords = DDGIGetProbeDataTexelCoords(probeIndex, volume);
-        int  probeState = ProbeData[probeDataCoords].w;
-        if (probeState == RTXGI_DDGI_PROBE_STATE_INACTIVE) return;
-    }
+    int probeState = DDGILoadProbeState(probeIndex, ProbeData, volume);
+    if (probeState == RTXGI_DDGI_PROBE_STATE_INACTIVE) return;
 
     // Visualize the probe index
 #if RTXGI_DDGI_BLEND_RADIANCE && RTXGI_DDGI_DEBUG_PROBE_INDEXING

rtxgi-sdk/shaders/ddgi/ProbeClassificationCS.hlsl

@@ -288,7 +288,7 @@ void DDGIProbeClassificationCS(uint3 DispatchThreadID : SV_DispatchThreadID)
     }
 
     // Get the probe's texel coordinates in the Probe Data texture
-    uint2 probeDataCoords = DDGIGetProbeDataTexelCoords(probeIndex, volume);
+    uint2 probeDataCoords = DDGIGetProbeTexelCoords(probeIndex, volume);
 
     // Early out: number of backface hits has been exceeded. The probe is probably inside geometry.
     if(((float)backfaceCount / (float)RTXGI_DDGI_NUM_FIXED_RAYS) > volume.probeBackfaceThreshold)
@@ -369,7 +369,7 @@ void DDGIProbeClassificationResetCS(uint3 DispatchThreadID : SV_DispatchThreadID
 #endif
 
     // Get the probe's texel coordinates in the Probe Data texture
-    uint2 probeDataCoords = DDGIGetProbeDataTexelCoords(DispatchThreadID.x, volume);
+    uint2 probeDataCoords = DDGIGetProbeTexelCoords(DispatchThreadID.x, volume);
 
     ProbeData[probeDataCoords].w = RTXGI_DDGI_PROBE_STATE_ACTIVE;
 }

rtxgi-sdk/shaders/ddgi/ProbeRelocationCS.hlsl

@@ -268,7 +268,7 @@ void DDGIProbeRelocationCS(uint3 DispatchThreadID : SV_DispatchThreadID)
 #endif
 
     // Get the probe's texel coordinates in the Probe Data texture
-    uint2 coords = DDGIGetProbeDataTexelCoords(probeIndex, volume);
+    uint2 coords = DDGIGetProbeTexelCoords(probeIndex, volume);
 
     // Read the current world position offset
     float3 offset = DDGILoadProbeDataOffset(ProbeData, coords, volume);
@@ -387,7 +387,7 @@ void DDGIProbeRelocationResetCS(uint3 DispatchThreadID : SV_DispatchThreadID)
 #endif
 
     // Get the probe's texel coordinates in the Probe Data texture
-    uint2 probeDataCoords = DDGIGetProbeDataTexelCoords(DispatchThreadID.x, volume);
+    uint2 probeDataCoords = DDGIGetProbeTexelCoords(DispatchThreadID.x, volume);
 
     // Write the probe offset
     ProbeData[probeDataCoords].xyz = float3(0.f, 0.f, 0.f);

rtxgi-sdk/shaders/ddgi/include/ProbeCommon.hlsl

@@ -62,7 +62,7 @@ float3 DDGIGetProbeWorldPosition(int3 probeCoords, DDGIVolumeDescGPU volume, Tex
         int probeIndex = DDGIGetScrollingProbeIndex(probeCoords, volume);
 
         // Find the texture coordinates of the probe in the Probe Data texture
-        uint2 coords = DDGIGetProbeDataTexelCoords(probeIndex, volume);
+        uint2 coords = DDGIGetProbeTexelCoords(probeIndex, volume);
 
         // Load the probe's world-space position offset and add it to the current world position
         probeWorldPosition += DDGILoadProbeDataOffset(probeData, coords, volume);
@@ -88,7 +88,7 @@ float3 DDGIGetProbeWorldPosition(int3 probeCoords, DDGIVolumeDescGPU volume, RWT
         int probeIndex = DDGIGetScrollingProbeIndex(probeCoords, volume);
 
         // Find the texture coordinates of the probe in the Probe Data texture
-        uint2 coords = DDGIGetProbeDataTexelCoords(probeIndex, volume);
+        uint2 coords = DDGIGetProbeTexelCoords(probeIndex, volume);
 
         // Load the probe's world-space position offset and add it to the current world position
         probeWorldPosition += DDGILoadProbeDataOffset(probeData, coords, volume);