godot-module-template/engine/thirdparty/amd-fsr2/patches/0002-godot-fsr2-options.patch

122 lines
5.4 KiB
Diff

diff --git a/thirdparty/amd-fsr2/ffx_fsr2.cpp b/thirdparty/amd-fsr2/ffx_fsr2.cpp
index 3970aa7f5b..ec571b9cd2 100644
--- a/thirdparty/amd-fsr2/ffx_fsr2.cpp
+++ b/thirdparty/amd-fsr2/ffx_fsr2.cpp
@@ -952,6 +952,8 @@ static FfxErrorCode fsr2Dispatch(FfxFsr2Context_Private* context, const FfxFsr2D
context->constants.lumaMipDimensions[0] = uint32_t(context->constants.maxRenderSize[0] / mipDiv);
context->constants.lumaMipDimensions[1] = uint32_t(context->constants.maxRenderSize[1] / mipDiv);
+ memcpy(context->constants.reprojectionMatrix, params->reprojectionMatrix, sizeof(context->constants.reprojectionMatrix));
+
// reactive mask bias
const int32_t threadGroupWorkRegionDim = 8;
const int32_t dispatchSrcX = (context->constants.renderSize[0] + (threadGroupWorkRegionDim - 1)) / threadGroupWorkRegionDim;
diff --git a/thirdparty/amd-fsr2/ffx_fsr2.h b/thirdparty/amd-fsr2/ffx_fsr2.h
index 2a1c74abb1..dfcd4caf35 100644
--- a/thirdparty/amd-fsr2/ffx_fsr2.h
+++ b/thirdparty/amd-fsr2/ffx_fsr2.h
@@ -146,6 +146,7 @@ typedef struct FfxFsr2DispatchDescription {
float autoReactiveScale; ///< A value to scale the reactive mask
float autoReactiveMax; ///< A value to clamp the reactive mask
+ float reprojectionMatrix[16]; ///< The matrix used for reprojecting pixels with invalid motion vectors by using the depth.
} FfxFsr2DispatchDescription;
/// A structure encapsulating the parameters for automatic generation of a reactive mask
diff --git a/thirdparty/amd-fsr2/ffx_fsr2_private.h b/thirdparty/amd-fsr2/ffx_fsr2_private.h
index 6b5fbc5117..8a9aec5778 100644
--- a/thirdparty/amd-fsr2/ffx_fsr2_private.h
+++ b/thirdparty/amd-fsr2/ffx_fsr2_private.h
@@ -44,6 +44,9 @@ typedef struct Fsr2Constants {
float deltaTime;
float dynamicResChangeFactor;
float viewSpaceToMetersFactor;
+
+ float pad;
+ float reprojectionMatrix[16];
} Fsr2Constants;
struct FfxFsr2ContextDescription;
diff --git a/thirdparty/amd-fsr2/shaders/ffx_fsr2_accumulate_pass.glsl b/thirdparty/amd-fsr2/shaders/ffx_fsr2_accumulate_pass.glsl
index 31d68292d4..2e98c8a6c5 100644
--- a/thirdparty/amd-fsr2/shaders/ffx_fsr2_accumulate_pass.glsl
+++ b/thirdparty/amd-fsr2/shaders/ffx_fsr2_accumulate_pass.glsl
@@ -35,7 +35,7 @@
#endif
#define FSR2_BIND_SRV_INTERNAL_UPSCALED 3
#define FSR2_BIND_SRV_LOCK_STATUS 4
-#define FSR2_BIND_SRV_INPUT_DEPTH_CLIP 5
+//#define FSR2_BIND_SRV_INPUT_DEPTH_CLIP 5
#define FSR2_BIND_SRV_PREPARED_INPUT_COLOR 6
#define FSR2_BIND_SRV_LUMA_INSTABILITY 7
#define FSR2_BIND_SRV_LANCZOS_LUT 8
@@ -52,6 +52,10 @@
#define FSR2_BIND_CB_FSR2 18
+#if FFX_FSR2_OPTION_GODOT_DERIVE_INVALID_MOTION_VECTORS
+#define FSR2_BIND_SRV_INPUT_DEPTH 5
+#endif
+
#include "ffx_fsr2_callbacks_glsl.h"
#include "ffx_fsr2_common.h"
#include "ffx_fsr2_sample.h"
diff --git a/thirdparty/amd-fsr2/shaders/ffx_fsr2_callbacks_glsl.h b/thirdparty/amd-fsr2/shaders/ffx_fsr2_callbacks_glsl.h
index 10da13fb81..b610037cc6 100644
--- a/thirdparty/amd-fsr2/shaders/ffx_fsr2_callbacks_glsl.h
+++ b/thirdparty/amd-fsr2/shaders/ffx_fsr2_callbacks_glsl.h
@@ -52,6 +52,9 @@
FfxFloat32 fDeltaTime;
FfxFloat32 fDynamicResChangeFactor;
FfxFloat32 fViewSpaceToMetersFactor;
+
+ FfxFloat32 fPad;
+ mat4 mReprojectionMatrix;
} cbFSR2;
#endif
@@ -317,7 +320,11 @@ FfxFloat32 LoadInputDepth(FfxInt32x2 iPxPos)
#if defined(FSR2_BIND_SRV_REACTIVE_MASK)
FfxFloat32 LoadReactiveMask(FfxInt32x2 iPxPos)
{
+#if FFX_FSR2_OPTION_GODOT_REACTIVE_MASK_CLAMP
+ return min(texelFetch(r_reactive_mask, FfxInt32x2(iPxPos), 0).r, 0.9f);
+#else
return texelFetch(r_reactive_mask, FfxInt32x2(iPxPos), 0).r;
+#endif
}
#endif
@@ -354,6 +361,16 @@ FfxFloat32x2 LoadInputMotionVector(FfxInt32x2 iPxDilatedMotionVectorPos)
{
FfxFloat32x2 fSrcMotionVector = texelFetch(r_input_motion_vectors, iPxDilatedMotionVectorPos, 0).xy;
+#if FFX_FSR2_OPTION_GODOT_DERIVE_INVALID_MOTION_VECTORS
+ bool bInvalidMotionVector = all(lessThanEqual(fSrcMotionVector, vec2(-1.0f, -1.0f)));
+ if (bInvalidMotionVector)
+ {
+ FfxFloat32 fSrcDepth = LoadInputDepth(iPxDilatedMotionVectorPos);
+ FfxFloat32x2 fUv = (iPxDilatedMotionVectorPos + FfxFloat32(0.5)) / RenderSize();
+ fSrcMotionVector = FFX_FSR2_OPTION_GODOT_DERIVE_INVALID_MOTION_VECTORS_FUNCTION(fUv, fSrcDepth, cbFSR2.mReprojectionMatrix);
+ }
+#endif
+
FfxFloat32x2 fUvMotionVector = fSrcMotionVector * MotionVectorScale();
#if FFX_FSR2_OPTION_JITTERED_MOTION_VECTORS
diff --git a/thirdparty/amd-fsr2/shaders/ffx_fsr2_tcr_autogen_pass.glsl b/thirdparty/amd-fsr2/shaders/ffx_fsr2_tcr_autogen_pass.glsl
index 7d6a66b8ac..5c042c332a 100644
--- a/thirdparty/amd-fsr2/shaders/ffx_fsr2_tcr_autogen_pass.glsl
+++ b/thirdparty/amd-fsr2/shaders/ffx_fsr2_tcr_autogen_pass.glsl
@@ -40,6 +40,10 @@
#define FSR2_BIND_CB_FSR2 11
#define FSR2_BIND_CB_REACTIVE 12
+#if FFX_FSR2_OPTION_GODOT_DERIVE_INVALID_MOTION_VECTORS
+#define FSR2_BIND_SRV_INPUT_DEPTH 13
+#endif
+
#include "ffx_fsr2_callbacks_glsl.h"
#include "ffx_fsr2_common.h"