From 00d36110d528991ebeae4d45d28be3dbb206bd9e Mon Sep 17 00:00:00 2001
From: Mathias Agopian <mathias@google.com>
Date: Wed, 1 Sep 2021 14:25:21 -0700
Subject: [PATCH] Integrate optimized EASU.

filament cannot use fp16, so we make the necessary changes for the
float version. Note that we are using mediump float, which ends up
using 16 bits floats anyways.

This provides 78% reduction in execution time (i.e. 4.6x performance
improvement). We see the EASU go from 4.2ms to 0.9ms.
The quality is slightly impacted.
---
 RELEASE_NOTES.md                             |   2 +
 filament/CMakeLists.txt                      |   2 +
 filament/src/materials/fsr/ffx_fsr1_mobile.h | 152 +++++++++++--------
 filament/src/materials/fsr/fsr_easu.mat      |  12 +-
 4 files changed, 101 insertions(+), 67 deletions(-)

diff --git a/RELEASE_NOTES.md b/RELEASE_NOTES.md
index 0f2d24ea2b..acd6da5238 100644
--- a/RELEASE_NOTES.md
+++ b/RELEASE_NOTES.md
@@ -8,6 +8,8 @@ A new header is inserted each time a *tag* is created.
 - engine: New night adaptation API on `ColorGrading`. This API can be used to create an effect that
   that simulates color and brightness shits in human vision in low-light conditions.
 
+- engine: improved performance of AMD FidelityFX FSR1 by 4.6x, it now runs in about 2ms in 4K.
+
 ## v1.12.3
 
 - engine: Support AMD FidelityFX Super Resolution for dynamic resolution scaling
diff --git a/filament/CMakeLists.txt b/filament/CMakeLists.txt
index 45dbd33278..1af9efef6c 100644
--- a/filament/CMakeLists.txt
+++ b/filament/CMakeLists.txt
@@ -165,6 +165,7 @@ set(PRIVATE_HDRS
         src/fsr.h
         src/materials/fsr/ffx_a.h
         src/materials/fsr/ffx_fsr1.h
+        src/materials/fsr/ffx_fsr1_mobile.h
         src/upcast.h
 )
 
@@ -362,6 +363,7 @@ add_custom_command(
         OUTPUT "${MATERIAL_DIR}/fsr_easu.filamat"
         DEPENDS src/materials/fsr/ffx_a.h
         DEPENDS src/materials/fsr/ffx_fsr1.h
+        DEPENDS src/materials/fsr/ffx_fsr1_mobile.h
         APPEND
 )
 
diff --git a/filament/src/materials/fsr/ffx_fsr1_mobile.h b/filament/src/materials/fsr/ffx_fsr1_mobile.h
index 010e152a81..4851cffe50 100644
--- a/filament/src/materials/fsr/ffx_fsr1_mobile.h
+++ b/filament/src/materials/fsr/ffx_fsr1_mobile.h
@@ -6,13 +6,19 @@
 // -- FsrEasuSampleH should be implemented by calling shader, like following:
 //    half3 FsrEasuSampleH(float2 p) { return MyTex.SampleLevel(LinearSampler, p, 0).xyz; }
 //==============================================================================================================================
+
+#define rcp(x) (1.0/(x))
+#define rsqrt inversesqrt
+
+AF3 FsrEasuSampleF(AF2 p);
+
 void FsrEasuL(
-        out AH3 pix,
-        AF2 ip,
-        AF4 con0,
-        AF4 con1,
-        AF4 con2,
-        AF4 con3){
+        out AF3 pix,
+        highp AF2 ip,
+        highp AF4 con0,
+        highp AF4 con1,
+        highp AF4 con2,
+        highp AF4 con3){
     //------------------------------------------------------------------------------------------------------------------------------
     // Direction is the '+' diff.
     //    A
@@ -20,83 +26,97 @@ void FsrEasuL(
     //    E
     AF2 pp=(ip)*(con0.xy)+(con0.zw);
     AF2 tc=(pp+AF2_(0.5))*con1.xy;
-    AH3 sA=FsrEasuSampleH(tc-AF2(0, con1.y));
-    AH3 sB=FsrEasuSampleH(tc-AF2(con1.x, 0));
-    AH3 sC=FsrEasuSampleH(tc);
-    AH3 sD=FsrEasuSampleH(tc+AF2(con1.x, 0));
-    AH3 sE=FsrEasuSampleH(tc+AF2(0, con1.y));
-    AH1 lA=sA.r*AH1_(0.5)+sA.g;
-    AH1 lB=sB.r*AH1_(0.5)+sB.g;
-    AH1 lC=sC.r*AH1_(0.5)+sC.g;
-    AH1 lD=sD.r*AH1_(0.5)+sD.g;
-    AH1 lE=sE.r*AH1_(0.5)+sE.g;
+    AF3 sA=FsrEasuSampleF(tc-AF2(0, con1.y));
+    AF3 sB=FsrEasuSampleF(tc-AF2(con1.x, 0));
+    AF3 sC=FsrEasuSampleF(tc);
+    AF3 sD=FsrEasuSampleF(tc+AF2(con1.x, 0));
+    AF3 sE=FsrEasuSampleF(tc+AF2(0, con1.y));
+    AF1 lA=sA.r*AF1_(0.5)+sA.g;
+    AF1 lB=sB.r*AF1_(0.5)+sB.g;
+    AF1 lC=sC.r*AF1_(0.5)+sC.g;
+    AF1 lD=sD.r*AF1_(0.5)+sD.g;
+    AF1 lE=sE.r*AF1_(0.5)+sE.g;
     // Then takes magnitude from abs average of both sides of 'C'.
     // Length converts gradient reversal to 0, smoothly to non-reversal at 1, shaped, then adding horz and vert terms.
-    AH1 dc=lD-lC;
-    AH1 cb=lC-lB;
-    AH1 lenX=max(abs(dc),abs(cb));
-    lenX=ARcpH1(lenX);
-    AH1 dirX=lD-lB;
-    lenX=ASatH1(abs(dirX)*lenX);
+    AF1 dc=lD-lC;
+    AF1 cb=lC-lB;
+    AF1 lenX=max(abs(dc),abs(cb));
+    lenX=ARcpF1(lenX);
+    AF1 dirX=lD-lB;
+    lenX=ASatF1(abs(dirX)*lenX);
     lenX*=lenX;
     // Repeat for the y axis.
-    AH1 ec=lE-lC;
-    AH1 ca=lC-lA;
-    AH1 lenY=max(abs(ec),abs(ca));
-    lenY=ARcpH1(lenY);
-    AH1 dirY=lE-lA;
-    lenY=ASatH1(abs(dirY)*lenY);
-    AH1 len = lenY * lenY + lenX;
-    AH2 dir = AH2(dirX, dirY);
-    //------------------------------------------------------------------------------------------------------------------------------
-    AH2 dir2=dir*dir;
-    AH1 dirR=dir2.x+dir2.y;
-    if (dirR<AH1_(1.0/64.0)) {
+    AF1 ec=lE-lC;
+    AF1 ca=lC-lA;
+    AF1 lenY=max(abs(ec),abs(ca));
+    lenY=ARcpF1(lenY);
+    AF1 dirY=lE-lA;
+    lenY=ASatF1(abs(dirY)*lenY);
+    AF1 len = lenY * lenY + lenX;
+    AF2 dir = AF2(dirX, dirY);
+    //------------------------------------------------------------------------------------------------------------------------------ 
+    AF2 dir2=dir*dir;
+    AF1 dirR=dir2.x+dir2.y;
+    if (dirR<AF1_(1.0/64.0)) {
         pix = sC;
         return;
     }
     dirR=rsqrt(dirR);
-    dir*=AH2_(dirR);
-    len=len*AH1_(0.5);
+    dir*=AF2_(dirR);
+    len=len*AF1_(0.5);
     len*=len;
-    AH1 stretch=(dir.x*dir.x+dir.y*dir.y)*rcp(max(abs(dir.x),abs(dir.y)));
-    AH2 len2=AH2(AH1_(1.0)+(stretch-AH1_(1.0))*len,AH1_(1.0)+AH1_(-0.5)*len);
-    AH1 lob=AH1_(0.5)+AH1_((1.0/4.0-0.04)-0.5)*len;
-    AH1 clp=rcp(lob);
+    AF1 stretch=(dir.x*dir.x+dir.y*dir.y)*rcp(max(abs(dir.x),abs(dir.y)));
+    AF2 len2=AF2(AF1_(1.0)+(stretch-AF1_(1.0))*len,AF1_(1.0)+AF1_(-0.5)*len);
+    AF1 lob=AF1_(0.5)+AF1_((1.0/4.0-0.04)-0.5)*len;
+    AF1 clp=rcp(lob);
     //------------------------------------------------------------------------------------------------------------------------------
     AF2 fp=floor(pp);
     pp-=fp;
-    AH2 ppp=AH2(pp);
+    AF2 ppp=AF2(pp);
     AF2 p0=fp*(con1.xy)+(con1.zw);
     AF2 p1=p0+(con2.xy);
     AF2 p2=p0+(con2.zw);
     AF2 p3=p0+(con3.xy);
     p0.y-=con1.w; p3.y+=con1.w;
-    AH4 fgcbR=FsrEasuRH(p0);
-    AH4 fgcbG=FsrEasuGH(p0);
-    AH4 fgcbB=FsrEasuBH(p0);
-    AH4 ijfeR=FsrEasuRH(p1);
-    AH4 ijfeG=FsrEasuGH(p1);
-    AH4 ijfeB=FsrEasuBH(p1);
-    AH4 klhgR=FsrEasuRH(p2);
-    AH4 klhgG=FsrEasuGH(p2);
-    AH4 klhgB=FsrEasuBH(p2);
-    AH4 nokjR=FsrEasuRH(p3);
-    AH4 nokjG=FsrEasuGH(p3);
-    AH4 nokjB=FsrEasuBH(p3);
+    AF4 bczzR=FsrEasuRF(p0);
+    AF4 bczzG=FsrEasuGF(p0);
+    AF4 bczzB=FsrEasuBF(p0);
+    AF4 ijfeR=FsrEasuRF(p1);
+    AF4 ijfeG=FsrEasuGF(p1);
+    AF4 ijfeB=FsrEasuBF(p1);
+    AF4 klhgR=FsrEasuRF(p2);
+    AF4 klhgG=FsrEasuGF(p2);
+    AF4 klhgB=FsrEasuBF(p2);
+    AF4 zzonR=FsrEasuRF(p3);
+    AF4 zzonG=FsrEasuGF(p3);
+    AF4 zzonB=FsrEasuBF(p3);
     //------------------------------------------------------------------------------------------------------------------------------
     // This part is different for FP16, working pairs of taps at a time.
-    AH2 pR=AH2_(0.0);
-    AH2 pG=AH2_(0.0);
-    AH2 pB=AH2_(0.0);
-    AH2 pW=AH2_(0.0);
-    FsrEasuTapH(pR,pG,pB,pW,AH2( 1.0, 0.0)-ppp.xx,AH2(-1.0,-1.0)-ppp.yy,dir,len2,lob,clp,fgcbR.zw,fgcbG.zw,fgcbB.zw);
-    FsrEasuTapH(pR,pG,pB,pW,AH2(-1.0, 0.0)-ppp.xx,AH2( 1.0, 1.0)-ppp.yy,dir,len2,lob,clp,ijfeR.xy,ijfeG.xy,ijfeB.xy);
-    FsrEasuTapH(pR,pG,pB,pW,AH2( 0.0,-1.0)-ppp.xx,AH2( 0.0, 0.0)-ppp.yy,dir,len2,lob,clp,ijfeR.zw,ijfeG.zw,ijfeB.zw);
-    FsrEasuTapH(pR,pG,pB,pW,AH2( 1.0, 2.0)-ppp.xx,AH2( 1.0, 1.0)-ppp.yy,dir,len2,lob,clp,klhgR.xy,klhgG.xy,klhgB.xy);
-    FsrEasuTapH(pR,pG,pB,pW,AH2( 2.0, 1.0)-ppp.xx,AH2( 0.0, 0.0)-ppp.yy,dir,len2,lob,clp,klhgR.zw,klhgG.zw,klhgB.zw);
-    FsrEasuTapH(pR,pG,pB,pW,AH2( 0.0, 1.0)-ppp.xx,AH2( 2.0, 2.0)-ppp.yy,dir,len2,lob,clp,nokjR.xy,nokjG.xy,nokjB.xy);
-    AH3 aC=AH3(pR.x+pR.y,pG.x+pG.y,pB.x+pB.y);
-    AH1 aW=pW.x+pW.y;
+
+    AF3 min4=min(AMin3F3(AF3(ijfeR.z,ijfeG.z,ijfeB.z),AF3(klhgR.w,klhgG.w,klhgB.w),AF3(ijfeR.y,ijfeG.y,ijfeB.y)),
+            AF3(klhgR.x,klhgG.x,klhgB.x));
+    AF3 max4=max(AMax3F3(AF3(ijfeR.z,ijfeG.z,ijfeB.z),AF3(klhgR.w,klhgG.w,klhgB.w),AF3(ijfeR.y,ijfeG.y,ijfeB.y)),
+            AF3(klhgR.x,klhgG.x,klhgB.x));
+
+    AF3 aC=AF3_(0.0);
+    AF1 aW=AF1_(0.0);
+    FsrEasuTapF(aC,aW,AF2( 0.0,-1.0)-pp,dir,len2,lob,clp,AF3(bczzR.x,bczzG.x,bczzB.x)); // b
+    FsrEasuTapF(aC,aW,AF2( 1.0,-1.0)-pp,dir,len2,lob,clp,AF3(bczzR.y,bczzG.y,bczzB.y)); // c
+    FsrEasuTapF(aC,aW,AF2(-1.0, 1.0)-pp,dir,len2,lob,clp,AF3(ijfeR.x,ijfeG.x,ijfeB.x)); // i
+    FsrEasuTapF(aC,aW,AF2( 0.0, 1.0)-pp,dir,len2,lob,clp,AF3(ijfeR.y,ijfeG.y,ijfeB.y)); // j
+    FsrEasuTapF(aC,aW,AF2( 0.0, 0.0)-pp,dir,len2,lob,clp,AF3(ijfeR.z,ijfeG.z,ijfeB.z)); // f
+    FsrEasuTapF(aC,aW,AF2(-1.0, 0.0)-pp,dir,len2,lob,clp,AF3(ijfeR.w,ijfeG.w,ijfeB.w)); // e
+    FsrEasuTapF(aC,aW,AF2( 1.0, 1.0)-pp,dir,len2,lob,clp,AF3(klhgR.x,klhgG.x,klhgB.x)); // k
+    FsrEasuTapF(aC,aW,AF2( 2.0, 1.0)-pp,dir,len2,lob,clp,AF3(klhgR.y,klhgG.y,klhgB.y)); // l
+    FsrEasuTapF(aC,aW,AF2( 2.0, 0.0)-pp,dir,len2,lob,clp,AF3(klhgR.z,klhgG.z,klhgB.z)); // h
+    FsrEasuTapF(aC,aW,AF2( 1.0, 0.0)-pp,dir,len2,lob,clp,AF3(klhgR.w,klhgG.w,klhgB.w)); // g
+    FsrEasuTapF(aC,aW,AF2( 1.0, 2.0)-pp,dir,len2,lob,clp,AF3(zzonR.z,zzonG.z,zzonB.z)); // o
+    FsrEasuTapF(aC,aW,AF2( 0.0, 2.0)-pp,dir,len2,lob,clp,AF3(zzonR.w,zzonG.w,zzonB.w)); // n
     //------------------------------------------------------------------------------------------------------------------------------
-    pix=aC*AH3_(ARcpH1(aW));} 
\ No newline at end of file
+
+#if defined(FILAMENT_FSR_DERINGING)
+    pix=min(max4,max(min4,aC*AF3_(ARcpF1(aW))));
+#else
+    pix=aC*AF3_(ARcpF1(aW));
+#endif
+}
diff --git a/filament/src/materials/fsr/fsr_easu.mat b/filament/src/materials/fsr/fsr_easu.mat
index afd0a27c1a..5341e62340 100644
--- a/filament/src/materials/fsr/fsr_easu.mat
+++ b/filament/src/materials/fsr/fsr_easu.mat
@@ -64,9 +64,13 @@ fragment {
     #define A_GPU 1
     #define A_GLSL 1
     #include "ffx_a.h"
+
     #define FSR_EASU_F 1
     #include "ffx_fsr1.h"
 
+    #define FILAMENT_FSR_DERINGING
+    #include "ffx_fsr1_mobile.h"
+
 #if defined(FILAMENT_HAS_FEATURE_TEXTURE_GATHER)
     #define gather textureGather
 #else
@@ -93,11 +97,17 @@ fragment {
         return gather(materialParams_colorNearest, p, 2);
     }
 
+    AF3 FsrEasuSampleF(AF2 p) {
+        return textureLod(materialParams_color, p, 0.0).rgb;
+    }
+
     void postProcess(inout PostProcessInputs postProcess) {
         highp vec2 p = floor(variable_vertex.xy);
 
         AF3 pix;
-        FsrEasuF(pix, p,
+
+        // just use FsrEasuF() for the original full quality EASU
+        FsrEasuL(pix, p,
             materialParams.EasuCon0,
             materialParams.EasuCon1,
             materialParams.EasuCon2,