Optimize and cleanup shaders

Change-Id: Ia27680e20f73ffe645d5a336fad7024e67a8389b
Reviewed-by: Janne Kangas <janne.kangas@qt.io>
Reviewed-by: Tomi Korpipää <tomi.korpipaa@qt.io>

1 files changed, 67 insertions, 139 deletions

diff --git a/res/effectlib/sampleProbe.glsllib b/res/effectlib/sampleProbe.glsllib
index 361aad5..4846ee3 100644
--- a/res/effectlib/sampleProbe.glsllib
+++ b/res/effectlib/sampleProbe.glsllib
@@ -1,7 +1,7 @@
 /****************************************************************************
 **
 ** Copyright (C) 2014 NVIDIA Corporation.
-** Copyright (C) 2017 The Qt Company Ltd.
+** Copyright (C) 2019 The Qt Company Ltd.
 ** Contact: https://www.qt.io/licensing/
 **
 ** This file is part of Qt 3D Studio.
@@ -31,6 +31,8 @@
 #ifndef SAMPLE_PROBE_GLSLLIB
 #define SAMPLE_PROBE_GLSLLIB 1
 
+#include "sampleProbePre.glsllib"
+
 #define USE_RGBE
 
 uniform sampler2D light_probe;
@@ -49,37 +51,37 @@ uniform vec4 light_probe_opts;
 
 float noise1d(vec2 n)
 {
-    return 0.5 + 0.5 * fract(sin(dot(n.xy, vec2(12.9898, 78.233)))* 43758.5453);
+    return 0.5 + 0.5 * fract(sin(dot(n.xy, vec2(12.9898, 78.233))) * 43758.5453);
 }
 
-mat3 orthoNormalize( in mat3 tanFrame )
+mat3 orthoNormalize(in mat3 tanFrame)
 {
    mat3 outMat;
-   outMat[0] = normalize( cross( tanFrame[1], tanFrame[2] ) );
-   outMat[1] = normalize( cross( tanFrame[2], outMat[0] ) );
+   outMat[0] = normalize(cross(tanFrame[1], tanFrame[2]));
+   outMat[1] = normalize(cross(tanFrame[2], outMat[0]));
    outMat[2] = tanFrame[2];
 
    return outMat;
 }
 
-mat3 tangentFrame( vec3 N, vec3 p )
+mat3 tangentFrame(vec3 N, vec3 p)
 {
     // get edge vectors of the pixel triangle
-    vec3 dp1 = dFdx( p );
-    vec3 dp2 = dFdy( p );
+    vec3 dp1 = dFdx(p);
+    vec3 dp2 = dFdy(p);
     // solve the linear system
-    vec3 dp2perp = cross( dp2, N );
-    vec3 dp1perp = cross( N, dp1 );
+    vec3 dp2perp = cross(dp2, N);
+    vec3 dp1perp = cross(N, dp1);
     vec3 T = normalize(dp1perp);
     vec3 B = normalize(dp2perp);
-    return mat3( T , B , N );
+    return mat3(T, B, N);
 }
 
-vec2 transformSample( vec2 origUV, vec4 probeRot, vec2 probeOfs )
+vec2 transformSample(vec2 origUV, vec4 probeRot, vec2 probeOfs)
 {
     vec2 retUV;
-    retUV.x = dot( vec3(origUV, 1.0), vec3(probeRot.xy, probeOfs.x) );
-    retUV.y = dot( vec3(origUV, 1.0), vec3(probeRot.zw, probeOfs.y) );
+    retUV.x = dot(vec3(origUV, 1.0), vec3(probeRot.xy, probeOfs.x));
+    retUV.y = dot(vec3(origUV, 1.0), vec3(probeRot.zw, probeOfs.y));
     return retUV;
 }
 
@@ -96,7 +98,7 @@ vec3 textureProbe(sampler2D lightProbe, vec2 coord, float lod)
 // This is broken out into its own routine so that if we get some other
 // format image than a lat-long, then we can account for that by changing
 // the code here alone.
-vec2 getProbeSampleUV( vec3 smpDir, vec4 probeRot, vec2 probeOfs )
+vec2 getProbeSampleUV(vec3 smpDir, vec4 probeRot, vec2 probeOfs)
 {
     vec2 smpUV;
 
@@ -118,86 +120,83 @@ vec2 getProbeSampleUV( vec3 smpDir, vec4 probeRot, vec2 probeOfs )
     return smpUV;
 }
 
-vec4 getTopLayerSample( vec3 inDir, float lodShift, vec3 lodOffsets )
+vec4 getTopLayerSample(vec3 inDir, float lodShift, vec3 lodOffsets)
 {
 #if QT3DS_ENABLE_LIGHT_PROBE_2
-    if ( light_probe2_props.w < 0.5 )
+    if (light_probe2_props.w < 0.5)
         return vec4(0.0, 0.0, 0.0, 0.0);
 
-    vec2 smpUV = getProbeSampleUV( inDir, vec4(1.0, 0.0, 0.0, 1.0), light_probe_props.xy );
+    vec2 smpUV = getProbeSampleUV(inDir, vec4(1.0, 0.0, 0.0, 1.0), light_probe_props.xy);
     smpUV.x -= 0.5;
     smpUV.x *= light_probe2_props.x;
     smpUV.x += light_probe2_props.y;
 
-    vec4 retVal = 0.4 * textureLod( light_probe2, smpUV , lodShift );
-    retVal += 0.2 * textureLod( light_probe2, smpUV , lodShift+lodOffsets.x );
-    retVal += 0.3 * textureLod( light_probe2, smpUV , lodShift+lodOffsets.y );
-    retVal += 0.1 * textureLod( light_probe2, smpUV , lodShift+lodOffsets.z );
+    vec4 retVal = 0.4 * textureLod(light_probe2, smpUV , lodShift);
+    retVal += 0.2 * textureLod(light_probe2, smpUV, lodShift+lodOffsets.x);
+    retVal += 0.3 * textureLod(light_probe2, smpUV, lodShift+lodOffsets.y);
+    retVal += 0.1 * textureLod(light_probe2, smpUV, lodShift+lodOffsets.z);
     return retVal;
 #else
     return vec4(0.0, 0.0, 0.0, 0.0);
 #endif
 }
 
-vec3 getProbeSample( vec3 smpDir, float lodShift, vec3 normal )
+vec3 getProbeSample(vec3 smpDir, float lodShift, vec3 normal)
 {
-    vec2 smpUV = getProbeSampleUV( smpDir, light_probe_rotation, light_probe_offset.xy );
-    return textureProbe( light_probe, smpUV , lodShift );
+    vec2 smpUV = getProbeSampleUV(smpDir, light_probe_rotation, light_probe_offset.xy);
+    return textureProbe(light_probe, smpUV , lodShift);
 }
 
-vec3 getProbeWeightedSample( vec3 smpDir, float lodShift, float roughness, vec3 normal )
+vec3 getProbeWeightedSample(vec3 smpDir, float lodShift, float roughness, vec3 normal)
 {
     // This gives us a weighted sum that approximates the total filter support
     // of the full-blown convolution.
-    vec2 smpUV = getProbeSampleUV( smpDir, light_probe_rotation, light_probe_offset.xy );
+    vec2 smpUV = getProbeSampleUV(smpDir, light_probe_rotation, light_probe_offset.xy);
     float wt = 1.0;
 
 #if QT3DS_ENABLE_IBL_FOV
     wt = min(wt, smoothstep(roughness * -0.25, roughness * 0.25, smpUV.x));
     wt = min(wt, smoothstep(roughness * -0.25, roughness * 0.25, smpUV.y));
-    wt = min(wt, 1.0 - smoothstep(1.0 - roughness*0.25, 1.0 + roughness*0.25, smpUV.x));
-    wt = min(wt, 1.0 - smoothstep(1.0 - roughness*0.25, 1.0 + roughness*0.25, smpUV.y));
+    wt = min(wt, 1.0 - smoothstep(1.0 - roughness * 0.25, 1.0 + roughness * 0.25, smpUV.x));
+    wt = min(wt, 1.0 - smoothstep(1.0 - roughness * 0.25, 1.0 + roughness * 0.25, smpUV.y));
 #endif
 
     vec3 lodOffsets;
     lodOffsets.x = mix(-2.0, -0.70710678, roughness);
-    lodOffsets.y = min( 2.0 * smoothstep(0.0, 0.1, roughness), 2.0 - 1.29289 * smoothstep(0.1, 1.0, roughness) );
-    lodOffsets.z = min( 6.0 * smoothstep(0.0, 0.1, roughness), 6.0 - 4.585786 * smoothstep(0.1, 1.0, roughness) );
+    lodOffsets.y = min(2.0 * smoothstep(0.0, 0.1, roughness), 2.0 - 1.29289 * smoothstep(0.1, 1.0, roughness));
+    lodOffsets.z = min(6.0 * smoothstep(0.0, 0.1, roughness), 6.0 - 4.585786 * smoothstep(0.1, 1.0, roughness));
 
-    ivec2 iSize = textureSize(light_probe, 0);
-    vec3 ddx = dFdx( smpDir ) * float(iSize.x);
-    vec3 ddy = dFdy( smpDir ) * float(iSize.y);
-//    vec2 ddxUV = dFdx( smpUV ) * float(iSize.x);
-//    vec2 ddyUV = dFdy( smpUV ) * float(iSize.y);
+    ivec2 iSize = lightProbeSize(light_probe, 0);
+    vec3 ddx = dFdx(smpDir) * float(iSize.x);
+    vec3 ddy = dFdy(smpDir) * float(iSize.y);
 
     vec2 deriv;
-    deriv.x = max( dot(ddx, ddx), dot(ddy, ddy) );
-//    deriv.y = max( dot(ddxUV, ddxUV), dot(ddyUV, ddyUV) );
-    deriv = clamp( deriv, vec2(1.0), vec2(iSize.x * iSize.y) );
-    vec2 lodBound = 0.5 * log2( deriv ) - vec2(1.0);
+    deriv.x = max(dot(ddx, ddx), dot(ddy, ddy));
+
+    deriv = clamp(deriv, vec2(1.0), vec2(iSize.x * iSize.y));
+    vec2 lodBound = 0.5 * log2(deriv) - vec2(1.0);
 
-//    float minLod = 0.5 * (lodBound.x + lodBound.y);
     float minLod = lodBound.x;
-    float maxLod = log2( max(float(iSize.x), float(iSize.y)) );
-    minLod = clamp( minLod / maxLod, 0.0, 1.0 );
+    float maxLod = log2(max(float(iSize.x), float(iSize.y)));
+    minLod = clamp(minLod / maxLod, 0.0, 1.0);
     minLod *= minLod * maxLod;
 
-    lodShift = max( lodShift, minLod );
+    lodShift = max(lodShift, minLod);
 
-    vec3 retVal = 0.4 * textureProbe( light_probe, smpUV , lodShift );
-    retVal += 0.2 * textureProbe( light_probe, smpUV , max(minLod, lodShift+lodOffsets.x) );
-    retVal += 0.3 * textureProbe( light_probe, smpUV , lodShift+lodOffsets.y );
-    retVal += 0.1 * textureProbe( light_probe, smpUV , lodShift+lodOffsets.z );
+    vec3 retVal = 0.4 * textureProbe(light_probe, smpUV , lodShift);
+    retVal += 0.2 * textureProbe(light_probe, smpUV , max(minLod, lodShift+lodOffsets.x));
+    retVal += 0.3 * textureProbe(light_probe, smpUV , lodShift+lodOffsets.y);
+    retVal += 0.1 * textureProbe(light_probe, smpUV , lodShift+lodOffsets.z);
 
 #if QT3DS_ENABLE_LIGHT_PROBE_2
-    vec4 topSmp = getTopLayerSample( smpDir, lodShift, lodOffsets );
-    vec3 tempVal = mix( retVal, topSmp.xyz, topSmp.w );
-    retVal = mix( retVal, tempVal, light_probe2_props.z );
+    vec4 topSmp = getTopLayerSample(smpDir, lodShift, lodOffsets);
+    vec3 tempVal = mix(retVal, topSmp.xyz, topSmp.w);
+    retVal = mix(retVal, tempVal, light_probe2_props.z);
 #endif
 
     if (light_probe_props.z > -1.0) {
         float ctr = 0.5 + 0.5 * light_probe_props.z;
-        float vertWt = smoothstep(ctr-roughness*0.25, ctr+roughness*0.25, smpUV.y);
+        float vertWt = smoothstep(ctr-roughness * 0.25, ctr+roughness * 0.25, smpUV.y);
         float wtScaled = mix(1.0, vertWt, light_probe_props.z + 1.0);
         retVal *= wtScaled;
     }
@@ -205,111 +204,40 @@ vec3 getProbeWeightedSample( vec3 smpDir, float lodShift, float roughness, vec3
     return retVal * wt;
 }
 
-vec3 getProbeAnisoSample( vec3 smpDir, float roughU, float roughV, mat3 tanFrame )
-{
-    float minRough = min(roughU, roughV);
-    float maxRough = max(roughU, roughV);
-
-    float lodMin = log2( (minRough*3.0 + maxRough)*0.25 ) + (light_probe_offset.w - 2.0);
-
-    float ratio = clamp( maxRough / minRough, 1.01, 27.0);
-    vec2 texSize = vec2( textureSize( light_probe, int(floor( lodMin )) ) );
-    texSize = mix( texSize, texSize * 0.5, fract(lodMin) );
-
-    // Boundary of 1.0..9.0 is just to keep the number of samples to within a
-    // reasonable number of samples in the filter.  Similarly, with the clamping
-    // of the ratio to a max of 27.0 is just to prevent the step size in the filter
-    // to be no bigger than 3 texels (beyond which, there are some artifacts at high
-    // roughness, aka low texture res).
-    float stepFig = clamp(floor( ratio ), 1.0, 9.0);
-
-    // numSteps is half the number of samples we need to take, which makes it
-    // the number of steps to take on each side.
-    int numSteps = int( floor(stepFig * 0.5) );
-
-    vec2 smpUV = getProbeSampleUV( smpDir, light_probe_rotation, light_probe_offset.xy );
-    vec4 result = vec4(0.0);
-
-    vec3 smpDirOfs = (maxRough == roughU) ? 0.01 * tanFrame[0] : 0.01 * tanFrame[1];
-    vec2 stepPos = getProbeSampleUV(normalize(smpDir + smpDirOfs), light_probe_rotation, light_probe_offset.xy);
-    vec2 stepNeg = getProbeSampleUV(normalize(smpDir - smpDirOfs), light_probe_rotation, light_probe_offset.xy);
-    stepPos -= smpUV;     stepNeg -= smpUV;
-    stepPos *= texSize;   stepNeg *= texSize;
-
-    // This ensures that we step along a size that makes sense even if one of the two
-    // sammpling directions wraps around the edges of the IBL texture.
-    smpDirOfs /= min( length(stepPos), length(stepNeg) );
-    smpDirOfs *= ratio / stepFig;
-
-    float sigma = mix(0.0, 2.0, ratio / 27.0);
-    sigma *= sigma;
-
-    float wt = (1.0 / (ratio - 1.0)) + 1.0;
-    result.xyz += wt * getProbeWeightedSample( smpDir, lodMin, minRough, tanFrame[2] );
-    result.w += wt;
-    for (int i = 0; i < numSteps; ++i)
-    {
-        wt = sigma / (sigma + float(i * i));
-        vec2 uv0 = getProbeSampleUV(normalize(smpDir + smpDirOfs * float(i)), light_probe_rotation, light_probe_offset.xy);
-        vec2 uv1 = getProbeSampleUV(normalize(smpDir - smpDirOfs * float(i)), light_probe_rotation, light_probe_offset.xy);
-        result.xyz += wt * textureProbe( light_probe, uv0 , lodMin );
-        result.w += wt;
-        result.xyz += wt * textureProbe( light_probe, uv1 , lodMin );
-        result.w += wt;
-    }
-
-    result /= result.w;
-    return result.xyz;
-}
-
-vec4 sampleDiffuse( mat3 tanFrame )
+vec4 sampleDiffuse(mat3 tanFrame)
 {
-    if ( light_probe_props.w < 0.005 )
-        return vec4( 0.0 );
+    if (light_probe_props.w < 0.005)
+        return vec4(0.0);
 
-    return vec4( light_probe_props.w * getProbeWeightedSample( tanFrame[2], light_probe_offset.w - 2.65149613, 1.0, tanFrame[2] ), 1.0 );
+    return vec4(light_probe_props.w * getProbeWeightedSample(tanFrame[2], light_probe_offset.w - 2.65149613, 1.0, tanFrame[2]), 1.0 );
 }
 
-vec4 sampleDiffuseCustomMaterial( vec3 normal, vec3 worldPos, float aoFactor  )
+vec4 sampleDiffuseCustomMaterial(vec3 normal, vec3 worldPos, float aoFactor)
 {
-
-    mat3 tanFrame = tangentFrame( normal, worldPos );
-    return sampleDiffuse( tanFrame );
+    mat3 tanFrame = tangentFrame(normal, worldPos);
+    return sampleDiffuse(tanFrame);
 }
 
-vec4 sampleGlossyAniso( mat3 tanFrame, vec3 viewDir, float roughU, float roughV )
+vec4 sampleGlossy(mat3 tanFrame, vec3 viewDir, float roughness)
 {
-    if ( light_probe_props.w < 0.005 )
-        return vec4( 0.0 );
+    if (light_probe_props.w < 0.005)
+        return vec4(0.0);
 
-    float sigmaU = smoothstep( 0.0, 1.0, clamp(roughU, 0.0001, 1.0) );
-    float sigmaV = smoothstep( 0.0, 1.0, clamp(roughV, 0.0001, 1.0) );
+    float sigma = smoothstep(0.0, 1.0, clamp(roughness, 0.0001, 1.0));
     vec3 ret = vec3(0, 0, 0);
 
-    vec3 smpDir = reflect( -viewDir, tanFrame[2] );
-    float sigma = sqrt(sigmaU * sigmaV);
+    vec3 smpDir = reflect(-viewDir, tanFrame[2]);
 
     // Compute the Geometric occlusion/self-shadowing term
-    float NdotL = clamp( dot( smpDir, tanFrame[2] ), 0.0, 0.999995);
+    float NdotL = clamp(dot(smpDir, tanFrame[2]), 0.0, 0.999995);
     float k = sigma * 0.31830988618;    // roughness / pi
-    float Gl = clamp( (NdotL / (NdotL*(1.0-k) + k) + (1.0 - k*k)) * 0.5, 0.0, 1.0 );
+    float Gl = clamp((NdotL / (NdotL*(1.0-k) + k) + (1.0 - k*k)) * 0.5, 0.0, 1.0 );
 
     vec3 outColor;
 
-    outColor = getProbeAnisoSample( smpDir, sigmaU, sigmaV, tanFrame );
-
-    return vec4( light_probe_props.w * Gl * outColor, 1.0 );
-}
+    outColor = getProbeSample(smpDir, sigma, tanFrame[2]);
 
-vec4 sampleGlossy( mat3 tanFrame, vec3 viewDir, float roughness )
-{
-    return sampleGlossyAniso( tanFrame, viewDir, roughness, roughness );
-}
-
-vec4 sampleGlossyCustomMaterial( vec3 normal, vec3 worldPos, vec3 viewDir, float roughness )
-{
-    mat3 tanFrame = tangentFrame( normal, worldPos );
-    return sampleGlossy( tanFrame, viewDir, roughness );
+    return vec4(light_probe_props.w * Gl * outColor, 1.0);
 }
 
 #endif