UE5 HLSL 详细学习笔记

这里的POSITION是变量Position的语义，告诉寄存器，此变量的保存位置，通常语义用于着色器的输入和输出，以冒号“：”的方式进一步说明此变量，COLOR也类似

还有什么语义呢？

HLSL核心函数：

float3 Pixels = Texture2DSample(Tex, TexSampler, myUV);
for(int i = 0; i < n; i++) {
    Pixels += Texture2DSample(Tex, TexSampler, myUV + float2(offset, 0));
    Pixels += Texture2DSample(Tex, TexSampler, myUV + float2(-offset, 0));
    Pixels += Texture2DSample(Tex, TexSampler, myUV + float2(0, offset));
    Pixels += Texture2DSample(Tex, TexSampler, myUV + float2(0, -offset));
    Pixels += Texture2DSample(Tex, TexSampler, myUV + float2(offset, -offset));
    Pixels += Texture2DSample(Tex, TexSampler, myUV + float2(-offset, offset));
    Pixels += Texture2DSample(Tex, TexSampler, myUV + float2(offset, offset));
    Pixels += Texture2DSample(Tex, TexSampler, myUV + float2(-offset, -offset));
}
Pixels /= 8 + 1;

return Pixels;

模糊

float2 myUV = floor(uv * 20) / 20;
float3 Pixels = Texture2DSample(Tex, TexSampler, myUV);

return Pixels;

像素化

float2 PixelsUV = floor(UV * GridNumber) / GridNumber;
float X = frac(UV.x * GridNumber);
float Y = frac(UV.y * GridNumber);
if((X < (Outline / GridNumber) || X > 1 - (Outline / GridNumber)) || (Y < (Outline / GridNumber) || Y > 1 - (Outline / GridNumber))){
     return float3(0, 0 , 0);
}
float3 Pixels = Texture2DSample(Tex, TexSampler, PixelsUV);
float Offset = 1 / GridNumber;
for(int i = 0; i < AverBounds; i++){
    Pixels += Texture2DSample(Tex, TexSampler, PixelsUV + float2(i * Offset, 0));
    Pixels += Texture2DSample(Tex, TexSampler, PixelsUV + float2(-Offset * i, 0));
    Pixels += Texture2DSample(Tex, TexSampler, PixelsUV + float2(0, i * Offset));
    Pixels += Texture2DSample(Tex, TexSampler, PixelsUV + float2(0, -Offset * i));
    Pixels += Texture2DSample(Tex, TexSampler, PixelsUV + float2(Offset * i, Offset * i));
    Pixels += Texture2DSample(Tex, TexSampler, PixelsUV + float2(Offset * i, -Offset * i));
    Pixels += Texture2DSample(Tex, TexSampler, PixelsUV + float2(-Offset * i, Offset * i));
    Pixels += Texture2DSample(Tex, TexSampler, PixelsUV + float2(-Offset * i, -Offset * i));
}
Pixels /= 8 * AverBounds + 1;

return Pixels;

像素化 + 模糊

struct Ns{
    float Noise(float2 uv, float2 Random, float RandomSeed) {
        return frac(sin(dot(uv, Random)) * RandomSeed);
    }
};
Ns n;
uv = floor(uv * noiseNumber);
return n.Noise(uv, Random, RandomSeed) < 0.5 ? 1 : 0;

struct Ns{
    float Noise(float2 uv, float2 Random, float RandomSeed) {
        return frac(sin(dot(uv, Random)) * RandomSeed);
    }
};
Ns n;
uv = floor(uv * noiseNumber);
return n.Noise(uv, Random, RandomSeed);

噪点
原理主要是先对UV里面所有内容进行dot，一个是每个uv的长度和方向合在一起具有唯一性与一个点位进行点乘就有随机数，但是这个随机数有个问题就是相邻的可能颜色很像，而噪点是黑白随机分配，不是一部分聚集在一起呈黑色，其他呈白色

float accumdens = 0;
LocCamVec = normalize(mul(Parameters.CameraVector, (float3x3)LWCToFloat(GetPrimitiveData(Parameters).WorldToLocal)));
for(int i = 0; i < MaxSteps; i++){
    float cursample = PseudoVolumeTexture(Tex, TexSampler, saturate(CurPos) + sin(Time * CloudSpeed) *  CloudSpeedIntensity, XYFrames, NumFrames).r;
    accumdens += cursample * StepSize;
    CurPos += -LocCamVec * StepSize;
}
return accumdens;

RayMarching做体积效果

---

float3 RayOrigin = viewDir - WorldPos;
float3 RayStep = viewDir * StepSize;
for(int i = 0; i < 256; i++){
    float dist = length(RayOrigin - SphereCenter) - SphereRadius;
    if(dist < 0) {
        return float3(1, 0, 0);
    }
    RayOrigin += RayStep;
}
return float3(0, 0, 0);

---

首先reflect(i, n)函数是以i为入射向量，n为法线向量的反射向量    float3 Reflection = reflect(LightDir, normal);

float3 RayOrigin = 1 - (viewDir - WorldPos);
float3 LightDir = normalize(LightPos);
float3 RayStep = viewDir * -StepSize;
for(int i = 0; i < 256; i++){
    float3 normal = normalize(RayOrigin - SphereCenter);
    float Diffuse = max(dot(normal, LightDir), 0);
    float3 Reflection = reflect(LightDir, normal);
    float Specular = pow(max(dot(Reflection, -viewDir),0), 128);
    float dist = length(RayOrigin - SphereCenter) - SphereRadius;
    if(dist < 0.01) {
        return float3(1, 0, 0) * Diffuse + (Specular) * LightColor;
    }
    OpacityMask = 1;
    RayOrigin += RayStep;
}
OpacityMask = 0;
return float3(0, 0, 0);

拿到了反射向量，就计算反射和射向相机的向量的点积来计算当点的高光亮度，如果刚好射到相机，那么就高亮越亮，反之，越暗

---

struct NoiseRotate {
    float2 RotateUV(float2 uv, float Angle) {
        float2x2 RotateMatrix = float2x2(cos(Angle), sin(Angle), -sin(Angle), cos(Angle));
       return  mul(uv - 0.5, RotateMatrix) + 0.5;
    }
};

NoiseRotate NR;

UV = (UV - 0.5) * Size + 0.5;

float UsingAngle = pow(length(UV - 0.5), sin(time));
UsingAngle = sin(UsingAngle) * 10;
float noiseUV = sin(atan2(UV.x - 0.5, UV.y - 0.5)) * length(UV - 0.5);

float3 Image = Texture2DSample(Tex, TexSampler, NR.RotateUV(UV, UsingAngle + time));
return Image;

利用旋转矩阵实现RotateUV的函数

然后根据离中心位置的距离，旋转的角度不同，做出该效果

---

float3 rayStep = viewDir * -1;
float4 inputTex = Texture2DSample(texObject, texObjectSampler, uv);
for(int i = 0; i < 70; i++){
    if(inputTex.r > 0.1 && inputTex.g > 0.1 && inputTex.b > 0.1){
         return float3(i, 0, 0);
    }
    uv += rayStep * 0.02;
    inputTex = Texture2DSample(texObject, texObjectSampler, uv.xy);
}
return inputTex;

思想很简单，其实就是如果cameraDir越垂直于上表面，那么CameraVector的xy值越小，z值越大，越偏离Z轴垂直点，xy越大，那么uv + rayStep的rayStep也小，越下层的偏离的就越厉害（因为for循环会导致uv += rayStep叠加，i值越大，加的rayStep越大，偏移的也就越大，这样产生了视差的效果）