关于C＃：HLSL修改像素着色器中的深度

HLSL modify depth in pixel shader

我需要渲染从外部获得的图像(具有深度)。我可以构造两个纹理，并将它们毫无问题地传递到着色器中(我可以验证在像素着色器中采样的值是正确的)。

我的HLSL如下所示：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49

// image texture
Texture2D m_TextureColor : register(t0);

// depth texture with values [0..1]
Texture2D<float> m_TextureDepth : register(t1);

// sampler to forbid linear filtering since we're dealing with pixels
SamplerState m_TextureSampler { Filter = MIN_MAG_MIP_POINT; };

struct VS_IN
{
float4 position : POSITION;
float2 texcoord : TEXCOORD;
};

struct VS_OUT
{
float4 position : SV_POSITION;
float2 texcoord : TEXCOORD0;
};

struct PS_OUT
{
float4 color : COLOR0;
float depth : DEPTH0;
};

VS_OUT VS(VS_IN input)
{
VS_OUT output = (VS_OUT)0;

output.position = input.position;
output.texcoord = input.texcoord;

return output;
}

PS_OUT PS(VS_OUT input) : SV_Target
{
PS_OUT output = (PS_OUT)0;

output.color = m_TextureColor.SampleLevel(m_TextureSampler, input.texcoord, 0);

// I want to modify depth of the pixel,
// but it looks like it has no effect on depth no matter what I set here
output.depth = m_TextureDepth.SampleLevel(m_TextureSampler, input.texcoord, 0);

return output;
}

我从那些(用PrimitiveTopology.TriangleStrip)构造顶点缓冲区的地方，其中第一个参数Vector4是位置，第二个参数Vector2是纹理坐标：

1
2
3
4
5
6
7

new[]
{
new Vertex(new Vector4(-1, -1, 0.5f, 1), new Vector2(0, 1)),
new Vertex(new Vector4(-1, 1, 0.5f, 1), new Vector2(0, 0)),
new Vertex(new Vector4(1, -1, 0.5f, 1), new Vector2(1, 1)),
new Vertex(new Vector4(1, 1, 0.5f, 1), new Vector2(1, 0)),
}

一切正常：我正在查看图像，可以从深度纹理中采样深度并从中构造出一些视觉效果(这就是我可以验证的方式
我采样的深度值是正确的)。但是我无法弄清楚如何修改像素的深度，以便在进行深度测试时可以正确地吃掉它。因为目前，这完全取决于我将哪种z值设置为顶点位置。

这就是我设置DirectX11的方式(我正在使用SharpDX和C#)：

1
2
3
4
5
6
7
8
9
10
11
12

var swapChainDescription = new SwapChainDescription
{
BufferCount = 1,
ModeDescription = new ModeDescription(bufferSize.Width, bufferSize.Height, new Rational(60, 1), Format.R8G8B8A8_UNorm),
IsWindowed = true,
OutputHandle = HostHandle,
SampleDescription = new SampleDescription(1, 0),
SwapEffect = SwapEffect.Discard,
Usage = Usage.RenderTargetOutput,
};
var swapChainFlags = DeviceCreationFlags.None | DeviceCreationFlags.BgraSupport;
SharpDX.Direct3D11.Device.CreateWithSwapChain(DriverType.Hardware, swapChainFlags, swapChainDescription, out var device, out var swapchain);

设置后缓冲区和深度/模板缓冲区：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25

// color buffer
using (var textureColor = SwapChain.GetBackBuffer<Texture2D>(0))
{
TextureColorResourceView = new RenderTargetView(Device, textureColor);
}

// depth buffer
using (var textureDepth = new Texture2D(Device, new Texture2DDescription
{
Format = Format.D32_Float,
ArraySize = 1,
MipLevels = 1,
Width = BufferSize.Width,
Height = BufferSize.Height,
SampleDescription = new SampleDescription(1, 0),
Usage = ResourceUsage.Default,
BindFlags = BindFlags.DepthStencil,
CpuAccessFlags = CpuAccessFlags.None,
OptionFlags = ResourceOptionFlags.None
}))
{
TextureDepthResourceView = new DepthStencilView(Device, textureDepth);
}

DeviceContext.OutputMerger.SetTargets(TextureDepthResourceView, TextureColorResourceView);

准备深度模具状态：

1
2
3
4
5

var description = DepthStencilStateDescription.Default();
description.DepthComparison = Comparison.LessEqual;
description.IsDepthEnabled = true;
description.DepthWriteMask = DepthWriteMask.All;
DepthState = new DepthStencilState(Device, description);

并使用它：

1	DeviceContext.OutputMerger.SetDepthStencilState(DepthState);

这是构造发送到着色器的颜色/深度纹理的方法：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

public static (ShaderResourceView resource, Texture2D texture) CreateTextureDynamic(this Device device, System.Drawing.Size size, Format format)
{
var textureDesc = new Texture2DDescription
{
MipLevels = 1,
Format = format,
Width = size.Width,
Height = size.Height,
ArraySize = 1,
BindFlags = BindFlags.ShaderResource,
Usage = ResourceUsage.Dynamic,
SampleDescription = new SampleDescription(1, 0),
CpuAccessFlags = CpuAccessFlags.Write,
};

var texture = new Texture2D(device, textureDesc);
return (new ShaderResourceView(device, texture), texture);
}

另外，由于我需要经常更新它们：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15

public static void UpdateResource(this Texture2D texture, int[] buffer, System.Drawing.Size size)
{
var dataBox = texture.Device.ImmediateContext.MapSubresource(texture, 0, MapMode.WriteDiscard, MapFlags.None, out var dataStream);
Parallel.For(0, size.Height, rowIndex => Marshal.Copy(buffer, size.Width * rowIndex, dataBox.DataPointer + dataBox.RowPitch * rowIndex, size.Width));
dataStream.Dispose();
texture.Device.ImmediateContext.UnmapSubresource(texture, 0);
}

public static void UpdateResource(this Texture2D texture, float[] buffer, System.Drawing.Size size)
{
var dataBox = texture.Device.ImmediateContext.MapSubresource(texture, 0, MapMode.WriteDiscard, MapFlags.None, out var dataStream);
Parallel.For(0, size.Height, rowIndex => Marshal.Copy(buffer, size.Width * rowIndex, dataBox.DataPointer + dataBox.RowPitch * rowIndex, size.Width));
dataStream.Dispose();
texture.Device.ImmediateContext.UnmapSubresource(texture, 0);
}

我对此也进行了很多搜索，发现了类似的帖子：https://www.gamedev.net/forums/topic/573961-how-to-set-depth-value-in-pixel-shader/但无法设法解决了这一点。

提前致谢！

要写入深度缓冲区，您需要定位SV_Depth系统值语义。因此，您的像素着色器输出结构将更像以下内容：

1
2
3
4
5

struct PS_OUT
{
float4 color : SV_Target;
float depth : SV_Depth;
};

并且着色器不会像您的示例中那样指定SV_Target(SV_输出在结构中定义)。因此，它看起来像：

1
2
3
4
5
6
7
8
9
10
11
12

PS_OUT PS(VS_OUT input)
{
PS_OUT output = (PS_OUT)0;

output.color = m_TextureColor.SampleLevel(m_TextureSampler, input.texcoord, 0);

// Now that output.depth is defined with SV_Depth, and you have depth-write enabled,
// this should write to the depth buffer.
output.depth = m_TextureDepth.SampleLevel(m_TextureSampler, input.texcoord, 0);

return output;
}

请注意，在显式写入深度(特别是在AMD硬件上)时，可能会导致性能下降，因为这会绕过其早期的深度硬件优化。将来所有使用该深度缓冲区的绘图调用都将禁用Early-Z优化，因此，通常最好尽可能晚地执行深度写入操作。