C# OpenGL-几何体着色器阴影映射过程

我正在使用方差阴影贴图计算许多点光源的阴影。立方体贴图的所有6个面都使用几何体着色器在一次过程中进行渲染，这对每个光源重复，整个批次存储在立方体贴图阵列中。这一切都运行良好，16灯在60帧/秒没有问题

为了进一步优化，我尝试将整个过程移动到一个几何体着色器过程，结果只达到硬件的113个顶点输出限制。出于好奇，我决定只渲染4个灯光（72个发射顶点），令我惊讶的是，它降到了24fps

那么，为什么16个灯光和16个渲染过程的性能要比单个过程中的4个灯光好得多呢

代码基本相同

#version 400 core

layout(triangles) in;
layout (triangle_strip, max_vertices=18) out;

uniform int lightID;
out vec4 frag_position;

uniform mat4 projectionMatrix;
uniform mat4 shadowTransforms[6];

void main()
{   
    for(int face = 0; face < 6; face++)
    {
        gl_Layer = face + (lightID * 6);

        for(int i=0; i<3; i++)
        {
            frag_position = shadowTransforms[face] * gl_in[i].gl_Position;
            gl_Position = projectionMatrix * shadowTransforms[face] * gl_in[i].gl_Position;

            EmitVertex();
        }
        EndPrimitive();
    }
}

#版本400核心
平面布置图（三角形）；
布局（三角形带，最大顶点=18）向外；
统一int-lightID；
输出vec4 FRAGU位置；
均匀mat4投影矩阵；
一致mat4阴影变换[6]；
void main（）
{   
用于（int face=0；face<6；face++）
{
gl_层=面+（lightID*6）；
对于（int i=0；i我猜第二种形式的并行代码执行的机会更少。几何体着色器的第一个版本生成18个顶点，必须执行4次，但这4次执行可以并行运行。第二个版本一个接一个地生成72个顶点。
啊，我明白了，这在总体上超过了draw调用艾尔毫秒。
#version 400 core

layout(triangles) in;
layout (triangle_strip, max_vertices=72) out;

out vec4 frag_position;

uniform mat4 projectionMatrix;
uniform mat4 shadowTransforms[24];

void main()
{   
    for (int lightSource = 0; lightSource < 4; lightSource++)
    {
        for(int face = 0; face < 6; face++)
        {
            gl_Layer = face + (lightSource * 6);

            for(int i=0; i<3; i++)
            {
                frag_position = shadowTransforms[gl_Layer] * gl_in[i].gl_Position;
                gl_Position = projectionMatrix * shadowTransforms[gl_Layer] * gl_in[i].gl_Position;
                EmitVertex();
            }
            EndPrimitive();
        }
    }
}

public void ShadowMapsPass(Shader shader)
{
    // Setup
    GL.UseProgram(shader.ID);
    GL.Viewport(0, 0, CubeMapArray.size, CubeMapArray.size);

    // Clear the cubemarray array data from the previous frame
    GL.BindFramebuffer(FramebufferTarget.Framebuffer, shadowMapArray.FBO_handle);
    GL.ClearColor(Color.White);
    GL.Clear(ClearBufferMask.ColorBufferBit | ClearBufferMask.DepthBufferBit);

    for (int j = 0; j < lights.Count; j++)
    {
        // Create the light's view matrices
        List<Matrix4> shadowTransforms = new List<Matrix4>();
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(1, 0, 0), new Vector3(0, -1, 0)));
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(-1, 0, 0), new Vector3(0, -1, 0)));
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(0, 1, 0), new Vector3(0, 0, 1)));
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(0, -1, 0), new Vector3(0, 0, -1)));
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(0, 0, 1), new Vector3(0, -1, 0)));
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(0, 0, -1), new Vector3(0, -1, 0)));

        // Send uniforms to the shader
        for (int i = 0; i < 6; i++)
        {
            Matrix4 shadowTransform = shadowTransforms[i];
            GL.UniformMatrix4(shader.getUniformID("shadowTransforms[" + i + "]"), false, ref shadowTransform);
        }
        GL.Uniform1(shader.getUniformID("lightID"), j);
        DrawScene(shader, false);
    }
}

public void ShadowMapsPass(Shader shader)
{
    // Setup
    GL.UseProgram(shader.ID);
    GL.Viewport(0, 0, CubeMapArray.size, CubeMapArray.size);

    // Clear the cubemarray array data from the previous frame
    GL.BindFramebuffer(FramebufferTarget.Framebuffer, shadowMapArray.FBO_handle);
    GL.ClearColor(Color.White);
    GL.Clear(ClearBufferMask.ColorBufferBit | ClearBufferMask.DepthBufferBit);

    // Create the light's view matrices
    List<Matrix4> shadowTransforms = new List<Matrix4>();
    for (int j = 0; j < lights.Count; j++)
    {
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(1, 0, 0), new Vector3(0, -1, 0)));
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(-1, 0, 0), new Vector3(0, -1, 0)));
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(0, 1, 0), new Vector3(0, 0, 1)));
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(0, -1, 0), new Vector3(0, 0, -1)));
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(0, 0, 1), new Vector3(0, -1, 0)));
        shadowTransforms.Add(Matrix4.LookAt(lights[j].position, lights[j].position + new Vector3(0, 0, -1), new Vector3(0, -1, 0)));
    }

    // Send uniforms to the shader
    for (int i = 0; i < shadowTransforms.Count; i++)
    {
        Matrix4 shadowTransform = shadowTransforms[i];
        GL.UniformMatrix4(shader.getUniformID("shadowTransforms[" + i + "]"), false, ref shadowTransform);
    }      
    DrawScene(shader, false);
}