GLSL:如何正确绑定数千个缓冲区? [英] GLSL : How to bind thousands of buffers properly?

问题描述

我想到了一个想法，要求将成千上万个缓冲区(原子计数器和Shader存储缓冲区)绑定到一个GLSL程序.我首先检查了这对openGL的限制是否有意义，并且似乎有两个原因:

I came up with an idea that requires to bind thousands of buffers (Atomic counters and Shader storage ones) to one GLSL program. I first checked if this would make any sense in the limitations of openGL and it seems possible for two reasons :

1. 在我的笔记本电脑上GL_MAX_ATOMIC_COUNTER_BUFFER_BINDINGS和GL_MAX_SHADER_STORAGE_BUFFER_BINDINGS都在32k左右.因此，openGL被允许我一次通过绑定数千个缓冲区.

2. openGL 4.4附带了:

1. On my laptop GL_MAX_ATOMIC_COUNTER_BUFFER_BINDINGS and GL_MAX_SHADER_STORAGE_BUFFER_BINDINGS both are around 32k. So openGL is enclined to let me bind thousands of buffers for one pass.

2. openGL 4.4 comes up with :

void BindBuffersBase(枚举目标，首先使用uint，sizei计数，使用const uint * buffers)；

void BindBuffersBase(enum target, uint first, sizei count, const uint *buffers);

在C代码上，这似乎很容易，但是我对如何进行着色器一无所知.当绑定索引缓冲区时，我们应该使用:

On the C code it seems easy, but I have no clue on how to do my shaders. When binding indexed buffers, we are supposed to use :

layout (std430, binding = 0) buffer Objects
{
  Object objects[];
};

绑定两个缓冲区时可以.但就我而言:

When binding a couple of buffers it is ok. But in my case :

1. 我事先不知道将绑定多少个缓冲区
2. 即使我知道，我是否应该为EACH缓冲区编写一些代码行并在布局绑定0、1、2，...，n中进行定义?

然后的问题是: 有没有办法使着色器不知道将要绑定到它的缓冲区的数量?假设它们全部包含相同类型的数据.

The question is then : Is there a way to make the shader unaware of the number of buffer that will be bound to it ? Assuming they are ALL containing the same type of data.

答案:

感谢安东·科尔曼，以获取快速，清晰的答案.

Thanks to Andon M. Coleman for the quick and clear answers.

首先，我以为我的硬件支持数千种绑定是错误的.新秀错误，我采用了定义的值，而不是运行时的值.

First of all, I was wrong to assume my hardware supported thousands of bindings. Rookie mistakes, I took the defined value instead of the runtime one.

在我的机器上:

1. GL_MAX_SHADER_STORAGE_BUFFER_BINDINGS = 96
2. GL_MAX_ATOMIC_COUNTER_BUFFER_BINDINGS = 8
3. GL_MAX_UNIFORM_BUFFER_BINDINGS = 84

与我提出问题的依据相去甚远.

Which is far from what I based my question on.

但是，有一种方法可以在着色器中定义多个缓冲区:

However, there is a mean to define multiple buffers in the shader :

A
{
  B b[];
}
layout (std430, binding = 0) buffer A a[32];

必须预先知道最大绑定数，但是这些缓冲区将占用0到31之间的绑定.

The amount of maximum bindings must be known in advance but those buffers will occupy the bindings from 0 to 31.

推荐答案

感谢安东·科尔曼(Andon M. Coleman)的快速明确的回答.

Thanks to Andon M. Coleman for the quick and clear answers.

首先，我以为我的硬件支持数千种绑定是错误的.新秀错误，我采用了定义的值，而不是运行时的值.

First of all, I was wrong to assume my hardware supported thousands of bindings. Rookie mistakes, I took the defined value instead of the runtime one.

在我的机器上:

1. GL_MAX_SHADER_STORAGE_BUFFER_BINDINGS = 96
2. GL_MAX_ATOMIC_COUNTER_BUFFER_BINDINGS = 8
3. GL_MAX_UNIFORM_BUFFER_BINDINGS = 84

与我提出问题的依据相去甚远.

Which is far from what I based my question on.

但是，有一种方法可以在着色器中定义多个缓冲区:

However, there is a mean to define multiple buffers in the shader :

A
{
  B b[];
}
layout (std430, binding = 0) buffer A a[32];

必须预先知道最大绑定数，但是这些缓冲区将占用0到31之间的绑定.

The amount of maximum bindings must be known in advance but those buffers will occupy the bindings from 0 to 31.

这篇关于GLSL:如何正确绑定数千个缓冲区?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！