Notes

GPU objects live in GPU memory. The CPU only holds a GLuint handle (an ID number) and asks the driver to act on “object N”. Treat it as a int number.

Handles

• A handle is just an integer, like a door number. The real object is on the GPU.
• Only gl calls hand out handles (glGenBuffers, glCreateShader). Never invent one.
• 0 means “no object”.
• Assigning a new handle to the same variable loses the old one; that GPU object still exists but can no longer be reached.
• The type does not say what kind of object it is. A buffer handle and a shader handle are both GLuint.

Type	Used for
GLuint	Object handles; index data (std::vector<GLuint> idx)
GLint	Shader variable locations; -1 means not found

Common GPU objects

Object	What it is
Buffer	Raw GPU memory (vertex data, indices)
Shader	One compiled stage (vertex or fragment)
Program	Vertex + fragment shader linked together
VAO	Remembers how buffers are wired to shader inputs
Texture	An image
Framebuffer	A canvas to draw into

A single vertex is not an object. One buffer holds all vertices.

State machine

• Context = the whole GL world: every object, every binding, the active program. One per window. No context, no gl calls.
• Bind = select. glBind* / glUseProgram put an object into a slot (target).
• Most calls take no “which object” argument. They act on whatever is bound to the slot they read, so binding the wrong object means the next call silently affects it.
• Each slot holds one object at a time. Different slots can be filled at once.
• Binding is CPU-side setup. One draw call still processes every vertex and pixel in parallel.

Core pattern

glGenBuffers(1, &buf);              // 1. allocate, get a handle
glBindBuffer(GL_ARRAY_BUFFER, buf); // 2. select it
glBufferData(GL_ARRAY_BUFFER, ...); // 3. operate on the selected object

Related

OpenGL Hub · OpenGL buffers and targets · OpenGL API naming and call order