Chapter 1: Introduction
Throughout this book we will be using C++. C++ is one of the primary languages used for computer graphics. If you’re already familiar with C or Java, you likely know enough to be able to do some basic graphics programming. To learn more about C++, you may want to read some of the chapters in the C++ part of this book.
Eventually, we are going to cover a variety of topics related to programming in OpenGL. OpenGL is a graphics API that allows you to use the dedicated graphics hardware in your computer. You can read more about OpenGL in the second part of this book - but first, we are going to cover some fundamentals of graphics. In particular, we’re going to cover everything necessary to write a software rasterizer. This makes it possible to study graphics in a more pure form, before diving into OpenGL programming, which carries a lot of OpenGL-related baggage. However, everything we cover in this part is relevant to OpenGL programming.
Before diving into graphics, you may want to take this fun quiz:
It asks you to guess whether or not a given image is a photograph or a computer generated image. It’s a fun exercise in seeing how realistic modern computer graphics is, and it also highlights one of the fundamental purposes of computer graphics: to create images that look like photographs. It is not always the case that we use computer graphics to make things that look like photographs, but it is common, so a lot of ideas and metaphors from photography leak in. For example, we’ll often discuss the idea of a “virtual camera” that we implement and simulate.
GPU vs. CPU
Nearly all computers have at least two processors. You may already be familiar with the CPU, or central processing-unit, which is responsible for carrying out many of the tasks performed by your compute. However, most modern computers have an additional processor (sometimes called a co-processor), the sole purpose of which is to perform graphics-related tasks.
This processor is the GPU, or graphics processing-unit. At its most basic level, the GPU is responsible for determining the color of pixels. It is also involved in the process of displaying pixels on display hardware (e.g. a monitor or screen). If you’re reading this on a screen, it’s very likely that a GPU is involved in the process of rendering the text and making those pixels visible to you.
However, you have to take some special steps in order to write programs that run on the GPU. If you write a program in Python, C, Java, or whatever, then compile and run it, you’re using the CPU. One way to make programs that utilize the GPU is by using OpenGL. However, as we will see, programs that run on a GPU are much more structured that CPU programs. Understanding the functionality that GPUs are designed for is essential to being able to write GPU programs.
GPUs started out as hardware rasterizers. That is, a processor that’s designed to be very fast at the process of rasterization. So before we dive into GPU programming, we’re first going to write a rasterizer. Such a rasterizer (one that runs on the CPU) is sometimes called a software rasterizer, since we are writing software to perform the rasterization process, instead of relying on dedicated hardware.
If you’re wondering a little bit more about what exactly a GPU is, you can think of it is a processor that is optimized for highly parallel tasks.
While a modern CPU might be good at running four dissimilar tasks at once, a GPU is good at running many hundreds of tasks at once (as long as those tasks are very similar).
A GPU is also a vector processor.
Typically each instruction refers to one or two registers.
So your CPU might run an add instruction in which a single register is added to another.
On a GPU, each instruction applies to multiple variables.
So each add instruction might add 16 numbers to 16 other numbers.
This type of instruction is called SIMD, which means
Same Instruction, Multiple Data.
SIMD instructions are also sometimes supported on CPUs,
but it’s what GPUs are designed for.
There is of course a lot more to GPU hardware than described here, but that should give you a basic idea of what we’ll be working with.
In This Part
The purpose of this part is to learn how to write a software rasterizer. Rasterization is the process of converting a primitive shape (such as a point, line, or triangle), into pixels that can show up on the screen. To understand how rasterization works, we’ll cover some geometry concepts (related to triangles) and some linear algebra (vectors and matrices).