A tiny machine: decoding

From here on you build analysis tools yourself, for a tiny machine we invented for these lessons. A virtual machine is a program that executes instructions encoded as bytes, called bytecode, just as Python itself does with your code. Understanding a small invented machine is the shortest path to understanding real ones.

Our machine has a single slot called the accumulator, which starts at zero, and six instructions. Each instruction starts with one byte, the opcode, and four of them are followed by a second byte, the operand. Each line below shows the opcode, then its name, then imm if it takes an operand, then its effect written in Python:

0x00 HALT      break
0x01 LOAD imm  acc = imm
0x02 ADD  imm  acc = (acc+imm) & 0xFF
0x03 XOR  imm  acc = acc ^ imm
0x04 SUB  imm  acc = (acc-imm) & 0xFF
0x05 OUT       out.append(acc)

So the program 01 48 05 00 reads like this: LOAD with the operand 0x48, then OUT, then HALT. The byte 48 isn't an instruction but the operand of the instruction before it. That's the most important idea in decoding: instructions vary in length, so you can't know where the next instruction starts until you know the length of the current one.

That's why you walk the bytes with an index i. You read the opcode at program[i], and if it has an operand you read it from program[i + 1] and move ahead two bytes; otherwise you move ahead one byte. When you reach HALT you stop, because the bytes after it aren't instructions.