Parallel Programming

Since the advent of microprocessors, Moore’s Law has said that the number of transistors on an integrated circuit would double about every eighteen months to two years, while the size of the circuit would stay about the same or get smaller, and the price of that integrated circuit would stay about the same or get cheaper. This meant that we’d have more powerful processors for about the same amount of money on a regular basis. This prediction worked really well until the beginning of the 2000s, when some things changed. First, Moore’s Law implies that to get twice as many transistors on a similarly sized chip, the size of the transistors would shrink and the distance between the transistors would also shrink. Clearly, physics being what it is, this couldn’t continue indefinitely and by the early 2000s we were beginning to see some problems with making the transistors a lot smaller. The second problem is heat. If you pack twice as many transistors on a chip and expect to run the chip at the same or a higher clock speed, you’ll need more electricity going through the circuit. More electricity means more heat. This is known as the “power wall.”