Nothing Is Free
We started this book by looking at a switch. A simple piece of metal that takes time to charge.
We have built logic, numbers, stacks, queues, caches, pipelines, and predictors on top of that switch.
But the physics of the switch never went away.
The Triangle of Doom
There is no such thing as "Optimal." There is only "Trade-off."
You can optimize for any two corners, but never all three.
- GPUs: High Speed, High Energy.
- Mobile: Low Energy, Low Speed.
- Compression: Low Memory, Low Speed.
Beyond Big-O
Big-O notation taught you to count operations. It assumed all operations were equal.
We now know that is false.
An operation in registers is free. An operation in L1 is cheap. An operation in RAM is expensive. An operation on Disk is death.
A great programmer does not just write
Will Quantum Computers fix this?
No. They solve specific problems (factoring, simulation). They do not make loops faster. The Triangle of Doom (Speed, Energy, Memory) applies to the universe, not just silicon.
What should I learn next?
Learn Computer Architecture (Hennessy & Patterson). Learn Compilers (Dragon Book). Learn the layers below you, so you can leverage them.
Is the "Lie" a bad thing?
No. Abstractions allow us to build operating systems and websites without going insane. The problem is not the Lie. The problem is believing the Lie when performance breaks.
The computer is not a black box.
It is a negotiation.
If you understand the physics, the latency, the bandwidth, and the predictions... you can negotiate well.
If you don't, the machine will win — silently.
End of File.