Big-O matters
We spend so much time drilling algorithmic complexity. Big-O and all that. But performance is so often about contention and memory, especially when working in parallel.
Source: Big-O matters, but it's often memory that's killing your performance, an article by Rob Napier.