Source: Hacker News
Article note: I'm not sold on some of the conclusions, but the motivating observations are really interesting, and, like Big-O analysis of memory-bound everything, another case of measuring badly for legacy reasons.
I've been eying doing some work on a platform with an insanely powerful DMA engine and a feeble CPU for the last few days, so my head is already tilted the right direction for the paradigm.
Comments