News for hackers. Stuff that matters.
Re:Speed memory access by arranging data to take advantage o... by Tiktaalik on Sunday April 06, 2014 @08:37PM (#7543970)
First I heard of this sort of thing was from DICE's presentation on "Data Oriented Design" http://dice.se/publications/introduction-to-data-oriented-de...

I think they'd differ from the author here that this isn't an optimization to go to when performance starts to suffer, but a design that you have to start using from the beginning, because it's a big change to refactor everything to this style.

Re:Speed memory access by arranging data to take advantage o... by AngusMcQuarrie on Monday April 07, 2014 @12:56PM (#7547694)
There was also a talk on this from Mike Action (Engine Director at Insomniac) at GDC this year. He echoed the same sentiment that cache misses are the main source of performance problems in software. I only partially agree, as network calls are several orders of magnitude slower, and in many services that's actually your bottle neck, not the cache misses.