News for hackers. Stuff that matters.
Re:A cache miss is not a cache miss by webaholic on Sunday May 01, 2016 @10:48AM (#11606374)
I think this is the reasoning behind preferring SOA vs AOS.
Re:A cache miss is not a cache miss by halomru on Sunday May 01, 2016 @11:18AM (#11606487)
I view SoA vs AoS (with SoA frequently being more performant) as a choice to maximise the efficiency of the cache, while this article is more about how reducing data dependency makes caches less important. Two different means to solve the same problem.
Re:A cache miss is not a cache miss by Sindisil on Sunday May 01, 2016 @11:07AM (#11606431)
AOS == Array of Structures

SOA == Structure of Arrays

Re:A cache miss is not a cache miss by Sindisil on Sunday May 01, 2016 @11:08AM (#11606442)
Oops, meant to reply to the sibling comment confusing the TLAs with web domain definitions.
Re:A cache miss is not a cache miss by userbinator on Sunday May 01, 2016 @11:02AM (#11606409)
How are Service-Oriented Architectures and... Architecture-Oriented Services(?) relevant to cache misses? There's more locality in the former, because each service is more specific?
Re:A cache miss is not a cache miss by Kronopath on Sunday May 01, 2016 @11:10AM (#11606452)
Structure of arrays vs. array of structures. I.e. Having one struct that contains several arrays as its members, vs. having one array that contains several structs as its elements. The former keeps like data together, so when you need to, for example, iterate over the position vectors of a bunch of objects in a 3D graphics program, all the position vectors for all the objects are tightly packed in the same array, increasing cache efficiency.
Re:A cache miss is not a cache miss by userbinator on Sunday May 01, 2016 @11:18AM (#11606488)
I've always called those "parallel arrays" --- and they were around long before Service Oriented Architectures: https://en.wikipedia.org/wiki/Parallel_array