Re:Why is it faster to process a sorted array than an unsort...
by Rychard on Saturday May 18, 2019 @09:46AM
(#19947352)
I wonder if the microcode changes to mitigate branch prediction vulnerabilities (i.e. Spectre) affect these results in any way.
[ Parent
| on HN ]
Re:Why is it faster to process a sorted array than an unsort...
by jcranmer on Saturday May 18, 2019 @10:03AM
(#19947418)
Unlikely. The microcode changes are largely about disabling indirect branch prediction--virtual functions, jump tables, function returns. The branch prediction here is generally driven by simple branch history predictors (i.e., at a fundamental level, the branch predictor boils down to "guess what the last branch did"), which are way too useful to disable even to mitigate Spectre.