It's only on mispredict though? So probably not much in the grand scheme of things?
-
-
-
That depends; plenty of software has branches that are entirely unpredictable (prediction statistically no better than coin toss).
-
at least on K10, the CPU perf would "eat it" on unpredictable branches, so it was often better to always evaluate both possible branches and then use bit-masking to select the desired answer. on my Piledriver based FX, this seems to be less of an issue...
-
though, in my testing, for small cases with "coin toss" probabilities, it is still often more effective to use branch-free logic, including any memory loads, ... in things like compressors or filters, branches are still sort of an enemy. just it is "smarter" here.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.