(L) [2006/08/14] [janos] [BranchPredictor help] Wayback!I wonder if any of you have tried the following:
if we can consider that branch prediction fails half of the time while traversing the tree (does it really?), then there must be a way of helping it through the first nodes:
Imagine that you code looks like this:
(L) [2006/08/14] [tbp] [BranchPredictor help] Wayback!I doubt complete unrolling is the silver bullet you make it to be, mainly because of the pollution it causes (instruction cache, btb etc) and the increased branch density (ie k8 have limitations in this regard). There was some related discussion here, [LINK http://ompf.org/forum/viewtopic.php?t=30]
In any case, feel free to provide real world benchmarks [SMILEY Wink]
_________________
May you live in interesting times.
[LINK https://gna.org/projects/radius/ radius] | [LINK http://ompf.org/ ompf] | [LINK http://ompf.org/wiki/ WompfKi]