|
|
|
|
|
|
by astrange
4080 days ago
|
|
|
If your data is unpredictable, the branchless instructions save time wasted trying to predict it. If you're doing any kind of data compression, either your compressed data is unpredictable or it's not compressed enough! Getting rid of branching in the unimportant parts of your program is good too; saves space in the cache for important branch predictions. This is the kind of thing where it's not worth doing by hand, but I don't think compilers really do it either… |
|
|