2 ms·
Using 'likely/unlikely' compiler directives, the instructions in the unlikely error branch can be hoisted out to another area and not significantly affect instr
by cma 8y ago
Using 'likely/unlikely' compiler directives, the instructions in the unlikely error branch can be hoisted out to another area and not significantly affect instruction cache. It still is a branch though.
- jcelerier 8y agoand exception-based code can have much less branches since you don't need to check every function call for error but instead let the exception propagate from where they can be thrown. e.g. in a hypothetic case where you load a settings file five layers deep, in an error-code-based solution you have potentially 5 branches while in an exception-based solution you only have one
- tpolzer 8y agoAt the cost of predictable performance though. With current implementations, the exceptional path will be several magnitudes (!) slower for cheap functions.