Since it's not supposed to know the numbers compile-time, I think that wouldn't be much different from an array lookup with a much larger array. In which case you might get cache misses? Not sure.
Though I realize the compiler probably caught the sequential numbers and compiled them differently, so I tried reading values from an array and adding the counter and got the following results:
I did some more testing, and it seems like doing %3)==0 and %5)==0 can be optimized way more than regular modulo, which probably causes it to be the true winner. This would also explain the different results I mention at the end of my post.
Ah, yeah, that's what I thought.
It doesn't fit the scope of this post since it requires compile-time knowledge of the target numbers, but it probably would be comstant.
Making everything compile-time would just be the same dilemma but with time needed to compile rather than time needed to run, so I don't have much interest in researching that.
For further actions, you may consider blocking this person and/or reporting abuse
We're a place where coders share, stay up-to-date and grow their careers.
Since it's not supposed to know the numbers compile-time, I think that wouldn't be much different from an array lookup with a much larger array. In which case you might get cache misses? Not sure.
Though I realize the compiler probably caught the sequential numbers and compiled them differently, so I tried reading values from an array and adding the counter and got the following results:
I did some more testing, and it seems like doing %3)==0 and %5)==0 can be optimized way more than regular modulo, which probably causes it to be the true winner. This would also explain the different results I mention at the end of my post.
Nice!
This is the kind of solution, I had in mind when I said compile-time:
arne-mertz.de/2019/12/constexpr-fi...
Ah, yeah, that's what I thought.
It doesn't fit the scope of this post since it requires compile-time knowledge of the target numbers, but it probably would be comstant.
Making everything compile-time would just be the same dilemma but with time needed to compile rather than time needed to run, so I don't have much interest in researching that.