The timer reports in whatever it feels like. More boringly, it'll be measuring std::chrono::high_resolution::period::type::Denom ticks per std::chrono::high_resolution::period::type::Num seconds - and helpfully it might go backwards, because it's not explicitly stable.
Also, yes - incremental caching is always a good plan with any expensive calculation (or other resource that takes time to obtain). That said, I think my "variables" version is sufficiently fast that you might not need to bother unless you needed the cached values an awful lot.
I thought we might further improve performance (in my vector version, too) by calling seq.reserve(n) just before the for loop - that ought to reduce the memory allocation workload significantly - but in my tests it didn't do much.
I also think the "variables" version is more than sufficient, just offering another look.
I've noticed that whenever I've used .reserve(n) with vectors nothing really seems to be gained time-wise. Perhaps I haven't used them in a scenario where it would be of real benefit.
The timer reports in whatever it feels like. More boringly, it'll be measuring
std::chrono::high_resolution::period::type::Denomticks perstd::chrono::high_resolution::period::type::Numseconds - and helpfully it might go backwards, because it's not explicitly stable.Also, yes - incremental caching is always a good plan with any expensive calculation (or other resource that takes time to obtain). That said, I think my "variables" version is sufficiently fast that you might not need to bother unless you needed the cached values an awful lot.
I thought we might further improve performance (in my vector version, too) by calling
seq.reserve(n)just before the for loop - that ought to reduce the memory allocation workload significantly - but in my tests it didn't do much.I also think the "variables" version is more than sufficient, just offering another look.
I've noticed that whenever I've used .reserve(n) with vectors nothing really seems to be gained time-wise. Perhaps I haven't used them in a scenario where it would be of real benefit.
No, I think you're right - in most cases, the allocators these days are simply good enough that the marginal benefit of reserving space isn't useful.