HigherOrderCO
Option to limit the number of threads ?
No matter how you run things, it will always run the maximum number of threads on your machine. This is not always the desired behaviour. It'd be nice if it could be changed, with perhaps an equivalent to OMP_NUM_THREADS for OpenMP ?
For now, we don't have the functionality of changing the number of max threads yet, however, you can change it in the code inside build.rs, changing let tpcl2 = (cores as f64).log2().floor() as u32; to let's say let tpcl2 = 3 for example, would run with 8 threads max (threads = 2^tpcl2)
BTW, it might be a good idea to make it so it doesn't use all the threads the system has, but rather all the cores. Especially since using too many threads seems to lead to some OOM errors ?
@Janiczek a lot of developper have a superstitious belief the having stuff be base two makes the code go faster. In the case of threads, it doesn't make much sense, you should be more concerned about thread placement (ie not putting multiples threads running on the same core) and not overpopulating (ie using more threads than you have cores). This can be tricky, because most cores nowadays have hyperthreading enabled, meaning they can run multiple threads at once, albeit at decreased efficiency.