3 ms·
How about running this on a 36+ core machine? Testing parallelism on a dual or quad core isn't that exciting. Maybe we'd realize this benchmark isn't really tes
by dxhdr 11y ago
How about running this on a 36+ core machine? Testing parallelism on a dual or quad core isn't that exciting. Maybe we'd realize this benchmark isn't really testing parallelism but primarily sequential thread creation/deletion. This could very well be fast on a scheduler optimized for a single core.
And just to be clear, the interesting property to test is parallelism, not concurrency.
- ovi256 11y agoExcept that schedulers optimized for single cores do not grow on trees, either. Python 3.5, which I think is a language with very high incentives for such single-core optimizations (because of the GIL), does almost two orders of magnitude worse than Golang (1.7s vs 63s) on a single core. I wrote a py3.5 asyncio coroutine implementation of this benchmark, and I hope it's just my inexperience with asyncio that makes it get such horrible results. Here it is: https://github.com/ociule/skynet/tree/master/python35-asyncio https://github.com/ociule/skynet/tree/master/python35-asynci...