4 ms·
This is an IO bound problem. Trying to "go faster" by using multiple threads or vectorisation isn't going to achieve much.
by Xophmeister 3y ago
This is an IO bound problem. Trying to "go faster" by using multiple threads or vectorisation isn't going to achieve much.
- frant-hartm 3y agoNo, it is not. The file is smaller than the available RAM, so will be cached in subsequent runs.
- brazzy 3y agoYou're not up to date with how fast IO can be these days. Nobody who cares primarily about performance uses spinning rust anymore.
- bufferoverflow 3y agoI'm pretty sure Hetzner servers (where the test is performed) have NVMe drives that can read at 4GB/s. So theoretically IO bound solution would be done in around 0.25 seconds. However the fastest current solution is around 12 seconds. So it's not IO-bound.