Dresdenboy
Golden Member
As former demo coder I somehow think, that throwing faster hardware at performance problems caused by incompetence is a way to solve it, but which feels wrong.
Last year I optimized two internal tools, who were developed by contractors, because it hurt watching them do their conversions on a fast Xeon. Adding <100 LOC (some prefetching and caching) sped them up by 31x and 51x respectively. No new CPU could do that.
But back to the topic: As I already said, measuring the perceived responsiveness/snappiness might be interesting.
Last year I optimized two internal tools, who were developed by contractors, because it hurt watching them do their conversions on a fast Xeon. Adding <100 LOC (some prefetching and caching) sped them up by 31x and 51x respectively. No new CPU could do that.
But back to the topic: As I already said, measuring the perceived responsiveness/snappiness might be interesting.