hello,
I would like to run benchmarks on p-cores vs e-cores, I suspect a difference of more than 50%
is it possible to choose your processor (P or E) to run your program?
nice day
hello,
I would like to run benchmarks on p-cores vs e-cores, I suspect a difference of more than 50%
is it possible to choose your processor (P or E) to run your program?
nice day
At a first "approximation" you shouldn't need to choose.
If your app would receive a 50% throughput boost from running on P cores — over some non-trivial period of time — the system is going to work that out on the fly and move your processing into a P core.
Conversely, if the system doesn't move your app out of E cores — over some non-trivial period of time — then it's decided that you wouldn't actually get meaningful benefits from a P core, so you may as well stay in the E core.
Benchmarking for its own sake would make this a more complicated scenario, but I think the question is whether comparing a "pure" P benchmark against a "pure" E benchmark gives any useful information. Probably not! What would you do with that information?
hi, I want to check the effecient of my algorithm.
only p-cores :
speedup :
2 cores : x2,03
4 cores : x3,86
6 cores : x5,65
seems very good
with E-cores
8 cores : x6,48
10 cores : x7,30
12 cores : x8,18
speedup is fall