search

tmds / Tmds.LinuxAsync

MIT License
25 stars 3 forks source link

Optimize ExecuteQueued to adding batched operation #62

Closed tmds closed 4 years ago

tmds commented 4 years ago

When looking at these traces, this catches my attention.

snip

The bottom yellow bar is: HandleEvents (1049 samples, 5.76%): epoll says this socket has data. The top yellow bar is: LinuxAIO.AddRead (117 samples): add a read for batching.

Given the 11% ratio, my feeling is this implementation is not very efficient.

@adamsitnik I wonder what you think about this. What would you do next?

tmds commented 4 years ago

Should be better thanks to reduced locking, and some refactorings. I'm not looking into this further.