for comparison: * sync using write_all_at/read_exact_at hits about 2gbps * io_uring hitting 6.5gbps reads and 5gbps writes (7th gen lenovo x1 carbon laptop w/ LUKS full disk encryption) I'm not even using SQPOLL or registered IO buffers or files yet...
-
-
Show this threadThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
still some little bugs to shake out but shit's getting realhttps://www.youtube.com/watch?v=CIdXPIN3j38 …
Show this threadThanks. Twitter will use this to make your timeline better. UndoUndo
-
-
-
What are you sending "over the wire" for testing? That throughput looks great
-
currently just running the O_DIRECT example that is in the readme, http://lib.rs , and examples directory. TODO: * registered misuse-resistant io buffers * registered files * SQPOLL mode for 0 syscall operation * scribble a little bit to quickly implement socket ops
- 1 more reply
New conversation -
-
-
Logo suggestion: Spix's Macaw. Or maybe something Patrick Nagel-esque...
-
I'm thinking maybe like a tropical bird soaring over a yellow waterfall because uring sounds like urine and this needs to be fully exploited and then a bunch of whips cracking because it's fast and awesome
End of conversation
New conversation -
-
-
Nice! Why FastLock instead of Mutex?
-
aint nobody got time for that... I had it laying around from sled and I'm going to convert the 2 parts that rely on spin_lock into a flatcombining queue + try_lock so nothing ever blocks, and cache utilization is better from the flat combining batching
End of conversation
New conversation -
-
-
What's with all the swearing in the http://README.md ...
-
it's narc repellent
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.