Performance improvement by not sorting every time. (#12)

Signed-off-by: Ed Warnicke <[email protected]>

Move to channel-less implementation. (#11)

* Working rough cut

Signed-off-by: Ed Warnicke <[email protected]>

* Move to channel-less implementation.

When multiple go routines queue to send into a channel, the order of their entry to the channel is unknowable:

https://groups.google.com/g/golang-nuts/c/PWt4r9b40bc/m/lC59KD5TQCwJ

Therefore, we cannot rely on a channel for ordering, as if we overrun its size, order will be non-deterministic.

This PR moves to using a strict ordering by atomically ordered ticket.

Signed-off-by: Ed Warnicke <[email protected]>

Minor performance improvement (#9)

Have the processing thread get the entire buffer
available to reduce its mutex locks

Signed-off-by: Ed Warnicke <[email protected]>

Minor performance improvement (#9)

Have the processing thread get the entire buffer
available to reduce its mutex locks

Signed-off-by: Ed Warnicke <[email protected]>

Revamped testing and benchmarking. (#8)

* Add benchmarking to CI (#7)

Signed-off-by: Ed Warnicke <[email protected]>

* Revamped testing and benchmarking.

Signed-off-by: Ed Warnicke <[email protected]>

Update README.md with info on deadlock avoidance. (#4)

Signed-off-by: Ed Warnicke <[email protected]>

Update README and fix CI (#1)

Signed-off-by: Ed Warnicke <[email protected]>