Tim Visée

Fantastic suggestion. Thanks so much for sharing this. I wasn't aware of the bucket queue structure. I'd like to focus on the other days first. I'll likely try to implement...

Interesting. I got slightly slower times while using A*. I actually started with A*, but switched to Dijkstra due to better timings. I wonder what may have caused this. Maybe...

Possibly. What did you use?

Ping @jamesmunns (due to draft PR)

> This uses fetch_or and fetch_add, do we need to implement this for thumbv6? Apparently, yes: https://github.com/jamesmunns/bbqueue/runs/1390103704 Edit: this is now implemented

> > Do we still need to zero the buffer when taking just the producer or consumer, like try_split? (currently commented-out) > > Yes, I would suggest doing either on...

@jamesmunns It now zeroes when the producer is taken. I amended this change to the last commit (03b97746e5af06c82f6a10841c54e2601a14ab5d).

I did attempt to use this on shared memory between processes, but failed. I think the problem is that only the data buffer (`buf`) is placed at the pointed-to memory...

I did some experimentation, and it seems to work quite well (easily reaching 280Gbps through it across processes if you're wondering). I'm just placing the `BBBuffer` struct on shared memory,...

> @timvisee sure! I have an issue for that already actually: #40 > > I'd definitely take a PR for this. I've submitted a draft PR for this, see: https://github.com/jamesmunns/bbqueue/pull/78...