Where do I go to follow the work on

[odd pine]

This is probably the best spot - agree that this is an area that needs some improvement.

@steady geyser do you know if this type of stuff already a part of our planned upcoming polish work?

[steady geyser]

I don't believe there are any actual issues in github yet, I will make a few since it's kind of a wide topic. @conor are you seeing the problem only on a fresh start with empty cache and/or new module? or is it every time you invoke a function? or both 😄

[odd pine]

My hot take is that in general if you run a module and non module side by side that does the same thing, the module version will be a bit slower every time.

Not sure if there are any easy wins or if this is the new normal but its worth addressing this out in the open IMO.

[formal patio]

Both, though tbh I don't mind the cold start as much on the initial run, it's the lack of caching between subsequent runs that is currently more painful

when you're developing on a module having to sometimes wait 45+ seconds between iteration loops is rough

I would be happy to file an issue specifically scoped to caching between different dagger call invocations of the same module/function with some wall clock measurements if it makes sense

[steady geyser]

Yeah that would be great thank you! We also have plans to make the initial run faster via cloud cache, which i'll make a public facing issue for

[steady geyser]

There's inherently going to be more overhead since calling a function involves rpc and creating a container, but we have not tried to optimize for latency very much at all yet so I am sure there's gotta be relatively low hanging fruit there. We just need to test to find what the bottlenecks are currently, which we can track via conor's issue.

[odd pine]

Thats great to hear, I am hoping we can find a few easy wins. I recognize its always going to have a bit more overhead.