James

@frebib If you could push (even a non-compiling) POC of your autoedges branch, that would be awesome!

A bit more data to be sure ``` main: new graph took: 913.653µs main: auto edges took: 9.273807153s main: auto grouping took: 28.690819ms main: send/recv building took: 566ns main: new...

> doesn't do shenanigans like sending more than one value downstream or waiting until it receives multiple inputs before sending that single output I'm not sure if those specific shenanigans...

> optimize the programs which we can guarantee are safe to optimize, and then some more complex programs won't be optimized even though in theory Sounds good to me =D...

Okay, I took your branch, and did the following: 1) Add a new patch in master with the missing Graph() method. Rebased everything on top of that. 2) Added some...

> Is it worth to add the sctx argument in order to address this case, or is this too small of a corner case? I do not know, but let's...

> is it worth implementing SpeculativeCall in order to support this use case? This case (or something that looks like it) seems more plausible, but still in the 20%. I...

> Does it still make sense to use the same Expr.Value() for obtaining the Value to give to a resource, or would it make more sense to split Expr.Value() into...

> right, right, you want me to make these kinds of decisions on my own You're in charge, I'm meant to be here to complain if my spider sense goes...

> if so, assume that calling Value() on the function would give the wrong result So currently we use Value() here for printf... Does this mean that this is currently...