Some Questions

[calebmkim]

Specifically, some questions I would love input on (although they might be better discussed synchronously) :

• How should we store the .json results for resource estimation? Right now, it’s just accumulated in a file called “{date + time the script was run}.json”
• How should we store the configuration settings/information about how to run fud? Right now I store it in a messy .toml file
• What sort of graphs should we generate? This is pretty broad/open ended, but just wanted to get the discussion going.

[sampsyo]

This honestly all sounds good! Big ol' JSON files with timestamps will be easy and simple and will give us plenty of flexibility in the future.

"What sort of graphs?" is of course a very big question… and ties back to designing the "selling points" of the paper. That is, we can first ask: What is the paper claiming? And then we can ask: What data will be necessary to support those claims?

Some very-top-level claims I can imagine are:

• Making things static leads to faster/smaller hardware because the control overhead is smaller than for dynamic hardware.
• Making things static enables novel optimizations that are not possible if everything is dynamic.
• Fluidly integrating static and dynamic pieces in a larger hardware design is good because [reasons].

These might point toward a direction about the graphs to start drawing?