I talk about nMigen a lot, so you might get the impression I think it's be-all and end-all of newHDLs. nope! it's just the one I picked to make as good as I can, while helping an established community around Migen.
lots of other newHDLs are cool too!
Conversation
Replying to
As it happens, I began a new Lava-like HDL (in Haskell) a few days ago. The implementation is just ~ 150 lines so far. It's intended to be dramatically simpler and fix some key issues with my old one.
2
4
In case anyone is interested, I've put Seq online as well, also not finished
2
1
I only had time to glance at the README but there’s much resonance. The York Lava that Reduceron uses, uses the Haskell typing system to, but it’s so forced that it become IMO a bigger burden of its own. I’m not a purist and that’s (probably going out) 1/2
1
I loved the sequential model of Handel C, rather `;` being the clock edge, I wanted it explict. FpgaC and host of other languages does it too (something calling `yield`). Anyway I don’t usual share things this half baked, but I’ll push my POC soon 2/2
1
Cool. I'm curious. I think I've learned enough with my own doodling that I can take in new ideas. And about not publishing half baked ideas: just don't write docs, everyone will ignore :)
1
1
Maybe it would be interesting to create a survey of ideas and see how these projects are subtly different in goals and implementation. There are a lot of approaches. I feel my goal is mostly towards a simple macro language, to not get lost in unbounded Haskell cleverness.
My goal would be a language in which straight forward logic looks ... straightforward. A more lofty and conflicting goal (maybe a different language) would would be one where you cannot so easily mix different time domains. Perhaps some CSP-like model (yes, Handel-C did it)
1
CSP is definitely high on my list. I started structuring code that runs in the same clock domain like this as well. For the rest I just want good composition, and substrate abstraction, e.g. run the same filter code in parallel, bit-serial, or sequential on CPU.
1
Show replies
Of course polymorphic algebraic types is a must in 2020. However before going down that rabbit hole, I want to study BlueSpec and see if it’s good enough.