Comment by gsmecher | Hacker News Reader

Comment by gsmecher | Hacker News Reader

These experiments are always unique, and each is interesting in its own way.

Can you comment on the use of asynchronous code rather than generators? For example, here's a throw-away example of a clock using generators and a syntax that's otherwise similar to yours:

  from more_itertools import take
  
  def drv_clock():
      while True:
          yield "1b1"
          yield "1b0"
  
  for x in take(10, drv_clock()):
      print(x)

I have written signal flow graphs (including feedback) using a generator-style approach, intended to prototype datapaths that are then manually translated into RTL. It's different, but not completely different.

As mentioned in another answer, deep down there's not much difference. Personally, I find async/await more elegant for concurrent algorithm design (e.g. https://github.com/cjdrake/seqlogic/blob/main/ipynb/dining_p...).

But control threads like a clock/reset aren't the main idea. Look at the example RISC V core for the design style we're attempting (https://github.com/cjdrake/seqlogic/tree/main/tests/riscv/co...). No async/await or yields anywhere. It's described in a structural way, and a couple layers of Python hide the details.

> Can you comment on the use of asynchronous code rather than generators?

I'm not the OP. IIUC, async functions are implemented using generators under the hood. cf: The asyncio.coroutine decorator from Python 3.4, before async await syntax was implemented in Python 3.5 (PEP 492).

Correct in this case. In src/seqlogic/sim.py there are four usages of 'yield' keyword that suspend the current coroutine and return control to the event loop.

Reply on news.ycombinator.com