Following up on the API topic: like many have pointed out, there are many pros with implementing a known API. However, we believe it needs to be both open source and a multi-algo miner API, which disqualifies some suggestions. Since we're AMD only, the natural target is sgminer, it makes it much easier for us to test things. Therefore, we're targeting the following setup in an upcoming release:
Would be nice if it will be compatible with Claymore's EthMan, for example like ethminer - https://github.com/ethereum-mining/ethminer/pull/253- Implement the sgminer API v4.0 (sgminer 5.5-gm is a good reference)
- We will not have time to add ADL/sysfs support for fans/temps, that will be added later.
- We will support both the ugly text-based and json-based formats.
- No privileged commands will be supported for now, the API is read-only.
- We will verify correctness against a few known monitoring tools, Awesome Miner is one of them. If you have other suggestions, let us know.
- PCIe bus id for each gpu will be displayed on startup and made available over the API to map cards against other tools.
Imho it's not a good fit, and it violates both of the conditions we set up (open source and a multi-algo miner API by design). It just doesn't sit right to use a ethash-centric miner API with a miner that doesn't do ethereum but rather a range of algos, we're just getting started on that front. For ethminer it's a very different ballgame though.
That said, if someone wants to write a little API adapter in e.g. Python, I think it's a great idea! I could even do it myself if I ever find a few hours to spare when we're done with the initial sgminer implementation. Building little adapters in a language who can massage json natively is just a so much better solution than pushing everything down into C code in the miner, it's so much more work writing the necessary boilerplate code down there.
. It’s quite possible that we will roll out CNv8 in a few weeks. The upside is limited due to the memory-bound characteristics of the algo, but we’ll see where we end up. For CN Heavy, I’m quite confident we will produce the fastest miner so far. Dev time is the limiting factor for us right now.