Adding a geth+wagon node to the testnet

[gballet]

[MaxGraey]

May be better life instead wagon? If you not yet starting progress in this direction

[lrettig]

Does this refer to adding a node to the testnet?

[MaxGraey]

I'm not sure. life just is better alternative of wagon VM. It already great tuning for blockchain scope, has JIT, validate and decline wasm op codes which contain non-deterministic behaviour like float point op codes and etc. Also it tested even on AssemblyScript binaries. Actually perlin-network also using AS for its own contacts. But anyway it's just suggestion. May be wagon better align for your needs

[MaxGraey]

Actually I personally am a big fan of VMs based on Cranelift (developed by Mozilla and use Rust) like wasmer, wasmtime and Nebulet (run on Ring-0). But they are not yet mature

[gballet]

@MaxGraey We already have a working implementation with wagon. Life sounds interesting, I will experiment with it next week. it shouldn't be too difficult to port the work I did on wagon.

[gballet]

So this should be good to go (not with life, that's something for the next milestone). Waiting feedback on https://github.com/ewasm/go-ethereum/issues/6 and https://github.com/ewasm/go-ethereum/issues/1

[jwasinger]

@gballet I'll deploy a geth+wagon node.

[gballet]

@jwasinger ok, just to confirm: you need to use --vm.ewasm="metering:true" for it to work

[gballet]

./geth+wagon init ewasm-testnet-geth-config.json && ./geth+wagon --etherbase 0x9AfA52330797885BC7f6Df7b7d87835995d0f4f1 --rpc --rpcapi "web3,net,eth,debug" --rpcvhosts="*" --rpcaddr "0.0.0.0" --rpccorsdomain "*" --networkid 66 --bootnodes "enode://53458e6bf0353f3378e115034cf6c6039b9faed52548da9030b37b4672de4a8fd09f869c48d16f9f10937e7398ae0dbe8b9d271408da7a0cf47f42a09e662827@23.101.78.254:30303" --vm.ewasm="metering:true" --syncmode full --mine --miner.threads 1

is the exact command line that I use

[jwasinger]

Thanks. I'll work on it in 10 hours (tomorrow).

[jwasinger]

Done here: https://github.com/ethereum/cluster/pull/97

[gballet]

Very nice, thank you

[MaxGraey]

Great bench for different VMs: https://github.com/fluencelabs/fluence/blob/master/bench/vm/README.md

Note: life in that research didn't use jit

[gballet]

thank you for this. This corresponds to what we're currently observing, although there are a lot of optimizations that we need to look into.