Block-stm initial commit

[cffls]

This PR enables parallel execution of transactions when validating blocks. A similar PR has been merged to bor.

The major change of existing component that needs attention is in "intra_block_state.go", where a multi-version hashmap is added between EVM and database.

[AskAlexSharov]

please take a look if you can use existing methods: ibs.SoftFinalise(), ibs.MakeWriteSet, txTask.ReadLists = rw.stateReader.ReadSet(), txTask.WriteLists = rw.bufferedWriter.WriteSet().
you can find them in:
cmd/state/exec3/state.go:RunTxTask

[cffls]

Thanks for the pointers. Will look into them.

[AlexeyAkhunov]

This is an interesting idea, and somewhat (or largely) overlaps with our own experiments on the parallel execution. And I am aware of block-stm ideas from the Polygon team.
At this point, I would rather not attempt on plugging this in, for two main reasons:

1. Parts of the code it is touching (and therefore would need extensive correctness and performance testing) are the same as we need to modify for Erigon3 project - and this extra disturbance and will potentially delay erigon3, which is critical for us.
2. We are considering replacing EVM implementation (including intra_block_state.go) with the code from Silkworm (evmone + some code around it in C++), so trying to port block-stm would also add to that effort.

[cffls]

This is an interesting idea, and somewhat (or largely) overlaps with our own experiments on the parallel execution. And I am aware of block-stm ideas from the Polygon team. At this point, I would rather not attempt on plugging this in, for two main reasons:

1. Parts of the code it is touching (and therefore would need extensive correctness and performance testing) are the same as we need to modify for Erigon3 project - and this extra disturbance and will potentially delay erigon3, which is critical for us.
2. We are considering replacing EVM implementation (including intra_block_state.go) with the code from Silkworm (evmone + some code around it in C++), so trying to port block-stm would also add to that effort.

Thanks @AlexeyAkhunov for the inputs. Agreed that it might be too early to plug this in. Do you think we should wait for both Erigon3 and EVM implementation replacement to complete before working on this PR, or just the completion of Erigon3?

[github-actions]

This PR is stale because it has been open for 40 days with no activity.

[github-actions]

This PR is stale because it has been open for 40 days with no activity.

[mh0lt]

This code has now been replaced by the full e3 STM implementation in this branch: https://github.com/erigontech/erigon/tree/async_tx