[evm] performance: since we only create a cacheCtx whenever a precompile is called, then reverting a precompile call can just get rid of the cacheCtx and reset the writeToCommitCtxFromCacheCtx function to nil. That would be a cleaner way to revert a PrecompileCalled journal entry rather than deep copy the multistore and save it in a snapshot, but I'd have to test that implementation out.

[Unique-Divine]

Note to future self: since we only create a cacheCtx whenever a precompile is called, then reverting a precompile call can just get rid of the cacheCtx and reset the writeToCommitCtxFromCacheCtx function to nil. That would be a cleaner way to revert a PrecompileCalled journal entry rather than deep copy the multistore and save it in a snapshot, but I'd have to test that implementation out.

Originally posted by @k-yang in #2094 (comment)

[k-yang]

I thought about this more and I don't think it works in the case of multiple precompile calls. If a transaction does Precompile Call A and Precompile Call B, but reverts Precompile Call B, you still want to restore the snapshot to before Precompile Call B and after Precompile Call A, so you'd need to keep a deep copy of the multistore after Precompile Call A in the journal.

Simply destroying the cacheCtx won't work because that only gets rid of the Precompile Call B changes, but doesn't restore the changes from Precompile Call A.