Issue created by jeff-aion (on Tuesday Sep 25, 2018 at 18:16 GMT)
Due to the in-memory transaction handling of AVM, we already need to do some basic interpretation of balance transfers. We need to expand this internal transaction support to handle all cases of balance transfer transactions.
Much of the rationale for this is that it allows us to further leverage our parallel executor. This means that we can not only overlap the read-write latency of the balance IO operations, but these transactions being included in runs means that our concurrent runs will be larger (since they will only be divided by FVM calls).
Issue created by jeff-aion (on Tuesday Sep 25, 2018 at 18:16 GMT)
Due to the in-memory transaction handling of AVM, we already need to do some basic interpretation of balance transfers. We need to expand this internal transaction support to handle all cases of balance transfer transactions. Much of the rationale for this is that it allows us to further leverage our parallel executor. This means that we can not only overlap the read-write latency of the balance IO operations, but these transactions being included in runs means that our concurrent runs will be larger (since they will only be divided by FVM calls).