now in testnet feature gate queue! it took me at least a half dozen prs lol a lot of stuff was designed assuming "no conflicts in entries" is a invariant but aside from one less blocker on async execution it shld also cut back few kinds of congestion
Anza
AnzaOct 13, 23:29
1/ SIMD-0083: Relax Entry Constraints by Hanako Mumei and @apfitzge , removes the constraints that entries cannot contain conflicting transactions. This simplifies the protocol and gives block producers more flexibility in transaction ordering. Here’s what changes 🧵
rn if any tx writes to an acct all users of that accnt have to be singlethreaded, but bc of the entry constraint they also cant be batched together. so if theres a lot of write demand on an account, you only get one per batch and lots of potentially lower fee txs in between
w simd83 the batching problem goes away. so if theres a popular mint or swap pair etc, and lots of ppl pay high fees, they can all go up front on that thread together. should also help program deploys (whch are 200+ txs writing to the same account so only 1 per entry pre-83)
38.12K