- From v1.10.14 onwards, validators will try to consume transactions in order of priority which is based on the prioritization fee paid per compute unit.
- In order to optimize priority / fee paid, transaction should set its compute unit limit as low as possible because leftover compute units are not refunded.
tx_wide_compute_cap (Implemented in v1.9.7, v1.10.0) (Activated on Devnet & Testnet)
- Updates compute unit limits to be transaction-wide rather than reset to 200k compute units per transaction instruction. Sets the maximum transaction compute unit limit to 1.4M compute units. The default compute unit limit is also set to 1.4M compute units to avoid breaking existing transactions which rely on consuming more than 200k compute units by spreading compute across multiple instructions.
- Adds the
ComputeBudget::RequestUnitsinstruction which has two parameters:- Compute unit limit
- Prioritization fee (lamports)