Q2 Performance

[Wodann]

I ran profiling on benchmarking scenarios when using a custom allocator. A summary of my findings:

Neptune Mutual

1. edr_evm::state::trie::account::clone (77.95%)
2. edr_evm::miner::mine_block (6.96%)

OpenZeppelin

1. edr_evm::blockchain::local::state_at_block_number (68.82%)
2. edr_evm::miner::mine_block (10.1%)
3. edr_evm::state::trie::account::clone (9.5%)

Rocketpool

1. edr_evm::blockchain::local::state_at_block_number (42.31%)
2. edr_evm::runtime::guaranteed_dry_run (15.8%)
3. edr_evm::blockchain::local::last_block (11.23%)
4. edr_evm::miner::mine_block (11.07%)
5. edr_evm::blockchain::local::block_by_number (6.53%)

Safe Contracts

1. edr_evm::miner::mine_block (48.6%)
2. edr_evm::runtime::guaranteed_dry_run (17.19%)
3. edr_evm::blockchain::local::state_at_block_number (8.78%)
4. edr_evm::blockchain::local::last_block (7.34%)

Seaport

1. edr_evm::miner::mine_block (38.19%)
2. edr_evm::state::trie::account::clone (19.8%)
3. edr_evm::runtime::guaranteed_dry_run (15.28%)
4. edr_evm::blockchain::local::last_block (9.23%)
5. edr_evm::blockchain::local::state_at_block_number (7.66%)

Uniswap

1. edr_evm::blockchain::local::state_at_block_number (74.19%)
2. edr_evm::miner::mine_block (11.26%)
3. edr_evm::state::trie::account::clone (6.68%)

Analysis

edr_evm::state::trie::account::clone and edr_evm::blockchain::local::state_at_block_number can be almost completely avoided by using a persistent data structure.

Flamegraphs

Neptune Mutual

OpenZeppelin

Rocketpool

Safe Contracts

Seaport

Uniswap