Some notes on things to improve in the future:
- Update the vendored intx library in
./3rdparty - Integrate mimalloc as default allocator via cmake; preloading shows some performance gains.
- Disable some of the de-optimizations in the intx library - they might not
be necessary anymore due to the opcode hooks in
tracecomp.hpp- non-EVM intx operators would gain performance (e.g., in the balances oracle). - Blockchain state snapshotting and access is not as fast as it could be for large imported states. Maybe use some of the actual Trie implementation from other Ethereum clients? Maybe use some other map implementation instead of STL (e.g., swisstable)? * We switched to a swisstable-based hashmap implementation and added a backing store mode to global state.
- Disable loading autodict - mostly useless because we do the same with evm2cpp one abstraction level higher and more accurate.
- Change the fuzzcase format to also include a byte-sized receiver selector
-> this would allow the fuzzer to choose the "to" field of a transaction.
Adapt the harness to select from a list of receivers.
Additionally, the mutator must be adapted accordingly to support mutating for multiple contracts at once.also done - Implement the crypto operations available through the special pre-compiled smart contracts. Should allow properly executing some of the more exotic contracts.