Conversation
e44ed59 to
16ceb83
Compare
| Opcodes marked with ``F``, ``H``, ``B``, ``C``, ``I``, ``L``, ``P``, ``N`` and ``O`` are present since | ||
| Frontier, Homestead, Byzantium, Constantinople, Istanbul, London, Paris, Cancun or Osaka respectively. | ||
| Frontier, Homestead, Byzantium, Constantinople, Istanbul, London, Paris, Cancun, Osaka or Amsterdam respectively. |
There was a problem hiding this comment.
Why add Amsterdam here if you are not adding the one-letter mark?
| uint8_t encodeDupSwapNImmediate(size_t _depth) | ||
| { | ||
| solAssert(_depth >= 17 && _depth <= 235); | ||
| return static_cast<uint8_t>(_depth + 111); |
There was a problem hiding this comment.
Isn't this overflowing uint8_t? What am I missing here?
There was a problem hiding this comment.
It should because according to spec we have to get % 256. Maybe explicit modulo would be more readable here.
def encode_single(n: int) -> int:
assert 17 <= n <= 235
return (n + 111) % 256
f730a8c to
2ebc101
Compare
matheusaaguiar
left a comment
matheusaaguiar
left a comment
There was a problem hiding this comment.
DupN and SwapN got their own AssemblyItemType while DUP and SWAP are under AssemblyItemType::Operation.
Both versions of Dup/Swap seem to be treated uniformly through the use of factory methods, however in SemanticInformation::breaksCSEAnalysisBlock and GasMeter::estimateMax they are treated differently because of their distinct AssemblyItemType.
They should be treated the same, no?
Also, Changelog entry needed.
Right. Fixed. I removed also |
AssemblyItem::dup()/swap() create the stack manipulation items from a depth argument, and AbstractAssembly::appendDup()/appendSwap() expose them uniformly to code generators. All call sites that constructed DUP1-16/SWAP1-16 from a computed depth are converted. No functional change. Co-authored-by: Francisco Giordano <fg@frang.io>
Introduce the DUPN and SWAPN instructions with the devnet-3 immediate encoding, available in legacy bytecode starting from the "amsterdam" EVM version. The AssemblyItem::dup()/swap() factories now emit them for depths above 16, extending the reachable stack depth to 235 in legacy codegen, inline assembly and the Yul code transforms. Includes assembler, disassembler, optimizer and gas metering support as well as assembly JSON import/export. Co-authored-by: Francisco Giordano <fg@frang.io>
Add assembler, disassembler, semantic and code transform tests for DUPN and SWAPN. Tests whose expectations depend on the stack being limited to 16 reachable slots are restricted to EVM versions before "amsterdam" via the new maxEVMVersionCheck() helper or the EVMVersion setting, with counterparts exceeding the new limit of 235 where applicable. Co-authored-by: Francisco Giordano <fg@frang.io>
All space-separated constraints in the `EVMVersion` setting have to be satisfied for the test to run. This allows expressing version ranges, e.g. `>homestead <=osaka` for tests whose expectations only hold before "amsterdam" but which cannot run on the oldest versions either.
3 participants
Description
EIP-8024
AI Disclosure
Claude Fable 5 (model ID claude-fable-5) was used to make automatic changes.