General purpose instructions for bytes

No labels

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

mighty-gerbils/gerbil-ethereum#49

kwannoel commented

2021-07-05 16:46:55 +00:00

(Migrated from github.com)

Part of https://gitlab.com/mukn/glow/-/issues/80.

Features

Batch 1 (Merged: https://github.com/fare/gerbil-ethereum/pull/46)

Methods for known-size bytes
- Compat with bytes larger than EVM Word
- Store at free memory pointer
- Load at free memory pointer
- Push for large bytes
Unit tests
Integration tests
Invariant assertions

Batch 2 (WIP: https://github.com/fare/gerbil-ethereum/pull/50)

Call by reference methods (by storing offset, length on stack)
Static offset methods for load, store. Glow compiler uses mstoreat, mloadat.
Support types with have variable sizes during runtime.
For example, given a Bytestring with a UInt64 length,
during compile time we do not know the actual length / contents of the bytestring.
instead we generate instructions to store the length:
- on stack is just takes up a slot.
- in memory it takes up 8bytes.
And use it to decode the contents:
- LOAD:
Use the offset and length on stack to retrieve the entire bytestring from memory
- STORE:
Use length on stack to store bytestring in memory
memcpy
Hashing
calldataload, calldatacopy.
Limit memory offset to 16/24 bytes due to gas limits, rather than 32.

Batch 3

Integrate with existing EVM-Types - Extend? Create new types?
- Either use .box? directly to handle boxed / unboxed types
- or use it implicitly, with a general .evm<- method for marshalling data to EVM.

Batch 4

Handle singletons.
Benchmark gas for methods
Support for marshalling? (further investigation needed)

Part of https://gitlab.com/mukn/glow/-/issues/80. ## Features ### Batch 1 (Merged: https://github.com/fare/gerbil-ethereum/pull/46) - [x] Methods for known-size bytes - [x] Compat with bytes larger than EVM Word - [x] Store at free memory pointer - [x] Load at free memory pointer - [x] Push for large bytes - [x] Unit tests - [x] Integration tests - [x] Invariant assertions ### Batch 2 (WIP: https://github.com/fare/gerbil-ethereum/pull/50) - [ ] Call by reference methods (by storing offset, length on stack) - [ ] Static offset methods for load, store. Glow compiler uses `mstoreat`, `mloadat`. - [ ] Support types with have variable sizes during runtime. For example, given a `Bytestring` with a `UInt64` length, during compile time we do not know the actual length / contents of the bytestring. instead we generate instructions to store the length: - on stack is just takes up a slot. - in memory it takes up 8bytes. And use it to decode the contents: - LOAD: Use the offset and length on stack to retrieve the entire bytestring from memory - STORE: Use length on stack to store bytestring in memory - [ ] memcpy - [ ] Hashing - [ ] calldataload, calldatacopy. - [ ] Limit memory offset to 16/24 bytes due to gas limits, rather than 32. ### Batch 3 - [ ] Integrate with existing EVM-Types - Extend? Create new types? - [ ] Either use .box? directly to handle boxed / unboxed types - [ ] or use it implicitly, with a general `.evm<-` method for marshalling data to EVM. ### Batch 4 - [ ] Handle singletons. - [ ] Benchmark gas for methods - [ ] Support for marshalling? (further investigation needed)

Rows
Columns

General purpose instructions for bytes #49

Features

Batch 1 (Merged: https://github.com/fare/gerbil-ethereum/pull/46)

Batch 2 (WIP: https://github.com/fare/gerbil-ethereum/pull/50)

Batch 3

Batch 4