Cadence Compiler - phase 2 (WIP)

[j1010001]

Goal (WIP):
Complete Cadence compiler & VM using custom bytecode. This is middle-ground between current stack (Interpreter) and "native" compiler +VM (platform-dependent, most performant). Estimating ~2x performance increase with concurrent on-chain Tx compilation and running the compiled custom bytecode in the VM. Estimated effort 27 dev/months, ~3 quarters with 3 dev capacity.

Outcomes of this phase (WIP):

• Complete compiler + VM using custom bytecode.
• Work towards specification of execution semantics, ensure correctness
• Continue evaluation of generating compiler components from specification.
• Address tech debt, make it easier to work on new functionality and reuse existing code for prototypes
• Proactive finding of security issues, improve security posture

Main Tasks (WIP)

Preview Give feedback

1. Add diagram of compilation “pipeline” to internal docs
2. Compare RISC-V/Swift to Custom ISA/Go #3697
   Performance Research +2
   

   
3. Add tracing to interpreter #3696
   +0
   
4. Standardize internal error messages #3688
   Feedback Improvement +2
   
5. Refactor interpreter package so we can reuse built-in operations in VM #3693
   Improvement Technical Debt +2
   
6. Add some benchmark programs #3702
   Research +1
   

Stretch-goals

Preview Give feedback

1. Avoid conversion from StaticType to sema.Type during runtime sub-type checking #3691
   Improvement Performance +2

Notes
Based on internal notes: https://www.notion.so/flowfoundation/Compilation-Discussion-1441aee1232480c897aed36c8f763e30

Phase 1 outcomes

• Better understanding of what target(s) to consider - Ruled out Move VM and WebAssembly.