Venom IR

SSA-based intermediate representation for Vyper. Inspired by LLVM IR, adapted for stack-based EVM.

Primary reference: vyper/venom/README.md — full instruction set, grammar, examples.

Structure

IRContext → IRFunction(s) → IRBasicBlock(s) → IRInstruction(s)

Variables: %-prefixed, immutable after assignment (SSA)
Basic blocks: non-branching, terminated by jmp/jnz/djmp/ret/return/stop/exit
Normalized form: no block has both multiple predecessors AND multiple successors

Key Files

Core data structures in vyper/venom/:

context.py — IRContext, top-level container
function.py — IRFunction
basicblock.py — IRBasicBlock, IRInstruction, IROperand
builder.py — helper for programmatic IR construction
venom_to_assembly.py — final pass: Venom → EVM assembly
parser.py — text format parser (for testing)

Passes and analysis live in venom/passes/ and venom/analysis/ respectively.

Pass Architecture

Passes inherit from base classes in venom/passes/base_pass.py. Three categories:

Analysis — venom/analysis/ (CFG, DFG, liveness, dominators, etc.)
Transformation — normalization, SSA construction, phi elimination
Optimization — DCE, SCCP, CSE, load elimination, mem2var, inlining, etc.

Pass ordering matters — e.g. assembly emission requires normalization, which requires CFG.

Browse venom/passes/ and venom/analysis/ for the full set. File names are self-descriptive.

Inspecting Venom Output

bash

vyper -f ir_runtime contract.vy        # Venom IR (runtime)
vyper -f ir contract.vy                # Venom IR (deploy)
vyper -f cfg_runtime contract.vy       # CFG as dot graph
vyper -f asm contract.vy               # final assembly

Enable Venom: --experimental-codegen flag or #pragma experimental-codegen in source.

Entry Path

AST → vyper/codegen_venom/ → Venom IR (direct translation, no legacy IR involved). Entry point: vyper/codegen_venom/module.py.