react/compiler/packages/babel-plugin-react-compiler/docs/passes/05-deadCodeElimination.md

deadCodeElimination

File

src/Optimization/DeadCodeElimination.ts

Purpose

Eliminates instructions whose values are unused, reducing generated code size. The pass performs mark-and-sweep analysis to identify and remove dead code while preserving side effects and program semantics.

Input Invariants

• Must run after InferMutationAliasingEffects because "dead" code may still affect effect inference
• HIR is in SSA form with phi nodes
• Unreachable blocks are already pruned during HIR construction

Output Guarantees

• All instructions with unused lvalues (that are safe to prune) are removed
• Unused phi nodes are deleted
• Unused context variables are removed from fn.context
• Destructuring patterns are rewritten to remove unused bindings
• StoreLocal instructions with unused initializers are converted to DeclareLocal

Algorithm

Two-phase mark-and-sweep with fixed-point iteration for loops:

Phase 1: Mark (findReferencedIdentifiers)

1. Detect if function has back-edges (loops)
2. Iterate blocks in reverse postorder (successors before predecessors) to visit usages before declarations
3. For each block:
   • Mark all terminal operands as referenced
   • Process instructions in reverse order:
     • If lvalue is used OR instruction is not pruneable, mark the lvalue and all operands as referenced
     • Special case for StoreLocal: only mark initializer if the SSA lvalue is actually read
   • Mark phi operands if the phi result is used
4. If loops exist and new identifiers were marked, repeat until fixed point

Phase 2: Sweep

1. Remove unused phi nodes from each block
2. Remove instructions with unused lvalues using retainWhere
3. Rewrite retained instructions:
   • Array destructuring: Replace unused elements with holes, truncate trailing holes
   • Object destructuring: Remove unused properties (only if rest element is unused or absent)
   • StoreLocal: Convert to DeclareLocal if initializer value is never read
4. Remove unused context variables

Key Data Structures

• State class: Tracks referenced identifiers
  • identifiers: Set<IdentifierId> - SSA-specific usages
  • named: Set<string> - Named variable usages (any version)
  • isIdOrNameUsed() - Checks if identifier or any version of named variable is used
  • isIdUsed() - Checks if specific SSA id is used
• hasBackEdge/findBlocksWithBackEdges: Detect loops requiring fixed-point iteration

Edge Cases

• Preserved even if unused:

  • debugger statements (to not break debugging workflows)
  • Call expressions and method calls (may have side effects)
  • Await expressions
  • Store operations (ComputedStore, PropertyStore, StoreGlobal)
  • Delete operations (ComputedDelete, PropertyDelete)
  • Iterator operations (GetIterator, IteratorNext, NextPropertyOf)
  • Context operations (LoadContext, DeclareContext, StoreContext)
  • Memoization markers (StartMemoize, FinishMemoize)

• SSR mode special case:

  • In SSR mode, unused useState, useReducer, and useRef hooks can be removed

• Object destructuring with rest:

  • Cannot remove unused properties if rest element is used (would change rest's value)

• Block value instructions:

  • Last instruction of value blocks (not 'block' kind) is never pruned as it's the block's value

TODOs

• "TODO: we could be more precise and make this conditional on whether any arguments are actually modified" (for mutating instructions)

Example

Input:

function Component(props) {
  const _ = 42;
  return props.value;
}

After DeadCodeElimination: The const _ = 42 assignment is removed since _ is never used:

function Component(props) {
  return props.value;
}

Array destructuring example:

Input:

function foo(props) {
  const [x, unused, y] = props.a;
  return x + y;
}

Output (middle element becomes a hole):

function foo(props) {
  const [x, , y] = props.a;
  return x + y;
}