Deep dive into the V8 JS Engine — Flow Diagram — Node.js

V8 engine — complete internal pipeline

flowchart TD
    SRC["<b>i · Source code</b><br/>var a = 10;<br/>function sum(x,y){...}"]
      --> LEX["<b>ii · Lexical Analysis</b><br/>Code → Tokens<br/>keyword · id · op · literal"]
    LEX --> SYN["<b>iii · Syntax Analysis</b><br/>Tokens → AST<br/>tree structure"]
    SYN --> IG["<b>iv · Ignition</b><br/>AST → Bytecode<br/>interprets line by line"]
    IG -- "hot code" --> TF["<b>v · TurboFan</b><br/>Bytecode → Machine code<br/>type assumptions"]
    TF -- "wrong assumptions<br/>(deoptimize)" --> IG
    IG --> BC["<b>vi · Bytecode</b><br/>intermediate representation"]
    TF --> MC["<b>vii · Optimized machine code</b><br/>native CPU instructions"]
    BC --> EX["<b>viii · Execution</b><br/>your code runs on the CPU"]
    MC --> EX
 
    GC["<b>ix · Garbage Collector</b><br/>Mark & Sweep<br/>Orinoco · Oil Pan ·<br/>Scavenger · MCompact"] -. "runs throughout" .-> IG
    GC -. "runs throughout" .-> TF
    GC -. "runs throughout" .-> EX
 
    classDef src fill:#eaf2f8,stroke:#3a6ea5,stroke-width:2px,color:#1a1915;
    classDef parse fill:#ece5f5,stroke:#6b46c1,stroke-width:2px,color:#1a1915;
    classDef ig fill:#fdecd3,stroke:#c2410c,stroke-width:2px,color:#1a1915;
    classDef tf fill:#f5efe1,stroke:#6a8a4f,stroke-width:2px,color:#1a1915;
    classDef exec fill:#e7efd9,stroke:#587640,stroke-width:2px,color:#1a1915;
    classDef gc fill:#fde8e8,stroke:#b91c1c,stroke-width:2px,color:#1a1915;
    class SRC src
    class LEX,SYN parse
    class IG,BC ig
    class TF,MC tf
    class EX exec
    class GC gc

Source code

Your raw .js text enters V8 as a string of characters.
Lexical analysis

V8 scans character-by-character and emits tokens: keywords, identifiers, operators, literals, punctuation.
Syntax analysis

Tokens are assembled into an Abstract Syntax Tree — a hierarchical grammar structure. Invalid tokens here → SyntaxError.
Ignition

V8's interpreter: walks the AST, emits bytecode, executes line by line for fast startup. Watches for hot functions.
TurboFan

Optimizing compiler: takes hot bytecode and compiles it to optimized machine code using type assumptions. Deopts back to Ignition if the assumptions break.
Bytecode

Intermediate representation emitted by Ignition — lower-level than JS but not yet machine code.
Optimized machine code

Native CPU instructions produced by TurboFan. Runs at compiled-language speed until an assumption breaks.
Execution

Both bytecode and machine code funnel into the CPU — your program actually runs here.
Garbage collector

Mark & Sweep via Orinoco (main), Oil Pan (C++), Scavenger (young gen), MCompact (old gen) — runs concurrently throughout.

V8 engine — complete internal pipeline

Comments