Bytecode is V8's intermediate instruction format for executing a parsed JavaScript function before specialized machine-code optimization. It is not designed for human readability; it gives the runtime a compact representation that can carry execution feedback. section 6.6.2
A function usually reaches bytecode only when called. V8 pre-parses function bodies to check syntax and identify declarations, then defers full parsing and bytecode generation for unused functions. This reduces initialization work but makes a function's first call slower. section 6.6.1 section 6.6.2
A small bytecode sequence
For `add(a, b) { return a + b; }`, Ignition emits operations that load argument `b`, add argument `a`, and return the accumulator. The add instruction includes a feedback slot where V8 records type information. Consistent numeric calls can support later optimization; a string call invalidates that feedback and requires new instructions. section 6.6.2
[generated bytecode for function: add (0x289efd349e91 <SharedFunctionInfo add>)] Bytecode length: 6 Parameter count 3 Register count 0 Frame size 0 20 S> 0x2da5e7fc16f0 @ 0 : 0b 04 Ldar a1 29 E> 0x2da5e7fc16f2 @ 2 : 3f 03 00 Add a0, [0] 33 S> 0x2da5e7fc16f5 @ 5 : b3 ReturnReturn
Inspect the generated bytecode for the addition function and map its load, add, feedback-slot, and return operations to the source. section 6.6.2
Bytecode is the first execution tier in V8's pipeline: lazy parsing produces the function representation, Ignition produces bytecode, and repeated execution can lead through baseline or optimizing machine-code tiers. Execution frequency therefore determines how far compilation proceeds. section 6.6.2

