BinaryEncoding.md

Binary Encoding

This document describes the portable binary encoding of the Abstract Syntax Tree nodes.

The binary encoding is designed to allow fast startup, which includes reducing download size and allow for quick decoding.

Reducing download size, is achieved through three layers:

• The raw binary encoding itself, natively decoded by the browser, and to be standardized in the MVP.
• Specific compression to the binary encoding, that is unreasonable to expect a generic compression algorithm like gzip to achieve.
  • This is not meant to be standardized, at least not initially, as it can be done with a downloaded decompressor that runs as web content on the client, and in particular can be implemented in a polyfill.
  • We can do better than generic compression because we are aware of the AST structure and other details:
    • For example, macro compression that deduplicates AST trees can focus on AST nodes + their children, thus having O(nodes) entities to worry about, compared to generic compression which in principle would need to look at O(bytes*bytes) entities. Such macros would allow the logical equivalent of #define ADD1(x) (x+1), i.e., to be parametrized. Simpler macros (#define ADDX1 (x+1)) can implement useful features like constant pools.
    • Another example is reordering of functions and some internal nodes, which we know does not change semantics, but can improve general compression.
• Generic compression, such as gzip, already supported in browsers. Other compression algorithms being considered and which might be standardized include: LZMA, LZHAM, Brotli.

Each of the three layers work to find compression opportunities to the best of its abilities, without encroaching upon the subsequent layer's compression opportunities.

Why a binary encoding instead of a text-only representation

Given that text is so compressible and it is well known that it is hard to beat gzipped source, is there any win from having a binary format over a text format? Yes:

• Large reductions in payload size can still significantly decrease the compressed file size.
  • Experimental results from a polyfill prototype show the gzipped binary format to be about 20-30% smaller than the corresponding gzipped asm.js.
• A binary format that represents the names of variables and functions with raw indices instead of strings is much faster to decode: array indexing vs. dictionary lookup.
  • Experimental results from a polyfill prototype show that decoding the binary format is about 23× faster than parsing the corresponding asm.js source (using this demo, comparing just parsing in SpiderMonkey (no validation, IR generation) to just decoding in the polyfill (no asm.js code generation).

Variable-length integers

• Polyfill prototype shows significant size savings before (31%) and after (7%) compression.
• LEB128 except limited to uint32_t payloads.

Global structure

• A module contains (in this order):
  • A header, containing:
    • The magic number
    • Other data TBD
  • A table (sorted by offset) containing, for each section:
    • A string literal section type name
    • 64-bit offset within the module
  • A sequence of sections
• A section contains:
  • A header followed by
  • The section contents (specific to the section type)
• A definitions section contains (in this order):
  • The generic section header
  • A table (sorted by offset) containing, for each type which has opcodes:
    • A standardized string literal type name. The index of a type name in this table is referred to as a type ID
    • 64-bit offset of its opcode table within the section
  • A sequence of opcode tables
  • An opcode table contains:
    • A sequence of standardized string literal opcode names, where order determines opcode index
• A code section contains (in this order):
  • The generic section header
  • A table (sorted by offset) containing, for each function:
    • Signature
    • Function attributes, valid attributes TBD (could include hot/cold, optimization level, noreturn, read/write/pure, ...)
    • 64-bit offset within the section
  • A sequence of functions
  • A function contains:
    • A table containing, for each type ID that has locals:
      • Type ID
      • Count of locals
    • The serialized AST
• A data section contains (in this order):
  • The generic section header
  • A sequence of byte ranges within the binary and corresponding addresses in the linear memory

All strings are encoded as null-terminated UTF8. Data segments represent initialized data that is loaded directly from the binary into the linear memory when the program starts (see modules).

Serialized AST

• Use a preorder encoding of the AST
  • Efficient single-pass validation+compilation and polyfill
• The data of a node (if there is any), is written immediately after the opcode and before child nodes
  • The opcode statically determines what follows, so no generic metadata is necessary.
• Examples
  • Given a simple AST node: struct I32Add { AstNode *left, *right; }
    • First write the opcode of I32Add (1 byte)
    • Then recursively write the left and right nodes.
  • Given a call AST node: struct Call { uint32_t callee; vector<AstNode*> args; }
    • First write the opcode of Call (1 byte)
    • Then write the (variable-length) integer Call::callee (1-5 bytes)
    • Then recursively write each arg node (arity is determined by looking up callee in table of signatures)

Backwards Compatibility

As explained above, for size- and decode-efficiency, the binary format will serialize AST nodes, their contents and children using dense integer indices and without any kind of embedded metadata or tagging. This raises the question of how to reconcile the efficient encoding with the backwards-compatibility goals.

Specifically, we'd like to avoid the situation where a future version of WebAssembly has features F1 and F2 and vendor V1 implements F1, assigning the next logical opcode indices to F1's new opcodes, and V2 implements F2, assigning the same next logical opcode indices to F2's new opcodes and now a single binary has ambiguous semantics if it tries to use either F1 or F2. This type of non-linear feature addition is commonplace in JS and Web APIs and is guarded against by having unique names for unique features (and associated conventions).

The current proposal is to maintain both the efficiency of indices in the serialized AST and the established conflict-avoidance practices surrounding string names:

• The WebAssembly spec doesn't define any global index spaces
  • So, as a general rule, no magic numbers in the spec (other than the literal magic number).
• Instead, a module defines its own local index spaces of opcodes by providing tables of names.
  • So what the spec would define is a set of names and their associated semantics.
• To avoid (over time) large index-space declaration sections that are largely the same between modules, finalized versions of standards would define named baseline index spaces that modules could optionally use as a starting point to further refine.
  • For example, to use all of the MVP plus SIMD the declaration could be "base" followed by the list of SIMD opcodes used.
  • This feature would also be most useful for people handwriting the text format.
  • However, such a version declaration does not establish a global "version" for the module or affect anything outside of the initialization of the index spaces; decoders would remain versionless and simply add cases for new names (as with current JS parsers).