Even before browsers ship native support for WebAssembly, developers can ship applications on the Web using a polyfill which converts WebAssembly to JavaScript. Aside from packaging the application in a forward-looking way—so that when browsers do support WebAssembly natively, the application will immediately benefit—this can provide additional value in other ways:
- Small download size due to the binary encoding;
- Minor impact on startup performance (decoding the binary into JavaScript is fast);
- The same high throughput as in existing approaches of compiling C/C++ applications to browsers such as Emscripten, asm.js and PNaCl.
This polyfill further allows us to experiment on the early binary encoding and get developer feedback before finalizing the format and supporting it natively as part of the MVP.
A working prototype to unpack a tentative binary format into JavaScript, as well as to convert asm.js into WebAssembly (useful for existing applications), is in the polyfill repo. We also leave open the possibility of multiple polyfills existing to meet different developers' needs.
An effective polyfill should reuse much of the Web platform's existing capabilities, as detailed in the browser embedding implementation details.
An efficient polyfill may purposely diverge from the specified WebAssembly semantics: a polyfill doesn't need to be 100% correct with respect to the WebAssembly specification to be useful in practice. There are corner cases (often undefined behavior in C/C++) where JavaScript and asm.js don't have ideal semantics to maintain correctness efficiently.
If needed, a polyfill could provide an option to ensure full correctness at the expense of performance, though this is not expected to be necessary for portable C/C++ code.
Some divergences that we've identified as potentially desirable:
- Misaligned heap access: Since misaligned
loads/stores are guaranteed to produce correct results and heap accesses in
asm.js force alignment (e.g.,
HEAP32[i>>2]masks off the low two bits), an asm.js polyfill would need to translate all loads/stores into byte accesses (regardless of specified alignment) to be correct. To achieve competitive performance, the polyfill prototype defaults to incorrect behavior by emitting full-size accesses as if the index was never misaligned. Providing correct alignment information is important for portable WebAssembly performance in general; that information also guarantees that the polyfill is both correct and fast. - Out of bounds heap access: Regardless of
WebAssembly behavior, an asm.js polyfill will follow standard asm.js behavior:
- Out of bound stores are ignored (treated as no-op);
- Out of bound loads return zero for integer loads or NaN for floating point.
- 32-bit integer operators:
Regardless of WebAssembly behavior, an asm.js polyfill will follow its
standard behavior:
- Division by zero returns zero;
INT32_MIN / -1returnsINT32_MIN;- Shift counts are implicitly masked.
- Datatype conversions:
Regardless of WebAssembly behavior, an asm.js polyfill will follow its
standard behavior:
- Return zero when conversion from floating point to integer fails;
- Optionally canonicalize NaN values.
- NaN bit-pattern propagation:
Regardless of WebAssembly behavior, an asm.js polyfill will follow its
standard behavior:
- Optionally canonicalize NaN values.
The MVP feature set is expected to be entirely polyfillable effectively and efficiently. As WebAssembly evolves past MVP the working group will strive to:
- Standardize features which can be polyfilled;
- Co-evolve with other web standards bodies, ensuring that upcoming WebAssembly features remain polyfillable.
There is, however, no hard requirement that WebAssembly features be polyfillable past the MVP. If a feature is highly desirable but cannot be effectively or efficiently polyfilled then WebAssembly may adopt the feature. This is not a decision that will be taken lightly, and the WebAssembly working group will attempt to make it possible for developers to avoid the feature or provide a fallback through feature detection.