Alexander Markov
9f49b47823
This is a major revamp of bytecode metadata format. Now bytecode has its own serialization mechanisms. This CL adds 'bytecode component' metadata, which contains bytecode object table and string table. All references from bytecode (constant pools) to libraries, classes, members, types and strings now have a new format. References to frequently used objects are represented as indices in object table, while rarely used objects are written inline. This allows VM to cache frequently used objects while reading bytecode. Representation of strings is aligned with VM - string characters are stored in separate pools of one-byte and two-byte strings. This allows VM to avoid UTF-8 decoding and extra copying. Closure declarations are now explicit. Type parameters no longer require enslosing scopes when reading/writing them. Benchmarks: GenKernelKernelReadAllBytecode (Intel Core i5) +29.84% GenKernelKernelReadAllBytecode (Intel Xeon) +28.74% Change-Id: I4b80009733a8f8c038264af74f97c4e094b9e311 Reviewed-on: https://dart-review.googlesource.com/c/85469 Commit-Queue: Alexander Markov <alexmarkov@google.com> Reviewed-by: Régis Crelier <regis@google.com> Reviewed-by: Zach Anderson <zra@google.com>
Dart Kernel
Dart Kernel is a small high-level language derived from Dart. It is designed for use as an intermediate format for whole-program analysis and transformations, and to be consumed by codegen and execution backends.
The kernel language has an in-memory representation in Dart and can be serialized as binary or text.
Both the kernel language and its implementations are unstable and are under development.
This package contains the Dart part of the implementation and contains:
- A transformable IR for the kernel language
- Serialization of kernel code
Note: The APIs in this package are in an early state; developers should be careful about depending on this package. In particular, there is no semver contract for release versions of this package. Please depend directly on individual versions.
See ast.dart for the in-memory IR, or binary.md for a description of the binary format. For now, the textual format is very ad-hoc and cannot be parsed back in.