| commit | 58af27c5e675822ffc3dfe8dc6a9f6f9d2a7cf99 | [log] [tgz] |
|---|---|---|
| author | Nate Biggs <natebiggs@google.com> | Mon Jun 13 22:34:27 2022 +0000 |
| committer | Commit Bot <commit-bot@chromium.org> | Mon Jun 13 22:34:27 2022 +0000 |
| tree | 0e4c4e7acb2cace3bc60fcd366968ac9da501b2c | |
| parent | 1219c6f33f19d12736ac3aa602a6df5a32781358 [diff] |
[dart2js] Update serialization indices to allow for deferred read/write. Results after this CL: Without the --deferred-serialization flag the linker phase finishes with ~10.2GB in ~120s. With the --deferred-serialization flag the linker phase finishes with ~8.4GB in ~118s. All other phases have negligible changes. This is a ~17.6% reduction in memory usage and a small improvement in runtime. Summary of this change: This introduces the idea of an UnorderedIndexed(Source/Sink) which removes the invariant that all data must be deserialized in the same order it is serialized. The ordered (current) version of Indexed(Source/Sink) writes a list index in place of previously indexed values. On deserialization it assumes that if it sees a new index the data must be written at that location since the read/write order are the same. The unordered (new) version of Indexed(Source/Sink) uses the data offset as the index for the data. This way whenever an index is read we know exactly where in the byte array to look it up. We move the "read head" to the specified index and perform the read operation. There is an extra complication in that indices are shared across serialized files. So an offset written in one file can refer to a location in a different file. To deal with this we offset all of the file offsets by the size of the previous files. This effectively treats all the files as chunks of contiguous memory and the offsets are within that shared address space. We must then allow for the DataSource being used by the DataSourceReader to be that of a different file for the duration of a read operation so that the data is read from the correct file. The "outside world" is unaware that the source of the data being read changed. Change-Id: I3896e3c839d8e4c4512cb5bf209dc7d2b8f60191 Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/246220 Reviewed-by: Joshua Litt <joshualitt@google.com> Reviewed-by: Sigmund Cherem <sigmund@google.com> Commit-Queue: Nate Biggs <natebiggs@google.com>
Dart is:
Optimized for UI: Develop with a programming language specialized around the needs of user interface creation.
Productive: Make changes iteratively: use hot reload to see the result instantly in your running app.
Fast on all platforms: Compile to ARM & x64 machine code for mobile, desktop, and backend. Or compile to JavaScript for the web.
Dart's flexible compiler technology lets you run Dart code in different ways, depending on your target platform and goals:
Dart Native: For programs targeting devices (mobile, desktop, server, and more), Dart Native includes both a Dart VM with JIT (just-in-time) compilation and an AOT (ahead-of-time) compiler for producing machine code.
Dart Web: For programs targeting the web, Dart Web includes both a development time compiler (dartdevc) and a production time compiler (dart2js).
Dart is free and open source.
See LICENSE and PATENT_GRANT.
Visit dart.dev to learn more about the language, tools, and to find codelabs.
Browse pub.dev for more packages and libraries contributed by the community and the Dart team.
Our API reference documentation is published at api.dart.dev, based on the stable release. (We also publish docs from our beta and dev channels, as well as from the primary development branch).
If you want to build Dart yourself, here is a guide to getting the source, preparing your machine to build the SDK, and building.
There are more documents on our wiki.
The easiest way to contribute to Dart is to file issues.
You can also contribute patches, as described in Contributing.