This gets the phi unboxing heuristic unstuck when its input comes from a parameter, which doesn't collect type feedback to propagate a speculative concrete cid.
TEST=ci
Change-Id: I939d64b41e4c3cefcff59d25b02afba8c40975ce
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/252040
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Previously there was a window between the decision to resume generator
and actually running the micro-task. The stream could be canceled
during that window, making running the generator incorrect.
The solution is to check if the stream was canceled right before
resuming the generator in yield/yield*.
This change also unifies yield and yield* to check if the stream
was canceled even before suspending the generator.
TEST=language/async_star/cancel_while_paused_test,
language/async_star/cancel_while_paused_at_yield_test
Fixes https://github.com/dart-lang/sdk/issues/49451
Change-Id: Ib8ff3a2da9a6a7da8766a7dbf28e2c9d618728f9
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/251562
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Change the mapping between SSA indices and virtual registers from 1:1 to 1:2.
This shrinks the size of bit vectors used during optimizations, and leaves the size of bit vectors used during register allocation the same.
TEST=ci
Change-Id: I0c82ca7972f7efb30559f7e4869396f1eed757c5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/250982
Reviewed-by: Slava Egorov <vegorov@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
In order to create a Future for the result of async closure,
closure needs to instantiate a type argument of its result type.
The result type may reference type arguments of enclosing function or
class and scope builder should visit closure's result type
in order to capture receiver or type arguments variable of parent
factory.
TEST=runtime/tests/vm/dart/regress_49424_test.dart
Fixes https://github.com/dart-lang/sdk/issues/49424
Change-Id: I1cd131251717fca43f8ca95856d13718eb9aca68
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/251320
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Slava Egorov <vegorov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
This change introduces separate stubs for suspending sync* functions
at start and at yield/yield*. Suspend stub for yield/yield*
no longer calls Dart callback (in order to make it faster).
Also, ReturnSyncStar stub is removed - sync* functions now directly
return false instead of going through the stub.
TEST=ci
Issue: https://github.com/dart-lang/sdk/issues/48378
Change-Id: Iee9a1f48cab2812cf0f9f0e4e6d8e847547e49f7
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/250420
Reviewed-by: Slava Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
The kernel builder relied on only seeing expressions emitted by the
finalizable transform (variable get and this) as arguments to the fence.
However, other transforms can run later. In this case the expression
was turned into a constant.
This CL changes the implementation to accept any expression rather than
accepting a subset of expressions.
TEST=tests/ffi/regress_49402_test.dart
Closes: https://github.com/dart-lang/sdk/issues/49402
Change-Id: I1a962a5b7a38099eb5c3bbf5a5a8145b16727d97
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-debug-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/250744
Commit-Queue: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Slava Egorov <vegorov@google.com>
It was calling EnsureIsNewOrRemembered on the slow path which
was forgetting to preserve registers around a runtime call.
Fixes https://github.com/flutter/flutter/issues/106510
TEST=vm/dart{,_2}/flutter_regress_106510
Cq-Include-Trybots: luci.dart.try:vm-kernel-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm-try
Change-Id: I621e392304fcd1fd643c009fbcde3f88b6f19b7f
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/250168
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Slava Egorov <vegorov@google.com>
If a library contained an entry point but was not reachable from the root library, it was not assigned to any loading unit and caused a null dereference in gen_snapshot. This is not possible with the standalone embedder, but is possible in Flutter because it passes multiple sources to frontend_server. E.g., `--source dart_plugin_registrant.dart`.
TEST=gallery
Bug: https://github.com/flutter/gallery/issues/545
Change-Id: I9c67f0e39f7509094ee873610d80851a702a0cf2
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/249640
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
In order to know whether a value is assignable to a given type,
we need the fully instantiated type. Thus, return false if the
type given to CompilerType::IsAssignableTo is uninstantiated,
similar to CompilerType::IsInstanceOf.
TEST=vm/dart{,_2}/regress_b_230945329_test.dart
Bug: b/230945329
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-dwarf-linux-product-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-nnbd-linux-release-x64-try
Change-Id: I7ff9eb2214debc274f09580d51c80f0921f8f77d
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/244200
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
This reverts commit 69f32d6ad7.
Reason for revert: We seem to have a number of tests failing with timeouts in CBUILD after this change, please see logs at
https://dart-in-g3-qa-prod.corp.google.com/dg3/Home#/cbuild/find/69f32d6ad7e724e3148cb2eb6601e63165e76ad3
Original change's description:
> Refactor `_Future`.
>
> This is a major rewrite of the `_Future` class,
> which is the default implementation of the `Future` interface.
>
> The main goal was to reduce the number of expensive type checks
> in the internal passing around of data.
> Expensive type checks are things like
> * `is _Future<T>` (more expensive than just `is _Future`, the latter
> can be a single class-ID check.
> * Covariant generic parameter checks (using `T` covariantly in a
> parameter forces a run-time type check).
>
> Also removed some plain unnecessary casts and turned some
> implicit casts from `dynamic` into `unsafeCast`s.
>
> This seems to be an success, at least on very primitive benchmarks, according to Golem:
> FutureCatchErrorTest 41.22% (1.9 noise)
> FutureValueTest 46.51% (2.8 noise)
> EmptyFutureTest 59.15% (3.1 noise)
> FutureWhenCompleteTest 51.10% (3.2 noise)
>
> A secondary goal was to clean up a very old and messy class,
> and make it clearer for other `dart:async` how to interact
> with the future.
>
> The change has a memory cost: The `_FutureListener<S,T>` class,
> which represents a `then`, `catchError` or `whenComplete`
> call on a `_Future`, now contains a reference to its source future,
> the one which provides the inputs to the callbacks,
> as well as the result future returned by the call.
> That's one extra memory slot per listener.
>
> In return, the `_FutureListener` now does not need to
> get its source future as an argument, which needs a covariant
> generic type check, and the methods of `_Future` can be written
> in a way which ignores the type parameters of both `_Future`
> and `_FutureListener`, which reduces complex type checks
> significantly.
>
> In general, typed code is in `_FutureListener`, which knows both
> the source and target types of the listener callbacks, and which
> contains the futures already at that type, so no extra type checking
> is needed.
> The `_Future` class is mostly untyped, except for its "public"
> API, called by other classes, which checks inputs,
> and code interacting with non-native futures.
> Invariants ensure that only correctly typed values
> are stored in the untyped shared `_resultOrListeners` field
> on `_Future`, as determined by its `_state` integer.
> (This was already partially true, and has simply been made
> more consistent.)
>
> Further, we now throw an error in a situation that was previously
> unhandled: When a `_Future` is completed with *itself*.
> That would ensure that the future would never complete
> (it waits for itself to complete before it can complete),
> and may potentially have caused weird loops in the representation.
> In practice, it probably never happens. Now it makes the error
> fail with an error.
> Currently a private `_FutureCyclicDependencyError` which presents
> as an `UnsupportedError`.
> That avoids code like
> ```dart
> import "dart:async";
> void main() {
> var c = Completer();
> c.complete(c.future); // bad.
> print("well!");
> var d = Completer();
> d.complete(c.future);
> print("shucks!");
> }
> ```
> from hanging the runtime by busily searching for the end of a cycle.
>
> See https://github.com/dart-lang/sdk/issues/48225
> Fixes#48225
>
> TEST= refactoring covered by existing tests, few new tests.
>
> Change-Id: Id9fc5af5fe011deb0af3e1e8a4ea3a91799f9da4
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/244241
> Reviewed-by: Martin Kustermann <kustermann@google.com>
> Commit-Queue: Lasse Nielsen <lrn@google.com>
TBR=lrn@google.com,kustermann@google.com,sra@google.com,sigmund@google.com,nshahan@google.com
Change-Id: I455be5a04b4c346df26d4ded0fa7388baccb0f8c
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/247762
Reviewed-by: Siva Annamalai <asiva@google.com>
Reviewed-by: Alexander Aprelev <aam@google.com>
Commit-Queue: Alexander Aprelev <aam@google.com>
This is a major rewrite of the `_Future` class,
which is the default implementation of the `Future` interface.
The main goal was to reduce the number of expensive type checks
in the internal passing around of data.
Expensive type checks are things like
* `is _Future<T>` (more expensive than just `is _Future`, the latter
can be a single class-ID check.
* Covariant generic parameter checks (using `T` covariantly in a
parameter forces a run-time type check).
Also removed some plain unnecessary casts and turned some
implicit casts from `dynamic` into `unsafeCast`s.
This seems to be an success, at least on very primitive benchmarks, according to Golem:
FutureCatchErrorTest 41.22% (1.9 noise)
FutureValueTest 46.51% (2.8 noise)
EmptyFutureTest 59.15% (3.1 noise)
FutureWhenCompleteTest 51.10% (3.2 noise)
A secondary goal was to clean up a very old and messy class,
and make it clearer for other `dart:async` how to interact
with the future.
The change has a memory cost: The `_FutureListener<S,T>` class,
which represents a `then`, `catchError` or `whenComplete`
call on a `_Future`, now contains a reference to its source future,
the one which provides the inputs to the callbacks,
as well as the result future returned by the call.
That's one extra memory slot per listener.
In return, the `_FutureListener` now does not need to
get its source future as an argument, which needs a covariant
generic type check, and the methods of `_Future` can be written
in a way which ignores the type parameters of both `_Future`
and `_FutureListener`, which reduces complex type checks
significantly.
In general, typed code is in `_FutureListener`, which knows both
the source and target types of the listener callbacks, and which
contains the futures already at that type, so no extra type checking
is needed.
The `_Future` class is mostly untyped, except for its "public"
API, called by other classes, which checks inputs,
and code interacting with non-native futures.
Invariants ensure that only correctly typed values
are stored in the untyped shared `_resultOrListeners` field
on `_Future`, as determined by its `_state` integer.
(This was already partially true, and has simply been made
more consistent.)
Further, we now throw an error in a situation that was previously
unhandled: When a `_Future` is completed with *itself*.
That would ensure that the future would never complete
(it waits for itself to complete before it can complete),
and may potentially have caused weird loops in the representation.
In practice, it probably never happens. Now it makes the error
fail with an error.
Currently a private `_FutureCyclicDependencyError` which presents
as an `UnsupportedError`.
That avoids code like
```dart
import "dart:async";
void main() {
var c = Completer();
c.complete(c.future); // bad.
print("well!");
var d = Completer();
d.complete(c.future);
print("shucks!");
}
```
from hanging the runtime by busily searching for the end of a cycle.
See https://github.com/dart-lang/sdk/issues/48225Fixes#48225
TEST= refactoring covered by existing tests, few new tests.
Change-Id: Id9fc5af5fe011deb0af3e1e8a4ea3a91799f9da4
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/244241
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Lasse Nielsen <lrn@google.com>
Anis Alibegić