When Dart invokes Isolate.spawn the Dart VM calls out to the embedder to
create an isolate (group) or initialize it (in case of lightweight
isolates).
* If the embedder doesn't implement this functionality or fails to
perform the creation/initialization, it will signal this failure
via the return value it gives to the VM in the call back
implementation (i.e. Dart_InitializeParams.create_group and
Dart_InitializeParams.initialize_isolate}).
* If the embeder sucessfully completed it's work, the VM owns the
isolate after the embedder call returns. The VM is then responsible
for running it.
It is an undocumented invariant atm that the embedder is responsible
to make the isolate runnable if it signals the VM that isolate was
sucessfully created/initialized.
=> If it does not do that, we have effectively an isolate leak.
=> Though right now seemingly all of our embedders to that correctly.
To avoid this unintentional isolate leak we make the isolate runnable
ourselves if embedder callback was successful but did not make it
runnable itself.
=> This also avoids unnecessary ceremony code in the embedders that
look like this:
Dart_ExitIsolate();
Dart_IsolateMakeRunnable(child);
Dart_EnterIsolate(child);
The implementation of `Dart_IsolateMakeRunnable()` had an untested and
unused code path that caused running the isolate (on a thread pool).
This is not documented and can lead to bugs because often the
embedders enter the isolate right after making it runnable (e.g. above
code sequence) - which would be racing with the spawned message handler
thread.
=> Furtunately our embedders are well behaved and make the isolate
runnable during the callback.
=> As a safeguard we'll make `Dart_IsolateMakeRunnable()` return an
error if it's called outside the callback (which we can detect by
the presense of a spawn state)
Issue https://github.com/dart-lang/sdk/issues/44088
Issue https://github.com/dart-lang/sdk/issues/36097
TEST=Changes internal details of implementation which is already covered by tests.
Change-Id: Ieef316cc0807d99f2fcb1fbd56bbc9c41e904ba8
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/170882
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Aprelev <aam@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
This reverts commit 5bf9163e1b.
Reason for revert: Tests failing on bots.
Original change's description:
> [ Service / dart:isolate ] Added getPorts RPC and 'debugName' optional
> parameter for ReceivePort and RawReceivePort
>
> This change collects additional information related to ReceivePort
> allocation locations and an optional debug name that will be displayed
> by tooling. ReceivePort is now a special InstanceKind and a ReceivePort
> @Instance will include the port ID, allocation stack trace, and debug
> name.
>
> Change-Id: I003cfff2b7649218e37d9b653c0e953df5d992e7
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/167902
> Reviewed-by: Ryan Macnak <rmacnak@google.com>
TBR=bkonyi@google.com,rmacnak@google.com,asiva@google.com
Change-Id: I39c3abb07c8c40c158eb4549749b076399bccce9
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/169160
Reviewed-by: Ben Konyi <bkonyi@google.com>
Commit-Queue: Ben Konyi <bkonyi@google.com>
parameter for ReceivePort and RawReceivePort
This change collects additional information related to ReceivePort
allocation locations and an optional debug name that will be displayed
by tooling. ReceivePort is now a special InstanceKind and a ReceivePort
@Instance will include the port ID, allocation stack trace, and debug
name.
Change-Id: I003cfff2b7649218e37d9b653c0e953df5d992e7
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/167902
Reviewed-by: Ryan Macnak <rmacnak@google.com>
This CL adds acquisiton of write lock and check for whether write lock is held when updating class functions, primarily populated during class finalization. Read locks will be added in successive CLs, so is extending of the locks coverage to all aspects of program structure changes.
Bug: https://github.com/dart-lang/sdk/issues/36097
Change-Id: I3dba6bc23db4e45599a20717226210f12e7fd2b3
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/168140
Commit-Queue: Alexander Aprelev <aam@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
This is to make sure that lookup can be done without getting a safepoint, which might be needed if class needs to be finalized.
Change-Id: I6ed52a56d53369327df0b85d7b04cf41a46ac690
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/159188
Commit-Queue: Alexander Aprelev <aam@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
- Don't use PC-relative calls when calling between loading units.
- Sort the contents of the code cluster by loading unit, then by text offset.
- Handle binding PC-relatives calls and inserting trampolines per loading unit.
- Create one code order table per loading unit to implement PC -> Code lookup.
- Read code order tables directly, instead of copying into malloc'd memory.
--use_table_dispatch still not yet supported.
This slightly shrinks non-split binaries (~2% clustered part, 0.4% total snapshot) due to the new delta encoding when Code references Instructions.
Bug: https://github.com/dart-lang/sdk/issues/41974
Change-Id: I51052ebc7b4968b554897d1710135a6c41821302
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157820
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Tess Strickland <sstrickl@google.com>
When class table is shared between isolates in isolate group it should be visited only once in (IsolateGroup/Isolate)::VisitObjectPointers.
Visiting it more than once degrades performance and potentially causes bugs since not all visitors are idempotent(GC-related).
Bug: b/163063081
Change-Id: If56fab2401ca7afe654754fead74cefa233327f6
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/158020
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Aprelev <aam@google.com>
(Assignment of libraries to loading units is already done in the kernel generation step.)
After compiling and before serializing, we walk the program and for each Code we assign its Instructions, CodeSourceMap and CompressedStackMap to the loading unit of that Code's defining library. Deduplication may cause Instructions, CodeSourceMaps and CompressedStackMaps to belong to more than one loading unit; in this case the objects are assigned to the root loading unit. Later they can be more precisely assigned to the dominating loading unit.
All objects except some Instructions, CodeSourceMaps and CompressedStackMaps belong to the root loading unit's snapshot. This snapshot is written like an unsplit snapshot, except that when serializing Code, we will write a reference to a stub or null when the Code's Instructions, CodeSourceMap or CompressedStackMap belongs to a non-root loading unit. The snapshots of non-root loading units contain these deferred objects and references to the corresponding Code objects to patch. The types of objects we defer (Instructions, CodeSourceMaps and CompressedStackMaps) usually represent 70+% of the snapshot size.
Bare instructions mode must be disabled when splitting because we cannot have PC-relative calls between loading units. Later we can re-enable this for calls within loading units.
Broken: Compactor probably crashes we can now have an unbounded number of image pages and the compactor assumes a fixed number.
Embedder's guide:
At compile-time, gen_snapshot should be passed --loading_unit_manifest with a path, which will enable splitting and output a mapping from loading unit ids to snapshot output paths.
At runtime, sometime during isolate startup, an embedder should call Dart_SetDeferredLoadHandler, probably near an existing call to Dart_SetLibraryTagHandler. The callback is given a loading unit id, and should eventually call Dart_DeferredLoadComplete[Error].
Bug: https://github.com/dart-lang/sdk/issues/41974
Change-Id: Ib597eb87c8cd634416d5ee1f00629c5550aebb00
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/152427
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Right now we assign class ids to top-level classes, abstract classes as
well as concrete classes. All of them have allocated from a 16-bit pool
of ids. The VM FATAL()s once it hits that limit.
Customers who run very large programs (significant amount of generated
code) on the Dart VM have started to hit this 16-bit class limit.
Concrete classes can have instances in the heap. Our current heap layout
only allows 16-bit class ids to be encoded in the header word. To avoid
increasing the size of heap objects or shrinking the size of the identity
hash code to 16-bit we keep class ids in object headers to be 16-bit.
Abstract classes cannot have instances in the heap. Though their class
ids are encoded in type objects. Furthermore we sort classes in
AOT/AppJIT mode to perform fast class-id range checks. To avoid impacting
this optimization we treat abstract classes the same way as concrete
classes.
Top-level classes cannot have instances in the heap. Their class ids are
only used in the runtime code, for example for hot-reload as well as
part of the service protocol.
=> We can allocate class ids outside the 16-bit range for top-level
classes, thereby freeing a significant amount of space in the 16-bit
range.
This CL does exactly that: We change classid_t to be int32_t. The
ClassLayout::id_ can now be assigned ids outside 16-bit range for
top-level classes. To do this we keep dart classes and top level classes
as separate arrays in the ClassTable.
Issue https://github.com/dart-lang/sdk/issues/42533
See also b/160229360
Change-Id: I6710a644e7b0ab2d4f4c792bef8e1f91cb117421
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/153607
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Previous version
(landed via https://dart-review.googlesource.com/c/sdk/+/135683)
only worked if the dill was loaded via
Dart_LoadScriptFromKernel where script_kernel_size was set,
and when the platform dill could be loaded from a file called the right
thing.
This CL makes it work for Dart_LoadLibraryFromKernel too and allows
the VM to send the platform along (explicitly or implicitly) and only
tries to load the platform from a file with the right name if no
platform is given in any of the inputs.
This should fix http://b/148776866
Change-Id: I62317400a932b7dcd9e126a5a88907d507f9d658
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/153609
Reviewed-by: Alexander Aprelev <aam@google.com>
Commit-Queue: Jens Johansen <jensj@google.com>
It relies on flutter copy of clang distribution, same one that is used to build flutter/engine.
It addressed several deprecated warnings from clang compiler for functions like strdup, unlink, etc.
It allows few warnings still since they are triggered in third_party sources.
Change-Id: Ieb13792c011438d46dbbc0fa030e1b5e4ea14315
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/142704
Commit-Queue: Alexander Aprelev <aam@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
- Detect null safety when not specified before isolate
initialization
- for source files by having CFE parse the source file
for @dart annotations
- for kernel files by sniffing the kernel file for
compilation mode
- for appJIT files by sniffing the feature string
- for AOT snapshots by sniffing the feature string
- Remove workaround of returning null safety to false during
bootstrapping
- Add a new Dart C API call for detecting null safety
Bug: 41766
Change-Id: Ia8cf264323a2d0d58c2855ce6491456aa6f1da07
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/150089
Commit-Queue: Siva Annamalai <asiva@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
In --enable-isolate group is
* turned off, then we have 1<->1 mapping between isolate and isolate
group, so moving the locks should have no impact
* turned on, then we have N<->1 mapping between isolates and isolate
group, mutators need to use the same locks to guarantee exclusive
access (e.g. to symbol table, type canonicalization, type arguments
canonicalization, ...)
This is a follow-up to share the program structure in AOT mode across
multiple isoltes.
In order to move the existing `Isolate::type_canonicalization_mutex_` to
`IsolateGroup` (and thereby make all mutators use the same lock), we need
to remove the `RunWithMutatorsStopped()` usage in the runtime entry.
=> Without this it can lead to deadlocks.
In order to remove usage of `RunWithMutatorsStopped()` we will change
access to the cached type arguments array to use load-acquire barriers
in generated code and store-release barriers in runtime code.
A nice side effect is that it reduces `sizeof(dart::Isolate)`.
Closes https://github.com/dart-lang/sdk/issues/41912
Issue https://github.com/dart-lang/sdk/issues/36097
Change-Id: Ifd42691524fe41ffe8bb4e2623c5b8c1151de973
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/148539
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Régis Crelier <regis@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
The most common case is to access the symbol table in read-only mode
outside a safepoint. This CL ensures that this use case is fast.
Furthermore it fixes a potential issue where one thread accesses the
symbol table using exclusive access via RW lock and is half-way-through
updating the symbol table while another thread requests a safepoint
operation, and updates the symbol table.
There is a remaining TODO item to not hold a safepoint while allocating
a symbol.
It increases performance by 10x for e.g. vm/dart/isolates/fibonacci_call_test
(26 seconds to 2.6 seconds).
Issue https://github.com/dart-lang/sdk/issues/36097
Issue https://github.com/dart-lang/sdk/issues/41943
Change-Id: I0b4e26aaffc613bed336a8845d12b7c7ee84b146
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/148323
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Aprelev <aam@google.com>
If a custom embedder does not use the VM's thread pool implementation
(1) but instead installs a custom message handler and invokes that
message handler on possibly unlimted number of threads, those mutator
threads will be constantly fighting to obtain TLABs.
To avoid this, the VM will ensure that only a limited number of mutators
can execute at the same time. The limit is the same as we use for
determining the VM's mutator thread pool.
In order to test this, we simulate such an embedder by disabling the
mutator pool limit, thereby allowing unlimited number of thread pool
workers.
=> The added test would timeout without without the new limit imposed
by this CL.
(1) Which has a maximum size adjusted based on TLAB size and new space size.
Issue https://github.com/dart-lang/sdk/issues/36097
Change-Id: I6d459f85fbeb3e4b1d756b752853157e40097271
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/148121
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
To avoid deadlock scenarios where an application runs N isolates on a
thread pool with T threads (T < N) where all scheduled T isolates will
do FFI call and block. That prevents any other isolates from executing.
To avoid such a scenario we add support for the isolate-group specific
thread pool to dynamically increase the maximum size if a worker is
running a isolate mutator thread, which calls to C which exits the
isolate.
=> While such an isolate is descheduled but still occupies a thread
pool worker we temporarily incrase the maximum size of the thread pool.
Issue https://github.com/dart-lang/sdk/issues/36097
Change-Id: Id61c39b06766da11f76d607ac7cbe8a8e623f250
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/148120
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Our TLAB sizes and maximum new space size constrain the number of
parallel mutator threads we can have. Having too many mutator threads
would cause constant races between threads to acquire TLABs.
In reality we should constrain the number of threads to be at most the
number of cores, since at most that many threads can run in parallel
(i.e. at the same time).
This CL extends the TreadPool implementation to be constrained by a
maximum size. Furthermore it makes each isolate group's have it's own
pool with constrained size and schedule all group member
mutator / message handler tasks on that pool.
Issue https://github.com/dart-lang/sdk/issues/36097
Change-Id: I095c749adad827ab892f33713a32be594d7606d1
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/145382
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Aprelev <aam@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
- change the null_safety flag to be a tri state flag (unspecified, no
null safety and null safety)
- added logic to set the null safety mode of an isolate based on the
value specified in the kernel file (.dill file) if the null safety
flag is not specified on the command line
- added logic to auto detect the null safety mode based on the language
version API provided by CFE
- added logic to pass the correct null safety option to CFE when
invoking it for compilation based on the null-safety flag setting
- Delete non-nullable-flag() function and adjust code that was using it.
- Remove 'nnbd-experiment' from the snapshot string.
https://github.com/dart-lang/sdk/issues/41206https://github.com/dart-lang/sdk/issues/41207
Change-Id: I006bf3c9229980fc7986faac6a5850d3722aec92
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/143160
Commit-Queue: Siva Annamalai <asiva@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Régis Crelier <regis@google.com>
Reviewed-by: Liam Appelbe <liama@google.com>
Reviewed-by: Johnni Winther <johnniwinther@google.com>
Martin Kustermann