Write barriers should not be eliminated for a StoreInstanceField
if there is a Dart call between store and object allocation.
This is implemented using Instruction::CanCallDart()
predicate, which is overridden for TemplateDartCall.
However, LoadField instructions for late fields with initializers can
also call initializer function directly without entering runtime.
This change introduces LoadFieldInstr::CanCallDart() to avoid
incorrect write barrier elimination across such LoadField instructions.
TEST=vm/cc/IRTest_WriteBarrierElimination_LoadLateField
Bug: b/204355807
Change-Id: Ibd9fe61e9723f1f4ecc9e4f5cedc3024d9d03c75
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/218583
Reviewed-by: Slava Egorov <vegorov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Instead of the background compiler task waiting forever for functions to show up in its queue, it performs a single compilation and schedules a new task if the queue is not empty.
This means an inactive isolate no longer consumes a thread. It also creates interrupt points and better fits with an embedder-provided task runner abstraction.
TEST=ci
Change-Id: I13960c776590b884fdee278d2f292bb0fa725fad
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/214242
Reviewed-by: Liam Appelbe <liama@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
This region is mostly unused after loading, but it may be accessed again if
- Isolate.spawn is invoke with isolate groups disabled
- A secondary snapshot produced by splitting is loaded
- An external typed data in the snapshot is accessed (usually a kernel file)
- Likely other cases
Even if these cases did not exist, the region is often part of a shared library and so unable to be released independently.
madvise(DONT_NEED) on this region will cause the OS to release the memory in this region but keep the address space reservation and mapping. If it is touched again, it will be brought back in from the file. If it is not backed by a file, such as malloc memory, it will be brought back in as zeros and the program will likely fail.
TEST=ci
Bug: https://github.com/dart-lang/sdk/issues/44019
Bug: https://github.com/flutter/flutter/issues/92120
Change-Id: I315a049b0f7d440e181d0a5e87fa6770a2fd4f79
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/216580
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Siva Annamalai <asiva@google.com>
Fixes a crash when SetObjectTypeAndName is called twice for the
same object id with different types.
Don't add Code objects to the profile for non-Code carrying snapshots,
as references to the Code object are replaced with references to null.
TEST=vm/dart{_2}/v8_snapshot_profile_writer_test
Bug: https://github.com/dart-lang/sdk/issues/47463
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-nnbd-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-nnbd-linux-debug-x64-try
Change-Id: Id9abf66509d422551ad5797744265eeb2c2ba2b5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/217643
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Now that Types are no longer created for top level classes, the type
class id is guaranteed to fit in ClassIdTagType. That means we can
reduce the size of Type objects slightly as well as avoid the need to
tag and untag the type class id.
Also fixes an issue uncovered in GenerateSubtype1TestCacheLookup where
the retrieved super type was not first checked to be non-null.
TEST=Refactoring, so existing tests.
Cq-Include-Trybots: luci.dart.try:vm-kernel-linux-release-ia32-try,vm-kernel-linux-release-x64-try,vm-kernel-nnbd-linux-release-simarm-try,vm-kernel-nnbd-linux-release-simarm64-try,vm-kernel-nnbd-linux-release-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm64-try,vm-kernel-precomp-linux-release-simarm_x64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-nnbd-linux-release-simarm64-try,vm-kernel-precomp-nnbd-linux-release-simarm_x64-try,vm-kernel-precomp-nnbd-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64c-try,vm-kernel-linux-debug-x64c-try,vm-kernel-precomp-linux-debug-simarm64c-try
Change-Id: I851a827843e34eac6961e0f06f888b40bf16df55
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/217221
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
The only time we create a Type in the VM for a top level class is as the
"receiver" for a no such method error where the method was either not
found in the top level class or the found method had incompatible
arguments. Unlike other cases, the Dart implementation doesn't directly
use the InvocationMirror receiver for no such method errors at the top
level, but instead just passes it back to the VM.
Now, for the InvocationMirror receiver in the top level class case, we
either
* pass a null receiver if the method wasn't found, or
* pass the signature of the found method as a string, if the arguments
were incompatible, since the only use of the top level class was for
printing the signature.
TEST=vm/cc/DartAPI, service{_2}, vm/cc/Eval, lib{_2}/mirrors
Cq-Include-Trybots: luci.dart.try:vm-kernel-linux-debug-x64-try,vm-kernel-linux-product-x64-try,vm-kernel-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-dwarf-linux-product-x64-try
Change-Id: I1ac701b0be2aaebbc6163e644262c5f3ead656c4
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/217220
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Otherwise a heap can accumulate just under 128MB (4 * max new-space capacity) worth of new-space external data without triggering a GC.
TEST=--verbose-gc
Change-Id: Ifaa5e4ea9277f1f4a09b722897b508a192c3297f
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/215142
Reviewed-by: Siva Annamalai <asiva@google.com>
Reviewed-by: Liam Appelbe <liama@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
This CL adds the `Bool` `NativeType` with support for loading/storing
from `Pointer<Bool>`, `Array<Bool>`, and `Struct` and `Union` fields,
and support for passing booleans through FFI calls and callbacks.
The assumption is that `bool` is always treated as `uint8_t` in the
native ABIs. Including: (1) whether there can be garbage in the upper
bytes in CPU registers, (2) stack alignment, and (3) alignment in
compounds.
The conversion from `bool` to `uint8_t` is implemented as follows:
- bool to int: `value ? 1 : 0`
- int to bool: `value != 0`
The conversion is implemented in Dart in patch files for memory loads
and stores (pointer, array, and struct fields) and kernel_to_il for
FFI call and callback arguments and return value.
TEST=runtime/vm/compiler/ffi/native_type_vm_test.cc
TEST=tests/ffi/bool_test.dart
TEST=tests/ffi/function_callbacks_structs_by_value_generated_test.dart
TEST=tests/ffi/function_structs_by_value_generated_test.dart
Closes: https://github.com/dart-lang/sdk/issues/36855
Change-Id: I75d100340ba41771abfb41c598ca92066a89370b
Cq-Include-Trybots: luci.dart.try:vm-kernel-linux-debug-x64-try,vm-ffi-android-debug-arm-try,vm-kernel-mac-debug-x64-try,vm-kernel-nnbd-linux-debug-x64-try,vm-kernel-linux-debug-ia32-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-reload-linux-debug-x64-try,vm-kernel-reload-rollback-linux-debug-x64-try,vm-kernel-win-debug-x64-try,vm-kernel-win-debug-ia32-try,vm-ffi-android-debug-arm64c-try,vm-kernel-mac-release-arm64-try,vm-precomp-ffi-qemu-linux-release-arm-try,vm-kernel-precomp-android-release-arm64c-try,vm-kernel-precomp-android-release-arm_x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/216900
Commit-Queue: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Clement Skau <cskau@google.com>
Avoid touching stubs in JIT mode when reading object pool
as stubs may not be initialized yet. This may happen if
using a core-jit snapshot.
TEST=ci
Change-Id: I598ef2ca4b6b22fb1bb560ad25b6ada9939799a0
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/216924
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Currently there is a strong link between dyn:-forwarder and
its target (through Function::data_ field). This link causes us
to retain some Function objects even if they are otherwise
considered dropped.
In reality this link is only ever used during compilation
or when accessing script information for the dyn:-forwarder.
As such when generating snapshots in DWARF stack traces mode
we should be able to consider this connection weak
and allow dropped Function objects to be truly discarded from
the snapshot.
This change puts WSR in between dyn:-forwarder and its target
when we know that target is not going to be retained.
We also clean how forwarding target represented to avoid
unnecessary indirection through a one-element array.
TEST=ci
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-dwarf-linux-product-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-x64-try
Change-Id: I4f0ec822ff15f78c72ff55a87d9dc52abd466cb3
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/216040
Commit-Queue: Slava Egorov <vegorov@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
The added field holds the interface signature type for a Procedure when
this is different from the class signature type (implied by the function
parameters and return type). This is needed for forwarding semi stubs
when its declared type is different from the called super member.
Closes https://github.com/dart-lang/sdk/issues/47072
TEST=pkg/front_end/testcases/general/implement_semi_stub.dart and more
Change-Id: I4a7cdc9d564fb0a59e7b06e83646b5410d8d82f5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/216420
Commit-Queue: Johnni Winther <johnniwinther@google.com>
Reviewed-by: Chloe Stefantsova <dmitryas@google.com>
Enabling the CPU profiler on OSX/iOS with a lldb connection would cause
applications to slow to a crawl due to lldb performing checks on each
SIGPROF. This change replaces the SIGPROF based thread interrupter for
MacOS with one based on the Mach Thread APIs.
Fixes https://github.com/dart-lang/sdk/issues/47139
TEST=Manual testing, CQ
Change-Id: Iedfd73a83f92d51e01b98bfa281440c7d1ba9e08
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/216220
Commit-Queue: Ben Konyi <bkonyi@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
UntaggedClass::host_type_arguments_field_offset_in_words_ is initialized
when class is finalized. When comparing generic types for the class
which was not finalized, we cannot rely on the value of that field.
Intrinsified Type.operator== has been using host_type_arguments_field_offset_in_words_
to check if the class is generic, and has been treating classes which
are not finalized as non-generic. This has been causing the wrong
result for generic types which are different only in type arguments.
This change fixes Type.operator== intrinsics to check if
type arguments match even if class was not finalized.
Also, EquivalentClassIds helper was revised to take 3 labels
explicitly for all possible outcomes (instead of 2 labels and
a fall-through).
TEST=runtime/tests/vm/dart/regress_47425_test.dart
Fixes https://github.com/dart-lang/sdk/issues/47425
Change-Id: I3f49114644d44808c668df7cbb8a702cd42b3673
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/216270
Reviewed-by: Tess Strickland <sstrickl@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Field::RecordStore inspects the object being pointed to. However,
it is not guaranteed that that objects has been initialized in
*DeserializationCluster::ReadEdges, that depends on the order of the
clusters.
In the reproduction, the TypedDataView cluster comes after the cluster
containing the instance. This leads to doing a field store while the
length field of the TypedDataView has not been read yet and is 0. This
causes a Field::RecordStore with value 0, which subsequently gets used
in compilation causing RangeErrors.
This CL delays the Field::RecordStore from ReadEdges to PostLoad so that
all objects are initialized before they are inspected by RecordStore.
TEST=tests/language/vm/regress_flutter_89584_test.dart
Fixes: https://github.com/flutter/flutter/issues/89584
Change-Id: I1d5fe25c4084d62d5e492446123987aee10200bb
Cq-Include-Trybots: luci.dart.try:vm-kernel-optcounter-threshold-linux-release-x64-try,vm-kernel-linux-debug-x64-try,vm-kernel-mac-debug-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-linux-debug-x64c-try,vm-kernel-reload-linux-debug-x64-try,vm-kernel-reload-rollback-linux-debug-x64-try,vm-kernel-nnbd-win-debug-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/216181
Commit-Queue: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
The existing lib/isolate/stacktrace_message_test was failing and also
incorrectly written due to "!" in front of the expected stringification
of the stacktrace.
Issue https://github.com/dart-lang/sdk/issues/46623
TEST=vm/dart{,_2}/isolates/fast_object_copy2_test, lib{,_2}/isolate/stacktrace_message_test
Change-Id: I169d8fd7a7c3cb3b8c57e00287565e3a5c622acf
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/216041
Reviewed-by: Alexander Aprelev <aam@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
This adds [sendMessage] and [receiveMessage] methods to [RawSocket] class. They are only supported on Linux at the moment as they require connection opened as unix domain socket connection.
This introduces [SocketControlMessage] class that represents a message sent via socket and also introduces [ResourceHandle] class that wraps [RandomAccessFile], [Socket]/[RawSocket]/[RawDatagramSocket], [Stdin]/[Stdout] for
marshalling/unmarshalling purposes.
Underlying OS implementation supports various kinds of control messages that can be passed via sockets, this CL only adds support for sending/receiving opened file description handles.
When receiving a message recipient can attempt to extract handles out of it via [tryExtractHandles]. It returns [null] if message has no handles to extract.
This is continuation of the work started on https://dart-review.googlesource.com/c/sdk/+/205067.
Bug: https://github.com/dart-lang/sdk/issues/46328
TEST=unix_socket_test
Change-Id: Ic9125b51dc80b677452e454366bae4118c298081
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/212036
Commit-Queue: Alexander Aprelev <aam@google.com>
Reviewed-by: Lasse R.H. Nielsen <lrn@google.com>
Reviewed-by: Zach Anderson <zra@google.com>
This makes parallel marking more like parallel scavenge and is preparation for allowing marking/scavenging to complete even if some workers haven't started.
Also improve work stealing of weak processing by processing the worker-local weak properties first.
TEST=ci
Change-Id: I73bfb930129e2d9848a3bbcf378303a51bd65a5d
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/215681
Reviewed-by: Liam Appelbe <liama@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Rename PointerKeyValueTrait<T>, which is used to create sets of pointers
to T instances where T <: Object and T has appropriate Hash and Equals
instance methods, to the more specific name PointerSetKeyValueTrait.
Create a PointerSet<T> alias for using this trait with
DirectChainedHashMap and use that alias in other code as a shorthand.
Remove PointerSetKeyValueTrait<const char> as a superclass of
CStringKeyValueTrait, as the only reuse from the former are the two
methods KeyOf and ValueOf which just return their argument, and having
this relationship is odd since const char is not a subtype of Object.
TEST=Renaming/refactoring, so existing tests.
Change-Id: I0274b16cb9fcb3939a28fb109fb8626c1ac8c0e9
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/215761
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
This reverts commit b5922e6db4.
Reason for revert: vm/cc/AllocateZone fails on Mac
Original change's description:
> [vm] Streamline Zones.
>
> - Don't put a HANDLESCOPE right after a StackZone, as there's no state worth restoring
> - Remove remaining zone bookkeeping dead since e3a9d70591.
> - Remove unnecessary resetting of handle blocks before zone deletion.
> - Shrink initial chunk size to make Zone's overall size fit in a faster malloc size class.
>
> TEST=ci
> Change-Id: I920e12f5af93aa71fb876435764c60df485215f9
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/212261
> Reviewed-by: Siva Annamalai <asiva@google.com>
> Commit-Queue: Ryan Macnak <rmacnak@google.com>
TBR=rmacnak@google.com,asiva@google.com
Change-Id: Id23dcebb781791616e48e54d2dcd7e58159178bc
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/215943
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
- Don't put a HANDLESCOPE right after a StackZone, as there's no state worth restoring
- Remove remaining zone bookkeeping dead since e3a9d70591.
- Remove unnecessary resetting of handle blocks before zone deletion.
- Shrink initial chunk size to make Zone's overall size fit in a faster malloc size class.
TEST=ci
Change-Id: I920e12f5af93aa71fb876435764c60df485215f9
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/212261
Reviewed-by: Siva Annamalai <asiva@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
pthread_setname_np ignores names that are too long rather than truncating. This was causing most dart:io threads to remain unnamed.
Cf. eec49f34c4.
TEST="thread info" in gdb
Change-Id: I12f47b06b81f1204f742d829e14784d792a4fb96
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/215800
Commit-Queue: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Alexander Aprelev <aam@google.com>
This change removes the extra pass over global object pool after AOT
snapshot is loaded by adding extra kSwitchableCallMissEntryPoint and
kMegamorphicCallEntryPoint object pool entry kinds which are handled
during ReadFill phase.
On a low-end phone and large Flutter app compiled in release mode
with dwarf_stack_traces, FullSnapshotReader::ReadProgramSnapshot time
Before: 232.41 ms
After: 202.43 ms (-12.8%)
Also, this change adds PrintTimeScope utility class which can be used
to measure and print time in release mode without timeline and profiling
tools:
ApiErrorPtr FullSnapshotReader::ReadProgramSnapshot() {
PrintTimeScope tm("FullSnapshotReader::ReadProgramSnapshot");
...
}
TEST=ci
Issue: https://github.com/dart-lang/sdk/issues/46116
Change-Id: I42bd46761eac8fc1e52ca695cacd2b86705034d4
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/215500
Reviewed-by: Slava Egorov <vegorov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Alexander Markov