This reverts commit 80102c981b.
Reason for revert: b/186359854
Original change's description:
> [vm] Switch datastream Write/WriteUnsigned to (S)LEB128.
>
> This reduces the number of variable-length integer encodings in our
> database from 2 to 1, and chooses the more standard one.
>
> TEST=Existing tests, in particular any that involve snapshots.
>
> Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-asan-linux-release-x64-try,vm-kernel-precomp-dwarf-linux-product-x64-try,vm-kernel-precomp-msan-linux-release-x64-try,vm-kernel-precomp-ubsan-linux-release-x64-try,vm-kernel-precomp-tsan-linux-release-x64-try,vm-kernel-precomp-nnbd-linux-debug-x64-try,vm-kernel-precomp-nnbd-linux-debug-simarm_x64-try,vm-kernel-precomp-linux-debug-simarm64c-try,vm-kernel-precomp-linux-debug-simarm_x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-debug-x64c-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-release-x64-try
> Change-Id: Ia700158ac873ad32ac28c1027a669895961bc715
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/196321
> Commit-Queue: Tess Strickland <sstrickl@google.com>
> Reviewed-by: Ryan Macnak <rmacnak@google.com>
> Reviewed-by: Martin Kustermann <kustermann@google.com>
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: Idf32bdd879cf8bb7407f6dae764312140ad6eeb2
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-asan-linux-release-x64-try,vm-kernel-precomp-dwarf-linux-product-x64-try,vm-kernel-precomp-msan-linux-release-x64-try,vm-kernel-precomp-ubsan-linux-release-x64-try,vm-kernel-precomp-tsan-linux-release-x64-try,vm-kernel-precomp-nnbd-linux-debug-x64-try,vm-kernel-precomp-nnbd-linux-debug-simarm_x64-try,vm-kernel-precomp-linux-debug-simarm64c-try,vm-kernel-precomp-linux-debug-simarm_x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-debug-x64c-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-release-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/196920
Reviewed-by: Ivan Inozemtsev <iinozemtsev@google.com>
Commit-Queue: Ivan Inozemtsev <iinozemtsev@google.com>
Right now the vm/cc/CodeRelocator_* tests are written in a strange way
to work around the conservative assumptions the relocator makes.
A future CL will make the relocator insert tramppolines precisely when
they are needed (not conservatively), which will allow writing precise
tests.
TEST=Fixes ARM64 vm/cc/CodeRelocator_* tests
Change-Id: I68f9bcce29f42a2264608f867f177e8f557804fb
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/196666
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
The tests generate & run relocated instructions. On ARM64 our macro
assembler uses X15 as stack pointer instead of the normal sp.
-> There was a missing RestoreCSP() in the generation of the call
instructions.
-> It's not an issue for running the test in simulator if CSP is wrong
when returning, though it is for HW.
Adding RestoreCSP() would require the code to be larger, so instead we
use lower level assembly instructions that operate on the normal ARM64
stack.
Closes https://github.com/dart-lang/sdk/issues/45807
TEST=Fixes vm/cc/RelocatinoTest_* on arm64 hw
Change-Id: I0c65e3d22897f536e39f0cf7e9af2be871051b0d
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/196498
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
This reduces the number of variable-length integer encodings in our
database from 2 to 1, and chooses the more standard one.
TEST=Existing tests, in particular any that involve snapshots.
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-asan-linux-release-x64-try,vm-kernel-precomp-dwarf-linux-product-x64-try,vm-kernel-precomp-msan-linux-release-x64-try,vm-kernel-precomp-ubsan-linux-release-x64-try,vm-kernel-precomp-tsan-linux-release-x64-try,vm-kernel-precomp-nnbd-linux-debug-x64-try,vm-kernel-precomp-nnbd-linux-debug-simarm_x64-try,vm-kernel-precomp-linux-debug-simarm64c-try,vm-kernel-precomp-linux-debug-simarm_x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-debug-x64c-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-release-x64-try
Change-Id: Ia700158ac873ad32ac28c1027a669895961bc715
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/196321
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
The following fields in Class objects
target_instance_size_in_words_
target_type_arguments_field_offset_in_words_
target_next_field_offset_in_words_
are only useful when cross-compiling.
This change removes these fields if DART_PRECOMPILER is not defined.
TEST=ci
Change-Id: I45e6e60ad6d0a738e2ac3afd45328209167393ff
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/196222
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Previously, compiler::target::Foo::InstanceSize(intptr_t length)
definitions were hand-written and compiler::target::Foo::InstanceSize()
were often defined using the size of UntaggedFoo for array-like objects.
This results in an unnecessary code duplication and discrepancy
between Foo::InstanceSize() which returns 0 and
compiler::target::Foo::InstanceSize() which returns size of the header.
This change adds ARRAY_SIZEOF variant to runtime_offsets_list.h
and switches array-like objects to auto-generated InstanceSize methods.
Also, this change strengthens assertions around host and target
values of InstanceSize() and NextFieldOffset() being the same
unless DART_PRECOMPILER is defined (which may do cross-compilation).
Stronger assertions flushed out a few discrepancies between
runtime_api.h and object.h, which are also fixed in this change.
TEST=ci
Change-Id: I9f67c6a71b9300af5e656fdc70f6a3e341c1d029
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/196345
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Tess Strickland <sstrickl@google.com>
This is a reland of 2852408881
Original change's description:
> [vm] Reduce size of Class objects in AOT and PRODUCT modes
>
> This change conditionally removes certain fields from Class objects
> in AOT and PRODUCT modes.
>
> Flutter gallery in release-sizeopt mode:
> Snapshot size -0.47% (arm64), -0.44% (arm)
> Heap size of snapshot objects -4.1% (arm64), -4.3% (arm)
>
> On a large Flutter application compiled in --dwarf-stack-traces mode:
> Heap size of Class objects -26%
> Heap size of all snapshot objects -3.6%
>
> TEST=ci
> Issue: https://github.com/dart-lang/sdk/issues/44020
> Change-Id: I795c625b71558cd2f336f92fc326c36fd339cd4b
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/195700
> Commit-Queue: Alexander Markov <alexmarkov@google.com>
> Reviewed-by: Ryan Macnak <rmacnak@google.com>
TEST=ci
Change-Id: Ib473a959a2c610b7a0e6f4000b2101e5256c6119
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/196103
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
This change conditionally removes certain fields from Class objects
in AOT and PRODUCT modes.
Flutter gallery in release-sizeopt mode:
Snapshot size -0.47% (arm64), -0.44% (arm)
Heap size of snapshot objects -4.1% (arm64), -4.3% (arm)
On a large Flutter application compiled in --dwarf-stack-traces mode:
Heap size of Class objects -26%
Heap size of all snapshot objects -3.6%
TEST=ci
Issue: https://github.com/dart-lang/sdk/issues/44020
Change-Id: I795c625b71558cd2f336f92fc326c36fd339cd4b
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/195700
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
There was a missing check for whether a backwards call is out-of-range.
The CL also removes leftover code related to [max_offset_into_target_] -
which wasn't used. The intended use is already over approximated via
[max_instructions_size_].
The CL makes the pc-relative call/tail-call distances pluggable so a
newly added test can modify them for testing in-range/out-of-range
forward/backwards calls.
TEST=vm/cc/CodeRelocator_*
Fixes https://github.com/flutter/flutter/issues/80043
Change-Id: Id4bdb7176108b61235dafb7ffc125da4a2bf07fa
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/195682
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
It is possible for 2 slots which correspond to different captured
variables to have the same offset, flags and name.
So Slot::Equals should also take the rest of the information into
account (static type) in order to prevent incorrect canonicalization
of Slot objects.
TEST=runtime/tests/vm/dart/regress_45631_test.dart
Fixes https://github.com/dart-lang/sdk/issues/45631
Change-Id: Ib64dc5b8a83f64efd06f4139949c637d7c6dd3a8
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/195423
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Previously, compiler canonicalized Phi instructions which reference
only one value (modulo redefinitions):
v1 = Phi(v2, v2, ..., v2)
This change extends the canonicalization rule to also allow the Phi
to reference itself:
v1 = Phi(v2, v2, ...., v2, v1, ...., v1)
Such Phi instructions may occur in loops when local variable is not
mutated (remains the same on the back-edge).
TEST=ci
Change-Id: I2838c4307a6dd3067d52624db66487ba4682538c
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/195043
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
In the codebase, we have several different interfaces for equality
instance methods on a type T:
- bool Equals(const T&) const;
- bool Equals(const T*) const;
- bool Equals(T*) const;
This CL standardizes on `bool Equals(const T&) const;`, and adjusts any
related functions to match.
TEST=Existing test suites, as this is an internal refactoring.
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-debug-x64-try
Change-Id: I2ebcf2b861732ce65c6f1c673b2df2a2c7f5dd4d
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/195071
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
In the codebase, we have several different interfaces for instance
methods that return a hash:
- uword Hash() const;
- intptr_t Hash() const;
- uint32_t Hash() const;
- intptr_t Hashcode() const;
This CL standardizes on `uword Hash() const` and adjusts any related
functions to match.
TEST=Existing test suites, as this is an internal refactoring.
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-debug-x64-try
Change-Id: If2cbce57f3fae0f0d24031b6e324f0323c965f41
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/195067
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
The StoreIndexedInstr::EmitNativeCode will use the "index" register
as is and not zero-extend it. That means the "index" register
cannot have kUnboxedInt32 representation (which allows upper 32
bits to contain garbage)
TEST=Existing test suite.
Change-Id: I413847ec9a43ff446a4f89c78dce7c4fdbceca10
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/195066
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
Introduces the class `_Compound extends NativeType`, for sharing between
`Struct` and `Union`.
Does minimal changes to CFE and IL construction. Follow up refactor CLs
rename everything.
Bug: https://github.com/dart-lang/sdk/issues/38491
tools/test.py ffi ffi_2
TEST=tests/ffi(_2)/(.*)by_value_(*.)_test.dart
Change-Id: I276ceb9249c20bd331c2f8b6ef64e35acb525e9c
Cq-Include-Trybots: luci.dart.try:vm-precomp-ffi-qemu-linux-release-arm-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-reload-linux-debug-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/194765
Reviewed-by: Aske Simon Christensen <askesc@google.com>
While deserializing AOT snapshot, Code objects which do not contain
valuable information besides entry point and stack maps are discarded
and not allocated on the heap (they are replaced with
StubCode::UnknownDartCode()).
PC -> Code/CompressedStackMaps lookup is implemented using a separate
table (InstructionsTable).
Flutter gallery in release-sizeopt mode:
Heap size of snapshot objects: arm -26.89%, arm64 -27.68%
Large Flutter application in release mode with --dwarf-stack-traces:
Heap size of snapshot objects: -24.3%.
Discarded Code objects: 72.5% of all Code objects.
Issue: https://github.com/dart-lang/sdk/issues/44852.
TEST=existing tests; "--dwarf_stack_traces --no-retain_function_objects
--no-retain_code_objects" mode is enabled for a few tests.
Change-Id: I5fe3e283630c8e8f4442319d5dcae38d174dd0d8
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/189560
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Adds a way to express native effects in Dart expressions in kernel.
This CL adds a `void _nativeEffect(Object)` to `dart:internal`.
The semantics of `_nativeEffect` are to not execute its arguments and
return `null`.
This CL uses this `_nativeEffect` to make sure that we never execute
the struct constructor invocations used to simulate the native behavior
of FFI trampolines.
Closes: https://github.com/dart-lang/sdk/issues/45607
TEST=tests/ffi(_2)/native_effect_test.dart
Change-Id: Ie06de145e49f8b1cae9e148c2d5d97d5cd8e6878
Cq-Include-Trybots: luci.dart.try:vm-precomp-ffi-qemu-linux-release-arm-try,vm-ffi-android-debug-arm64-try,vm-kernel-precomp-asan-linux-release-x64-try,vm-kernel-precomp-dwarf-linux-product-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-obfuscate-linux-release-x64-try,analyzer-analysis-server-linux-try,analyzer-linux-release-try,dart-sdk-linux-try,vm-kernel-reload-rollback-linux-debug-x64-try,vm-kernel-reload-linux-debug-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/194421
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Daco Harkes <dacoharkes@google.com>
This Splits the NativeCompoundType into NativeStructType and
NativeUnionType.
The calling conventions themselves did not have to be updated at all,
they rely on the 'queries' on NativeTypes.
Some interesting corner cases in the native calling conventions are:
1. Arm64 and arm hardfp still regard unions as homogeneous floats, even
if the union members are different sizes.
2. Unions can be larger than their members if their largest member has
a smaller alignment than a smaller member.
3. When a specific range in a struct contains floating points in one
member, but integers in another, the integers take precedence.
Bug: https://github.com/dart-lang/sdk/issues/38491
tools/build.py run_ffi_unit_tests && tools/test.py ffi_unit
TEST=runtime/vm/compiler/ffi/native_calling_convention_test.cc
TEST=runtime/vm/compiler/ffi/native_type_test.cc
These tests are exercised on vm-precomp-ffi-qemu-linux-release-arm-try.
Change-Id: Ib97769457d1ad0fce47601e9e4a3008bd837167f
Cq-Include-Trybots: luci.dart.try:vm-precomp-ffi-qemu-linux-release-arm-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/194422
Commit-Queue: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Aske Simon Christensen <askesc@google.com>
This change removes code duplication between
LoadIndexedInstr::representation() and
StoreIndexedInstr::RequiredInputRepresentation(), multiplied by 4
target architectures.
This change also removes duplicated logic from FFI load and store
intrinsics.
TEST=ci
Change-Id: I9f10cc8c70288ecaea33a337a880c181ecbfcd6a
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/194281
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
This CL enables tree shaking of `Struct` sub classes by simulating the
native behavior in Dart code.
We call the struct constructors in let expressions where FFI trampolines
are created which allocate these struct objects in native code. This way
TFA is instructed about the native behavior. The VM recognizes these
constructor calls as dead code and removes them.
For more info see go/dart-ffi-struct-treeshaking.
Closes: https://github.com/dart-lang/sdk/issues/38721
This CL fixes the types on the generated #fromTypedDataBase constructors.
Also, this CL also cleans up the struct naming:
* _addressOf -> _typedDataBase
* _fromPointer -> _fromTypedDataBase
* #pointer -> #typedDataBase
These cleanups are not split into a separate CL to prevent updating
the .expect files multiple times.
Finally, this CL enables running a single transformer test through:
`dart pkg/vm/test/transformations/type_flow/transformer_test.dart name`
TEST=pkg/vm/testcases/transformations/type_flow/transformer/ffi_struct_constructors.dart
TEST=tests/ffi(_2)/function_callbacks_structs_by_value_generated_test.dart
TEST=tests/ffi(_2)/function_structs_by_value_generated_test.dart
Change-Id: I418d0d73bc86b234dfe5b7b04ae726c33d2b8aeb
Cq-Include-Trybots: luci.dart.try:vm-precomp-ffi-qemu-linux-release-arm-try,vm-ffi-android-debug-arm64-try,vm-kernel-precomp-asan-linux-release-x64-try,vm-kernel-precomp-dwarf-linux-product-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-obfuscate-linux-release-x64-try,analyzer-analysis-server-linux-try,analyzer-linux-release-try,dart-sdk-linux-try,vm-kernel-reload-rollback-linux-debug-x64-try,vm-kernel-reload-linux-debug-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/193661
Commit-Queue: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Previously, FFI store could use kUnboxedUint32 for value being stored
via 8-bit or 16-bit StoreIndexed instruction. However, such
StoreIndexed instructions require kUnboxedIntPtr representation.
Due to the mismatch in the representations, SelectRepresentations
pass inserts a speculative (deoptimizing) IntConverter instruction,
which cases crash in AOT mode. Similar problem exists for FFI loads.
This change corrects representation when unboxing value in the body
of FFI store intrinsics and when boxing the value in FFI loads,
so representation of the value matches representation required by
StoreIndexed / returned by LoadIndexed.
TEST=ffi/regress_flutter79441_test
Fixes https://github.com/flutter/flutter/issues/79441
Change-Id: Ida144e8d2e7a69d6767c9d4447bb20e79d847d48
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/193824
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Before value is stored into Float32List, it is converted to 32-bit
float using DoubleToFloat instruction. If allocation sinking
eliminated the allocation of Float32List but we need to deoptimize,
the list is materialized and elements are filled. In such case,
we shouldn't perform double->float conversion as it already happened.
This change also updates the assertion in DoubleToFloatInstr::Canonicalize
which verifies that DoubleToFloat instruction can be used only
in certain cases.
TEST=tests/language/vm/allocation_sinking_arrays_test.dart
Fixes https://github.com/dart-lang/sdk/issues/45547
Change-Id: I9a1bd28a9fc09bccad6aa3c91a7880abd002f7c9
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/193831
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
This change extends EqualityCompare IL instruction to handle case
when operands are nullable ints.
Unlike CheckedSmiComparison, EqualityCompare performs comparison
without making a dynamic call on a slow path. Dynamic call in
CheckedSmiComparison is a problem because it is not present in
the original program and not properly tracked in TFA and precompiler,
so precompiler is not able to drop Code objects (needed for
dynamic calls) reliably.
This change removes the remaining case when CheckedSmiComparison
instruction was used. The cleanup will follow in a separate CL.
TEST=ci
Issue: https://github.com/dart-lang/sdk/issues/32166
Issue: https://github.com/dart-lang/sdk/issues/44852
Change-Id: I8114a2616becf632462d4cb3bfc255691f24e1ef
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/193447
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Function type parameters in occurring in constants were using an offset
based on the context in which the constant first occurred. This meant
that function type parameters occurring in a generic context would have
a different offset that those occurring in a non-generic context.
The loading of .dill in ast_from_binary would read all constants outside
the context and would therefore get the indices wrong on generic
function types.
This CL changes the encoding of these type parameters to always use
a fresh context.
Closes#45415
TEST=pkg/front_end/testcases/generic_metadata/from_dill/main.dart
Change-Id: Ifdaedb4581f1e022c908fc58a58ceac3d7b0900e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/193481
Commit-Queue: Johnni Winther <johnniwinther@google.com>
Reviewed-by: Dmitry Stefantsov <dmitryas@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Ivan Inozemtsev