The inliner code is called from AOT/JIT call specializer on functions
where it knows the target. The callsite might be dynamic, in which case
the target function is a dyn:* forwarder (for calls which have one or
more non-implicit arguments).
The dyn:* forwarders inherit the recognized kind from their target. That
means the inliner has to take special care when trying to inline
functions: It can only rely on the static type safety for interface
calls, not for dynamic calls.
This CL fixes an existing bug where the call specializer knows the
target is a list and method recognizer builds custom graph for the list
indexing operator without ensuring that the index is actually an
integer.
The existing test will start failing on a dependent CL:
tests/language_2/list/double_index_in_loop2_test.dart
Issue https://github.com/dart-lang/sdk/issues/37737
Change-Id: Ic107fe736c1b6151562880919c39d1c650c119fd
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/151849
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Tess Strickland <sstrickl@google.com>
After constant propagation is performed on the callee function body
during inlining, there are extra opportunities for specializing
calls. Unspecialized calls are not inlined, so it is important to
specialize those calls during inlining to open further opportunities
for inlining.
With this extra pass inliner is able to inline get:length after
specializing inlined List.generate constructor, and eliminate
bounds check from the initialization loop.
This change improves micro-benchmark creating 100,000 lists using
List<int>.generate(1000, (int x) => x, growable: false);
in AOT mode on x64 from 308ms to 167ms.
This makes it on par with pre-NNBD fixed-size list initialization:
var result = List<int>(1000);
for (var i = 0; i < 1000; i++) {
result[i] = i;
}
In fact, after all optimizations final flow graph (and generated
code) for List.generate and explicit loop become very similar.
This change has no effect on Flutter gallery size in release mode.
Issue: https://github.com/dart-lang/sdk/issues/42283
Change-Id: I3566386608733226e1c11cb5d6892cd869e564d7
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/151466
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Instead of having identical cases for every instruction that implements
RedefinedValue, just have one generic piece of code.
Change-Id: If3f4905a578eb345ae3d740d1f539d346c6223af
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/150985
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Aske Simon Christensen <askesc@google.com>
This includes support for calling Dart_PropagateError in native code
when doing FFI calls, and catching uncaught exceptions with Dart_IsError
when doing FFI callbacks.
The support for Dart_PropagateError adds a catch entry to the FFI
trampoline, which prevents inlining these trampolines in AOT. This
regresses the FfiCall benchmarks by 1-2% in AOT.
In addition, Dart_PropagateError requires maintaining a bit whether we
entered native/VM code from generated code through FFI or not. That way
we can do the proper transition on the exception path. When entering
generated code, we store this bit on the stack, right after the entry
frame.
Design: http://go/dart-ffi-handles
Issue: https://github.com/dart-lang/sdk/issues/36858
Issue: https://github.com/dart-lang/sdk/issues/41319
Change-Id: Idfd7ff69132fb29cc730931a4113d914d4437396
Cq-Include-Trybots: luci.dart.try:vm-ffi-android-debug-arm-try,vm-ffi-android-debug-arm64-try,app-kernel-linux-debug-x64-try,vm-kernel-linux-debug-ia32-try,vm-kernel-win-debug-x64-try,vm-kernel-win-debug-ia32-try,vm-kernel-precomp-linux-debug-x64-try,vm-dartkb-linux-release-x64-abi-try,vm-kernel-precomp-android-release-arm64-try,vm-kernel-asan-linux-release-x64-try,vm-kernel-linux-release-simarm-try,vm-kernel-linux-release-simarm64-try,vm-kernel-precomp-android-release-arm_x64-try,vm-kernel-precomp-obfuscate-linux-release-x64-try,dart-sdk-linux-try,analyzer-analysis-server-linux-try,analyzer-linux-release-try,front-end-linux-release-x64-try,vm-kernel-precomp-win-release-x64-try,vm-kernel-mac-debug-x64-try,vm-precomp-ffi-qemu-linux-release-arm-try,vm-kernel-nnbd-linux-debug-x64-try,analyzer-nnbd-linux-release-try,front-end-nnbd-linux-release-x64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/145591
Commit-Queue: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Operator== has parameter type Object, which is non-nullable in NNBD
strong mode. This is useful, as null should be handled magically by
the implementation, and the body of operator== should never receive
null.
However, VM handles comparison of an object with null inside
operator==, so its parameter type should not be taken into account
when inferring type of ParameterInstr. Otherwise implicit null
handling code is removed by the optimizer.
Fixes https://github.com/dart-lang/sdk/issues/42287
Change-Id: I856a2ce47a11309e8c32d4d9bb481d7cc03bab85
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/151021
Reviewed-by: Régis Crelier <regis@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Clean up signatures of runtime functions in object.{cc,h} using the trail to avoid cycles. Move the trail to the last parameter of these functions.
Introduce a trail in IsSubtypeOf() to avoid cycles introduced by bounds of F-bounded types. Issue uncovered by the new normalization tests.
Change-Id: I3241c7e4023a09c122e1594b7aff90b5b103f4f4
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/150180
Commit-Queue: Régis Crelier <regis@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Initializer functions are created lazily at compile time when generating
LoadField which calls initializer. This may happen in the background
compiler thread, where Field objects are cloned.
EnsureInitializerFunction should be called from original field, and it
should be properly synchronized to make sure there are no races with
mutator thread.
Change-Id: Icc339afa2ee410385019c72c717737719767d367
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/150166
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
When replacing an unboxed phi with its contant value we should not
try to insert unboxing instruction before the phi - as phis exist
outside of the normal stream of instructions and InsertBefore does
not work on them.
Instead we should insert unboxing into the block which contains
the phi.
Original test is expanded to cover this case as well now.
Fixes https://github.com/dart-lang/sdk/issues/42164
TEST=vm/cc/ConstantPropagator_Regress35371
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try
Fixed: 42164
Change-Id: I53f984216e53195206fb8247b0ed8999590415a3
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/149842
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
This addresses an issue that AOT snapshot, disassembly and flowgraph are different when generated on Windows vs Linux.
Difference is caused by different order of targets in CallTargets, caused by the fact that qsort doesn't have to preserve the order of equal elements in the original list.
While qsort on Linux preserves original (by id) order, on Windows it does not.
See https://github.com/flutter/flutter/issues/44425 for context.
Change-Id: I74c17fd83e5ae21f9733712a87de8f55a1ba94fd
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/149496
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Aprelev <aam@google.com>
Load*Field and Init*Field IL instructions are fused into load
instructions with 'calls_initializer' flag.
A new, more powerful elimination of lazy initializer calls uses
data flow analysis and load forwarding (place numbering).
In addition to improved elimination of initializer calls,
fusing instructions avoids extra load after field is already
loaded for lazy initialization check.
Fixes https://github.com/dart-lang/sdk/issues/41417
Change-Id: Ibb2a63f84b9b0c970db67b3d6684ec15384a64e7
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/148283
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Before this change we would replace unboxed definitions (those with
representation not equal to kTagged) which is too loose and causes issues
in subsequent passes, which assume that replacements have normalised
representation.
After this change we make sure that replacement has a matching
representation.
Closes https://github.com/dart-lang/sdk/issues/35371
Fixes https://github.com/dart-lang/sdk/issues/41971
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try
Change-Id: I48543893562acf4f8516651196eba7c0c0769a22
Fixed: 35371,41971
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/103139
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Previous variant of the code did not handle FutureOr<T> and Comparable<U>
correctly - variables of these types can contain a smi value if T can be a
subclass of num or if U can be num.
Rename HasNonSmiAssignableInterface to CanReceiverBeSmiBasedOnInterfaceTarget
to better conway its meaning.
Fixes https://github.com/flutter/flutter/issues/57398
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-debug-x64-try
Change-Id: I67b29908e057632e62c46fb327a5c805fcc78d74
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/149063
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
It might be too late to remove redundant phis in EliminateDeadPhis
because EliminateDeadPhis runs after OptimizeBranches, and
OptimizeBranches may be able to eliminate more branches if redundant
phis are cleaned up.
This change moves elimination of redundant phis into canonicalization
pass which is repeated more frequently.
Fixes recent regressions on several micro-benchmarks after
https://dart-review.googlesource.com/c/sdk/+/139314
Change-Id: Ib0148564615dbd0adb88e93b938ca9f3f59cabfd
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/148792
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
This saves one instruction for every BoxInt64 overflow check, reducing
the instructions size for Flutter benchmarks by between 0.26% and 1.03%
(0.26% for Flutter Gallery).
Also add OVERFLOW and NO_OVERFLOW condition aliases to arm and arm64.
Change-Id: I82990419b448f21a22ea2cc7a15a9497d3275943
Cq-Do-Not-Cancel-Tryjobs: true
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/145860
Commit-Queue: Aske Simon Christensen <askesc@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Previously, AssertSubtype instructions could only be created with a
compile-time AbstractType value for sub_type and super_type. Relax this
so that AssertSubtype instructions can be created with possibly
non-constant Values instead.
Currently, all AssertSubtype uses still have constant sub- and
supertypes, so for now we check at code generation that we indeed
have constants for these values and only handle the constant case. Thus,
there should be no impact on size or speed of the generated code from
these changes. Follow-up work will add code generation for non-constant
sub- and supertypes and add uses the generalized form.
Bug: https://github.com/dart-lang/sdk/issues/40813
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm_x64-try,vm-kernel-precomp-linux-release-simarm64-try
Change-Id: Ib1512b3e07a016d68a8e0c670ce857b8ac1b2777
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/147520
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Also base these and the corresponding methods in _OneByteString on
functions in dart:_internal, so they can be used in other patches.
Change-Id: Ibced31758db2959c111a01cdaa46df2971ae4a6c
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/146784
Reviewed-by: Martin Kustermann <kustermann@google.com>
Previously, AssertAssignable instructions could only be created with a
compile-time AbstractType value for dst_type. Relax this so that
AssertAssignable instructions can be created with a possibly
non-constant Value instead.
This is only an IL-level change for the instruction. Code generation for
AssertAssignable still requires the dst_type Value to be bound to a
constant and is otherwise unchanged from before, thus this should have
no impact on size or speed of generated code. (Generating the
non-constant case will be done in follow-up work.)
Bug: https://github.com/dart-lang/sdk/issues/40813
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm_x64-try,vm-kernel-precomp-linux-release-simarm64-try,vm-dartkb-linux-release-simarm64-try,vm-dartkb-linux-release-x64-try
Bug: https://github.com/dart-lang/sdk/issues/40813
Change-Id: I1f984039d8c2695c66161a69a3d80cfbe7110f19
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/146801
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
There was a code in the register allocator that recognized whether
destination is a spill slot or not by checking if it is one of
different location types. By accident QuadSpillSlot was missing
from that list - and as a result we were emitting redundant moves
when connecting split siblings after register allocation.
Change the code to just ask parent live which spill slot it was assigned
and use that to drive decisions.
https://github.com/dart-lang/sdk/issues/41805
Change-Id: I73e208acd0bf86dceeaad970b8e5cbcb9c3f8eaa
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/147550
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Additionally make it possible to filter the output of this tracing
using --print-flow-graph-filter=...
Expose should_filter() setting through CompilerState - to avoid
plumbing FlowGraph or Function through various methods that don't
already have access to either of these two.
Change-Id: Ia1b8bcff5702e869db23ea4407191741662b2417
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/147544
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Originally, the assert here required the location be a machine register,
invalid, or a pair location if always_calls() was true, which meant
that SameAsFirstInput() could not be used.
Extend the assert to allow SameAsFirstInput(), but only if the location
of the first input is a machine register, invalid, or a pair location.
Also make sure any further changes to the first input remain consistent
with this expectation.
Fixes: https://github.com/dart-lang/sdk/issues/32787
Cq-Include-Trybots: luci.dart.try:vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-debug-simarm_x64-try
Change-Id: If43eb5399283d23f62a502edb0a9ce9c82813b3b
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/147101
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
The following options are removed:
--use_strong_mode_types
--experimental_unsafe_mode_use_at_your_own_risk
The following flags and predicates are removed:
Dart_IsolateFlags::unsafe_trust_strong_mode_types
Isolate::argument_type_checks()
Isolate::can_use_strong_mode_types()
Isolate::should_emit_strong_mode_checks()
Also, everything depending on these flags is cleaned up.
Change-Id: I9328009ad5a42ea2173842386d612c465e3ebec1
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/147325
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
IsTopTypeForAssignability is renamed to IsTopTypeForSubtyping to better
reflect that it works both for subtyping and assignability.
Most uses of IsTopType() are changed to IsTopTypeForSubtyping().
The remaining uses of IsTopType() are checking for top types in
instance-of ('is') tests, so IsTopType is renamed to IsTopTypeForInstanceOf.
Also, Instance::IsFutureOrInstanceOf is renamed to RuntimeTypeIsSubtypeOfFutureOr
to align it with RuntimeTypeIsSubtypeOf and reflect that it is
calculating subtyping and not instance-of.
Change-Id: Ib8d3574edc1b90d80f5528330bd84b099b37d2b8
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/145446
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Régis Crelier <regis@google.com>
When using C++ value types to represent tagged pointers, we cannot use forward declarations for tagged pointers. This helps to break include cycles when attempting to keep IsXYZ predicates as tagged pointer member functions.
class_id.h also seems like a more natural place for these predicates, which were written before there was a class_id.h.
Change-Id: I0677560a794ed084d10f844606e202feb0c3820a
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144321
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
For some reason constexpr seems to need a definition, because the linker
complains about not being able to find the symbol:
% tools/gn.py -ax64 -mdebug -d0
% tools/build.py -ax64 -mdebug
...
ld.lld: error: undefined symbol: dart::FlowGraphBuilderHelper::kPhiSelfReference
>>> referenced by constant_propagator_test.cc:53 (../../runtime/vm/compiler/backend/constant_propagator_test.cc:53)
...
We may also want to consider passing `-fmerge-all-constants` explicitly
in the future - though it is not standards compliant.
See also http://crbug.com/829795.
Change-Id: I1cc125a01751c8a83092b033e80afdcc591c7b92
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/144287
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Martin Kustermann <kustermann@google.com>
Ryan Macnak