We create the parameter names array when loading the
FunctionDeclaration, but the required parameter flags are only set later
when we load the Code declaration. If there's an intermediate use of the
Function object that copies the required parameter flags, like creating
an ImplicitClosureFunction when loading a tearoff stored in the object
pool, there will be no set flags at that point.
Add a copy of the parameter flags to the FunctionDeclaration so the
required flags can be set at the same time as the parameter name and
type to avoid this scenario.
To avoid adding unneeded flags to the kernel bytecode, we only serialize
the parameter flags needed for each declaration, so FunctionDeclaration
and ClosureDeclaration only contain the required flag, and Code contains
all flags _but_ the required flag.
There are no trybots for strong mode + bytecode, but manually checked
by running the following on the main branch (cherry-picking the
test_matrix.json changes) and this branch:
python tools/test.py \
-n dartkb-mixed-strong-linux-release-x64,dartkb-interpret-strong-linux-release-x64 \
tests/language/nnbd/required_named_parameters/missing_required_argument_dynamic_test.dart
Cq-Include-Trybots: luci.dart.try:vm-dartkb-linux-release-simarm64-try,vm-dartkb-linux-release-x64-try
Change-Id: Ic75b831d5ceed08154e2c61b8f64461705558653
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/162501
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Current users of InstantiateTypeArguments have the uninstantiated type
arguments available at compile time. After upcoming work on #40813,
there are cases where the generated code needs to instantiate type
arguments that are only available at runtime.
This work changes the uninstantiated type arguments for the
InstantiateTypeArguments instruction to a runtime Value pointer instead
of a compile time TypeArguments reference. The changes beyond that are
limited to only what is needed to get current uses working, with ASSERTs
checking that current invariants still hold. The rest of the changes
needed for dynamic uninstantiated type arguments will come later.
Bug: https://github.com/dart-lang/sdk/issues/40813
Cq-Include-Trybots: luci.dart.try:vm-kernel-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm_x64-try,vm-kernel-linux-release-simarm-try,vm-kernel-linux-release-simarm64-try
Change-Id: Iafe836cfd8744d3e835e60112a4f099e631f4782
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/162183
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Leaf runtime calls a silently clobbering R23 and R25 registers, which
causes bugs when these calls happen from write barriers (which are not
expected to clobber anything outside of R25).
We introduce a special helper class to handle saving and restoring
necessary registers.
The bug originally reported by Joe Lin <zuojian.lzj@alibaba-inc.com>
who proposed a fix[1]. This CL builds on top of that fix and tries to
make it more safe by forcing users to use a more strict API.
[1] https://dart-review.googlesource.com/c/sdk/+/162300
TEST=vm/dart/write_barrier_register_clobber_test
Change-Id: I93e0a13a3c4c38ad28210b35750a66e615e3e44a
Cq-Include-Trybots: luci.dart.try:vm-kernel-linux-release-simarm64-try,vm-kernel-precomp-linux-release-simarm64-try
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/162191
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Instead of possibly looping over the closure function parameter names
array multiple times, just do it once. Check each entry against the
provided argument names. If none match, check the required bit (if in
null-safe mode). Also keep a count of matched names to check that all
provided names matched after the iteration.
Now that optional name checking is part of the fragment built by
BuildClosureCallArgumentsValidCheck, that method builds checks for
exactly the same things as its namesake in Function.
Also refactor the variables used by the checker into read/write ones
that need to be allocated in the parsed function and read-only ones
that can just be temporaries since no Phi nodes are needed.
Cq-Include-Trybots: luci.dart.try:vm-kernel-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm_x64-try,vm-kernel-linux-release-simarm-try,vm-kernel-linux-release-simarm64-try
Bug: https://github.com/dart-lang/sdk/issues/40813
Change-Id: I3cb421dd538629d7f5499f3bbf0653d34b850dce
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/160725
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Right now AssertAssignable will not call the TTS of a TypeParameter but
instead load the value for the type parameter inline and then call it's
TTS.
To prepare for future changes where we want to call TTS of an arbitrary
type object, we need to be able to install a functioning TTS on
TypeParameter, which this CL does.
Issue https://github.com/dart-lang/sdk/issues/40813
Change-Id: I0792cc1a796fe8afb2167b7665580b3f1385005b
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/162007
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Tess Strickland <sstrickl@google.com>
Also replace any null flag entries left in the parameter names array to
zero Smi values, so that after truncation, no null checks are needed for
retrieved flag entries.
Now that all appropriate places truncate the parameter names array after
setting any required bits, dynamic closure call dispatchers can simply
check for any post-name flag entries to detect missing required named
parameters if no named arguments were passed.
Without the added truncate calls, then
vm/dart/minimal_kernel_bytecode_test fails when dynamic closure call
dispatchers use that shortcut as seen on patchset 1 of CL 160725.
Bug: https://github.com/dart-lang/sdk/issues/40813
Change-Id: I25abd23ad3ce42b03ec5ca29b4e067c162826066
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/160860
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Code objects might be connected by static calls and these references
were not included into the profile. This lead to some objects being
unreachable from the root.
This CL also adds a test which computes dominators for all nodes
in the snapshot (this in turn verifies that there are no
unreachable nodes).
Cq-Include-Trybots: luci.dart.try:pkg-linux-release-try,pkg-win-release-try,pkg-mac-release-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-linux-product-x64-try,vm-kernel-precomp-linux-debug-x64-try
Change-Id: I2af4107fdb7f875624192e892ce1cec78cbf0dd0
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/161709
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
The bodies provided to COMPILER_PASS macro uses are put into
a DoBody method of a generated class that has the following signature:
virtual bool DoBody(CompilerPassState* state);
CompilerState::Run uses this return value to decide whether to repeat a
given pass.
Previously, if no precompiler was provided in the compiler state, the
SerializeGraph compiler pass returned the compiler state. Since the
compiler state itself was not nullptr, it would implicitly convert to
true, meaning that CompilerState::Run would always run SerializeGraph
twice in this case. Since SerializeGraph was marked as an AOT pass,
though, this case never happened as the precompiler is always provided
for AOT passes.
In the meantime, others have used it as a basis for new compiler passes
with early returns that _could_ happen and then were confused by seeing
the pass happen twice in those cases. Fix the return value so it avoids
such confusion in the future.
Change-Id: I3a91e84abe2fbcc8437c64ce1b5b91f3d4808c65
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/161784
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Arguments descriptors for invocation dispatchers are loaded as
constants. However, we only optimize type argument length field loads
from constant argument descriptors, so the prologue code is less
efficient than it could be in these cases.
Add the appropriate cases for pulling out the count, positional count,
and size as constants from constant argument descriptors in
LoadFieldInstr::TryEvaluateLoad, which is enough to let the optimizing
compiler tune this code appropriately.
Also remove the code for handling dynamic closure call dispatchers
specially within the prologue builder and instead explicitly generate
any checks that use the runtime closure value in the dynamic closure
call fragment builders after the prologue.
Bug: https://github.com/dart-lang/sdk/issues/40813
Change-Id: Ie9bd4c01c3b604f282e9b2a72327b7fc81dc4eaf
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/159905
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
When resolving control flow check detect a case when all
predecessors would contain the same move and try to
emit this move at the entry to the block if possible
(i.e. when the source of the move is not destroyed by other
moves at predecessors).
Optimization discovered as part of go/dart-llvm-assessment
flutter_gallery_total_size (sizeopt) -0.7127%
R=alexmarkov@google.com
Cq-Include-Trybots:dart/try:vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-x64-try
Change-Id: Ic8ead8ec3974a896ac58dfe97d652ffe497dfcce
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/160100
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Sometimes we might decide to inline a function which
was never executed itself - which causes us to divide
by zero when computing scaled edge weights.
Instead use 1.0 scaling factor when function was
never executed.
Minor cleanup in SetEdgeWeight helper function.
Bug: b/166297352
Change-Id: Ibe6922d2986ba42ec1b81b2f82263c7609782ce5
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/160282
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
This adds a mechanism similar to that used in async functions where
exceptions are caught in the synthetic code and passed into the
generated body to be rethrow'n.
This ensures the exception is throw'n from the same place as the
original yield*, as per the spec.
Bug: https://github.com/dart-lang/sdk/issues/42466
Change-Id: I553dd89141985f6f30751913a1c8a62b1229af04
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/158543
Commit-Queue: Clement Skau <cskau@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Unify the order of checks in both LoadFieldInstr::MakeLocationSummary()
and LoadFieldInstr::EmitNativeCode() so that all architectures first
check the slot representation, then for unboxed fields when optimizing,
then for potential unboxed fields.
Add kUnboxedInt64 handling to IA32, so that all architectures support
the same set of unboxed slot representations.
For 32-bit architectures, use RepresentationUtils::ValueSize to
determine when to use single registers vs. pairs of registers for
unboxed integers.
Move (and dedup) any asserts in LoadFieldInstr::EmitNativeCode() into
LoadFieldInstr::MakeLocationSummary(), since the latter is always called
first, making the ASSERTs fail as soon as possible, and there's no need
to re-check the duplicated ASSERTS.
Change-Id: Ia0f8fddd5a3cda036f9e4c51e3e9a374be121e7f
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/159667
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
So far inference of LoadIndexed type was based on the reaching type
of array. However, in certain cases array definition type may provide
somewhat less accurate but more useful information about generic type of
array. This change improves inference of LoadIndexed type to also take
definition type into account.
Benchmarks in JIT mode with null safety:
MegaFieldHasGetterOnce +8-13%
MegaFieldIndirect +14-31%
MegaFieldHasGetterOnceIndirect +11-32%
MegaOnlyOverriddenOnceIndirect +11-26%
MegaOnlyOverriddenOnce +13-29%
MegaNeverOverriddenIndirect +40-55%
MegaNeverOverridden +36-52%
MegaField +47-59%
MegaOverriddenEvery2nd -23-24%
(This regression is due to an additional method inlining.)
Change-Id: I8ec3342ab05543f810f8415d6fea7cf2e2cdb3c2
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/159521
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@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>
Previously, all boxed native field slots were marked not nullable.
However, several boxed native fields are in fact nullable, which means
that even generated IL that appropriately checks these fields for null
can break in optimized code when the null checks are removed.
Also add some helper utilities for working with representations.
Change-Id: I22703d14cfbadbabeae74179476dea47ace12161
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/159280
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Clement Skau <cskau@google.com>
For recognized list factories, which take constant type arguments,
compiler can infer full generic type in addition to cid.
This helps remove AssertAssignable when creating List<List<...>> with
List.generate constructor.
Havlak +3.7-10% in JIT mode with null safety.
Change-Id: I0a050c9a0d3604ffcfd1a0afa0f123a07e9ac03a
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/159185
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Régis Crelier <regis@google.com>
Previously, empty list literals were generated as
List<E>._fromLiteral(const [])
List._fromLiteral checks if the argument is an empty list and
calls _GrowableList<E>(0) in such case.
If List._fromLiteral is not inlined, this adds an unnecessary overhead.
Now empty list literals are generated more directly as
_GrowableList<E>(0)
Havlak +15.58% in AOT mode with null safety.
Change-Id: I6723e1e912cd0bbcbbb622f928d912c554217e5e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/159201
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Régis Crelier <regis@google.com>
This CL performs the following checks in the invoke field dispatcher for
dynamic closure calls when lazy dispatchers are enabled:
* The provided function type arguments vector (if any) has the correct
length.
* No function type arguments should be provided if the closure has
delayed type arguments.
* All required arguments (positional in all modes, named in appropriate
null safety modes) have been provided by the caller.
* If there are optional positional arguments, an appropriate number
has been provided.
* If there are optional named arguments, their names are valid.
Since the runtime already handles checking the argument shapes when lazy
dispatchers are disabled, these checks are now completely removed from
closure bodies in all cases. Thus, the only remaining checks in closure
bodies are the type checks performed by AssertSubtype and
AssertAssignable when lazy dispatchers are enabled.
Changes in the Flutter Gallery:
* ARM7, release: -3.61% instructions, -2.19% total
* ARM7, sizeopt: -3.62% instructions, -2.55% total
* ARM8, release: -3.66% instructions, -1.98% total
* ARM8, sizeopt: -3.65% instructions, -2.37% total
Most of these changes are already exercised by existing tests such as
(but not limited to):
* corelib{,_2}/dynamic_nosuchmethod_test
* language{,_2}/call/call_test
* language{,_2}/closure/tearoff_dynamic_test
* language{,_2}/generic/function_bounds_test
* language{,_2}/parameter/named_with_conversions_test
* language{,_2}/vm/no_such_args_error_message_vm_test
I've added one test to specifically check the interaction between
dynamic calls and required named parameters. There is some coverage in
other NNBD tests, but those are not directly focused on testing this
specifically.
Other changes:
* Adds initial cached ranges for certain BinarySmiOp and ShiftIntegerOp
instructions when the RHS is a constant, to avoid false negatives for
deoptimization and throw checks prior to range analysis.
* Adds new slots for various Function fields.
* Adds the ability to define unboxed native slots, which are always
unboxed after retrieval even in unoptimized code. In the first
iteration, the backend only handles loads from Uint32 unboxed native
slots. Part of https://github.com/dart-lang/sdk/issues/42793.
* Removed the special handling for loading from non-nullable int fields
in AOT compilation. Instead, their treatment is unified with the
treatment of the new unboxed native fields, since the source field is
always unboxed and the result of the load is also always unboxed, as
code involving them is always optimized.
Bug: https://github.com/dart-lang/sdk/issues/40813
Change-Id: Ia02aa3e872c1fefd906fd67b55021ea1797556e4
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/155604
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
This is a reland of 501a7ae7b0
The change is re-landed as is, without any extra changes.
The change was reverted due to timeout on internal test (b/163882044),
but that timeout was actually not related to this particular change.
Original change's description:
> [vm/compiler] Infer type of LoadIndexed from array type
>
> After inlining we may have more accurate information about array being
> accessed. This change adds computation of type of LoadIndexed
> instruction based on the type of array being accessed.
>
> The computation is able to skip loading of GrowableArray.data.
> If array was loaded from a field, it takes static type of a field
> into account.
>
> Benchmarks with null safety in JIT mode on x64:
> ForInLoop +80-106%
> DeltaBlueIterators +15-31%
>
> Change-Id: Ia61aa761f4ccd4692e05fdac043f8a2a46f88f21
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157823
> Reviewed-by: Ryan Macnak <rmacnak@google.com>
> Commit-Queue: Alexander Markov <alexmarkov@google.com>
Change-Id: Iae4fc5132f62fa2635159cbf9ec47636826e1402
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/158526
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
This reverts commit d4b12d6b03.
Reason for revert: Breaks mockito mocks mocking null-safe classes from non-null-safe tests. I am not intending to submit it, just need a commit id of a revert
Original change's description:
> [VM/compiler] Do not load member signatures from kernel file in JIT mode.
>
> The number of CFE-generated member signatures can be quite large (see buganizer #162073826).
> CFE now provides a reference from member signatures to their origin (see https://dart-review.googlesource.com/c/sdk/+/157498).
>
> This CL bypasses loading of member signatures in JIT mode and uses their origin as target in interface calls. AOT mode is not modified.
>
> Fix the implementation of --print-classes flag.
>
> Change-Id: Ief3e4e58a67bc3321a55d184b3c62ce62cd78689
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/158145
> Reviewed-by: Alexander Markov <alexmarkov@google.com>
TBR=rmacnak@google.com,alexmarkov@google.com,regis@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: I775a11244cc971b7aade594aadf2394636c95a2e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/158387
Reviewed-by: Régis Crelier <regis@google.com>
Reviewed-by: Ivan Inozemtsev <iinozemtsev@google.com>
Commit-Queue: Régis Crelier <regis@google.com>
With --no-use-bare-instructions or with splitting, not all calls are pc-relative. We must replace the Function with its Code in the static call table before detaching the Function's code as part of removing Function objects.
Also handle WeakSerializationReferences in AssignLoadingUnitsCodeVisitor.
Bug: https://github.com/dart-lang/sdk/issues/41974
Change-Id: I29f9258c240f5a2b1b8dca52f74146dfe44d6401
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/158165
Reviewed-by: Tess Strickland <sstrickl@google.com>
Commit-Queue: Ryan Macnak <rmacnak@google.com>
This reverts commit 501a7ae7b0.
Reason for revert: timeout on internal test, b/163882044
Original change's description:
> [vm/compiler] Infer type of LoadIndexed from array type
>
> After inlining we may have more accurate information about array being
> accessed. This change adds computation of type of LoadIndexed
> instruction based on the type of array being accessed.
>
> The computation is able to skip loading of GrowableArray.data.
> If array was loaded from a field, it takes static type of a field
> into account.
>
> Benchmarks with null safety in JIT mode on x64:
> ForInLoop +80-106%
> DeltaBlueIterators +15-31%
>
> Change-Id: Ia61aa761f4ccd4692e05fdac043f8a2a46f88f21
> Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157823
> Reviewed-by: Ryan Macnak <rmacnak@google.com>
> Commit-Queue: Alexander Markov <alexmarkov@google.com>
TBR=vegorov@google.com,rmacnak@google.com,alexmarkov@google.com,askesc@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: I10f6488b85b310fae394b372a2682a8ef802316c
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/158401
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@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>
The number of CFE-generated member signatures can be quite large (see buganizer #162073826).
CFE now provides a reference from member signatures to their origin (see https://dart-review.googlesource.com/c/sdk/+/157498).
This CL bypasses loading of member signatures in JIT mode and uses their origin as target in interface calls. AOT mode is not modified.
Fix the implementation of --print-classes flag.
Change-Id: Ief3e4e58a67bc3321a55d184b3c62ce62cd78689
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/158145
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Add information about the saved arguments descriptor for dispatcher
functions to Function::ToCString and information about both the
function kind and the saved arguments descriptor (when appropriate)
to the flow graph printer and the disassembler.
Change-Id: I41f5db2d80039cb397447eb9b9db47e60469b284
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/158007
Reviewed-by: Daco Harkes <dacoharkes@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Adds a member signature origin to member signatures and includes
an additional reference to these in MethodInvocation, PropertyGet,
PropertySet, etc.
This enables the VM to bypass the member signatures when loading the
AST.
Change-Id: I5cae11f4434bc052d0ba0acf0670dd002c6f3df1
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157498
Reviewed-by: Jens Johansen <jensj@google.com>
Commit-Queue: Johnni Winther <johnniwinther@google.com>
After inlining we may have more accurate information about array being
accessed. This change adds computation of type of LoadIndexed
instruction based on the type of array being accessed.
The computation is able to skip loading of GrowableArray.data.
If array was loaded from a field, it takes static type of a field
into account.
Benchmarks with null safety in JIT mode on x64:
ForInLoop +80-106%
DeltaBlueIterators +15-31%
Change-Id: Ia61aa761f4ccd4692e05fdac043f8a2a46f88f21
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157823
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Generally, methods that take a *TextBuffer pointer do not care how the
internal buffer is allocated, and so they could be used for either if
both were subclasses of a base class that contained the printing
methods. This CL makes that base class, and now TextBuffer and
ZoneTextBuffer now share the exact same set of methods for printing to
the internal buffer.
Since the base class is in platform, this does mean dropping the
overload of AddString for Dart String objects that was part of
ZoneTextBuffer. Instead, this CL just adds an intermediate call to
ToCString() for the small number of callers that used the overload,
keeping the printing interface the same for both.
In addition, one use of TextBuffer that then re-allocated the buffer
contents into the zone manually has been replaced with a ZoneTextBuffer
instead.
Change-Id: I438a085e7e20d55d93987fd7f36afd636f95955f
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157741
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Tess Strickland <sstrickl@google.com>
Concrete cids are 16-bit unsigned values, but Slot::cid_ was erroneously
declared as a signed value - which causes issues in program with large
amount of classes once it flows into variables of larger signed types
(classid_t or intptr_t).
Fixes b/162922506
Fixed: 162922506
Change-Id: Id116273aa34dbf6886c352427cf4565fc5dc2293
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157487
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Before null safety
try {
...
} catch (e) {
...
}
was translated to a try-catch block with 'dynamic' catch type.
VM has a special, more efficient handling of such catch-all try blocks.
Those try blocks were detected by comparing catch type with 'dynamic'.
With null safety front-end started to translate those try blocks
using non-nullable Object as a catch type. As a result, this disabled
all optimizations for catch-all try blocks in the VM.
This change extends detection of catch-all try blocks to handle both
dynamic and Object as catch types.
Improves ParserCombinators benchmark with null safety 12x in JIT mode,
15x in AOT mode. This benchmark is now on par with legacy (pre-NNBD)
version.
Change-Id: I128aa1599d8a6f979fc2e8535d0f5c934bf3a5ba
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157565
Reviewed-by: Régis Crelier <regis@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
When converting InstanceCall to LoadField don't adjust deoptimization
environment because when we deoptimize from LoadField back to
InstanceCall we would expect to have receiver on the stack.
This is a different situation from LoadField to LoadField lazy
deoptimization in which case receiver should not be on the stack.
Regression test provided by Ryan.
Fixed: 42900
Change-Id: I6041ff02abf22a2bae0fdcd65d5b49ef8aec6204
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157182
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
In AOT mode, TFA turns optional arguments (named or positional) that are
always provided into required positional arguments and increases the
number of required arguments accordingly. Thus, from the standpoint of
the VM, whether it's reading in bytecode or kernel, these are just
regular fixed positional arguments. However, TFA originally left the
parameter flags alone for these parameters, so required named parameters
that were converted to positional arguments still had the required flag
set.
This CL does two things:
1. It fixes TFA to clear the required bit on a parameter if it is
converted from a named parameter to a required positional parameter.
2. It adds ASSERT()s and RELEASE_ASSERT()s on the VM side to ensure that
only named parameters have the required bit set.
Change-Id: I1c36d777a7e9e7d703bcc23ee1b798bffeffa29e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157381
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Fixes the change in d7feab5a, which created a base definition for
BinaryIntegerOpInstr::InferRange from BinaryInt32OpInstr (which used
GetSmiRange to get the operand ranges) instead of BinaryInt64OpInstr
(which used the operand ranges directly).
Instead, just use GetInputRange to get an appropriate range based on
the representation of each input.
Change-Id: Ib62199ecc3ea3246fcf5deffc69285a15124f462
Bug: https://github.com/dart-lang/sdk/issues/42948
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/156906
Commit-Queue: Tess Strickland <sstrickl@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Aske Simon Christensen <askesc@google.com>
This assignment appears to have the intention of performing a comparison
and asserting that the opcode encountered was as expected, rather than
assigning it to be that particular opcode.
Change-Id: I21f3c00f414d4b191f85e9dcf4391a6dd7023495
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/157105
Auto-Submit: Drew Fisher <zarvox@google.com>
Reviewed-by: Aske Simon Christensen <askesc@google.com>
Commit-Queue: Aske Simon Christensen <askesc@google.com>
Tess Strickland