As number of bytecode generation options grows, it becomes cumbersome
to add them and propagate from the place where they are parsed to
the place where they are used. In order to make it easier to
change and add new bytecode generation options, this CL introduces
BytecodeOptions class which consolidates all options for bytecode
generation. Also, command line options --emit-bytecode-*** are gathered
into a single multi-option --bytecode-options=opt1,opt2,...
Also, unused --use-future-bytecode-format option is cleaned up. If needed,
it could be easily re-introduced in the new BytecodeOptions.
Change-Id: I637bf28ceb4233ead2562afe7ad51c69a99f2d60
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/106965
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Size of a large application:
Before: 26628600
After: 21480120 (-19.3%)
Size of snapshots:
isolate_snapshot_framework.bin
Before: 9322496
After: 6782976 (-27.2%)
isolate_snapshot_product_framework.bin
Before: 9166848
After: 6602752 (-27.9%)
Regressions in tests:
1) Test language_2/type_alias_equality_test/04 fails similarly to default mode, as VM
does not implement comparison of function types according to the specification.
Previously this test was passing as function types were canonicalized in bytecode,
which was not always correct. This CL fixes the problem with canonicalization of
function types in bytecode and the test starts failing again.
2) Tests standalone_2/entrypoints_verification_test, standalone_2/io/test_extension_test,
standalone_2/io/test_extension_fail_test fail as native extensions are not supported
in bytecode yet. These tests start passing after df5e7aac17,
which switched bytecode tests to kernel service (on x64), because kernel service doesn't
drop ASTs. This CL switches from reading AST library declarations to bytecode even if
AST is not removed, so tests fail again.
Change-Id: I8b7ba44bfa49d0b1599b2509553ff7c831a4e244
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/104700
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Régis Crelier <regis@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
This change replaces kernel AST declarations of fields and functions
with bytecode declarations.
Size of dilp files is reduced by 11-12%.
Startup latency:
Time to the first full frame: 1.945s -> 1.687s
FinalizeClass: 554ms -> 277ms
FinishClassLoading: 296ms -> 156ms
There are following regressions in bytecode mode, which will be fixed
in future:
* dart:mirrors are not supported yet (implementation of mirrors relies
on reading kernel AST in certain cases).
As the result, lib_2/mirrors/* tests fail.
* native extensions are not supported yet (annotations on libraries
and classes in AST are cleaned up as they could reference members
which are now removed from AST).
As the result, standalone_2/entrypoints_verification_test test fails.
* language_2/spread_collections/const_error_test/* tests fail
due to https://github.com/dart-lang/sdk/issues/36286.
Change-Id: I5130f401fd7b84038b136136e7ccc1a6e51b6cea
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/97561
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
The constant evaluator now generates all errors as Fasta diagnostic
messages. The ErrorReporter is simplified to just accept a diagnostic
message, or a notification that the constant evaluator encountered an
invalid expression (presumably put there due to an earlier error).
Also, the flow of control between the error reporter and the internal
abort exceptions is reversed, so the error reporter is now called as a
result of an abort exception being caught by the evaluate method.
Reland of https://dart-review.googlesource.com/c/sdk/+/96300
Change-Id: I7d32b6e98962b6ee781a6c96b593b00ee7fd8a89
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/97225
Commit-Queue: Aske Simon Christensen <askesc@google.com>
Reviewed-by: Kevin Millikin <kmillikin@google.com>
TBR=askesc@google.com
Revert "[CFE] Move constant evaluator to Fasta."
This reverts commit 845b5b2df1.
Revert "[CFE] Always call the constant evaluator by the evaluate method."
This reverts commit 91bc4ec2b9.
Revert "[CFE] Use Fasta diagnostics in the constant evaluator."
This reverts commit c7b572aa29.
Revert "[CFE] Check for null in constant evaluation"
This reverts commit e6d2751e9c.
Revert "Rename import after moving file."
This reverts commit a6e2c5eb4c.
Change-Id: Iadfe087c0110f6f331b82d990213f95d3ef4541b
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/97223
Reviewed-by: William Hesse <whesse@google.com>
The constant evaluator now generates all errors as Fasta diagnostic
messages. The ErrorReporter is simplified to just accept a diagnostic
message, or a notification that the constant evaluator encountered an
invalid expression (presumably put there due to an earlier error).
Also, the flow of control between the error reporter and the internal
abort exceptions is reversed, so the error reporter is now called as a
result of an abort exception being caught by the evaluate method.
Change-Id: I66f148cc4e202e328f895ae0b770f9b68c9f3c8e
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/96300
Reviewed-by: Kevin Millikin <kmillikin@google.com>
This is in preparation for recognizing Fasta-specific nodes in the
constant evaluator, and for using the Fasta diagnostics framework.
Change-Id: I8535fbb68e622f1814a1d577c348d87e573b6b34
Reviewed-on: https://dart-review.googlesource.com/c/sdk/+/96081
Reviewed-by: Kevin Millikin <kmillikin@google.com>
* Type arguments are moved to global object table (and de-duplicated).
* Finalize non-recursive generic types at bytecode generation time.
* Constants are moved to global object table (and de-duplicated).
* ICData creation is avoided for direct calls.
* Interface calls reference target member instead of selector name.
Size of a dilp files: 17868K => 15896K (-11%).
Change-Id: I0c9bf338137a0fae6ed90ab6b125ed2b24a1a8ad
Reviewed-on: https://dart-review.googlesource.com/c/91108
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Zach Anderson <zra@google.com>
Reviewed-by: Régis Crelier <regis@google.com>
When creating a closure, receiver is loaded in order to obtain
instantiator type arguments. That access was not capturing
receiver, if a closure was created inside another closure.
Change-Id: I9ec6d9f2c90643e6deb9c4efde1b070a3255a387
Reviewed-on: https://dart-review.googlesource.com/c/91442
Commit-Queue: Régis Crelier <regis@google.com>
Auto-Submit: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Zach Anderson <zra@google.com>
Reviewed-by: Régis Crelier <regis@google.com>
This achieves consistency with similar getters in the API.
This is technically a breaking change, since it changes a published
part of the Kernel API. Since the constants API is relatively new and
so far only used internally in the AOT compiler, the change is
expected to be unproblematic.
Closes https://github.com/dart-lang/sdk/issues/35696
Change-Id: I3ca30922580d226ccbdb6f77496983c21ef2102b
Reviewed-on: https://dart-review.googlesource.com/c/90220
Commit-Queue: Aske Simon Christensen <askesc@google.com>
Reviewed-by: Kevin Millikin <kmillikin@google.com>
An invalid constant is (currently) represented by an unevaluated
invalid expression. Using this instead of null fixes 63 out of 125
CFE constant-evaluation crashes with constant-update-2018 and
correctly signals 26 more previously-missed compile-time errors.
Change-Id: I5b4de3995b3a59978dfa08fc542ef0f027572eb6
Reviewed-on: https://dart-review.googlesource.com/c/89506
Commit-Queue: Kevin Millikin <kmillikin@google.com>
Reviewed-by: Peter von der Ahé <ahe@google.com>
Reviewed-by: Aske Simon Christensen <askesc@google.com>
In package-split mode, bytecode generation is performed separately for
each package. Previously, dropping AST was done right after generating
bytecode. However, dropping AST for a package makes it impossible to do
constant evaluation in other packages which import the package with dropped
AST. This breaks bytecode generation for subsequent packages.
To work around this problem, in package-split mode AST is removed
temporary until dillp file is written. After that, removed AST is restored
back.
Change-Id: I3d8b6a8ad98f2fe88b57f7b6393bbbe87b046c21
Reviewed-on: https://dart-review.googlesource.com/c/89822
Auto-Submit: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Zach Anderson <zra@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Make the constant evaluator take an explicit error reporter so we have
to opt in to using the "simple" one that reports errors in an ad hoc
way. This is the start of a change to use Fasta-controlled error
messages throughout and eventually get rid of the simple error
handler, and to continue constant evaluation after the first constant
error.
Change-Id: If6b1801edab6063754b642cf4a603abf9d63103a
Reviewed-on: https://dart-review.googlesource.com/c/89501
Commit-Queue: Kevin Millikin <kmillikin@google.com>
Auto-Submit: Kevin Millikin <kmillikin@google.com>
Reviewed-by: Peter von der Ahé <ahe@google.com>
Reviewed-by: Aske Simon Christensen <askesc@google.com>
Dispatch interface calls via hashtable rather than inline cache.
InterfaceCall doesn't need to take arguments descriptor into account
when doing method lookup.
Change-Id: I30eae6ea638d1d2ad2cf3ff073c653fee3377f31
Reviewed-on: https://dart-review.googlesource.com/c/86106
Reviewed-by: Zach Anderson <zra@google.com>
Commit-Queue: Régis Crelier <regis@google.com>
This includes:
* Selecting front-end target.
* Support for multi-root virtual file system.
* Not linking platform into resulting kernel file.
* Specifying input as URI (instead of file path) on command line.
* Automatically converting input script URI to package URI.
* Writing ninja dependencies file.
* Writing package-split kernel binaries.
After this change Fuchsia's compiler.dart will become a small wrapper
over pkg/vm, sharing most logic and even most command line options
with pkg/vm gen_kernel tool.
Also, this CL attempts to share some pieces of code between frontend
server and gen_kernel.
In addition, seperate bytecode generation for package-split binaries
is implemented (needed for https://dart-review.googlesource.com/c/sdk/+/85469).
Corresponding Fuchsia CL: https://fuchsia-review.googlesource.com/c/topaz/+/229964
Change-Id: I12d7b2f6401357b3c9df2e31bc736af5a9dc5fd2
Reviewed-on: https://dart-review.googlesource.com/c/85721
Reviewed-by: Alexander Aprelev <aam@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
This is a major revamp of bytecode metadata format. Now bytecode has
its own serialization mechanisms.
This CL adds 'bytecode component' metadata, which contains bytecode
object table and string table. All references from bytecode (constant
pools) to libraries, classes, members, types and strings now have a new
format. References to frequently used objects are represented as indices
in object table, while rarely used objects are written inline.
This allows VM to cache frequently used objects while reading bytecode.
Representation of strings is aligned with VM - string characters are
stored in separate pools of one-byte and two-byte strings. This allows
VM to avoid UTF-8 decoding and extra copying.
Closure declarations are now explicit. Type parameters no longer require
enslosing scopes when reading/writing them.
Benchmarks:
GenKernelKernelReadAllBytecode (Intel Core i5) +29.84%
GenKernelKernelReadAllBytecode (Intel Xeon) +28.74%
Change-Id: I4b80009733a8f8c038264af74f97c4e094b9e311
Reviewed-on: https://dart-review.googlesource.com/c/85469
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Régis Crelier <regis@google.com>
Reviewed-by: Zach Anderson <zra@google.com>
This CL adds context ID operand to AllocateContext, CloneContext,
LoadContextVar and StoreContextVar bytecode instructions.
The context ID will be used to create distinct Slots and disambiguate
accesses to context objects corresponding to different scopes.
Change-Id: I98850ab763017b71c1dcacfccaffc085bd850e00
Reviewed-on: https://dart-review.googlesource.com/c/84681
Auto-Submit: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Régis Crelier <regis@google.com>
Reviewed-by: Régis Crelier <regis@google.com>
Number of AssertBoolean instructions in the 100Mb dynamic trace of
Richards benchmark in pure interpreted mode (platform with bytecode):
Before: 1812005
After: 251451
Change-Id: I9fe9d37499d73ffa7636c9ca5fb9c2c18dbc1397
Reviewed-on: https://dart-review.googlesource.com/c/84643
Auto-Submit: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Zach Anderson <zra@google.com>
Reviewed-by: Zach Anderson <zra@google.com>
Make StoreIntanceField and LoadField instructions use a single uniform
abstraction: Slot (used to be called NativeFieldDesc), which represents
either a real Dart field (i.e. a field that has a corresponding Field object)
or a native VM field that does not have a corresponding Field object.
This refactoring eliminates raw stores/loads that were just using offsets
before - now we always know what kind of slots we are accessing and
this yields better aliasing information.
Change-Id: I2f48332d58258219565bd961764e8cc9dd4d75ce
Reviewed-on: https://dart-review.googlesource.com/c/74582
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
In order to enable OSR, CheckStackOverflow IL instructions are attributed
with a non-zero loop depth. The original loop depth is passed as operand
of CheckStack bytecode instruction.
Change-Id: I771f59ba9f8d071680a3b7156be380e0f606b7b2
Reviewed-on: https://dart-review.googlesource.com/c/84081
Reviewed-by: Régis Crelier <regis@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Alexander Markov