Test Plan:
Behavioral correctness should be ensured by existing tests. Tests in vm/dart/entrypoints
ensure that the unchecked entrypoint is used in cases where the optimization should trigger.
Bug: https://github.com/dart-lang/sdk/issues/31798
Change-Id: Id25ecba86e20c22f0678c12986ad620db312ddaa
Cq-Include-Trybots: luci.dart.try:vm-kernel-win-release-x64-try,vm-kernel-optcounter-threshold-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-win-release-x64-try
Reviewed-on: https://dart-review.googlesource.com/69743
Commit-Queue: Samir Jindel <sjindel@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Test Plan:
Behavioral correctness should be ensured by existing tests. Tests in vm/dart/entrypoints
ensure that the unchecked entrypoint is used in cases where the optimization should trigger.
Bug: https://github.com/dart-lang/sdk/issues/31798
Change-Id: I5b880b2dfa6343b4bb0a96ad23562facff73e41f
Cq-Include-Trybots: luci.dart.try:vm-kernel-win-release-x64-try,vm-kernel-optcounter-threshold-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-win-release-x64-try
Reviewed-on: https://dart-review.googlesource.com/69741
Commit-Queue: Samir Jindel <sjindel@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Rationale:
Two improvements
(1) using class hierarchy during union operation
provides a better type progation lattice, e.g.
taking the union of B,C now yields A.
A
/ \
B C
this immediately benefits propagation over
phis, but also helps the new (2).
(2) check class with cid ranges now take union
too (in contrast before only singletons were
handled by propagation).
Impact:
About 40-80% improvement on covariant typecheck benchmarks.
https://github.com/dart-lang/sdk/issues/33755
Change-Id: Ie84bc266a52a26bed9e7709371c638cabe3863b8
Reviewed-on: https://dart-review.googlesource.com/70300
Commit-Queue: Aart Bik <ajcbik@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
- Don't add dependency on fields that are either unitialized or already marked
as non guarded (guarded cid is kIllegalCid in the first case and kDynamicCid
in the second case).
- Exit early from FieldStore if guarded_cid is kDynamicCid (or we are storing
null into a nullable field).
- Don't track exactness of fields which have function type type.
Also move logic for starting to track fields into class_finalizer because we
need resolved field types to make our decisions.
Change-Id: I164b42abe3108a70b0769f4c78d92459b4e78883
Reviewed-on: https://dart-review.googlesource.com/70620
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
* Make sure to use Original field when generating code;
* Make sure to keep invariant that field with known cid should have
known fixed length or no fixed length. This invariant was always
violated by the precompiler - but it only surfaced now because
we started hitting the code path with assertion during disassembly
test.
TBR=kustermann@google.com
Change-Id: Ide86a8b9c334e6b2c75c67f466f9cf145ab9bdae
Reviewed-on: https://dart-review.googlesource.com/70510
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
We say that field's static type G<T0, ..., Tn> is exact if for
any value that can be loaded from this field, its runtime type
T is such that T at G has type arguments exactly equal to
<T0, ..., Tn>.
Know if field's static type is exact allows us to apply optimizations
that require knowing type arguments e.g.
- we can fold o.f.:type_arguments
to a constant value if we know that o.f is trivially exact;
- for method invocations o.f.m(...) we can skip argument type checks
on the callee if we know that o.f is invariant (this optimization will
be enabled once multiple entry points CLs will land).
Bug: https://github.com/dart-lang/sdk/issues/31798
Change-Id: Id565046d45a842625d41feb002b65db48451034c
Reviewed-on: https://dart-review.googlesource.com/69969
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Test Plan:
Will be tested by following revisions which introduce optimizations that invoke the
new entrypoint.
Change-Id: I56d714bceb92c44a13c46d52457154b82577aa8c
Cq-Include-Trybots: luci.dart.try:vm-kernel-win-release-x64-try,vm-kernel-optcounter-threshold-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-win-release-x64-try
Reviewed-on: https://dart-review.googlesource.com/67345
Commit-Queue: Samir Jindel <sjindel@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
This change improves type flow analysis to infer int types, in addition
to concrete classes (int type is abstract and has 2 concrete subclasses,
_Smi and _Mint).
Also, this CL enables devirtualization of dynamic calls (previously, only
calls with known interface target were devirtualized).
DartMicroBench.PrimeNumber:
x64: +221%
armv8: +461%
armv7h: +47.98%
No measurable impact on Flutter gallery build time in release mode.
Issue: https://github.com/flutter/flutter/issues/19677
Change-Id: I34db195f52f2a434218ee11eee7f308d4661738b
Reviewed-on: https://dart-review.googlesource.com/69520
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Number of checked arguments is corrected in ICData objects created when
reading bytecode, both for instance and static calls.
InstanceCall1 and InstanceCall2 bytecode instructions are replaced with
InstanceCall which works for any number of checked arguments. This makes
decision on number of checked arguments internal to VM and it is no longer
exposed to bytecode.
Change-Id: I0bba01eca6347336f3832de863b2ce4715fda04a
Reviewed-on: https://dart-review.googlesource.com/69421
Reviewed-by: Régis Crelier <regis@google.com>
Reviewed-by: Zach Anderson <zra@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Rationale:
Running an extra call specialization pass a bit later
in the compiler passes stream (after types are propagated)
recognizes more operators, which execute more efficiently.
In addition, unboxing phis seems useful on all archs, not
just 32-bit. These minor tweeks combined improve the prime
number benchmark 3 fold.
https://github.com/flutter/flutter/issues/19677
Change-Id: Ib2102ce807c2f0a9f801542e0c4bc6a7673f552b
Reviewed-on: https://dart-review.googlesource.com/69240
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Aart Bik <ajcbik@google.com>
Rationale:
This improves JIT and AOT performance of unary minus
and also improves constant folding and range analysis
on negative constant (viz. x / -3 is often x / - (3)).
The SHIFT operator needed some special treatment, since
we have to avoid converting a NON-speculative shifts
back into a deopt.
https://github.com/dart-lang/sdk/issues/34072
Change-Id: I230c9cfda98297f683bbba53688e57c2cc659360
Reviewed-on: https://dart-review.googlesource.com/68434
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Aart Bik <ajcbik@google.com>
After this change, ICData map is always unpacked in CompileParsedFunctionHelper::Compile
regardless of the optimizing mode. The unpacked ICData array is passed
both to FlowGraphBuilder and to FlowGraphCompiler.
FlowGraphCompiler::FlowGraphCompiler no longer unpacks ICData array, and
the duplicated code for unpacking of ICData map is replaced with debug asserts.
Change-Id: Ib6864acdb6bcd02fe7acd1cb5bcf70d6c69808af
Reviewed-on: https://dart-review.googlesource.com/67505
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
We permit CSE for loads from final static fields because - however we should
not be permitting LICM for such loads unless field is already initialized.
This bug was only affecting programs running from app-jit snapshots because
under normal circumstances a static get would not be inlined.
Fixes https://github.com/dart-lang/sdk/issues/34016
Bug: 34016
Change-Id: I48f30013d1541d14812b8e17b6dd75037945227d
Reviewed-on: https://dart-review.googlesource.com/67322
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Our IL graphs can contain portions that can not be reached due to execution
and all optimizations need to take that into account.
In this particular case AssertAssignable::Canonicalize should have taken
into account that constant instantiator type arguments might not necessarily
correspond to the correct class - because AssertAssignable is on the
path that will never be reached during execution.
Fixes https://github.com/dart-lang/sdk/issues/33999
Bug: http://dartbug.com/33999
Change-Id: Ia151a3902b2bc7f0f8d4aa6dcdb302fe6b1d626d
Reviewed-on: https://dart-review.googlesource.com/67321
Commit-Queue: Vyacheslav Egorov <vegorov@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
This CL introduces, similar to RawClass::direct_subclasses_, a
RawClass::direct_implementors field which is kept up-to-date.
The generation of type testing stubs uses this cached
reverse-relationship information for faster cid-range calculation (time
spent in type testing stub generation is decreased by an order of
magnitude).
This CL also enable type testing stubs for any instantiated type.
This seems to improve
* analysis-server-cold-analysis by 6%
* analysis-server-warm-analysis by 11%
as well as other non-analysis benchmarks in JIT mode.
Issue https://github.com/dart-lang/sdk/issues/33257
Change-Id: If01ee06ac08251b2ed27f79d6b82ad7354c8336e
Reviewed-on: https://dart-review.googlesource.com/66576
Commit-Queue: Martin Kustermann <kustermann@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
There is no difference from the previous version of this CL.
This depends on https://dart-review.googlesource.com/c/sdk/+/67220/3 which fixes a bug
in `CompileType` that was causing `ShouldEmitStoreBarrier()` to return different results
across multiple calls on the same instruction.
The buildbot on which this caused a failure is run as a tryjob below.
Change-Id: I078b587f7e23e88d95f3b05c966019a433ba57d9
Reviewed-on: https://dart-review.googlesource.com/67221
Commit-Queue: Samir Jindel <sjindel@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Test Plan:
Previously this caused the following assert to fail in
`StoreInstanceFieldInstr::MakeLocationSummary` on some tests:
```
ASSERT(ShouldEmitStoreBarrier() == ShouldEmitStoreBarrier())
```
I've confirmed that the assert now succeeds.
Change-Id: I5efb1d89ab1685c136d1947ae01ccb21136061a5
Reviewed-on: https://dart-review.googlesource.com/67220
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Samir Jindel <sjindel@google.com>
This reverts commit 79ed0553e9.
Reason for revert: Issue with register allocator on SIMARM.
Original change's description:
> Re-land "[vm] Factor out more of the slow-path of the store barrier into the stubs."
>
> When compiling instrinsic graphs on ARM and ARM64 we need to save LR when emitting
> the store barrier because the intrinsics don't have frames and won't restore it before
> returning.
>
> Original revision is in Patchset 1.
>
> Change-Id: I58cee3941c82efa22bd6ddbd00e00c489de53898
> Cq-Include-Trybots: luci.dart.try: vm-kernel-optcounter-threshold-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-win-release-x64-try,vm-kernel-linux-release-simarm-try,vm-kernel-linux-release-simarm64-try
> Reviewed-on: https://dart-review.googlesource.com/66382
> Commit-Queue: Samir Jindel <sjindel@google.com>
> Reviewed-by: Alexander Markov <alexmarkov@google.com>
TBR=alexmarkov@google.com,sjindel@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Change-Id: If8f8bc1a508f300ab83b03a0be7e0c318e5f99fe
Cq-Include-Trybots: luci.dart.try: vm-kernel-optcounter-threshold-linux-release-x64-try, vm-kernel-precomp-linux-debug-x64-try, vm-kernel-precomp-linux-release-simarm-try, vm-kernel-precomp-linux-release-simarm64-try, vm-kernel-precomp-linux-release-x64-try, vm-kernel-precomp-win-release-x64-try, vm-kernel-linux-release-simarm-try, vm-kernel-linux-release-simarm64-try
Reviewed-on: https://dart-review.googlesource.com/67140
Reviewed-by: Samir Jindel <sjindel@google.com>
Reviewed-by: Martin Kustermann <kustermann@google.com>
Commit-Queue: Samir Jindel <sjindel@google.com>
When compiling instrinsic graphs on ARM and ARM64 we need to save LR when emitting
the store barrier because the intrinsics don't have frames and won't restore it before
returning.
Original revision is in Patchset 1.
Change-Id: I58cee3941c82efa22bd6ddbd00e00c489de53898
Cq-Include-Trybots: luci.dart.try: vm-kernel-optcounter-threshold-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-win-release-x64-try,vm-kernel-linux-release-simarm-try,vm-kernel-linux-release-simarm64-try
Reviewed-on: https://dart-review.googlesource.com/66382
Commit-Queue: Samir Jindel <sjindel@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
This reverts commit b3e289a95e.
Reason for revert: multiple failures on *-simarm, *-simarm64 and vm-precomp-android-release-2-3-be bots.
Original change's description:
> [vm] Factor out more of the slow-path of the store barrier into the stubs.
>
> This delivers a 2.1% code size reduction on Flutter Gallery, and roughly 1.6% speed
> improvement on dart2js(dart2js).
>
> Cq-Include-Trybots: luci.dart.try: vm-kernel-optcounter-threshold-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-win-release-x64-try
> Change-Id: I2d2ceca8707d5e6fba79c558624d611f9e17d21a
> Reviewed-on: https://dart-review.googlesource.com/65060
> Commit-Queue: Samir Jindel <sjindel@google.com>
> Reviewed-by: Alexander Markov <alexmarkov@google.com>
TBR=alexmarkov@google.com,sjindel@google.com
Change-Id: Iae83a5108d56cbd30e5727217b5ee043449f82ff
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Cq-Include-Trybots: luci.dart.try: vm-kernel-optcounter-threshold-linux-release-x64-try, vm-kernel-precomp-linux-debug-x64-try, vm-kernel-precomp-linux-release-simarm-try, vm-kernel-precomp-linux-release-simarm64-try, vm-kernel-precomp-linux-release-x64-try, vm-kernel-precomp-win-release-x64-try
Reviewed-on: https://dart-review.googlesource.com/66360
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
This CL adds experimental --experimental-unsafe-mode-use-at-your-own-risk
VM option which does the following:
* Dart 2 strong mode type and bool checks are omitted.
* VM compiler optimizations which rely on strong mode types are disabled.
Applications which do not fail any strong mode checks at run time
should behave exactly the same in strong and weak modes.
Applications with failing strong mode checks will not see corresponding
errors but VM should not crash.
This option can be used for experiments, or as a temporary remedy for
regressions caused by expensive strong mode type checks.
Change-Id: I042cbccba83c105b61b3e11c659a35c20e0329cd
Reviewed-on: https://dart-review.googlesource.com/65484
Reviewed-by: Siva Annamalai <asiva@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
If the left-hand side of the store is an allocation and no GC-triggering
instructions have been executed since, it is safe to elide the store barrier.
This gives a 1.4% code size reduction on Flutter Gallery ARM32.
Change-Id: Ib7227d3ef9d798d5e30f238f3f789f9a2e637d6a
Cq-Include-Trybots: luci.dart.try: vm-kernel-optcounter-threshold-linux-release-x64-try,vm-kernel-precomp-linux-debug-x64-try,vm-kernel-precomp-linux-release-simarm-try,vm-kernel-precomp-linux-release-simarm64-try,vm-kernel-precomp-linux-release-x64-try,vm-kernel-precomp-win-release-x64-try
Reviewed-on: https://dart-review.googlesource.com/64687
Commit-Queue: Samir Jindel <sjindel@google.com>
Reviewed-by: Alexander Markov <alexmarkov@google.com>
Reviewed-by: Ryan Macnak <rmacnak@google.com>
After previous refactoring, StreamingScopeBuilder no longer depends on
StreamingFlowGraphBuilder.
This CL moves StreamingScopeBuilder to a separate file
vm/compiler/frontend/scope_builder{.h, .cc}.
Also:
* StreamingScopeBuilder is renamed to ScopeBuilder.
* IntMap is moved to hash_map.h.
* Minor cleanup of unused forward declarations and friend classes.
Change-Id: I5e780a53f37935c0a1d563d4fbaec3a92ae13d6a
Reviewed-on: https://dart-review.googlesource.com/63692
Reviewed-by: Zach Anderson <zra@google.com>
Reviewed-by: Vyacheslav Egorov <vegorov@google.com>
Commit-Queue: Alexander Markov <alexmarkov@google.com>
Samir Jindel