Design doc: bit.ly/36MfD6Y
We introduce simplified operators LoadImmutableFromObject and
InitializeImmutableInObject. These are lowered to Loads and Stores like
LoadFromObject and StoreToObject.
We split CsaLoadElimination::AbstractState in two HalfStates,
which represent the mutable and immutable component of the state.
Immutable operators in the effect chain modify the immutable half-state,
and plain operators modify the mutable half-state. The immutable part is
maintained through write effects and loop headers. Immutable
initializations do not lookup and kill previous overlapping stores,
assuming each offset cannot be initialized more than once.
Bug: v8:11510
Change-Id: I0f5feca3354fdd3bdc1f511cc5214ec51e1407ad
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3268728
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78325}
This reverts commit 91f08378bc.
Reason for revert: It's a fairly big change, and the clusterfuzz
found some bugs. Will reland with the fix after M98 branch point.
Original change's description:
> [class] implement reparsing of class instance member initializers
>
> Previously, since the source code for the synthetic class instance
> member initializer function was recorded as the span from the first
> initializer to the last initializer, there was no way to reparse the
> class and recompile the initializer function. It was working for
> most use cases because the code for the initializer function was
> generated eagarly and it was usually alive as long as the class was
> alive, so the initializer wouldn't normally be lazily parsed. This
> didn't work, however, when the class was snapshotted with
> v8::SnapshotCreator::FunctionCodeHandling::kClear,
> becuase then we needed to recompile the initializer when the class
> was instantiated. This patch implements the reparsing so that
> these classes can work with FunctionCodeHandling::kClear.
>
> This patch refactors ParserBase::ParseClassLiteral() so that we can
> reuse it for both parsing the class body normally and reparsing it
> to collect initializers. When reparsing the synthetic initializer
> function, we rewind the scanner to the beginning of the class, and
> parse the class body to collect the initializers. During the
> reparsing, field initializers are parsed with the full parser while
> methods of the class are pre-parsed.
>
> A few notable changes:
>
> - Extended the source range of the initializer function to cover the
> entire class so that we can rewind the scanner to parse the class
> body to collect initializers (previously, it starts from the first
> field initializer and ends at the last initializer). This resulted
> some expectation changes in the debugger tests, though the
> initializers remain debuggable.
> - A temporary ClassScope is created during reparsing. After the class
> is reparsed, we use the information from the ScopeInfo to update
> the allocated indices of the variables in the ClassScope.
>
> Bug: v8:10704
> Change-Id: Ifb6431a1447d8844f2a548283d59158742fe9027
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2988830
> Reviewed-by: Leszek Swirski <leszeks@chromium.org>
> Reviewed-by: Toon Verwaest <verwaest@chromium.org>
> Commit-Queue: Joyee Cheung <joyee@igalia.com>
> Cr-Commit-Position: refs/heads/main@{#78299}
Bug: v8:10704
Change-Id: I039cb728ebf0ada438a8f26c7d2c2547dbe3bf2d
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3325328
Auto-Submit: Joyee Cheung <joyee@igalia.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Reviewed-by: Marja Hölttä <marja@chromium.org>
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78315}
Previously, since the source code for the synthetic class instance
member initializer function was recorded as the span from the first
initializer to the last initializer, there was no way to reparse the
class and recompile the initializer function. It was working for
most use cases because the code for the initializer function was
generated eagarly and it was usually alive as long as the class was
alive, so the initializer wouldn't normally be lazily parsed. This
didn't work, however, when the class was snapshotted with
v8::SnapshotCreator::FunctionCodeHandling::kClear,
becuase then we needed to recompile the initializer when the class
was instantiated. This patch implements the reparsing so that
these classes can work with FunctionCodeHandling::kClear.
This patch refactors ParserBase::ParseClassLiteral() so that we can
reuse it for both parsing the class body normally and reparsing it
to collect initializers. When reparsing the synthetic initializer
function, we rewind the scanner to the beginning of the class, and
parse the class body to collect the initializers. During the
reparsing, field initializers are parsed with the full parser while
methods of the class are pre-parsed.
A few notable changes:
- Extended the source range of the initializer function to cover the
entire class so that we can rewind the scanner to parse the class
body to collect initializers (previously, it starts from the first
field initializer and ends at the last initializer). This resulted
some expectation changes in the debugger tests, though the
initializers remain debuggable.
- A temporary ClassScope is created during reparsing. After the class
is reparsed, we use the information from the ScopeInfo to update
the allocated indices of the variables in the ClassScope.
Bug: v8:10704
Change-Id: Ifb6431a1447d8844f2a548283d59158742fe9027
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/2988830
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Toon Verwaest <verwaest@chromium.org>
Commit-Queue: Joyee Cheung <joyee@igalia.com>
Cr-Commit-Position: refs/heads/main@{#78299}
Treat all stack pointers as roots:
- Maintain a global linked-list of StackMemories
- Update StackFrameIterator to process inactive stacks
- Visit roots in all inactive stacks (including root marking and root
pointer updating).
Drive-by:
- Fix some issues uncovered by the test
- Refactor the builtin constants
R=mlippautz@chromium.org,ahaas@chromium.org
Bug: v8:12191
Change-Id: I5b6381f9818166e2eabf80dd59135673dddb2afc
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3310932
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Thibaud Michaud <thibaudm@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78291}
This reverts commit 863bc2b88a.
Reason for revert: https://crbug.com/1276923
Original change's description:
> [turbofan] Improve StoreStoreElimination
>
> Previously, StoreStoreElimination handled allocations as
> "can observe anything". This is pretty conservative and prohibits
> elimination of repeated double stores to the same field.
> With this CL allocations are changed to "observes initializing or
> transitioning stores".
> This way it is guaranteed that initializing stores to a freshly created
> object or stores that are part of a map transition are not eliminated
> before allocations (that can trigger GC), but allows elimination of
> non-initializing, non-transitioning, unobservable stores in the
> presence of allocations.
>
> Bug: v8:12200
> Change-Id: Ie1419696b9c8cb7c39aecf38d9f08102177b2c0f
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3295449
> Commit-Queue: Patrick Thier <pthier@chromium.org>
> Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
> Reviewed-by: Maya Lekova <mslekova@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#78230}
Bug: chromium:1276923
Change-Id: I43dc3572ce1ef1fda42b7551ce8210d9f03e36ca
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3318666
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Commit-Queue: Patrick Thier <pthier@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78277}
This CL adds the following specific cases, to make sure they are
correctly handled by the slow path:
- the backing store of a TypedArray gets detached after optimisation
- passing null instead of a TA doesn't lead to a deopt
Bug: v8:11739, chromium:1052746
Change-Id: I7dfd3da9f535831901998ca6fad854af6e93e9f9
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3320425
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Maya Lekova <mslekova@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78274}
This is a reland of 3ee4804f83.
The CL was originally reverted for blink test failures. Since the
revert, the blink top-level await flag has been removed.
Original change's description:
> [top-level-await] Remove --harmony-top-level-await
>
> TLA has been shipped since v8.9.
>
> Bug: v8:9344, chromium:1271114
> Change-Id: Ibebf21da8bacb1f0d212390133847495ad8553e5
> Reviewed-on:
https://chromium-review.googlesource.com/c/v8/v8/+/3307103
> Commit-Queue: Shu-yu Guo <syg@chromium.org>
> Reviewed-by: Camillo Bruni <cbruni@chromium.org>
> Reviewed-by: Marja Hölttä <marja@chromium.org>
> Reviewed-by: Igor Sheludko <ishell@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#78169}
Bug: v8:9344, chromium:1271114
Change-Id: I96a9641967a23a12ba2467a69e5859ad8647f3e3
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3318717
Commit-Queue: Shu-yu Guo <syg@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78261}
Previously, StoreStoreElimination handled allocations as
"can observe anything". This is pretty conservative and prohibits
elimination of repeated double stores to the same field.
With this CL allocations are changed to "observes initializing or
transitioning stores".
This way it is guaranteed that initializing stores to a freshly created
object or stores that are part of a map transition are not eliminated
before allocations (that can trigger GC), but allows elimination of
non-initializing, non-transitioning, unobservable stores in the
presence of allocations.
Bug: v8:12200
Change-Id: Ie1419696b9c8cb7c39aecf38d9f08102177b2c0f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3295449
Commit-Queue: Patrick Thier <pthier@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78230}
If multiple isolates are running concurrently and one of them calls
`quit`, we should not delete the counters map, because another isolate
might still access it.
R=mlippautz@chromium.orgCC=nikolaos@chromium.org
Bug: v8:12453
Change-Id: I6d41478f188f0043b7d6055b0872574c28fd3039
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3310807
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78226}
This fixes data races when lazily creating counters (and populating the
{counter_map_}, and when concurrently adding samples to the counters.
It also ensures that the Wasm engine is stopped (via {V8::Dispose})
before printing and deleting counters, as background threads might still
try to update the counters otherwise.
R=mlippautz@chromium.org
CC=nikolaos@chromium.org
Bug: v8:12453, chromium:1275117
Change-Id: Ie6beea6cc74eea52143d12f9921597da4a250f2a
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3308710
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78191}
When emitting code, character ranges must only specify ranges which
the actual subject string (one- or two-byte) may contain.
This was not always the case, specifically for ranges with
`from <= kMaxUint8` and `to > kMaxUint8`.
The reason this is so tricky: 1. not all parts of the pipeline know
whether we are compiling for one- or two-byte subjects; 2. for
case-insensitive regexps, an out-of-bounds CharacterRange may have an
in-bounds case equivalent (e.g. /[Ÿ]/i also matches 'ÿ' == \u{ff}),
which only gets added somewhere in the middle of the pipeline.
Our current solution is to clamp immediately before code emission. We
also keep the existing handling/dchecks of the 0x10ffff marker value
which may occur in the two-byte subject case.
Bug: v8:11069
Change-Id: Ic7b34a13a900ea2aa3df032daac9236bf5682a42
Fixed: chromium:1275096
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3306569
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78186}
Add suppose for compiling non-eager, non-top-level inner functions in
parallel, using the compiler dispatcher. This behaviour can be enabled
with --parallel-compile-tasks-for-lazy.
There are a couple of consequences:
* To support this we need support for off-thread ScopeInfo
deserialization, so this adds that too.
* The previous --parallel-compile-tasks flag is renamed to the more
descriptive --parallel-compile-tasks-for-eager-toplevel.
* Both parallel-compile-tasks flags are moved onto
UnoptimizedCompileFlags so that they can be enabled/disabled on a
per-compile basis (e.g. enabled for streaming, disabled for
re-parsing).
* asm.js compilations can now happen without an active Context (in
the compiler dispatcher's idle finalization) so we can't get a
ContextId for metric reporting; we'd need to somehow fix this if we
wanted asm.js UKM but for now it's probably fine.
* Took the opportunity to clean up some of the "can preparse" logic in
the parser.
Change-Id: I20b1ec6a6bacfe268808edc8d812b92370c5840d
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3281924
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Toon Verwaest <verwaest@chromium.org>
Reviewed-by: Emanuel Ziegler <ecmziegler@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78183}
Try blocks (whether catch or finally) will unconditionally create
handler tables and start a new basic block for the exception handler.
This can accidentally resurrect a dead block when the entire try block
is dead (and hence can never enter the exception handler in the first
place).
Add a deadness check to BuildTryCatch/Finally to fix this.
Bug: chromium:1273677
Change-Id: Icda9deb1459e47de5cb83e7b636299e24c3ebe77
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3306555
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Commit-Queue: Marja Hölttä <marja@chromium.org>
Reviewed-by: Marja Hölttä <marja@chromium.org>
Auto-Submit: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78179}
Currently, JSFinalizationRegistry has a BodyDescriptor that iterates
next_dirty as a custom weak field, and it has a WeakListVisitor that
cleans up any items from the list that should be removed. However, none
of that code is used, because JSFinalizationRegistry objects are created
with visitor ID kVisitJSObjectFast. This change gives them a custom
visitor ID so that next_dirty can be treated as weak.
Bug: v8:12430
Change-Id: I31c1935257ad508b13a3e684662d2ca406d8ed19
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3307096
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78167}
--experimental-wasm-mv has been removed, and --wasm-loop-unrolling is on
by default.
Change-Id: I094870765031ef46fcec95e9071333e4e80f407b
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3302805
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78140}
{WasmInternalFunction::external} might return a function that is not
a WasmExportedFunction, at which point the code in ProcessTypeFeedback
fails.
See crrev.com/c/3277878 for context.
Bug: v8:12436, v8:12166
Change-Id: I09ef96df3fc051586044dd9c2ce88d6aeeb34b9f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3306391
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78139}
To trigger speculative inlining, a function needs to run a few times on
Liftoff-tier, and then tier-up to Turbofan. We make sure this happens
by enabling --wasm-dynamic-tiering with a small budget, and running
the critical functions until {%IsLiftoffFunction} is false.
We also add an additional test.
Bug: v8:12166
Change-Id: I6c5e0d3f1e3ba0df8510ae6850d8e9af1d01c179
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3306372
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78123}
See crrev.com/c/3277878 for context.
We should only transform extenral to internal function references when
passing a function value to a function-typed global or table. For their
externref counterparts, we should preserve the reference unchanged.
Bug: v8:11510, chromium:1273705
Change-Id: Ic1719c4d31e175f3a37ced6e4e4dfcd61a19ae57
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3302790
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78108}
Design doc: bit.ly/3jEVgzz
We separate the internal representation of function references in Wasm
from their JSFunction-based (external) representation. This improves
performance of call_ref by requiring less indirections to load the
context and call target from a function reference. In the boundary
between wasm and JS/the C API, we add transformations between the two
representations.
Detailed changes:
- Introduce WasmInternalFunction, containing fields required by
call_ref, as well as a reference to the corresponding
WasmExternalFunction. Add a reference to the WasmInternalFunction in
WasmFunctionData. The {WasmInternalFunction::FromExternal} helper
extracts the internal out of an external function.
- Change {WasmInstanceObject::external_functions()} to internal
functions.
- Change wasm function tables to contain internal functions.
- Change the following code to use internal functions:
- call_ref in liftoff and Turbofan
- function type checks in liftoff and Turbofan
- CallRefIC and GenericJSToWasmWrapper builtins
- {InitExprInterface::RefFunc}
- module-compiler.cc in {ProcessTypeFeedback}
- In module-instantiate.cc, in function-rtt creation.
- Add transformations between internal and external functions in:
- WasmWrapperGraphBuilder::{ToJS, BuildUnpackObjectWrapper, FromJS,
BuildJSToJSWrapper}.
- debug-wasm-objects.cc in {FunctionProxy::Get},
{WasmValueObject::New} and {AddWasmTableObjectInternalProperties}.
- runtime-wasm.cc in ReplaceWrapper
- the C and JS APIs
- module-instantiate.cc, in import and export processing, as well as
{InitializeIndirectFunctionTables}
- WasmTableObject::{IsValidElement, SetFunctionTableEntry}
- {WasmGlobalObject::SetFuncRef}
- Simplify body descriptors of WasmExternalFunction variants.
- Adjust tests.
Bug: v8:11510
Change-Id: I8377f46f55c3771391ae1c5c8201a83854ee7878
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3277878
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78068}
In the final version, we might parse the parameters when deserializing
instead, but this approach is more suitable for prototyping.
Bug: v8:11525
Change-Id: I000869877b03fd1909acf602ab5190951b1939e5
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3295456
Reviewed-by: Camillo Bruni <cbruni@chromium.org>
Commit-Queue: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78065}
Private methods use a property with symbol name that stores a context.
This is then loaded to perform the brand check. Since this uses
JSLoadProperty in Turbofan, we should not type JSLoadProperty as
NonInternal.
Bug: chromium:1269063, v8:12359
Change-Id: I920ccf46e939ab0477ff2fdb3cda6d7d94bab0af
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3293089
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78042}
It is possible to inline a JSToWasm wrapper that directly calls a
WasmToJS wrapper. In this case we need to make sure that the instance
we pass is a WasmApiFunctionRef, not a WasmInstanceObject.
Bug: chromium:1271456
Change-Id: I684a769922895860a43c73ba43c2598c9bda4b64
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3293423
Commit-Queue: Paolo Severini <paolosev@microsoft.com>
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78031}
We prevent unrolling of loops with indirect calls. We expand the set of
permitted wasm builtins in unrolled loops.
Bug: v8:11298
Change-Id: I70b8ff3b16d9b0d3a4ea2d103f8ffb74083fd2a9
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3289152
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77965}
On 32-bit architectures, we need to run Int64Lowering on the inlinee
code to make it compatible with the caller code.
Since Int64Lowering now runs while a GraphReducer is active, only one of
them can use node marks to store node states. Therefore, we move the
Int64Lowering node states to an internal map.
Bug: v8:12166
Change-Id: I53b85442d503e71fa533e06568f4b9db572a4401
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3283072
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77941}
The stored value might be an allocation that can be removed once the
Store node is removed. We need to revisit this node manually because
inputs in a node removed with ReplaceWithValue are not revisited
automatically.
Bug: v8:11510
Change-Id: I57cb8955a3e2f7143474ad7ced9d946e6d1cc18e
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3277880
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77899}
In each wasm CallDescriptor, we store the signature of the call based on
the real parameters passed to the call. This signature is more precise
than the formal function signature. We use this signature in inlining
to enable more optimizations.
Changes:
- Add wasm_sig_ field to CallDescriptor.
- Construct the real signature in {DoCall} and {DoReturnCall} in
graph-builder-interface, and pass it to all call-related functions in
WasmGraphBuilder.
- Update {ReplaceTypeInCallDescriptorWith} to use ValueType over
MachineType. Construct the updated function signature.
- In wasm-inlining, kill the Call node after inlining.
- Add two tests.
Bug: v8:11510
Change-Id: Ica711b6b4d83945ecb7201be26577eab7db3c060
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3270539
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77889}
FLAG_wasm_dynamic_tiering is still off by default. When enabled,
it now uses the technique previously behind --new-wasm-dynamic-tiering.
Bug: v8:12281
Change-Id: I365c2c066e62418cd6abc7830f95d1fe0f950e33
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3275570
Commit-Queue: Jakob Kummerow <jkummerow@chromium.org>
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77883}
Similar to previous bug v8:11771, this test needs deterministic GC
behavior so it is incompatible with concurrent inlining.
Bug: v8:12374, v8:4578
Change-Id: Ib3667744d1032524a0c2e697a970876dfc1677ae
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3272882
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77851}
Loop headers in the interpreter would start a new basic block, which
among other things would reset the liveness of that block. This meant
that a loop created after dead code, without a check for whether the
code is currently dead or not, would "resurrect" that block's liveness,
making the inside of the loop live even though the loop itself is
unreachable.
This works fine, since the loop is still unreachable, but can breaks
DCHECKs in bytecode liveness analysis for cases where a register is
supposed to be initialised before the loop, in the dead code, and is
then used inside the loop, in the resurrected code.
Normally this wouldn't be a problem, since blocks are normally killed on
the statement level and we check for deadness during statement
iteration, but `foo() = x` introduces an expression-level block killer
(being re-written to `foo[throw ReferenceError] = x`) and we don't check
for deadness after assignment Lhs preparation.
This does mean that we have to fix the InterpreterJumps test, to not try
to jump into the middle of a loop (since this could revive the loop).
This can only happen when manually creating bytecode, bytecode generated
from JavaScript is always reducible.
Bug: chromium:1230597
Change-Id: I8403ccdeae7e5450adf629026e2ca8a134c81877
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3275557
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77846}
Changes:
- Enable allocation folding for wasm-gc graphs.
- Improve structure of wasm escape analysis code. Kill dead nodes.
- Revisit object node after eliminating a load or a store to that node.
- Add a couple of tests, rename one test file.
Bug: v8:11510
Change-Id: I8b3c5186cd0a8827744a05eba366ff79bc7bc975
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3264215
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Commit-Queue: Manos Koukoutos <manoskouk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77840}
The GC already treats some embedded object pointers in Code as weak,
based on Code::IsWeakObject. If one of those embedded objects ends up
unmarked during a full mark-collect GC, then the Code is marked for lazy
deoptimization and the embedded objects are cleared. However, many of
those same objects are often held strongly by the deoptimization literal
array for the Code, which causes memory leaks. This change updates the
deoptimization literals array to store those objects weakly. Any Code
currently executing on the stack might need those deoptimization
literals in order to deoptimize, so the deoptimization literal array is
marked strongly in that case.
Design document:
https://docs.google.com/document/d/1gFRBYCeqz9Mysx8CVYQkldBbk3AZLo8UX0DMLZV_7qw/edit?usp=sharing
Bug: v8:4578
Change-Id: I02e86683c59371e9f88ecf523750c9c6afebdb39
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3160299
Reviewed-by: Yang Guo <yangguo@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/main@{#77805}
In assignments the lhs should be evaluated first and shouldn't be
re-evaluated when the value of the rhs is available. Fix it by
saving the receiver and the key registers into AssignmentLhsData
before building the assignment and use them later, instead of visiting
the AST again to retrieve the receiver.
In addition, now that we save the receiver register, use it to
perform the brand check even when we know for sure that it's
going to fail later because it's a write to a private
method or accessing the accessor in the wrong way (v8:11364),
so that the brand check error always appears first if it is present,
as specified in
https://tc39.es/proposal-private-methods/#sec-privatefieldget
Drive-by: unify the brand check error messages, and replace "Object"
with "Receiver" in the messages for clarity. The instance private
brand check now throws "Receiver must be an instance of class <name>"
and the static private brand check now throws "Receiver must be
class <name>". Also always set the expression position to the
property load position, because the brand check failure comes from
the load operation.
Bug: v8:12352, v8:11364
Change-Id: I61a8979b2e02b561dd5b2b35f9e0b6691fe07599
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3266964
Commit-Queue: Joyee Cheung <joyee@igalia.com>
Reviewed-by: Marja Hölttä <marja@chromium.org>
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77797}
In order to construct the optimal source code combination for functions
(and especially the positions inside the optimal source code for each
function), we need to know all functions upfront.
Rewrite the d8-based snapshotter so that it discovers all objects first
(+ assigns ids), then does the actual snapshotting.
Bug: v8:11525
Change-Id: I4f4e5589575ee9738e8a83aa6cda2c8164803b4c
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3268915
Reviewed-by: Camillo Bruni <cbruni@chromium.org>
Commit-Queue: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77794}
Capture group names were extended in
https://github.com/tc39/ecma262/pull/1869/fileshttps://github.com/tc39/ecma262/pull/1932/files
RegExpIdentifierName now explicitly enables unicode (+U) for
unicode escape sequences; likewise, surrogate pairs are now allowed
unconditionally.
The implementation simply switches on unicode temporarily while
parsing a capture group name.
Good news everyone, /(?<𝒜>.)/ is now a legal pattern.
Bug: v8:10384
Change-Id: Ida805998eb91ed717b2e05d81d52c1ed61104e3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3233234
Auto-Submit: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77722}
This is a reland of 0446ab7ce1
Additional fix:
Manually set the host-defined options on deserialised scripts in d8.
Original change's description:
> [d8] Verify host-defined options
>
> d8 never checked what the actual value of the host-defined options are.
> We now properly very that the host-defined options is a specific object
> so we we don't end up accidentally ignoring a wrong options object.
>
> Drive-by-fix:
> - Convert %AbortJS argument to string
>
> Bug: chromium:1244145
> Change-Id: If0ed128d215682bcf066592418420548b06eb6a1
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3259655
> Commit-Queue: Camillo Bruni <cbruni@chromium.org>
> Reviewed-by: Shu-yu Guo <syg@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#77699}
Bug: chromium:1244145
Change-Id: I8ddfdba27d84c36862323ab9e1aba14b2ff932a4
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3259539
Auto-Submit: Camillo Bruni <cbruni@chromium.org>
Reviewed-by: Shu-yu Guo <syg@chromium.org>
Commit-Queue: Shu-yu Guo <syg@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77716}
The wasm serialization format only contains TurboFan code. All other
functions are only represented by placeholders. With this CL
serialization fails if the serialized module does not contain any
TurboFan functions and would therefore consist only of placeholders.
This is a defense in depth approach, because ideally serialization
only gets triggered when TurboFan code is available. However, in some
scenarios like debugging it can happen that modules without TurboFan
code get serialized.
Bug: v8:12281
Change-Id: Ib05430ff89eb2317da80fc0d086ce1d7ab0e919d
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3212510
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77715}
The old "gc-safe" implementation to get the off-heap type information
wasn't quite as gc-safe as it needs to be.
Due to parallel compaction, we shouldn't check for forwarding pointers;
instead we should rely on the old location of the Foreign, but make sure
not to look at its Map (which might be a forwarding pointer).
Bug: v8:12185
Change-Id: I4570b00a5300a0d7ed8c042fa21d355373e0e691
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3260513
Commit-Queue: Jakob Kummerow <jkummerow@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77707}
This reverts commit 0446ab7ce1.
Reason for revert: Lots of failures https://ci.chromium.org/ui/p/v8/builders/ci/V8%20Linux%20-%20debug/37355/overview
Original change's description:
> [d8] Verify host-defined options
>
> d8 never checked what the actual value of the host-defined options are.
> We now properly very that the host-defined options is a specific object
> so we we don't end up accidentally ignoring a wrong options object.
>
> Drive-by-fix:
> - Convert %AbortJS argument to string
>
> Bug: chromium:1244145
> Change-Id: If0ed128d215682bcf066592418420548b06eb6a1
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3259655
> Commit-Queue: Camillo Bruni <cbruni@chromium.org>
> Reviewed-by: Shu-yu Guo <syg@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#77699}
Bug: chromium:1244145
Change-Id: I267f4bdbd8afce81934f4e813dbe1ec09ebdc1ae
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3259538
Auto-Submit: Leszek Swirski <leszeks@chromium.org>
Owners-Override: Leszek Swirski <leszeks@chromium.org>
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Cr-Commit-Position: refs/heads/main@{#77705}
This runtime function behaves like StoreDataPropertyInLiteral, except it
can throw, since it's also used for defining public class fields. Unlike
the literal use case, class field can end up throwing due to field
initializers doing things like freezing the instance.
Bug: chromium:1264828
Change-Id: I3ea4d15ad9b906c26763f022c8e22b757fa80b6c
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3252558
Commit-Queue: Igor Sheludko <ishell@chromium.org>
Auto-Submit: Shu-yu Guo <syg@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77704}
d8 never checked what the actual value of the host-defined options are.
We now properly very that the host-defined options is a specific object
so we we don't end up accidentally ignoring a wrong options object.
Drive-by-fix:
- Convert %AbortJS argument to string
Bug: chromium:1244145
Change-Id: If0ed128d215682bcf066592418420548b06eb6a1
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3259655
Commit-Queue: Camillo Bruni <cbruni@chromium.org>
Reviewed-by: Shu-yu Guo <syg@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77699}
If index > JSObject::kMaxElementIndex, we have to perform a prototype
chain lookup for a named property. The corresponding check was missing
for string receivers.
Fixed: chromium:1265043
Change-Id: Ibccd058a4bd108eeee235762bea0bc4163aaa0b3
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3257704
Auto-Submit: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77683}
Unfortunately, CharacterRanges may use 0x10ffff as a marker value
signifying 'highest possible code unit' irrespective of whether the
regexp instance has the unicode flag or not. This value makes it
through RegExpCharacterClass::ToNode unmodified (since no surrogate
desugaring takes place without /u). Correctly mask out the 0xffff
value for purposes of building our uint16_t range array.
Note: It'd be better to never introduce 0x10ffff in the first place,
but given the irregexp pipeline's lack of hackability I hesitate to
change this - we are sure to rely on it implicitly in other spots.
Drive-by: Refactors.
Fixed: chromium:1264508
Bug: v8:11069
Change-Id: Ib3c5780e91f682f1a6d15f26eb4cf03636d93c25
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3256549
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Mathias Bynens <mathias@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77673}
When a Promise-Reject handler throws an unhandled exception, we should
use that promise's context for reporting the exception to the runtime.
This avoids a null-pointer deref.
Fixed: chromium:1263994
Change-Id: I3792a1884af4a83991249d612caf15588ea77dad
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3250912
Commit-Queue: Jakob Kummerow <jkummerow@chromium.org>
Auto-Submit: Jakob Kummerow <jkummerow@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77652}
In stress runs a GC may kick in before assertOptimized() is called
on the optimized function, causing it to deoptimize and the test
would fail even if the function is optimized as expected.
Adding --no-stress-flush-code --no-flush-bytecode fixes the timing
issue.
Bug: v8:12332
Change-Id: If43c2a06e84c0e3b5e21f7bde160340a0fc742f4
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3237326
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Commit-Queue: Joyee Cheung <joyee@igalia.com>
Cr-Commit-Position: refs/heads/main@{#77645}
Drive-by:
* don't create proto handlers for DefineOwnIC and StoreOwnIC,
* make sure that none of the DefineOwnIC and StoreOwnIC handlers are
leaked into StoreIC's megamorphic stub cache.
Bug: v8:9888, chromium:1259950
Change-Id: I9db538e6ed14bc578aa80df037ffebd9e8c3c649
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3250641
Commit-Queue: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Shu-yu Guo <syg@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77632}
There is a bit of a contradictory register requirement in the
instruction selector for i64x2.mul. We want dst == lhs (when AVX not
supported), but we also want lhs and rhs to be unique (to ensure that
that they don't alias the temp).
We remove the requirement for dst == lhs, since the code gen can handle
both cases (dst == lhs, dst != lhs), at the expense of 1 movaps.
Bug: chromium:1264462
Change-Id: Ia48572412b1f6e0da3551880d8b68a03f42fe2a3
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3253661
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77625}
This is a reland of 45227ffdb4
Differences:
- Handle one more flags conflict in variants.py.
- Disallow %VerifyType without --concurrent-recompilation.
Original change's description:
> [turbofan] extend type asserts to cover all JS types
>
> Extend type assertions to all types covering JavaScript values.
> This is achieved by allocating type representations on the heap using
> newly defined HeapObject subclasses. To allocate these in the compiler,
> we disable concurrent compilation for the --assert-types flag for now.
>
> Fix two type errors that came up with the existing tests:
> 1. JSCreateKeyValueArray has type Array (i.e., a JSArray) instead of
> OtherObject.
> 2. OperationTyper::NumberToString(Type) can type the result as the
> HeapConstant Factory::zero_string(). However, NumberToString does
> not always produce this string. To avoid regressions, the CL keeps
> the HeapConstant type and changes the runtime and builtin code to
> always produce the canonical "0" string.
>
> A few tests were failing because they check for truncations to work
> and prevent deoptimization. However, AssertType nodes destroy all
> truncations (which is by design), so these tests are incompatible
> and now disabled for the assert_types variant.
>
> Drive-by fix: a few minor Torque issues that came up.
>
> Change-Id: If03b7851f7e6803a2f69edead4fa91231998f764
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3234717
> Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
> Reviewed-by: Omer Katz <omerkatz@chromium.org>
> Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#77565}
Change-Id: I5b3c6745c6ad349ff8c2b199d9afdf0a9b5a7392
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3247035
Auto-Submit: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Nico Hartmann <nicohartmann@chromium.org>
Reviewed-by: Omer Katz <omerkatz@chromium.org>
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77596}
The stack-switching test is not expected to pass yet if a GC happens
in the runtime call or in the wasm call.
R=ahaas@chromium.org
Bug: v8:12191, v8:12344
Change-Id: Iba66be58c1abd2ffbb22bbd7d34f8df0246a2a92
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3250900
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Commit-Queue: Thibaud Michaud <thibaudm@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77594}
A "store own" slow runtime was missing, and the slow handler on the
StoreOwnIC was using the non-own slow runtime function, incorrectly
causing setters to be called.
For baseline, [1] invalidates the invariant that StoreOwnIC is only used
for storing properties already in the literal boilerplate, since it's
also used when the new literal is cloned from an object spread.
[1] https://chromium-review.googlesource.com/c/v8/v8/+/3224666
Bug: chromium:1263389, v8:11429
Change-Id: I0284396f306f937d1b8ff96adda6cc133c19726a
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3244308
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Commit-Queue: Shu-yu Guo <syg@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77591}
Bug: v8:12329
Change-Id: I51c38d70537889b7534fb7e6b4066e6ab440234a
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3248163
Auto-Submit: Shu-yu Guo <syg@chromium.org>
Commit-Queue: Adam Klein <adamk@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77582}
This reverts commit 45227ffdb4.
Reason for revert: Breaks on gc_stress mode, see https://ci.chromium.org/ui/p/v8/builders/ci/V8%20Linux%20-%20gc%20stress/35988/overview
Original change's description:
> [turbofan] extend type asserts to cover all JS types
>
> Extend type assertions to all types covering JavaScript values.
> This is achieved by allocating type representations on the heap using
> newly defined HeapObject subclasses. To allocate these in the compiler,
> we disable concurrent compilation for the --assert-types flag for now.
>
> Fix two type errors that came up with the existing tests:
> 1. JSCreateKeyValueArray has type Array (i.e., a JSArray) instead of
> OtherObject.
> 2. OperationTyper::NumberToString(Type) can type the result as the
> HeapConstant Factory::zero_string(). However, NumberToString does
> not always produce this string. To avoid regressions, the CL keeps
> the HeapConstant type and changes the runtime and builtin code to
> always produce the canonical "0" string.
>
> A few tests were failing because they check for truncations to work
> and prevent deoptimization. However, AssertType nodes destroy all
> truncations (which is by design), so these tests are incompatible
> and now disabled for the assert_types variant.
>
> Drive-by fix: a few minor Torque issues that came up.
>
> Change-Id: If03b7851f7e6803a2f69edead4fa91231998f764
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3234717
> Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
> Reviewed-by: Omer Katz <omerkatz@chromium.org>
> Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#77565}
Change-Id: Ia779a11fc811846194c7a8d1e40b372b265e7ea4
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3247034
Auto-Submit: Maya Lekova <mslekova@chromium.org>
Owners-Override: Maya Lekova <mslekova@chromium.org>
Commit-Queue: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Cr-Commit-Position: refs/heads/main@{#77566}
Extend type assertions to all types covering JavaScript values.
This is achieved by allocating type representations on the heap using
newly defined HeapObject subclasses. To allocate these in the compiler,
we disable concurrent compilation for the --assert-types flag for now.
Fix two type errors that came up with the existing tests:
1. JSCreateKeyValueArray has type Array (i.e., a JSArray) instead of
OtherObject.
2. OperationTyper::NumberToString(Type) can type the result as the
HeapConstant Factory::zero_string(). However, NumberToString does
not always produce this string. To avoid regressions, the CL keeps
the HeapConstant type and changes the runtime and builtin code to
always produce the canonical "0" string.
A few tests were failing because they check for truncations to work
and prevent deoptimization. However, AssertType nodes destroy all
truncations (which is by design), so these tests are incompatible
and now disabled for the assert_types variant.
Drive-by fix: a few minor Torque issues that came up.
Change-Id: If03b7851f7e6803a2f69edead4fa91231998f764
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3234717
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Reviewed-by: Omer Katz <omerkatz@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77565}
Recompute the spill offsets for values in the merge region, instead of
reusing the offsets of the source. This ensures that spill slots stay
contiguous (modulo alignment).
This also solves a correctness issue where the spill offsets in the
merge region could move up, thereby overwriting the source of another
move.
With this change, the spill offsets always move down (to fill the gap)
or stay the same, such that processing them from bottom to top
can only overwrite sources of already-processed moves.
Since we do not reuse the current state's offsets, this might generate
extra stack moves and regress generated code performance a bit.
Drive-by: print spill offsets in the Liftoff trace
R=clemensb@chromium.org
Bug: v8:12270
Change-Id: I8d20df8fc1e80dd36b6f651de457686e9935a628
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3245115
Commit-Queue: Thibaud Michaud <thibaudm@chromium.org>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77556}
According to the spec, the three parameters (stdlib, foreign, and heap)
must be mutually distinct. We did not check this yet, which led to
observable differences between asm validation and standard JavaScript
semantics.
R=thibaudm@chromium.org
Bug: chromium:1068355
Change-Id: I451f63d10ea50474aeb6e8a547918b5af769343b
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3244408
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Thibaud Michaud <thibaudm@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77554}
We used to apply an invalid optimization which skips `length` writes
if the JSArray is 'fast' and the old value equals the new value. This
optimization is not valid if e.g. `length` is non-writable.
Fixed: chromium:1262478
Change-Id: I49ef50de293dae5c3a62c64b303ec34b9c0f6cbc
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3236720
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Auto-Submit: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77552}
This CL fixes an issue in EffectControlLinearizer, where the primitive
type of the argument of a fast C call was used to compute its
MachineType even when the argument was actually a TypedArray, which
should always be treated as a Tagged type. This resulted in Float32/64
typed arrays being passed in FP registers, leading to a crash in the
register allocator.
Drive-by fix: Fixed output from --trace-turbo-alloc so that all of its
sub-parts are printed with PrintF, avoiding interleaved log lines.
Bug: chromium:1260954
Change-Id: I249c8629daae3af437fb52f53f45211f3a214222
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3231341
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Commit-Queue: Maya Lekova <mslekova@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77521}
Character class handling in the irregexp pipeline is quite complex;
codepoints outside the BMP (basic multilingual plane) are only
translated into surrogate pairs when needed, e.g. when the subject
string is two-byte. If not needed, the codepoints simply stay part of
the list of CharacterRanges.
In EmitCharClass, we determine the valid subset of ranges through
ranges_length; until this CL, we forgot to pass that information on to
MakeRangeArray. Do that now by truncating the list of CharacterRanges.
Fixed: chromium:1262423
Change-Id: I5bb5b839e9935890ca2d10908ad66d72c3217178
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3240782
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Auto-Submit: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Mathias Bynens <mathias@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77514}
To get more details about the failure when it flakes.
Bug: v8:12332
Change-Id: I1bf5188efe13a75e34dd9832e6275fd985ee8dbd
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3233974
Commit-Queue: Joyee Cheung <joyee@igalia.com>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77498}
If an output operand has "same as input" policy, we cannot assume that
it's input 0. Instead we should look at the {input_index}.
The bug manifests on Wasm select instructions, where the input index is
actually 2 and not 0.
In order to test this better, we introduce the a new
--turbo-force-mid-tier-regalloc flag, which always uses the mid-tier
register allocator. Otherwise the bug would only manifest on huge
functions.
R=mslekova@chromium.org
CC=thibaudm@chromium.org
Bug: v8:12330
Change-Id: I6a005a48bbd2aba354dc99fed587bffce24c8839
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3234722
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77495}
Calling class constructors without new is a spec violation and should
raise an exception.
In [1] calling class constructors with spread (without new) was handled
by reducing the node to a call to runtime to raise the exception.
However, arguments of the call have to be evaluated first ([2]).
This CL changes the reduction of JSCallWithSpread/JSCallWithArrayLike to
a no-op in JSCallReducer if the target is a class constructor, delaying
raising of the exception to the call builtin.
[1] https://crrev.com/c/3229369
[2] https://tc39.es/ecma262/#sec-evaluatecall
Bug: chromium:1262007
Change-Id: I2ef504d4ce6e51d582b5951beb6debb983cefba6
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3236348
Commit-Queue: Patrick Thier <pthier@chromium.org>
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77492}
--no-stress-flush-bytecode doesn't exist and should be
--no-stress-flush-code. Not supressing it means a tester could pass
--stress-flush-code and --no-flush-bytecode, which are contradictory and
will assert.
Bug: v8:12331
Cq-Include-Trybots: luci.v8.try:v8_numfuzz_dbg_ng,v8_numfuzz_ng
Change-Id: I6490271bcb11f5ea925eb8b65fbe0455c2dafeaf
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3233952
Reviewed-by: Zhi An Ng <zhin@chromium.org>
Commit-Queue: Shu-yu Guo <syg@chromium.org>
Auto-Submit: Shu-yu Guo <syg@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77483}
Handling of class constructors was moved from CallFunction to Call
in [1].
When reducing calls with spread we forward varargs directly to
CallFunction, if we are spreading to inlined arguments or arguments of
the outermost function.
In that case we didn't check for class constructors and therefore didn't
raise an exception.
This CL adds checks for class constructors to all JSCall* nodes in
JSCallReducer that missed them before.
[1] https://crrev.com/c/3186434
Bug: chromium:1260623
Change-Id: Id39cdfd09ff5aae804ae30d96909518e408c9613
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3229369
Commit-Queue: Patrick Thier <pthier@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77472}
