Many safepoint tables do not contain any deoptimization info and/or no
callee-saved registers. Do not emit empty fields for all entries in this
case.
This often shrinks the size of the encoded safepoint table by more than
50%.
Drive-by cleanups:
- Rename fields of the safepoint table entries to clarify their meaning
("tagged slots" instead of "bits", "tagged register indexes" instead of
"register bits").
- Include the PC in the decoded {SafepointEntry} to make it the single
source of truth.
R=jkummerow@chromium.org
Bug: v8:12401
Change-Id: If5c24a688a434842ed3b6427f5f1f3ea9232173a
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3289173
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78021}
With dynamic tiering, not all functions get compiled with TurboFan,
and not all functions can therefore get cached. With this metric we
want to keep track of the time needed to compile missing functions with
Liftoff.
R=jkummerow@chromium.org
Bug: v8:12281
Change-Id: I1f2de3e1f6b004477985df5d2116f69371528c3f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3289178
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78020}
This reverts commit d7c3f1cd8a.
Reason for revert: https://ci.chromium.org/ui/p/v8/builders/ci/V8%20Android%20Arm64%20-%20debug%20builder/22043/overview
Original change's description:
> Reland "[fastcall] Enable float support on arm64 simulator"
>
> This is a reland of b9ddcbc86f
>
> The original CL was reverted due to an MSAN issue, that is fixed by
> moving the signature mapping onto the Isolate (instead of having
> per-thread storage, which got invalid on multithreaded compilation).
>
> This CL also contains fixes for the Bazel config and for a data race
> when obtaining the PerIsolateSimulatorData.
>
> Original change's description:
> > [fastcall] Enable float support on arm64 simulator
> >
> > This CL adds support for handling calls to C functions with arbitrary
> > signatures on the arm64 simulator. It adds infrastructure for
> > encoding the signature data from CallDescriptor and FunctionInfo
> > classes into a compact representation, stored in the simulator and
> > called EncodedCSignature.
> >
> > Design doc:
> > https://docs.google.com/document/d/1ZxOF3GSyNmtU0C0YJvrsydPJj35W_tTJZymeXwfDxoI/edit
> >
> > This CL is a follow up on the native support added in
> > https://chromium-review.googlesource.com/c/v8/v8/+/3182232
> > and is partially based on the previous attempt:
> > https://chromium-review.googlesource.com/c/v8/v8/+/2343072
> >
> > Bug: chromium:1052746
> > Change-Id: I0991b47bd644b2fc2244c5eb923b085261f04765
> > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3060486
> > Commit-Queue: Maya Lekova <mslekova@chromium.org>
> > Reviewed-by: Camillo Bruni <cbruni@chromium.org>
> > Reviewed-by: Jakob Gruber <jgruber@chromium.org>
> > Cr-Commit-Position: refs/heads/main@{#77744}
>
> Bug: chromium:1052746, chromium:1267854
> Change-Id: I89bbd01e33fb1080543d98bcfd4c2d17b5c76861
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3270541
> Reviewed-by: Jakob Gruber <jgruber@chromium.org>
> Reviewed-by: Camillo Bruni <cbruni@chromium.org>
> Commit-Queue: Maya Lekova <mslekova@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#78018}
Bug: chromium:1052746, chromium:1267854
Change-Id: Ia8f10d085d13990b331f306957f95ecf3e003cfd
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3295453
Owners-Override: Nico Hartmann <nicohartmann@chromium.org>
Bot-Commit: Rubber Stamper <rubber-stamper@appspot.gserviceaccount.com>
Commit-Queue: Nico Hartmann <nicohartmann@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78019}
This is a reland of b9ddcbc86f
The original CL was reverted due to an MSAN issue, that is fixed by
moving the signature mapping onto the Isolate (instead of having
per-thread storage, which got invalid on multithreaded compilation).
This CL also contains fixes for the Bazel config and for a data race
when obtaining the PerIsolateSimulatorData.
Original change's description:
> [fastcall] Enable float support on arm64 simulator
>
> This CL adds support for handling calls to C functions with arbitrary
> signatures on the arm64 simulator. It adds infrastructure for
> encoding the signature data from CallDescriptor and FunctionInfo
> classes into a compact representation, stored in the simulator and
> called EncodedCSignature.
>
> Design doc:
> https://docs.google.com/document/d/1ZxOF3GSyNmtU0C0YJvrsydPJj35W_tTJZymeXwfDxoI/edit
>
> This CL is a follow up on the native support added in
> https://chromium-review.googlesource.com/c/v8/v8/+/3182232
> and is partially based on the previous attempt:
> https://chromium-review.googlesource.com/c/v8/v8/+/2343072
>
> Bug: chromium:1052746
> Change-Id: I0991b47bd644b2fc2244c5eb923b085261f04765
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3060486
> Commit-Queue: Maya Lekova <mslekova@chromium.org>
> Reviewed-by: Camillo Bruni <cbruni@chromium.org>
> Reviewed-by: Jakob Gruber <jgruber@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#77744}
Bug: chromium:1052746, chromium:1267854
Change-Id: I89bbd01e33fb1080543d98bcfd4c2d17b5c76861
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3270541
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Camillo Bruni <cbruni@chromium.org>
Commit-Queue: Maya Lekova <mslekova@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78018}
There are cases where DeclareEvalVar can fail to create a variable, in
particular when the declaration scope is not extensible. Throw an error
in such cases.
Fixed: chromium:1145119
Change-Id: I6260f570570c4d7801231bdb5f6724f3734444fd
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3295348
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Auto-Submit: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78017}
Once CppHeap moves away from EmbedderHeapTracer, it can no longer be
returned by Isolate::GetEmbedderHeapTracer. Blink currently uses
GetEmbedderHeapTracer to trigger GCs. This CL introduces a variant of
Isolate::RequestGarbageCollectionForTesting to replace that.
Bug: v8:12407
Change-Id: I099989f55291a0daab7d2051072dedddef58dd40
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3293413
Commit-Queue: Omer Katz <omerkatz@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78015}
Instead of just DCHECKing that we are in a safepoint in
Snapshot::Create, pass the GlobalSafepointScope as argument.
Bug: v8:11708, v8:12377
Change-Id: Ia856c27b92223bf3a96fe6946b86e4625f3a1e0f
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3295342
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Dominik Inführ <dinfuehr@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78014}
This is a reland of 90a9d6cb13
The original CL got reverted because of two different issues:
* The DCHECK failure on AllowGarbageCollection::IsAllowed() got fixed
in https://crrev.com/c/3289625.
* The crash with the incremental marking job were because of a nested
GC started from a SafepointScope. This CL adds IgnoreLocalGCRequests
scopes to SafepointScopes in src/heap.
In addition this CL prevents shared GCs during isolate deserialization
by locking the clients_mutex_ until the isolate is fully deserialized.
The original GC used a DisallowSafepoints scope to prevent shared GCs
from interrupting isolate deserialization.
Original change's description:
> [heap] Support multiple clients in shared GC
>
> Add support for safepointing multiple isolates as described in the
> design doc (link is below). A safepoint across multiple isolates is
> considered a global safepoint to distinguish it from regular safepoints.
>
> The basic idea behind the implementation is that we reach a
> safepoint for each client. What's new is that now also main threads
> need to participate in the safepointing protocol and need to give up
> control in time. The slow paths of Park(), Unpark() and Safepoint() on
> the main thread need to be adjusted for this reason as well.
>
> This CL introduces GlobalSafepoint and GlobalSafepointScope to mirror
> IsolateSafepoint and IsolateSafepointScope.
>
> This CL adds the type IgnoreLocalGCRequests, it is used to prevent
> Park() and Unpark() from honoring the request from background threads
> to perform a local GC. This is used heap-internally to not have GCs
> (or even nested GCs) in certain locations. E.g. when initiating a
> safepoint to perform a GC we don't want a "recursive" GC to occur.
>
> Design doc: https://docs.google.com/document/d/1y6C9zAACEr0sBYMIYk3YpXosnkF3Ak4CEuWJu1-3zXs/edit?usp=sharing
>
> Bug: v8:11708
> Change-Id: I5aca8f5f24873279271a53be3bb093fc92a1a1eb
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3009224
> Commit-Queue: Dominik Inführ <dinfuehr@chromium.org>
> Reviewed-by: Jakob Gruber <jgruber@chromium.org>
> Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
> Cr-Commit-Position: refs/heads/main@{#77812}
Bug: v8:11708, v8:12375, v8:12377
Change-Id: I9d1af6fbc06a3a8b6f216ec5e9027665ad071809
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3283067
Commit-Queue: Dominik Inführ <dinfuehr@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78013}
Starting an incremental GC when it is not supported is not allowed but
trying to finalize one when it might be unknown whether it's supported
or not should be legal.
In such cases, finalization would immediately bail out since there is no
active gc.
Bug: chromium:1272076
Change-Id: If5aaaa130dc55474d3fc3afcd7039880a2c11333
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3293411
Auto-Submit: Omer Katz <omerkatz@chromium.org>
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78010}
This is a partial revert of https://crrev.com/c/2033171 to bring back
WasmModuleObject::Compile, that compiles Wasm bytes into a Wasm module.
Also adding a simple test to make sure the compile works.
There are some users, e.g. Cloudflare, Halide, see bug for more details.
Bug: v8:10461
Change-Id: I199abea108311025decad793d2232a08bba05d38
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3292088
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78003}
PPC and S390 implement vsel slightly differently which
is also reflected on the simulator.
Change-Id: Iec5e49bfae09d8bf15149e5ab149b82971271b24
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3293409
Reviewed-by: Junliang Yan <junyan@redhat.com>
Commit-Queue: Milad Fa <mfarazma@redhat.com>
Cr-Commit-Position: refs/heads/main@{#78002}
`foo ## __LINE__` just emits foo__LINE__ because of how preprocessor
expansion works.
The typical solution for this is to use a CONCAT macro, but we in fact
already have a helper for what this is trying to solve,
UNIQUE_IDENTIFIER, so just use that instead.
Change-Id: Icea3f01db458c5d557e0affd3b004f4478c6c315
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3293084
Auto-Submit: Leszek Swirski <leszeks@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Reviewed-by: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77998}
Previously, an allocation from a separate thread could grab the
just-released region and make it accessible before the regions
permissions are changed to kNoAccess at the end of ReleasePages.
Bug: v8:12414
Change-Id: I98c8f8e3df76d4a44c357ddab107cfeff20049b8
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3293083
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Commit-Queue: Samuel Groß <saelo@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77997}
When starting incremental marking from an allocation observer, the
DCHECK for AllowGarbageCollection::IsAllowed() fails. We need to
explicitly allow this here in order to use SafepointScope.
Bug: chromium:1271659, v8:11708, v8:12413
Change-Id: I8607a82faed50f8dfffce1e495fb5e715a47bc1b
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3293086
Commit-Queue: Dominik Inführ <dinfuehr@chromium.org>
Auto-Submit: Dominik Inführ <dinfuehr@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77996}
In such a case, we must post-process all aborted pages. The early
bailout assumed that not pages are aborted before trying compaction.
Bug: chromium:1271229
Change-Id: I4c5998b9d1c7974f7d1d25a6729f42d5690a5759
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3289171
Reviewed-by: Hannes Payer <hpayer@chromium.org>
Reviewed-by: Dominik Inführ <dinfuehr@chromium.org>
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77987}
The entries contain more than just deoptimization info. Thus rename them
to "entries" and use "EntryBuilder" instead of "DeoptimizationInfo".
Drive-by: Remove the redundant {emitted_} field, just check if
{offset_} was initialized.
R=jkummerow@chromium.org
Bug: v8:12401
Change-Id: Ifedc265dc27dd6817b731d9e24a1d8654edc99de
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3291310
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77985}
Verify certain properties when emitting safepoint tables:
- entries are ordered by PC,
- trampoline PCs are also in order,
- all trampoline PCs come after all regular PCs, and
- an entry with a deopt index also has a trampoline PC set.
Drive-by: Slightly simplify logic in {UpdateDeoptimizationInfo}.
Drive-by 2: Inline IsIdenticalExceptForPc into the only caller, and
further simplify the logic there.
R=jkummerow@chromium.org
Bug: v8:12401
Change-Id: I8ea41b1628dceb914191f23165c8dccb79f9204c
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3289162
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77983}
We should not assume that the embedder can run with concurrent marking
even though we just finalize the GC at this point.
Bug: chromium:1271371
Change-Id: I3d3a45e4283dcae9c6fdd1a467158254914868c3
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3289169
Reviewed-by: Omer Katz <omerkatz@chromium.org>
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77982}
A SafepointScope might need to block for a shared GC initiated from
another client isolate. This means that anytime we create a
SafepointScope a shared GC may run. This CL adds a DCHECK to ensure
AllowGarbageCollected::IsAllowed() holds for each SafepointScope.
So far this DCHECK was only run in the less likely event that a
SafepointScope actually runs a shared GC. Which is technically good
enough but it is easy to miss use cases of SafepointScope where this
does not hold.
Bug: v8:11708, v8:12377
Change-Id: I30cc33c05ebe4835430e1d699a86079810523858
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3289625
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Dominik Inführ <dinfuehr@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77976}
Many callers already pass an int, so there is no point in storing as an
uint32_t internally. The style guide also recommends int as the general
data type for integers, even if it's statically known that only positive
values will occur.
R=jkummerow@chromium.org
Bug: v8:12401
Change-Id: I6067139f514895f925d1c536112b4cb5c2c24a36
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3289157
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77975}
In order to determine which module contains V8, v8windbg has been
looking for the known symbol v8::Script::Run. However, that symbol might
not be found if the function was inlined. To fix, we should instead
choose a symbol that is declared with V8_NOINLINE.
Change-Id: Ib73dfb27f35ab81ab69f92ffab944f6f1096b895
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3290107
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Commit-Queue: Seth Brenith <seth.brenith@microsoft.com>
Cr-Commit-Position: refs/heads/main@{#77974}
Refactor s.t. we now compact the DependentCode weak fixed array both
when adding new entries and when marked for deopts.
Bug: v8:12397
Change-Id: I842f3369644e88cc0b98a1f1371259c920cc8bbf
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3291320
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Dominik Inführ <dinfuehr@chromium.org>
Reviewed-by: Dominik Inführ <dinfuehr@chromium.org>
Auto-Submit: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77973}
The cpu profiler tests require a precise trace and become
flaky when running with concurrent sparkplug.
Bug: v8:12054
Change-Id: I62a20d15530c610bc1091371781c28687c718720
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3289158
Auto-Submit: Victor Gomes <victorgomes@chromium.org>
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/main@{#77972}
