In wasm code, we sometimes see the pattern
<some 64 bit expression>
i32.wrap/i64
i32.load
where we generate an instruction to extend the 32 bit offset into a zero
extended 64 bit value for the actual load. However, the preceeding
truncate already yields a zero extended 32 bit value, so the extra
instruction is not needed. Even more, it might get in the way of
munching more computation into the final load.
This change adds information about the zero extending behavior to
the existing optimization that avoids the zero extension.
Bug: chromium:853685
Change-Id: Iab9179379923ecb88651df6091b3d9408341cf4c
Reviewed-on: https://chromium-review.googlesource.com/c/1421839
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Stephan Herhut <herhut@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58998}
{Isolate::per_isolate_thread_data_key()} is not even defined, and
{Isolate::isolate_key()} is unused.
R=mstarzinger@chromium.org
Bug: v8:8562
Change-Id: I490989510865903c702158e33621c9990052c2a8
Reviewed-on: https://chromium-review.googlesource.com/c/1425907
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58997}
The decision as to whether to optimize an IIFE as oneshot depends on
whether it's outer scope is the script scope. During lazy compile, we
might have discarded scopes which don't need a context between the IIFE
and the script scope, which means we might treat an IIFE as oneshot,
even though initial eager compile treated it as non-oneshot. Both
bytecode flushing and lazy source positions rely on us generating the
same bytecode during lazy compile as eager compile, so we move the
decision into the parser where it happens once and is then stored in
the SFI for any future lazy compiles.
BUG=v8:8395,v8:8510
Change-Id: I88f1e74ad95d47a2636c393ceb1318d7d610055d
Reviewed-on: https://chromium-review.googlesource.com/c/1421841
Reviewed-by: Toon Verwaest <verwaest@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58996}
This adds missing cases for exception handling opcodes to the stack
effect helper {WasmDecoder::StackEffect}. It is a first step towards
adding exception handling support to the {WasmInterpreter}.
R=clemensh@chromium.org
BUG=v8:8091
Change-Id: Idacf440a894e5c71a180502c1d2f10fa15c8f5fa
Reviewed-on: https://chromium-review.googlesource.com/c/1425911
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58995}
Reworking and adding a node type would require also adding
parallelization support for minor mc. Since this is unused and not
benchmarked right now, just remove it.
Bug: chromium:923361
Change-Id: Iaf67a743d76d2b37ffff9961b510bfd8a1bd15ff
Reviewed-on: https://chromium-review.googlesource.com/c/1425900
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58993}
This saves about 80,000 LoC after preprocessor expansion.
Bug: v8:8562
Change-Id: I67b20edb73b801ddcc2937b84468241e3076535f
Reviewed-on: https://chromium-review.googlesource.com/c/1425906
Reviewed-by: Peter Marshall <petermarshall@chromium.org>
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58989}
This is part of an effort to improve the performance of TA#subarray.
Bug: v8:7161
Change-Id: I6f4b0f01e498d48e0fce11fbf7dcd7a0ad1ae748
Reviewed-on: https://chromium-review.googlesource.com/c/1425002
Commit-Queue: Peter Wong <peter.wm.wong@gmail.com>
Reviewed-by: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58988}
This prevents the bytecode generator generating loads that look for
extensions in the global context, which can never succeed and means
that lazy and eager bytecode compilation will match.
Bug: v8:8510
Change-Id: I51dca62b5d1ee34f8dea82260cf27295ddf427d9
Reviewed-on: https://chromium-review.googlesource.com/c/1425520
Reviewed-by: Toon Verwaest <verwaest@chromium.org>
Commit-Queue: Dan Elphick <delphick@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58983}
Currently the memory reducer is activated only after the first mark-
compact GC, which triggered after the old generation reaches 8 MB.
That threshold is too large for mobile. This patch adds a heuristic
to activate the memory reducer if the old generation expands by more
than 1 MB after the bootstrap.
Change-Id: Ic38bc6e2fe8887677f764246c45e38d237e49a94
Reviewed-on: https://chromium-review.googlesource.com/c/1425898
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58982}
Everything was including log.h through heap-inl.h, so remove that
include by moving the one user into heap.cc, and then fix all the
include errors.
This reduces the log.h include ball from ~550 to ~100.
Change-Id: I6d09bc2f365b48645fcfdc695a68ea12539a745d
Reviewed-on: https://chromium-review.googlesource.com/c/1424198
Commit-Queue: Georg Neis <neis@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58981}
This reverts commit d11a0648af.
Reason for revert: <INSERT REASONING HERE>
Original change's description:
> [torque] Implement safe initialization of classes through hidden structs
>
> Initialization of classes now happens atomically at the end of the
> class constructor only once all of the values for the class' fields
> have been fully computed. This makes Torque constructors completely
> GC safe, e.g. hardened against allocations or exceptions in
> constructors.
>
> As part of this change, make the 'this' parameter for method calls
> explicit rather than implicit.
>
> Drive by: add validation to check for duplicate field declarations
>
> Bug: v8:7793
> Change-Id: I8b5e85980d6a103ef9fc3262b76f6514f36ebf88
> Reviewed-on: https://chromium-review.googlesource.com/c/1411252
> Commit-Queue: Daniel Clifford <danno@chromium.org>
> Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#58979}
TBR=danno@chromium.org,tebbi@chromium.org
Change-Id: Id6c46c175f53c5a77db1e6ca242586fba34cd02e
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: v8:7793
Reviewed-on: https://chromium-review.googlesource.com/c/1426121
Reviewed-by: Daniel Clifford <danno@chromium.org>
Commit-Queue: Daniel Clifford <danno@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58980}
Initialization of classes now happens atomically at the end of the
class constructor only once all of the values for the class' fields
have been fully computed. This makes Torque constructors completely
GC safe, e.g. hardened against allocations or exceptions in
constructors.
As part of this change, make the 'this' parameter for method calls
explicit rather than implicit.
Drive by: add validation to check for duplicate field declarations
Bug: v8:7793
Change-Id: I8b5e85980d6a103ef9fc3262b76f6514f36ebf88
Reviewed-on: https://chromium-review.googlesource.com/c/1411252
Commit-Queue: Daniel Clifford <danno@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58979}
This includes generalizing the notion of hints to allow for
unallocated functions, represented by a pair of SFI and FeedbackVector.
Bug: v8:7790
Change-Id: I7887665e1981b2039ecd626b82aebd5b5b64263c
Reviewed-on: https://chromium-review.googlesource.com/c/1424946
Commit-Queue: Georg Neis <neis@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58978}
When inlining based on CreateClosure, we don't have a JSFunction but
only the SharedFunctionInfo and FeedbackVector.
Bug: v8:7790
Change-Id: I7a3cf50710273c7175e43e969d2364cff11c3d93
Reviewed-on: https://chromium-review.googlesource.com/c/1421357
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58977}
Many values stored in the preparse data for the skippable functions
fit in one byte most of the time. The varint encoding uses a single
continue bit per byte to tell whether there is a following byte.
Change-Id: Ia0a622ba42a338fc91eea1e0c1a72d2582d9f867
Reviewed-on: https://chromium-review.googlesource.com/c/1400842
Commit-Queue: Camillo Bruni <cbruni@chromium.org>
Reviewed-by: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58972}
This current fails, since {mmap} fails with EINVAL for empty mappings.
The destructor already has special handling for a {nullptr} mapping, so
we can just use {nullptr} for empty files. We get a similar error on
windows, and can fix it the same way.
On order to make presubmit checks happy, we have to skip copyright
checking and checking for terminating newlines for empty files.
R=mlippautz@chromium.org
Change-Id: I2b73da7ff6df72d8bdd40df1fff6422e0a46881e
Reviewed-on: https://chromium-review.googlesource.com/c/1424861
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58970}
This removes the unused "argument count" field from the safepoint table
as the field was unused by now and always contained the value zero.
Also note that associating a callee's argument count with the call-site
is not compatible with tail-call support. When tail-calling a function
with a different number of arguments, the information associated with
the call-site becomes stale. The number of arguments is a property of
the callee, not of the call-site in the caller. For this reason the
field in question is not usable to support reference types in function
arguments (at least when tail-calls are also supported).
R=ahaas@chromium.org
Change-Id: If667d729267f2dd2642b755c54235cc08ca9b141
Reviewed-on: https://chromium-review.googlesource.com/c/1402548
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58969}
The purpose of the {WireBytesStorage} (typically held in a shared_ptr
itself) is to keep the actual wire bytes alive. Thus implement it this
way for the {NativeModuleWireBytesStorage}.
R=mstarzinger@chromium.org
Bug: v8:8689
Change-Id: I8f171b4fa8b80b517badb1b1d3228503a32830dd
Reviewed-on: https://chromium-review.googlesource.com/c/1421362
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58966}
The change is just a refactoring without functional changes.
Bug: chromium:923361
Change-Id: Ie633c56122ff72658f0238dc40db698350a7b8e7
Reviewed-on: https://chromium-review.googlesource.com/c/1421363
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58963}
The temp registers might be needed by the assembler, so avoid using them
in LiftoffAssembler. Use Liftoff cache registers instead. This might
introduce additional spills if all registers are in use, but this is
unlikely.
This also simplifies the logic to ensure non-aliasing of certain
registers.
R=ahaas@chromium.org
Bug: chromium:922933, v8:6600
Change-Id: Ie929d9de0b6f4f41c6117d820b6a367dd0a342f7
Reviewed-on: https://chromium-review.googlesource.com/c/1424862
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58961}
This allows us to stop tracking variables_ in the preparser.
This currently makes us track slightly more variables than neccessary in the
case `for (var ...` since `var ... of` needs to check conflicts with out simple
catch variables. We should probably track the names through a ScopedPtrList
instead of a ZonePtrList anyway. Then it won't matter anymore.
Change-Id: I64e3f9ab13af8269456439cf15b0bc4d5b9e5380
Reviewed-on: https://chromium-review.googlesource.com/c/1421360
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Commit-Queue: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58960}
This is a reland of 4e1d7c87b9.
Failure on arm and arm64 is fixed by https://crrev.com/c/1411885.
Original change's description:
> [wasm] Split compilation in three stages
>
> In order to refactor ownership between objects in wasm compilation, the
> compilation (executed by background tasks) is split in three stages:
> getting a compilation unit (while holding a mutex), executing the work
> (without any mutex and without keeping the NativeModule alive), and
> submitting the work (with a mutex again).
>
> This CL prepares this design by splitting compilation from submission.
> Both steps are still executed right after each other. This will be
> changed in a follow-up CL.
>
> R=titzer@chromium.org
> CC=mstarzinger@chromium.org
>
> Bug: v8:8689
> Change-Id: I2f92aee8e2f2d45470d8c63314ed026341630902
> Reviewed-on: https://chromium-review.googlesource.com/c/1414920
> Reviewed-by: Ben Titzer <titzer@chromium.org>
> Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#58929}
TBR=titzer@chromium.org
Bug: v8:8689
Change-Id: I58ff07d0e0ac8df0f6ee23c416f992954f4673d2
Reviewed-on: https://chromium-review.googlesource.com/c/1422748
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58959}
This extends the CodeReference class to be able to hold a pointer to
a CodeDesc struct which represents finished but unpublished wasm
code.
For readability, it refactors the implementation of the dispatch for
the different CodeReference kinds.
R=mstarzinger@chromium.org
Bug: v8:8689
Change-Id: Idd031dffaa9d80740c073b6cca0fc3561c5df0c1
Reviewed-on: https://chromium-review.googlesource.com/c/1411885
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58958}
- Use unique ptrs for owned objects
- Remove friendship with CpuProfiler and replace with public API
- Remove unused method LogFailure()
- Remove StopProfiler() which was only used by LogFailure() (removed)
and one test, which can use StopProfilerThread() instead
- Remove 'paused' state which was only used by the above
- Remove 'engage' state. There is no reason we need this as along as
users keep track of Engage/Disengage calls
Drive-by cleanup:
- Remove import of log.h from profile-generator.h
- Remove unnecessary includes of log.h
Change-Id: Ifc4ca156bef038c40953f8361ffea17788e3a59b
Reviewed-on: https://chromium-review.googlesource.com/c/1424338
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58957}
Fix fixes a minor DCHECK error, which fired because usually only
LiftoffRegisters of the same type should be compared against each
other. In the RegisterReuseMap we store both register pairs and single
registers, hence check for pair vs non-pair comparisons explicitly.
R=ahaas@chromium.org
Bug: v8:6600, chromium:922670
Change-Id: I18beb61b6b1906cc42bcf6e7dfdd5eb803e874d7
Reviewed-on: https://chromium-review.googlesource.com/c/1421921
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58956}
Simplify the internal state of Sampler a bit. There are basically two
users of Sampler - the CpuSampler used by the CpuProfiler and the
Ticker used by log.cc. Ticker calls Start/Stop to manage the Sampler
lifetime, but CpuProfiler does not. This leads to much confusion and
overlap of functionality.
Fix that here by removing the distinction between active, registered
and isProfiling states. These are now all the same thing and are
represented by IsActive(). The state is set to active when Start is
called, and set inactive when Stop is called. Both users of Sampler
now call Start and Stop at appropriate times.
The concept of profiling depth was not used - each Sampler would
only ever have a sampling depth of 1. We still need to call
SignalHandler::IncreaseSamplerCount(), so we do that in Start
and the corresponding DecreaseSamplerCount() in Stop.
Change-Id: I16a9435d26169a7dd00b1c7876e66af45f12e4b0
Reviewed-on: https://chromium-review.googlesource.com/c/1424337
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58955}
Use variable tracking from ExpressionScopes rather than the PatternRewriter and
PreParserExpression::variables_ to declare variables.
We only figure out that variables are non-simple parameters once we see the
first non-simple parameter. This still uses the pattern rewriter to make
variables non-simple (kLet instead of kVar).
Change-Id: I4a4ee4852d667c26806bb24896722cfea3e093f2
Reviewed-on: https://chromium-review.googlesource.com/c/1417630
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Commit-Queue: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58954}
Since the PreparseData now directly contains the byte data inline it can
grow very large as well.
Bug: chromium:923264
Change-Id: I456d5bcbfb40587b283584f726d9e084061fd30f
Reviewed-on: https://chromium-review.googlesource.com/c/1421321
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Commit-Queue: Camillo Bruni <cbruni@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58953}
This CL adds a stack check to the TFS builtin "FlattenIntoArray" as it
is called recursively and can cause a SEGV with a large enough
"depth" argument.
R=jgruber@chromium.org
Bug: v8:8708
Change-Id: I833506531bcff1c4703b9a21678028cf0e63638d
Reviewed-on: https://chromium-review.googlesource.com/c/1424858
Commit-Queue: Simon Zünd <szuend@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58952}
With and without infra_staging flag testrunner behaves the same for old
features. This CL removes duplicate tests testing the same behavior.
R=machenbach@chromium.orgCC=yangguo@chromium.org,sergiyb@chromium.org
No-Try: true
Bug: v8:8174
Change-Id: Icf7bea15b2343b90697016d050fa0d918a99997d
Reviewed-on: https://chromium-review.googlesource.com/c/1424859
Reviewed-by: Sergiy Belozorov <sergiyb@chromium.org>
Commit-Queue: Tamer Tas <tmrts@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58951}
There's no reason to use our self-baked atomics anymore. Also
- Changes two boolean values to use a boolean instead of an int
- Uses a unique ptr for data_
- Removes has_processing_thread_ which is not used
- Moves most initialization inline into the class
- Removes SetUp/TearDown which weren't needed
Change-Id: I8f50133636961502d56351abd2fb17196603a01a
Reviewed-on: https://chromium-review.googlesource.com/c/1422918
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#58950}
