kFpRegPair requires a bit of special case logic in allocation and spill
logic, since it requires adjacent fp registers, with the low register
being an even-numbered one.
We make use of a shift+and trick with the reg list to check this. Given
a bit mask of registers available, AND this with a right-shift-by-one
version of itself, and then AND it again with with a bit mask of even
numbered fp registers, we can get a bit mask where all available even
registers are set.
For example, given 00011011, where 1 means register is available:
00011011
& 0001101 (right shift by 1)
----------
00001001
& 01010101 (mask of even registers)
----------
00000001
We see that only the first fp register is available, based on the
adjacency requirement.
The mask of even registers is defined to be uint64_t, to workarond shift
overflow issues when non-unsigned values are used, and also to make this
work on all platforms and compilers without implicit conversion warnings.
It is a bit wasteful, since we really only need storage_t, but since it
is a constexpr it should be okay.
The spilling case tries to be a bit smart, since there are three possible
cases (described in the code), two of which we can spill one register,
and in the last case we have to spill two registers. We try to
distinguish between these cases in order to minimize spills. The dumb
thing here is in the last case, we arbitrarily pick the first two
registers (d0, d1) to spill.
Bug: v8:9909
Change-Id: I408f9f79b1f3c505f2fd73bad15923188ae7aaf4
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1975033
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65546}
Add support for storing values of type kWasmS128 to ia32, x64, arm,
arm64. These stores are never hit right now, since Liftoff does not
handle any S128 values yet.
Bug: v8:9909
Change-Id: Ib74b9db158428186757086711b8249866026dab5
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1974834
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65545}
These worklists are needed for accounting of objects retained by
native contexts for the new memory measurement API.
Bug: chromium:973627
Change-Id: I354c5ebbbac11da4d01800164e15b94a93aa654c
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1943158
Commit-Queue: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Hannes Payer <hpayer@chromium.org>
Reviewed-by: Dominik Inführ <dinfuehr@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65542}
This reverts commit 91e3243d60.
Reason for revert: This deopts to the wrong point.
Original change's description:
> Extend GetIterator bytecode to perform JSReceiver check on object[Symbol.iterator]()
>
> Current GetIterator bytecode loads and calls @@iterator property on a
> given object. This change extends the bytecode functionality to check
> whether the value returned after calling @@iterator property is a valid
> JSReceiver. The bytecode throws SymbolIteratorInvalid exception if the
> returned value is not a valid JSReceiver. This change absorbs the
> functionality of additional two bytecodes - JumpIfJSReceiver and
> CallRuntime, that are part of the iterator protocol in the GetIterator
> bytecode.
>
> Bug: v8:9489
> Change-Id: I9e84cfe85eeb9a1b8a97ca0595375ac26ba1bbfd
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1792905
> Reviewed-by: Leszek Swirski <leszeks@chromium.org>
> Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
> Commit-Queue: Swapnil Gaikwad <swapnilgaikwad@google.com>
> Cr-Commit-Position: refs/heads/master@{#63704}
TBR=rmcilroy@chromium.org,leszeks@chromium.org,tebbi@chromium.org,swapnilgaikwad@google.com
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: v8:9489
Change-Id: I9324b5b01ead29912ad793a1e7b4d009643d7901
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1960288
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65541}
This adds a respective test, even though Liftoff frames currently do not
show scope information. The output will change with future CLs.
R=jkummerow@chromium.org
Bug: v8:10019
Change-Id: I015863768c02678e97404f2edb538e3b4320d134
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1975754
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65540}
This uses the Smi-fast-path that was added for i32 values also for f32
and f64. The code was generalized to do the smi-to-wasm-type and
float64-to-wasm-type conversion dependent on the type, and keep the rest
generic.
R=jkummerow@chromium.org
Bug: v8:10070
Change-Id: I42b021dcc99755883cf003579c730675e6d93ff6
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1965577
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65539}
This allows the definition of classes with several arrays and ports
SmallOrderedHashTable subclasses to Torque as an example, including
the existing CSA allocation functions for them.
Overview of changes:
- Introduce ResidueClass to encapsulate the modulo-arithmetic
necessary to do alignment checks.
- Add MachineOperatorReducer to the CSA pipeline to address now
missing CSA ad-hoc constant folding that got blocked by a
temporary phi.
- Allow assignments to references to structs. This is needed to
initialize the data_table part of SmallOrderedHashMap.
- Make the NumberLiteralExpression AST-node store a double instead
of a string. This is necessary to detect arrays with constant size
used for padding.
- Turn offsets into base::Optional<size_t> to ensure we don't use
an invalid or statically unknown offset.
- Remove CreateFieldReferenceInstruction since it doesn't work for
complex offset computations and the logic can be expressed better
in ImplementationVisitor.
- Validate alignment of structs embedded in classes.
Bug: v8:10004 v8:7793
Change-Id: Ifa414b42278e572a0c577bf9da3d37f80771a258
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1958011
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Nico Hartmann <nicohartmann@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65538}
This is a reland of 53308bf7c0
Original change's description:
> [csa] use JSGraph to create constants in CodeAssembler
>
> Now that CodeAssembler uses optimizing TurboFan passes, creating
> constants without using the caching implemented in JSGraph leads to
> problems, since value numbering only works properly if all constants
> in the graph were introduced through the cache.
> To mitigate this, this CL creates the JSGraph earlier so that
> CodeAssembler can already use the same JSGraph used by later TurboFan
> optimizations.
> For other uses of RawMachineAssembler, everything stays as before.
>
> This issue is creating bot failures in
> https://chromium-review.googlesource.com/c/v8/v8/+/1958011
>
> Change-Id: Ife017876b19cb2602694279ef1da75f23e18a031
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1967329
> Reviewed-by: Maya Lekova <mslekova@chromium.org>
> Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#65477}
TBR=mslekova@chromium.org
Change-Id: I5c8218ce22470b3efa06d872176c910a4c5325a4
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1977858
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65537}
This CL limits the total number of hints of each type (constants, maps,
etc.) to 50. It also adds a limit to the number of seen functions with
the same SFI and feedback vector to 200. Octane already hits those
limits in DeltaBlue and TypeScript, but that doesn't affect the scores.
Bug: v8:7790
Change-Id: I644519955115c09bfb8ba6d98cf21087b153668d
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1975757
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Michael Stanton <mvstanton@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65535}
The instruction selector for these two operations use VisitRROFloat
which uses and operand for the second input (operand1). The codegen
assumes it is a register, which is wrong. This fixes it to be an
operand.
Change-Id: Ic27ba2c0520ba6e15963b01af7593411ec2a34e9
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1958052
Reviewed-by: Deepti Gandluri <gdeepti@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65533}
Change unregister tokens to be held weakly instead of strongly. This
enables the use case for an object to be used as its own unregister
token.
To avoid using an ephemeron table, FinalizationGroup's key_map is
changed to key off unregister tokens' identity hashes. Because hashes
may collide, a single key list may rarely contain multiple tokens. When
a FinalizationGroup WeakCell's token becomes unreachable, during GC, it
is removed from the the doubly linked key list and removed from the key
map if it had a unique key.
Bug: v8:8179
Change-Id: If88fd2ab196e3f9a287990ae345117a0abb2f04d
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1970493
Commit-Queue: Shu-yu Guo <syg@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65532}
This change includes templatization of the test helper to allow the
same function to be reused for both signed and unsigned data types.
We implement a new function RoundingAverageUnsigned in overflowing-math,
rather than in base/utils, since the addition could overflow.
SIMD scalar lowering and implementation for other backends will follow
in future patches.
Bug: v8:10039
Change-Id: I70735f7b6536f197869ef1afbccaf5649e7e8448
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1958007
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Deepti Gandluri <gdeepti@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65531}
This is the first part of switching between Liftoff and Turbofan in
debugging Wasm. In this CL, we implemented the logic to tier down/up all
functions in module.
Bug: v8:9654
Change-Id: Ia25103ca29963afa103c124ff5f159f197c2b2b0
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1970470
Commit-Queue: Z Nguyen-Huu <duongn@microsoft.com>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65529}
It would be nice if unused {RegClass}es just equal {kNoReg} on all
platforms.
This CL adds the static asserts for that, and slightly modifies the
definition of the {RegClass} enum to fulfil that. Interestingly, this
only changes the values for the one configuration that is currently
unused :)
R=zhin@chromium.org
Change-Id: Ib1b2a663183ad34822f514d6168c65450c1d7f2e
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1975753
Reviewed-by: Zhi An Ng <zhin@chromium.org>
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65527}
For functions which check if a register pair is used, change them from
is_gp_pair to is_pair, since they do not care if the pair of register is
a gp pair or a fp pair.
Bug: v8:9909
Change-Id: I262108caadd616c64019b39ebbf6972a548112df
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1974833
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65526}
For Smi, we can avoid calling the ToNumber builtin, and also the
conversion to float and back. This adds some complexity, but greatly
improves performance for parameters that fit in a Smi.
R=jkummerow@chromium.org
Bug: v8:10070
Change-Id: Ic55f17e42ce52811296293f3cd218c850af42961
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1965576
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65525}
This new API uses the code pages rather than code ranges approach.
It's supported on arm32, as well as the previous two supported
platforms, x64 and arm64.
Deprecate the old API which only works on x64 and arm64 to reduce the
maintenance overhead of keeping both. Users of the old API should
migrate to the new one as it can be used all on supported platforms.
We keep the tests for the old API by ignoring deprecation warnings so
that we don't accidentally break it while it is still in the codebase.
Design doc:
https://docs.google.com/document/d/1VGwUult5AHLRk658VetwEHMOmDDxA2eDQs9lDFMZTE0
Bug: v8:8116
Change-Id: I1de8246a48fc1b4991603501ea6087db6b43fdd9
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1969900
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65521}
After calling the ToNumber builtin, we are guaranteed to get either a
Smi or a HeapNumber. Thus we can skip the check for undefined, which
removes quite some unneeded code from wasm-to-js and js-to-wasm
wrappers.
R=ahaas@chromium.org
Bug: v8:10070
Change-Id: I0e2ad19f1575fb6979222e324fbdb71aaf85e6fd
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1962279
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65520}
The spec was normatively changed to simplify var scopes for parameter
expressions. Previously there was a per-parameter var scope in sloppy
mode so direct evals could introduce vars that did not escape the
parameter position. That semantics is complex both for the programmer
and implementation and has resulted in bugs in the past. Furthermore, it
has never been fully interoperable (with Safari in particular). The spec
was instead changed to be simpler: to have a single var scope for
sloppy evals in parameters that encloses the parameter scope and body
scope.
This simplification lets us remove expression-scope-reparenter.
Drive-by removal of stale reference to PatternRewriter.
Bug: v8:7532
Change-Id: Iade5594abe0009f7f3f6a1adad18628b17e1e779
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1962471
Reviewed-by: Toon Verwaest <verwaest@chromium.org>
Commit-Queue: Shu-yu Guo <syg@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65517}
Also port a few smaller functions.
Bug: v8:9838
Change-Id: I2245abe648378970a89331baa19af49f7f49359b
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1961942
Commit-Queue: Joshua Litt <joshualitt@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65510}
This introduces a new constant kNeedS128RegPair, which is set for
architecture where a pair of FP registers (DoubleRegister) is mapped to
single Simd128Register (ARM).
In this case, a new RegClass, kFpRegPair, is defined to represent
LiftoffRegister of this type. kFpRegPair will be kNoReg on all other
architectures.
We add 1 more bit to the encoding on ARM, so now the top bit is set for
kFpRegPair, the second most top bit is set for kGpRegPair. When the top
bit is set, we encode the FP register code into the bottom bits of
code_. Note that this is directly encoded, i.e. not added to kMaxGpRegCode,
so we can save an add/subtract when converting to/from DoubleRegister.
We only need to store 1 FP register, the low register, since the high
other register is implicity the next register. Note that the stored
register is *always* an even-numbered register.
Bug: v8:9909
Change-Id: I78d603c9938c3d0add9bd3ca77ddebbfa7abbc05
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1973276
Commit-Queue: Zhi An Ng <zhin@chromium.org>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65508}
In setting breakpoint in wasm, we can find wasm script from location but
in removing a breakpoint, only breakpoint id is provided. For wasm, we
have a list of all BreakPointInfo objects attached to the Script. From
breakpoint id, we iterates all scripts to find the targeted breakpoint
and remove it.
Bug: chromium:837572
Change-Id: Ia5d0fb7d804fb98270b2103232bc10eb5d4f93a3
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1959749
Commit-Queue: Z Nguyen-Huu <duongn@microsoft.com>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Reviewed-by: Simon Zünd <szuend@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65505}
This CL implements enums in Torque in three steps:
1.) It implements necessary changes to Torque's type system. In
particular, the constraints on constexpr types are relaxed such that
constexpr types can exist without a corresponding non-constexpr
version. Furthermore, constexpr and their non-constexpr counterpart
need not be of the same kind of type. This allows an AbstractType to
have a UnionType as its non-constexpr counterpart.
2.) The enum feature itself is realized as a pure desugaring in the
parser, where all required types, constants and macro specializations
(like FromConstexpr<>) are generated from a simple enum declaration,
such that enum entries are not just constants, but are namespace
scoped and have distinct types so that they can be used within
typeswitch constructs.
3.) Almost all of the existing constants defined in torque
(.tq files) are ported to new enum definitions.
Bug: v8:10053
Change-Id: I72426d3b1434f301fd690847e15603de0dc1021b
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1964392
Commit-Queue: Nico Hartmann <nicohartmann@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65503}
This is similar to the change made to IsImmLSPair in
9f7ae50aa8.
Change-Id: I17a7cc95661542efb5711df0639cc11ac7926702
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1971950
Commit-Queue: Joey Gouly <joey.gouly@arm.com>
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65501}
This is a reland of c509bb8c55
Original change's description:
> Cache native modules in the wasm engine by their wire bytes. This is to
> prepare for sharing {Script} objects between multiple {WasmModuleObject}
> created from the same bytes. This also saves unnecessary compilation
> time and memory.
>
> R=clemensb@chromium.org
>
> Bug: v8:6847
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1916603
> Commit-Queue: Thibaud Michaud <thibaudm@chromium.org>
> Reviewed-by: Clemens Backes <clemensb@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#65296}
R=clemensb@chromium.org
Bug: v8:6847
Change-Id: I8839c9ec96dc4141cf3c30916a62ccf86f5463ff
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1960287
Commit-Queue: Thibaud Michaud <thibaudm@chromium.org>
Reviewed-by: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65500}
This reverts commit 5377e72c0e.
Reason for revert: Looks like the relevant graphs didn't recover after
this revert, which suggests that the regression was an unrelated
secondary effect. Relanding the original change since the revert did
cause some microbenchmark regressions.
Original change's description:
> Revert "[ic] Load name/context lazily in LdaNamedProperty"
>
> This reverts commit 347092ac2b.
>
> Not a clean revert, since other changes got baked on top, but rather
> a manual removal of LoadLazyICParameters.
>
> Reason for revert: Seems to actually regress bindings perf tests (see
> bugs and https://chromeperf.appspot.com/group_report?rev=62539), doesn't
> seem to improve performance elsewhere, and increases complexity.
>
> Original change's description:
> > [ic] Load name/context lazily in LdaNamedProperty
> >
> > Introduces LazyLoadICParameters which allow a LazyNode for context and
> > name. These aren't used on the fast path, so we want to avoid reading
> > them for both performance and register pressure reasons.
> >
> > Change-Id: Ifb637cf4782ce984feee9af503998e7539beb823
> > Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1686665
> > Commit-Queue: Leszek Swirski <leszeks@chromium.org>
> > Reviewed-by: Toon Verwaest <verwaest@chromium.org>
> > Cr-Commit-Position: refs/heads/master@{#62539}
>
> # Not skipping CQ checks because original CL landed > 1 day ago.
>
> Bug: chromium:981797
> Bug: chromium:982630
> Change-Id: I88af764d17afb76d6e64b95a3d1e4aaa1c6c8978
> Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1934327
> Auto-Submit: Leszek Swirski <leszeks@chromium.org>
> Reviewed-by: Toon Verwaest <verwaest@chromium.org>
> Commit-Queue: Toon Verwaest <verwaest@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#65205}
TBR=leszeks@chromium.org,verwaest@chromium.org
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: chromium:981797, chromium:982630, v8:10059
Change-Id: I13754de06c83439e03e22cfaa7a14ce454076db9
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1973730
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65499}
For supporting use of dictionaries during GC, such as in the JS WeakRef
implementation.
Bug: v8:8179
Change-Id: Ide3f5c45d2602f13a1bcb1968b36f08881067090
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1972427
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Commit-Queue: Shu-yu Guo <syg@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65498}
