Rather check expressions used as patterns directly. Check parentheses by
tagging parenthesized expressions as parenthesized.
This allows us to drop UnexpectedPatternToken and makes it clear why a specific
token is unexpected (because it's invalid in a binding pattern).
This also more uniformly restores messages like "Invalid destructuring
assignment target".
Change-Id: Idd98e9116c85de4c2304cf1fef1baa097b67149d
Reviewed-on: https://chromium-review.googlesource.com/c/1349572
Commit-Queue: Toon Verwaest <verwaest@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57792}
Instead, simply track it as a valid binding pattern. To do this in the case of
parenthesized formals, we delay throwing the binding pattern error for
parenthesized (and async "calls") until we know it's not an arrow function head
by itself.
This guarantees that if an arrow head is a valid binding pattern, it's either a
valid parenthesized head or a valid identifier, or invalid pattern ("array" or
"object" literal style). We can detect the latter case by checking that the
current token is not a RPAREN and the expression isn't an identifier.
(Alternatively we could check that the curren token is RBRACE or RBRACK...)
Bug: chromium:907575
Change-Id: Ie40cc3235d3188f2620b6c089a0f49d93604dda6
Reviewed-on: https://chromium-review.googlesource.com/c/1348078
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Commit-Queue: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57743}
Load elimination is running together with to dead code elimination, the
latter of which might eliminate allocations (in particular FinishRegion
nodes). These are treated as alias nodes by load elimination, and load
elimination does not immediatelly learn that a node has been disconnected.
This causes load elimination to access the inputs of dead code eliminated
nodes while resolving renames, which causes nullptr dereferences.
This CL modifies load elimination to not resolve to a nullptr alias but
simply stop before that.
Change-Id: If4cef061c7c0e25f353727c9e27f790439b0beb5
Bug: chromium:906406
Reviewed-on: https://chromium-review.googlesource.com/c/1346491
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57688}
Previously the simplified operation `Number.min(x,y)` would lower to
`Select(Float64LessThan(x, y), x, y)` which would yield `y` when both
`x` and `y` are zeros, specifically when `x` was -0 and `y` was +0.
For `NumberMin` we need to use `Float64LessThanOrEqual` since we
generally allow -0 on the left hand side (in SimplifiedLowering).
Bug: chromium:906870
Change-Id: I25ae8fb19608b77c90ed130e69d9d9fa93fcea9d
Reviewed-on: https://chromium-review.googlesource.com/c/1342920
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57633}
With just five cache registers, Liftoff can run out of memory on a
64bit shift. This CL solves this by using a parallel register move and
pinning less registers.
R=ahaas@chromium.org
Bug: chromium:894307
Change-Id: I91ed0fee00ceb452841e5d1bb10905be6702dcce
Reviewed-on: https://chromium-review.googlesource.com/c/1337580
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57552}
When one of the inputs to NumberMin or NumberMax is NaN we need to
return NaN, ignoring whatever else was passed. Specifically we cannot
lower `NumberMin(x,y)` to `(x < y) ? x : y` if `x` can be NaN. So
limit this optimization to only perform the above lowering if we
know that `x` is an OrderedNumber and `y` is a PlainNumber (or if
the difference between zeros doesn't matter, an OrderedNumber as
well).
Bug: chromium:905457
Change-Id: If05f19255e14789ab0e277e072469c40e161b85b
Reviewed-on: https://chromium-review.googlesource.com/c/1337576
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57535}
An oversight in my previous change (3b64764b1d) could
cause a CHECK failure.
Bug: chromium:904707
Change-Id: Ie5f1c500bddc00741b889f78ae9ecd9af581ba5c
Reviewed-on: https://chromium-review.googlesource.com/c/1333409
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57459}
It's not sufficient to check the NoElements protector because that
doesn't guard against the array having a custom prototype.
Bug: v8:8449
Change-Id: I843815466a1e4ae197a2b76eec62d04cdc2d619d
Reviewed-on: https://chromium-review.googlesource.com/c/1332232
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57457}
This matches the pre-torque behavior when the receiver's length
was greater than the max array length.
Bug: chromium:902672
Change-Id: Icf8ae3a1a4acc0680ce1b709f5b3372892337203
Reviewed-on: https://chromium-review.googlesource.com/c/1330921
Commit-Queue: Peter Wong <peter.wm.wong@gmail.com>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57456}
Now that we have no more do-expressions, we don't need to reparent variables
and declarations anymore. However, it's still possible that temporaries were
implicitly allocated. We still need to move those.
Bug: chromium:904255
Change-Id: Ia8a90eb822b9db123ffb0bad58e4b720c1452d9f
Reviewed-on: https://chromium-review.googlesource.com/c/1329685
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Commit-Queue: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57424}
Split the Install methods into PrepareInstall and Install, such that
all heap mutations (besides the actual installation) are done in
PrepareInstall and only the actual installation in Install. This
ensures that the code object in question doesn't get deoptimized while
we're still installing its dependencies.
Bug: chromium:903697
Change-Id: I4da97d89d0707fa3c00c97c092af0d0faa7a4946
Reviewed-on: https://chromium-review.googlesource.com/c/1329162
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57419}
This test was adapted from a repro, and thus it's rather complex.
It takes over seven minutes to run on the arm64 sim debug bot,
and nearly five minutes on arm.
Given that it was originally accompanied by a very targeted fix in
Crankshaft, it strikes me that this probably isn't worth our CPU
time to continue running.
Bug: v8:7783, chromium:85177
Change-Id: Ibe85cc254aa754365404b5fbbf80bcb1f5a09c68
Reviewed-on: https://chromium-review.googlesource.com/c/1327188
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57408}
The flag is on by default, so we don't need to specify it. More
importantly, the tests are expected to work for any value of that flag.
So don't force the flag but use whatever the test variant chooses.
Note that in streaming-compile.js, the flag was accidentally specified
as '-async-compilation'. I also removed that one.
R=ahaas@chromium.org
Change-Id: Ifad31160d266dda38cdd9dd1d73dad69bd2c2f2c
Reviewed-on: https://chromium-review.googlesource.com/c/1325961
Reviewed-by: Andreas Haas <ahaas@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57406}
Previously we'd check `x` for -0 by testing `(1.0 / x) == -Infinity`,
but this will yield the wrong results when `x` is a subnormal, i.e.
really close to 0.
In CSA we already perform bit checks to test for -0, so teach TurboFan
to do the same for comparisons to -0 (via `Object.is`). We introduce a
new NumberIsMinusZero simplified operator to handle the case where
SimplifiedLowering already knows that the input is a number.
Bug: chromium:903043, v8:6882
Change-Id: I0cb7c568029b461a92fc183104d5f359b4bfe7f4
Reviewed-on: https://chromium-review.googlesource.com/c/1328802
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57382}
Since we use a ScopedPtrList to track cover grammar expressions we don't know
the position of the commas anymore. The position of the commas was used to
demark the initializer, which is needed to figure out whether we need hole
checks for variable references. (Typically only references within the
initializer need hole checks for the initialized variable.) Since we didn't
have the comma position, we simply used the position of the first expression as
the position of any subsequent comma, which would make it seem as if the
initializer body wasn't in the initializer. Now instead we simply use the
position of the subsequent parameter as the end of the initializer, which is
close enough.
Bug: chromium:902810
Change-Id: I8d2bc7a2dc9f59db16ce56ccef01e263a18a3b7a
Reviewed-on: https://chromium-review.googlesource.com/c/1326022
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Commit-Queue: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57357}
The Ignition statement list visitor will skip the rest of the
statements in the list if it hits a jump statement (like a return
or break), as the rest of the code in the list can be considered
dead.
return;
dead_call(); // skipped
However, since this is at an AST node level, it does not take into
account condition shortcutting:
if(2.2) return;
dead_call(); // not skipped
There is also a second dead code elimination in Ignition compilation, at
the bytecode array writer level, where a bytecodes are not emitted if an
"exit" bytecode (Return, Jump, or a few others) has been written, until
the next basic block starts (i.e. a Bind).
This can cause an issue with statements that resurrect the bytecode
array writer part-way through their visit. An example is try-catch
statements, which save the context to a register, and then Bind to start
the try region.
For the case:
if (2.2) return;
try { // try statement not skipped
...
}
the bytecode writer is called with
OutputReturn() // exit bytecode seen
OutputMove(<context>, r1) // not emitted
Bind(&try_begin) // starts new basic block
// try body
So, the try is emitted, but without saving the context to a register.
This means that the liveness analysis sees the read of that register
(as the output liveness of throwing bytecodes), but does not have a
write to the register, which means that the liveness escapes.
This patch fixes this by using the bytecode array writer dead-code
elimination (i.e. "exit bytecode seen") to inform the statement list
visitor, so that in this example the try statement is not visited at
all.
Bug: chromium:902395
Change-Id: Ieb8e46a4318df3edbac0ae17235e0ce8fba12ee3
Reviewed-on: https://chromium-review.googlesource.com/c/1322951
Reviewed-by: Mythri Alle <mythria@chromium.org>
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57350}
Installation of the PrototypePropertyDependency, as well as GC, can
invalidate dependencies.
Bug: chromium:902552
Change-Id: Iabcce026c7475c722d19ac0b80758b22d9fbcfda
Reviewed-on: https://chromium-review.googlesource.com/c/1322450
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57343}
regress-336820 is testing that joining a very sparse
array to create a too-big string results in a RangeError,
rather than a crash. Reducing the largest index by
two orders of magnitude speeds this up (on x64 debug)
by 8x (from 8 seconds down to 1). Given that this test
takes nearly 9 minutes on arm64 sim debug, I hope to
see big ones there too.
Bug: v8:7783, chromium:336820
Change-Id: I74c22cf451a892eb039efc7f1259152921bf8530
Reviewed-on: https://chromium-review.googlesource.com/c/1323915
Reviewed-by: Jakob Kummerow <jkummerow@chromium.org>
Commit-Queue: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57335}
The array length is modifiable by user code that is called as a
side-effect during the sorting algorithm. We thus cannot base any
guarantees on the current length, but must reference the initially-read
array length instead.
Note that even though the algorithm may read and write from beyond
the current array length value, this adheres to the spec, which only
requires accesses to be within the original array dimensions (i.e.: 0
<= i < original array length).
Bug: chromium:901633
Change-Id: Id765e80d4231ff6f2a73e040ec94c2b07f8c5b0f
Reviewed-on: https://chromium-review.googlesource.com/c/1317814
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Daniel Clifford <danno@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57279}
- Avoid allocating AstRawString in the preparser
- Use fast LiteralEquals to compare the directive.
Bug: chromium:901250
Change-Id: I178aca812f6c0ffa28d7f48b707316a5a99a2ac0
Reviewed-on: https://chromium-review.googlesource.com/c/1314570
Commit-Queue: Toon Verwaest <verwaest@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57217}
Restructure the code a little, and change how we detect sloppy block function
redeclaration so we don't dereference a possibly nullptr function.
Bug: chromium:900786
Change-Id: Ief124fe767603ca36f4dc8865c4aeb3e0635b4cf
Reviewed-on: https://chromium-review.googlesource.com/c/1314331
Reviewed-by: Camillo Bruni <cbruni@chromium.org>
Commit-Queue: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57206}
The fast-path in the `ArrayPrototypeLastIndexOf` torque implementation
didn't check that the `fromIndex` is within the bounds of the JSArray
_AFTER_ the call to ToInteger, which can have arbitrary side-effects,
i.e. it can change the length of the array.
R=yangguo@chromium.org
Bug: chromium:898785
Change-Id: I7ef84143ec8c33148f6e9d451bd52769d5074fb4
Reviewed-on: https://chromium-review.googlesource.com/c/1314329
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57204}
CanCover is not transitive. The counter example are Nodes A,B,C such
that CanCover(A, B) and CanCover(B,C) and B is pure. In this case the
effect level of A and B might differ.
This CL adds a missing CanCover check to a case of shift reduction where
we assumed transitivity.
Change-Id: I9f368ffa6907d2af21bbc87b3e6570d0d422e125
Bug: v8:8384
Reviewed-on: https://chromium-review.googlesource.com/c/1307419
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57157}
For memory limit checks, we should use the minimum of the
--wasm-max-mem-pages flag and kV8MaxWasmMemoryPages. The former is a
limit set by the user, the latter is the maximum we can handle
internally.
R=titzer@chromium.org
Bug: chromium:898677
Change-Id: I3c549f4e90dd016b5d07475d9353f30134f76dcc
Reviewed-on: https://chromium-review.googlesource.com/c/1305274
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Reviewed-by: Ben Titzer <titzer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57127}
The "grow_memory" opcode was renamed to "memory.grow", and the spec
repo was updated to use kExprMemoryGrow internally instead of
kExprGrowMemory (https://github.com/WebAssembly/spec/pull/720).
This CL does the same change for v8.
Drive-by: Rename "current_size" to "memory.size", and a minor cleanup
in wasm-graph-builder.js to bring it in line with the version in the
js-api tests in the spec repo.
R=titzer@chromium.org
Change-Id: If525dba898b2c248890a616d3392c22b45f698ef
Reviewed-on: https://chromium-review.googlesource.com/c/1302057
Reviewed-by: Ben Titzer <titzer@chromium.org>
Commit-Queue: Clemens Hammacher <clemensh@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57089}
This fixes the fall-back case when parsing a multiplicative expression
where the lookahead found a '-' token followed by an unsigned token, but
no '*' token is following. We cannot rewind both tokens, but still need
to make sure that a full multiplicative expression is parsed.
R=clemensh@chromium.org
TEST=mjsunit/regress/regress-8377
BUG=v8:8377
Change-Id: I20ce6267445b32bdaf03f41f11d9ef4be66cb636
Reviewed-on: https://chromium-review.googlesource.com/c/1304317
Reviewed-by: Clemens Hammacher <clemensh@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57084}
This allows very large arrays being joined to incrementally,
on-demand allocate the internal buffer. Previously, join
would allocate the buffer upfront and all at once. Large,
sparse arrays will use less memory.
Bug: chromium:897404
Change-Id: Id914b14a7c55a62834f63ad602bdb45363249075
Reviewed-on: https://chromium-review.googlesource.com/c/1303538
Commit-Queue: Peter Wong <peter.wm.wong@gmail.com>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57075}
While not strictly necessary, this is consistent with how
SlowFlagGetter behaves. It adds an additional shift operation (which
we could fold into the smi untagging if needed).
Drive-by: Typify flag accessors.
Bug: chromium:899464
Change-Id: Ib154d626e522ed723e2c19b1ab7f68560ac414bc
Reviewed-on: https://chromium-review.googlesource.com/c/1304315
Reviewed-by: Peter Wong <peter.wm.wong@gmail.com>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57063}
When replacing a LoadElement with variable index with its known fields,
only do it if the types match, otherwise we end up with a graph that
representation selection cannot handle. That can only happen in dead
code, but TurboFan would nevertheless crash in representation selection.
Bug: chromium:893982, chromium:899524, v8:5267, v8:6200
Change-Id: I01e645d5e01bffb911d216d37d923792d9d0beab
Reviewed-on: https://chromium-review.googlesource.com/c/1303721
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57059}
