This fixes several problems with instanceof and constant field tracking
in the compiler:
- properly bailout on numbers and non-functions at @@hasInstance.
- deopt on changes of @@hasInstance property.
Bug: v8:8361
Change-Id: I4a1cf9e29d72076f2d37a7c703f18cb2fb8f4040
Reviewed-on: https://chromium-review.googlesource.com/c/1322449
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57532}
As opposed to the register.
For subtle reasons, this fixes a deoptimizer bug with handling return
values in lazy deopt. Since the return values can now only overwrite
the accumulator, there is no danger of overwriting a captured object
that might be later used (since there is no "later").
Bug: chromium:902608
Change-Id: I3a7a10bb1c7a6f4303a01d60f80680afcb7bc942
Reviewed-on: https://chromium-review.googlesource.com/c/1325901
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57349}
This test currently takes nearly 10 minutes on the arm64 debug builder.
Bug: v8:7783
Change-Id: I500fc026b01873e666f32062d790eca3f34455b9
Reviewed-on: https://chromium-review.googlesource.com/c/1318495
Commit-Queue: Adam Klein <adamk@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57298}
BinaryNumberOpTyper was not monotonic: if one input changes
from Number to Numeric, while the other input stays BigInt,
the result would change from Number to BigInt.
We have some fuzzing tests for monotonicity but unfortunately
they never generated the inputs required for triggering this bug.
We'll look into improving our tests.
Bug: v8:8380
Change-Id: I7320d9ae4b89ad8798bf9e97cc272edba2162a77
Reviewed-on: https://chromium-review.googlesource.com/c/1307418
Commit-Queue: Hai Dang <dhai@google.com>
Reviewed-by: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57125}
This adds appropriate LoopExit nodes for the JSCallReducer lowerings of
the following higher order Array builtins:
- Array.prototype.every()
- Array.prototype.find()
- Array.prototype.findIndex()
- Array.prototype.some()
Loop peeling allows TurboFan to make loop invariant operations in the
callback passed to the higher order builtin fully redundant, and thus
completely eliminate the loop invariant code from the subsequent loop
iterations. This can have a huge performance impact, depending on what
kind of code runs inside of the callback. For example, on the micro-
benchmarks outlined in http://crbug.com/v8/8273 we go from
forLoop: 364 ms.
every: 443 ms.
some: 432 ms.
find: 522 ms.
findIndex: 437 ms.
to
forLoop: 369 ms.
every: 354 ms.
some: 348 ms.
find: 419 ms.
findIndex: 360 ms.
which is 20% improvement, and essentially brings the Array builtins (the
appropriate ones Array#some() and Array#every() in this case) on par
with the hand-written `for`-loop.
Bug: v8:1956, v8:8273
Change-Id: I9d32736e5402807b4ac79cd5ad15ceacd1945681
Reviewed-on: https://chromium-review.googlesource.com/c/1305935
Reviewed-by: Daniel Clifford <danno@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57110}
This introduces Word64 support for the CheckBounds operator, which now
lowers to either CheckedUint32Bounds or CheckedUint64Bounds after the
representation selection. The right hand side of CheckBounds can now
be any positive safe integer on 64-bit architectures, whereas it remains
Unsigned31 for 32-bit architectures. We only use the extended Word64
support when the right hand side is outside the Unsigned31 range, so
for everything except DataViews this means that the performance should
remain the same. The typing rule for the CheckBounds operator was
updated to reflect this new behavior.
The CheckBounds with a right hand side outside the Unsigned31 range will
pass a new Signed64 feedback kind, which is handled with newly introduced
CheckedFloat64ToInt64 and CheckedTaggedToInt64 operators in representation
selection.
The JSCallReducer lowering for DataView getType()/setType() methods was
updated to not smi-check the [[ByteLength]] and [[ByteOffset]] anymore,
but instead just use the raw uintptr_t values and operate on any value
(for 64-bit architectures these fields can hold any positive safe
integer, for 32-bit architectures it's limited to Unsigned31 range as
before). This means that V8 can now handle huge DataViews fully, without
falling off a performance cliff.
This refactoring even gave us some performance improvements, on a simple
micro-benchmark just exercising different DataView accesses we go from
testDataViewGetUint8: 796 ms.
testDataViewGetUint16: 997 ms.
testDataViewGetInt32: 994 ms.
testDataViewGetFloat64: 997 ms.
to
testDataViewGetUint8: 895 ms.
testDataViewGetUint16: 889 ms.
testDataViewGetInt32: 888 ms.
testDataViewGetFloat64: 890 ms.
meaning we lost around 10% on the single byte case, but gained 10% across
the board for all the other element sizes.
Design-Document: http://bit.ly/turbofan-word64
Bug: chromium:225811, v8:4153, v8:7881, v8:8171, v8:8383
Change-Id: Ic9d1bf152e47802c04dcfd679372e5c85e4abc83
Reviewed-on: https://chromium-review.googlesource.com/c/1303732
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57095}
For NumberMin and NumberMax we don't need to go to Float64 when the
inputs are known to be in SafeInteger range, instead we can go to
Word64 on 64-bit architectures. This is preliminary work for the
huge DataView support, since we'll utilize NumberMax in that case
to clamp the limit for the bounds check.
Bug: v8:8178, v8:8383
Change-Id: I414114229c5c86b92749d30d645cedc641541ae4
Reviewed-on: https://chromium-review.googlesource.com/c/1304535
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#57090}
This changes the ReceiverOrOddball feedback on JSStrictEqual to
ReceiverOrNullOrUndefined feedback, which can also safely be
consumed by JSEqual (we cannot generally accept any oddball here
since booleans trigger implicit conversions, unfortunately).
Thus we replace the previously introduced CheckReceiverOrOddball
with CheckReceiverOrNullOrUndefined, and drop CheckOddball, since
we will no longer collect Oddball feedback separately.
TurboFan will then turn a JSEqual[ReceiverOrNullOrUndefined] into
a sequence like this:
```
left = CheckReceiverOrNullOrUndefined(left);
right = CheckReceiverOrNullOrUndefined(right);
result = if ObjectIsUndetectable(left) then
ObjectIsUndetectable(right)
else
ReferenceEqual(left, right);
```
This significantly improves the peak performance of abstract equality
with Receiver, Null or Undefined inputs. On the test case outlined in
http://crbug.com/v8/8356 we go from
naive: 2946 ms.
tenary: 2134 ms.
to
naive: 2230 ms.
tenary: 2250 ms.
which corresponds to a 25% improvement on the abstract equality case.
For regular code this will probably yield more performance, since we
get rid of the JSEqual operator, which might have arbitrary side
effects and thus blocks all kinds of TurboFan optimizations. The
JSStrictEqual case is slightly slower now, since it has to rule out
booleans as well (even though that's not strictly necessary, but
consistency is key here).
This way developers can safely use `a == b` instead of doing a dance
like `a == null ? b == null : a === b` (which is what dart2js does
right now) when both `a` and `b` are known to be Receiver, Null or
Undefined. The abstract equality is not only faster to parse than
the tenary, but also generates a shorter bytecode sequence. In the
test case referenced in http://crbug.com/v8/8356 the bytecode for
`naive` is
```
StackCheck
Ldar a1
TestEqual a0, [0]
JumpIfFalse [5]
LdaSmi [1]
Return
LdaSmi [2]
Return
```
which is 14 bytes, whereas the `tenary` function generates
```
StackCheck
Ldar a0
TestUndetectable
JumpIfFalse [7]
Ldar a1
TestUndetectable
Jump [7]
Ldar a1
TestEqualStrict a0, [0]
JumpIfToBooleanFalse [5]
LdaSmi [1]
Return
LdaSmi [2]
Return
```
which is 24 bytes. So the `naive` version is 40% smaller and requires
fewer bytecode dispatches.
Bug: chromium:898455, v8:8356
Change-Id: If3961b2518b4438700706b3bd6071d546305e233
Reviewed-on: https://chromium-review.googlesource.com/c/1297315
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56948}
This CL introduces proper Oddball and ReceiverOrOddball states for the
CompareOperationFeedback, and updates the StrictEqual IC to collect this
feedback as well. Previously it would not collect Oddball feedback, not
even in the sense of NumberOrOddball, since that's not usable for the
SpeculativeNumberEqual.
The new feedback is handled via newly introduced CheckReceiverOrOddball
and CheckOddball operators in TurboFan, introduced by JSTypedLowering.
Just like with the Receiver feedback, it's enough to check one side and
do a ReferenceEqual afterwards, since strict equal can only yield true
if both sides refer to the same instance.
This improves the benchmark mentioned in http://crbug.com/v8/8356 from
naive: 2950 ms.
tenary: 2456 ms.
to around
naive: 2996 ms.
tenary: 2192 ms.
which corresponds to a roughly 10% improvement in the case for the
tenary pattern, which is currently used by dart2js. In real world
scenarios this will probably help even more, since TurboFan is able
to optimize across the strict equality, i.e. there's no longer a stub
call forcibly spilling all registers that are live across the call.
This new feedback will be used as a basis for the JSEqual support for
ReceiverOrOddball, which will allow dart2js switching to the shorter
a==b form, at the same peak performance.
Bug: v8:8356
Change-Id: Iafbf5d64fcc9312f9e575b54c32c631ce9b572b2
Reviewed-on: https://chromium-review.googlesource.com/c/1297309
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56925}
When InferReceiverMaps doesn't provide us with reliable maps for the
resolution, we can still utilize the information if all the maps that
are found are stable - aka leaf - maps. But in that case we need to
make sure that we add proper dependencies on the stability of these
maps.
Bug: v8:7253
Change-Id: I6f5825583acc3f2575e83a244d55609ac64d04d3
Reviewed-on: https://chromium-review.googlesource.com/c/1288633
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56789}
The NumberMultiply typing rule gave up in the presence of NaN inputs,
but we can still infer useful ranges here and just union the result
of that with the NaN propagation (similar for MinusZero propagation).
This way we can still makes sense of these ranges at the uses.
Bug: v8:8015
Change-Id: Ic4c5e8edc6c68776ff3baca9628ad7de0f8e2a92
Reviewed-on: https://chromium-review.googlesource.com/c/1261143
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56539}
This change introduces new intrinsics used to desugar async functions
in the Parser and the BytecodeGenerator, namely we introduce a new
%_AsyncFunctionEnter intrinsic that constructs the generator object
for the async function (and in the future will also create the outer
promise for the async function). This generator object is internal
and never escapes to user code, plus since async functions don't have
a "prototype" property, we can just a single map here instead of tracking
the prototype/initial_map on every async function. This saves one word
per async function plus one initial_map per async function that was
invoked at least once.
We also introduce two new intrinsics %_AsyncFunctionReject, which
rejects the outer promise with the caught exception, and another
%_AsyncFunctionResolve, which resolves the outer promise with the
right hand side of the `return` statement. These functions also perform
the DevTools part of the job (aka popping from the promise stack and
sending the debug event). This allows us to get rid of the implicit
try-finally from async functions completely; because the finally
block only called to the %AsyncFunctionPromiseRelease builtin, which
was used to inform DevTools.
In essence we now turn an async function like
```js
async function f(x) { return await bar(x); }
```
into something like this (in Parser and BytecodeGenerator respectively):
```
function f(x) {
.generator_object = %_AsyncFunctionEnter(.closure, this);
.promise = %AsyncFunctionCreatePromise();
try {
.tmp = await bar(x);
return %_AsyncFunctionResolve(.promise, .tmp);
} catch (e) {
return %_AsyncFunctionReject(.promise, e);
}
}
```
Overall the bytecode for async functions gets significantly shorter
already (and will get even shorter once we put the outer promise into
the async function generator object). For example the bytecode for a
simple async function
```js
async function f(x) { return await x; }
```
goes from 175 bytes to 110 bytes (a ~38% reduction in size), which
is in particular due to the simplification around the try-finally
removal.
Overall this seems to improve the doxbee-async-es2017-native test by
around 2-3%. On the test case mentioned in v8:8276 we go from
1124ms to 441ms, which corresponds to a 60% reduction in total
execution time!
Tbr: marja@chromium.org
Bug: v8:7253, v8:7522, v8:8276
Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;master.tryserver.blink:linux_trusty_blink_rel
Change-Id: Id29dc92de7490b387ff697860c900cee44c9a7a4
Reviewed-on: https://chromium-review.googlesource.com/c/1269041
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Maya Lekova <mslekova@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56502}
This adds support to the escape analysis to allow scalar replacement
of (small) FixedArrays with element accesses where the index is not a
compile time constant. This happens quite often when inlining functions
that operate on variable number of arguments. For example consider this
little piece of code:
```js
function sum(...args) {
let s = 0;
for (let i = 0; i < args.length; ++i) s += args[i];
return s;
}
function sum2(x, y) {
return sum(x, y);
}
```
This example is made up, of course, but it shows the problem. Let's
assume that TurboFan inlines the function `sum` into it's call site
at `sum2`. Now it has to materialize the `args` array with the two
values `x` and `y`, and iterate through these `args` to sum them up.
The escape analysis pass figures out that `args` doesn't escape (aka
doesn't outlive) the optimized code for `sum2` now, but TurboFan still
needs to materialize the elements backing store for `args` since there's
a `LoadElement(args.elements,i)` in the graph now, and `i` is not a
compile time constant.
However the escape analysis has more information than just that. In
particular the escape analysis knows exactly how many elements a non
escaping object has, based on the fact that the allocation must be
local to the function and that we only track objects with known size.
So in the case above when we get to `args[i]` in the escape analysis
the relevant part of the graph looks something like this:
```
elements = LoadField[elements](args)
length = LoadField[length](args)
index = CheckBounds(i, length)
value = LoadElement(elements, index)
```
In particular the contract here is that `LoadElement(elements,index)`
is guaranteed to have an `index` that is within the valid bounds for
the `elements` (there must be a preceeding `CheckBounds` or some other
guard in optimized code before it). And since `elements` is allocated
inside of the optimized code object, the escape analysis also knows
that `elements` has exactly two elements inside (namely the values of
`x` and `y`). So we can use that information and replace the access
with a `Select(index===0,x,y)` operation instead, which allows us to
scalar replace the `elements`, since there's no escaping use anymore
in the graph.
We do this for the case that the number of elements is 2, as described
above, but also for the case where elements length is one. In case
of 0, we know that the `LoadElement` must be in dead code, but we can't
just mark it for deletion from the graph (to make sure it doesn't block
scalar replacement of non-dead code), so we don't handle this for now.
And for one element it's even easier, since the `LoadElement` has to
yield exactly said element.
We could generalize this to handle arbitrary lengths, but since there's
a cost to arbitrary decision trees here, it's unclear when this is still
beneficial. Another possible solution for length > 2 would be to have
special stack allocation for these backing stores and do variable index
accesses to these stack areas. But that's way beyond the scope of this
isolated change.
This change shows a ~2% improvement on the EarleyBoyer benchmark in
JetStream, since it benefits a lot from not having to materialize these
small arguments backing stores.
Drive-by-fix: Fix JSCreateLowering to properly initialize "elements"
with StoreElement instead of StoreField (which violates the invariant
in TurboFan that fields and elements never alias).
Bug: v8:5267, v8:6200
Change-Id: Idd464a15a81e7c9653c48c814b406eb859841428
Reviewed-on: https://chromium-review.googlesource.com/c/1267935
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56442}
This change adds predicates to check whether a given JavaScript operator
needs the "current context" or if any surrounding context (including the
"native context") does it. For example JSAdd doesn't ever need the
current context, but actually only the native context. In the
BytecodeGraphBuilder we use this predicate to check whether a given
operator needs the current context, and if not, we just pass in the
native context.
Doing so we improve the performance on the benchmarks given in the
tracking bug significantly, and go from something around
arrayMap: 476 ms.
arrayFilter: 312 ms.
arrayEvery: 241 ms.
arraySome: 152 ms.
to
arrayMap: 377 ms.
arrayFilter: 296 ms.
arrayEvery: 191 ms.
arraySome: 91 ms.
which is an up to 40% improvement. So for idiomatic modern JavaScript
which uses higher order functions quite a lot, not just the builtins
provided by the JSVM, this is going to improve peak performance
noticably.
This also makes it possible to completely eliminate all the allocations
in the aliased sloppy arguments example
```js
function foo(a) { return arguments.length; }
```
concretely we don't allocate the function context anymore and we also
don't allocate the arguments object anymore (the JSStackCheck was the
reason why we did this in the past, because it was holding on to the
current context, which also kept the allocation for the arguments
alive).
Bug: v8:6200, v8:8060
Change-Id: I1db56d00d6b510ce6337608c0fff16af96e95eef
Design-Document: bit.ly/v8-turbofan-context-sensitive-js-operators
Reviewed-on: https://chromium-review.googlesource.com/c/1267176
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56441}
As identified in the web-tooling-benchmark, there are specific code
patterns involving array indexed property accesses and subsequent
comparisons of those indices that lead to repeated Smi checks in the
optimized code, which in turn leads to high register pressure and
generally bad register allocation. An example of this pattern is
code like this:
```js
function f(a, n) {
const i = a[n];
if (n >= 1) return i;
}
```
The `a[n]` property access introduces a CheckBounds on `n`, which
later lowers to a `CheckedTaggedToInt32[dont-check-minus-zero]`,
however the `n >= 1` comparison has collected `SignedSmall` feedback
and so it introduces a `CheckedTaggedToTaggedSigned` operation. This
second Smi check is redundant and cannot easily be combined with the
earlier tagged->int32 conversion, since that also deals with heap
numbers and even truncates -0 to 0.
So we teach the RedundancyElimination to look at the inputs of these
speculative number comparisons and if there's a leading bounds check
on either of these inputs, we change the input to the result of the
bounds check. This avoids the redundant Smi checks later and generally
allows the SimplifiedLowering to do a significantly better job on the
number comparisons. We only do this in case of SignedSmall feedback
and only for inputs that are not already known to be in UnsignedSmall
range, to avoid doing too many (unnecessary) expensive lookups during
RedundancyElimination.
All of this is safe despite the fact that CheckBounds truncates -0
to 0, since the regular number comparisons in JavaScript identify
0 and -0 (unlike Object.is()). This also adds appropriate tests,
especially for the interesting cases where -0 is used only after
the code was optimized.
Bug: v8:6936, v8:7094
Change-Id: Ie37114fb6192e941ae1a4f0bfe00e9c0a8305c07
Reviewed-on: https://chromium-review.googlesource.com/c/1246181
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56428}
This reverts commit 4fd92b252b.
Reason for revert: Significant tankage on the no-mitigations bots (bad timing on the regular bots)
Original change's description:
> [turbofan] Do not consume SignedSmall feedback in TurboFan anymore.
>
> This changes TurboFan to treat SignedSmall feedback similar to Signed32
> feedback for binary and compare operations, in order to simplify and
> unify the machinery.
>
> This is an experiment. If this turns out to tank performance, we will
> need to revisit and ideally revert this change.
>
> Bug: v8:7094
> Change-Id: I885769c2fe93d8413e59838fbe844650c848c3f1
> Reviewed-on: https://chromium-review.googlesource.com/c/1261442
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#56411}
TBR=jarin@chromium.org,bmeurer@chromium.org
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: v8:7094
Change-Id: I9fff3b40e6dc0ceb7611b55e1ca9940089470404
Reviewed-on: https://chromium-review.googlesource.com/c/1267175
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56427}
This changes TurboFan to treat SignedSmall feedback similar to Signed32
feedback for binary and compare operations, in order to simplify and
unify the machinery.
This is an experiment. If this turns out to tank performance, we will
need to revisit and ideally revert this change.
Bug: v8:7094
Change-Id: I885769c2fe93d8413e59838fbe844650c848c3f1
Reviewed-on: https://chromium-review.googlesource.com/c/1261442
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56411}
The CheckSmi in String.fromCodePoint() is unnecessary and even leads to
unnecessary deoptimizations, since the CheckBounds already does the
right thing, plus it also handles HeapNumbers (in Signed32 range) and
properly identifies zeros.
Bug: v8:8238
Change-Id: I73bf7a70c3cd718c987f112ceb928188c0534cd5
Reviewed-on: https://chromium-review.googlesource.com/c/1262675
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56395}
For NumberModulus and SpeculativeNumberModulus there's no observable
difference between 0 and -0 for the right hand side, since both of them
result in NaN (in general the sign of the right hand side is ignored
for modulus in JavaScript). For the left hand side we can just propagate
the zero identification part of the truncation, since we only care about
-0 on the left hand side if the use nodes care about -0 too.
This further improves the Kraken/audio-oscillator test from around 67ms
to 64ms.
Bug: v8:8015, v8:8178
Change-Id: I1f51d42f7df08aaa28a9b0ddd3177df6b76be98c
Reviewed-on: https://chromium-review.googlesource.com/c/1260024
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56372}
This is a follow-up cleanup to treat NumberRound like the other rounding
operations (NumberFloor, NumberCeil and NumberTrunc).
Bug: v8:8015
Change-Id: I2b2fbc7f0319497d16ccb7472595eeb68be1f51d
Reviewed-on: https://chromium-review.googlesource.com/c/1260403
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56371}
The slow-path of CheckedInt32Mod(x,y) when x is found to be negative
still had the power of two right hand side optimization, and thus would
perform a dynamic check on y. Now the same dynamic check was done for
the fast-path, and the word operations for this check were pure, leading
to weird bit materialization in TurboFan (due to sea of nodes). But
there's not really a point to be clever for the slow-path, so we just
insert the Uint32Mod operation directly here, which completely avoids
the problem.
This improves the Kraken/audio-oscillator test from around 73ms to 69ms.
Bug: v8:8069
Change-Id: Ie8ea667136c95df2bd8c5ba56ebbc6bd2442ff23
Reviewed-on: https://chromium-review.googlesource.com/c/1259063
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56370}
Following up on the earlier work regarding redundant Smi checks in
https://chromium-review.googlesource.com/c/v8/v8/+/1246181, it was
noticed that the handling of the 0 and -0 and how some operations
identify these is not really consistent, but was still rather ad-hoc.
This change tries to unify the handling a bit by making sure that all
number comparisons generally pass truncations that identify zeros, since
for the number comparisons in JavaScript there's no difference between
0 and -0. In the same spirit NumberAbs and NumberToBoolean should also
pass these truncations, since they also don't care about the differences
between 0 and -0.
Adjust NumberCeil, NumberFloor, NumberTrunc, NumberMin and NumberMax
to pass along any incoming kIdentifiesZeros truncation, since these
operations also don't really care whether the inputs can be -0 if the
use nodes don't care.
Also utilize the kIdentifiesZeros truncation for NumberModulus with
Signed32 inputs, because it's kind of common to do something like
`x % 2 === 0`, where it doesn't really matter whether `x % 2` would
eventually produce a negative zero (since that would still be considered
true for the sake of the comparison).
This also adds a whole lot of tests to ensure that not only are these
optimizations correct, but also that we do indeed perform them.
Drive-by-fix: The `NumberAbs(x)` would incorrectly lower to just `x` for
PositiveIntegerOrMinusZeroOrNaN inputs, which was obviously wrong in
case of -0. This was fixed as well, and an appropriate test was added.
The reason for the unification is that with the introduction of Word64
for CheckBounds (which is necessary to support large TypedArrays and
DataViews) we can no longer safely pass Word32 truncations for the
interesting cases, since the index might be outside the Signed32 or
Unsigned32 ranges, but we still identify 0 and -0 for the sake of the
bounds check, and so it's important that this is handled consistently
to not regress performance on TypedArrays and DataViews accesses.
Bug: v8:8015, v8:8178
Change-Id: Ia1d32f1b726754cea1e5793105d9423d84a6393a
Reviewed-on: https://chromium-review.googlesource.com/1246172
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56325}
It was shipped in Chrome 67.
Bug: v8:6791, v8:8238
Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel;luci.chromium.try:linux_chromium_rel_ng;luci.v8.try:v8_linux_noi18n_rel_ng;master.tryserver.blink:linux_trusty_blink_rel
Change-Id: I94d8f0aa18570452403a35dea270b18f155c970a
Reviewed-on: https://chromium-review.googlesource.com/1253604
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Mathias Bynens <mathias@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56310}
Currently, we call the MapRef::AsElementsKind method on an initial
map multiple times (from JSCreateLowering::ReduceJSCreateArray).
However, this does not does not play well with the heap copier/broker,
which only expectes AsElementsKind to be called on initial maps.
This CL makes sure we only call AsElementsKind once (on the initial map).
Bug: chromium:890620
Change-Id: If44421d3900abb7629ea8f789a005b8d8ebaf881
Reviewed-on: https://chromium-review.googlesource.com/1253105
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56307}
Bug: chromium:890057
Change-Id: I98bc278ebc202c3d8f6417367bd1c592e4824011
Reviewed-on: https://chromium-review.googlesource.com/1250481
Commit-Queue: Georg Neis <neis@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56279}
Properly test the abstract equality - both JSEqual and JSNotEqual - for
the case of symbols. Also add tests for the corner cases of the
JSObjectIsArray operator, which is used to implement Array.isArray()
builtin.
Bug: v8:8015
Change-Id: Ib008e85553d04527a5992a904ec77774761f872e
Reviewed-on: https://chromium-review.googlesource.com/1238237
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56121}
Improve the lowering of CheckedInt32Div and CheckedUint32Div for the
case that the right hand side is a known (positive) power of two, as
in that case it's sufficient to just check the relevant bits on the
left hand side and then shift by the appropriate amount of bits.
This is significantly faster than what TurboFan is able to generate
from the general lowering, even with all the MachineOperatorReducer
magic (it even shows as a steady ~1.5% overall improvement on the
Kraken crypto ccm benchmark).
Also turn the general CheckedInt32Div lowering into readable code again,
and make sure that all the bailout cases are properly covered by mjsunit
tests (i.e. the "division by zero" bailout was not covered properly).
Bug: v8:8015
Change-Id: Ibfdd367a6ee5d70dcaa48801858042c5029b7004
Reviewed-on: https://chromium-review.googlesource.com/1236954
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56115}
The previous tests didn't cover the case Number.isSafeInteger(x)
where TurboFan was unable to tell that `x` is always a Number and
thus had to use the ObjectIsSafeInteger operator instead.
Bug: v8:8015
Change-Id: I9bdbfa602fe0bf8c5fb2bc6c160ace7ab0bc0aaa
Reviewed-on: https://chromium-review.googlesource.com/1238234
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56114}
Again in the spirit of https://chromium-review.googlesource.com/1226033
we can simplify the handling of NumberDivide and decide the lowering
based on the feedback type.
Drive-by-fix: Add test coverage for the relevant corner cases of the
NumberDivide handling in SimplifiedLowering.
Bug: v8:8015
Change-Id: I0edaca0fddb31d64d2c269268e87a32a687a0b26
Reviewed-on: https://chromium-review.googlesource.com/1236262
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56113}
The ObjectIsArrayBuffer simplified operator, which is used to implement
the ArrayBuffer.isView() builtin, didn't have any test coverage.
Bug: v8:8015
Change-Id: Ia15e35bc4ae61627137f7a89976560a8d3db771f
Reviewed-on: https://chromium-review.googlesource.com/1238215
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56112}
In the spirit of https://chromium-review.googlesource.com/1226033 we can
also unify the handling of NumberModulus based on feedback types.
Drive-by-fix: Add appropriate tests for the corner cases of the
NumberModules with (surrounding) feedback integration.
Bug: v8:8015
Change-Id: I5e3207d2f6e72f9ea1d7658014b7272075088d63
Reviewed-on: https://chromium-review.googlesource.com/1236260
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56094}
The coverage bot figured out that there's missing test coverage
for the SpeculativeNumberModulus corner cases inside of the
SimplifiedLowering logic.
Bug: v8:8015
Change-Id: Id32aa545dc43adae5e67c66574ccea5f2b3db846
Reviewed-on: https://chromium-review.googlesource.com/1236259
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56093}
This adds missing test coverage for corner cases of SpeculativeNumberAdd
and SpeculativeNumberSubtract inside of SimplifiedLowering. This was
discovered to be untested by the coverage bot.
Bug: v8:8015
Change-Id: I7355b1b840a76bc12bd911adb6c2d88f05d816c5
Reviewed-on: https://chromium-review.googlesource.com/1236256
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56090}
Part of https://chromium-review.googlesource.com/1231994 that landed
earlier, but was reverted due to breakage. Landing this cleanup
separately instead.
Drive-by-fix: Also add test coverage for the cases that weren't covered
properly (according to the test coverage bot).
Bug: chromium:225811, v8:8015
Change-Id: I9c13ed5fcf0ba9e6b190489e15df86970eafdc13
Reviewed-on: https://chromium-review.googlesource.com/1236213
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56087}
According to the coverage bot, there's some lack of test coverage for
corner cases of Math.imul(). Add the missing test coverage and also
add some coverage for the generally interesting cases.
Bug: v8:8015
Change-Id: I2a917283b4777510fb5db421a039ff0de9b2a25f
Reviewed-on: https://chromium-review.googlesource.com/1235577
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56077}
Remove the NumberConstant right hand side limitation for the speculative
number operation optimization, and extend the logic to also deal with
SpeculativeToNumber, which is common when dealing with postfix increment
and array operations.
Also add appropriate tests for all the relevant cases, specifically we
mjsunit tests to increase the general coverage for the various cases
here (in addition to dedicated unittests).
Bug: v8:8015
Change-Id: I8c92f98490c63b07eb19686efd404322979e57c4
Reviewed-on: https://chromium-review.googlesource.com/1235919
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#56072}
If type checks in simplified lowering produced dead value (i.e., of
type Type::None()), we have only propagated deadness along value
edges. With this CL, we also insert an Unreachable node after every
effectful node that produces dead value.
This is more consistent with dead code elimination, which also inserts
unreachable nodes after effectful nodes with value output None.
Bug: chromium:884052
Change-Id: Idcb168461f05f1811b2c9c16ab8ff179b259fbd3
Reviewed-on: https://chromium-review.googlesource.com/1228125
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#55987}
For NumberAdd/Subtract/Multiply we currently onlt consult the upper
bound to decide whether to compute using Int32 or Float64 operations,
whereas for NumberModulus, NumberEqual, etc. we do decide based on
the feedback types, where the only significant difference is that we
cannot promise Word32 truncations on the inputs.
This change unifies the handling for NumberAdd/Subtract/Multiply as
well, which triggers surprisingly often in our core benchmark suites..
Bug: v8:8015
Change-Id: If8ec1bc82d1e1b71285c829262a0d343a4eb2af7
Reviewed-on: https://chromium-review.googlesource.com/1226033
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#55943}
This change introduces the necessary conversion operators to convert
from Word64 to other representations (Tagged, Word32, Float64, etc.),
and plugs in the Word64 representation for NumberAdd/NumberSubtract,
such that TurboFan will go to Int64Add/Sub on 64-bit architectures
when the inputs and the output of the operation is in safe integer
range. This includes the necessary changes to the Deoptimizer to be
able to rematerialize Int64 values as Smi/HeapNumber when going back
to Ignition later.
This change might affect performance, although measurements indicate
that there should be no noticable performance impact.
The goal is to have TurboFan support Word64 representation to a degree
that changing the TypedArray length to an uint64_t (for 64-bit archs)
becomes viable and doesn't have any negative performance implications.
Independent of that we might get performance improvements in other areas
such as for crypto code later.
Bug: v8:4153, v8:7881, v8:8171, v8:8178
Design-Document: bit.ly/turbofan-word64
Change-Id: I29d56e2a31c1bae61d04a89d29ea73f21fd49c59
Cq-Include-Trybots: luci.chromium.try:linux_chromium_headless_rel
Reviewed-on: https://chromium-review.googlesource.com/1225709
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#55937}
The typing rules for NumberMax and NumberMin didn't properly deal with
-0 up until now, leading to suboptimal typing, i.e. for a simple case
like
Math.max(Math.round(x), 1)
TurboFan was unable to figure out that the result is definitely going
to be a positive integer in the range [1,inf] or NaN (assuming that
NumberOrOddball feedback is used for the value x).
Bug: v8:8015
Change-Id: I06e14a9c9b0b813eb214ace7749fcc6ab36bb66a
Reviewed-on: https://chromium-review.googlesource.com/1199304
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#55570}
The new node is introduced for literal string addition and calling
String.prototype.concat in the typed lowering phase. It later might get optimized
away during redundancy elimination, keeping the performance of already existing
benchmarks with string addition. In case the operation is about to throw
(due to too long string being constructed) we just deoptimize, reusing
the interpreter logic for creating the error.
Modify relevant mjsunit and unit tests for string concatenation.
Bug: v8:7902
Change-Id: Ie97d39534df4480fa8d4fe3ba276d02ed5e750e3
Reviewed-on: https://chromium-review.googlesource.com/1193342
Commit-Queue: Maya Lekova <mslekova@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#55482}
Direct call to CloneFastJSArray is used to improve performance in that
case. Tests are also added.
Bug: v8:7980
Change-Id: Ifca34f3e182b776cd9862da8bf529fc13f6be9ed
Reviewed-on: https://chromium-review.googlesource.com/1172782
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Hai Dang <dhai@google.com>
Cr-Commit-Position: refs/heads/master@{#55154}
This fixes the bug where the reducer ignores a prototype that is not
initial. Tests are also added.
Bug: v8:8056
Change-Id: I428eed2d2790fffa22f67a051f7d1f1e4d3ce947
Reviewed-on: https://chromium-review.googlesource.com/1174542
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Hai Dang <dhai@google.com>
Cr-Commit-Position: refs/heads/master@{#55149}
This affects map, filter, every, some, indexOf and includes.
Tests for those cases and more are also added.
Bug: v8:8049
Change-Id: I196abd8e7e2419a2bb465f44caf4de52990ffced
Reviewed-on: https://chromium-review.googlesource.com/1172346
Reviewed-by: Georg Neis <neis@chromium.org>
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Commit-Queue: Hai Dang <dhai@google.com>
Cr-Commit-Position: refs/heads/master@{#55103}
The DataView access methods can use the neutering protector to avoid
introducing an explicit check into the optimized code to see if the
backing store was neutered. Instead the optimized code has an implicit
dependency on the global neutering protector which gets invalidated
when the first array buffer is neutered (globally). We use the same
trick for typed arrays already.
Bug: chromium:225811
Change-Id: I9b3c95b3113b8fa00dcbba216ef29c84c0056951
Reviewed-on: https://chromium-review.googlesource.com/1172779
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Cr-Commit-Position: refs/heads/master@{#55097}
This CL simplifies the implementation of inlined DataView
methods in TurboFan. It removes the explicit exception handling,
and just deopts and relies on the baseline code to handle
exceptions instead.
It also adapts the DataView test files in mjsunit/compiler/
accordingly.
Change-Id: I013c76970e1480df2b755d17d397bd0f9f26f0ec
Reviewed-on: https://chromium-review.googlesource.com/1148207
Commit-Queue: Théotime Grohens <theotime@google.com>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#54648}
This is a reland of 9eca23e9ed
Adds a deopt continuation, which fixes JavaScript stack traces
to contain the number constructor after inlining.
Original change's description:
> [turbofan] Inline Number constructor in certain cases
>
> This CL adds inlining for the Number constructor if new.target is not
> present. The lowering is BigInt compatible, i.e. it converts BigInts to
> numbers.
>
> Bug: v8:7904
> Change-Id: If03b9f872d82e50b6ded7709069181c33dc44e82
> Reviewed-on: https://chromium-review.googlesource.com/1118557
> Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
> Reviewed-by: Georg Neis <neis@chromium.org>
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#54454}
Bug: v8:7904
Change-Id: Ic416e5ba81fa3a0f59ae4afa80df83c46a759487
Reviewed-on: https://chromium-review.googlesource.com/1146581
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#54609}
This CL completes the implementation of DataView prototype methods
in TurboFan, by implementing the Uint8, Int8, Uint16, Int16,
Uint32, Int32, Float32 and Float64 setters.
DataView performance is now ahead of the equivalent TypedArray wrapper,
and is now expected to at least match TypedArray performance in
the general case as well.
This CL also adds a test file in the compiler directory, to make
sure that the setters actually behave correctly.
Change-Id: I4ad4341c6b9b9d461348b62216f37a73abe321e8
Reviewed-on: https://chromium-review.googlesource.com/1128867
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Commit-Queue: Théotime Grohens <theotime@google.com>
Cr-Commit-Position: refs/heads/master@{#54331}
This CL implements Reduction and Lowering for the DataView Int32,
Uint32, Float32 and Float64 getters.
This makes DataView getters fully supported in TurboFan (except for
BigInts), and should bridge the performance gap with TypedArrays.
Change-Id: Ifa98df9cf13e44d6468ad9ec8a19c86b41c6d2b1
Reviewed-on: https://chromium-review.googlesource.com/1127360
Commit-Queue: Théotime Grohens <theotime@google.com>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#54288}
This CL adds a Reduction for the DataViewGetInt16 and -Uint16 builtins,
and the corresponding handling in LoadDataViewElement node in the
effect control linearizer.
It also adds tests for the new getters.
Change-Id: I5101755e47657c25f10be1417f105e3ae72a3c39
Reviewed-on: https://chromium-review.googlesource.com/1126919
Commit-Queue: Théotime Grohens <theotime@google.com>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#54271}
This CL adds code to inline the Int8 and Uint8 getters for DataView
objects in TurboFan in js-call-reducer.cc, as well as a new test file.
It already improves execution speed compared to the Torque baseline
implementation, and implements most of the architecture needed
for inlining the other DataView getters and setters as well.
Change-Id: I0e62b98fd6ec995f7db5ec42ea1eff1f03572f97
Reviewed-on: https://chromium-review.googlesource.com/1119909
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Théotime Grohens <theotime@google.com>
Cr-Commit-Position: refs/heads/master@{#54157}
Also add a DCHECK to a branch that can only be taken for the null
prototype.
R=sigurds@chromium.org
Change-Id: Ib94fe8f25ecfd1a4baa576915e6edfa60bcd771b
Reviewed-on: https://chromium-review.googlesource.com/1109961
Commit-Queue: Georg Neis <neis@chromium.org>
Reviewed-by: Sigurd Schneider <sigurds@chromium.org>
Cr-Commit-Position: refs/heads/master@{#53939}
This is not web compatible, so let's delete the code.
Bug: v8:5536
Change-Id: I50506d37dcdff1f7f95577c47adcec653cc1f06e
Reviewed-on: https://chromium-review.googlesource.com/1064740
Commit-Queue: Sathya Gunasekaran <gsathya@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#53264}
... in order to be able to use it in other constants definitions in the header.
Bug: v8:7570
Cq-Include-Trybots: luci.chromium.try:linux_chromium_rel_ng
Change-Id: Id5d6ae34ab401ecf063bf5897b87b6bb87c24960
Reviewed-on: https://chromium-review.googlesource.com/1032782
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Igor Sheludko <ishell@chromium.org>
Cr-Commit-Position: refs/heads/master@{#53097}
This mjsunittest assumed specific internal types (i.e. Smi)
for certain fields; it generates some dozens of variants of
the test using new Function, but used the same property names
in all of them. This causes V8 to sometimes learn more general
types for fields (i.e. unboxed double), which the test did not
expect. This commit uses unique field names for each of the test
variants.
Change-Id: Ib1ecb3ae33a57c8a1293a29a2233dad4e16a39fb
Reviewed-on: https://chromium-review.googlesource.com/1004897
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#52528}
This CL allows builtin continuations to handle pending exceptions.
This implements exception handling for the promise constructor in
case of deoptimization.
Bug: v8:7584
Change-Id: Ib5df5eb6606abb3f9690f294397981858dbdbf25
Reviewed-on: https://chromium-review.googlesource.com/983912
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#52340}
This also introduces two new simplified operators,
NumberIsFinite and ObjectIsFiniteNumber; the latter
handles all values, and the former is a fast-path
of the fast-path that is inserted by typed optimization
if we know the input has Type::Number.
Bug: v8:7340, v8:7250
Change-Id: I1b4812c01bf470bbff40fb3da6e11da543a22cd2
Reviewed-on: https://chromium-review.googlesource.com/951244
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51980}
This introduces a new JSCreateTypedArray operator, backed by a dedicated
CreateTypedArray builtin, and adds support to lowering new TypedArray
calls to this operator. This way we avoid the overhead of going through
the generic construct stub machinery for hot code. This not only
recovers the performance regression on the typed array constructor
benchmarks, but even improves slightly beyond what we had in 6.6.
We might in the future try to fully inline the TypedArray constructor
into optimized code for certain cases.
Bug: chromium:820726, v8:7503, v8:7518
Change-Id: Ied465924d5695db576d533792f1db68456b9b5ea
Reviewed-on: https://chromium-review.googlesource.com/959010
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Peter Marshall <petermarshall@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51973}
The analysis phase used to skip TypeGuard nodes, which are
normally re-introduced by the reduction phase. However, phi nodes
are created during the analysis phase already, and so it could happen
that a phi input skips a TypeGuard.
This CL solves the problem by not removing TypeGuard nodes in the first
place, but only forwarding the VirtualObject. This is analogous to how
we already treat FinishRegion nodes, which are similar in that they are
a renaming too.
Bug: chromium:741225
Change-Id: Icf8aa2d40a30d89788d875b37b9986111f9c966f
Reviewed-on: https://chromium-review.googlesource.com/958442
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51863}
Folding _ + NaN => NaN can widen type None to a constant type, which leads to floating DeadValue nodes. This CL fixes this by removing the optimization. Alternatively, we should consider removing all nodes of type None in simplified lowering.
Bug: chromium:817225
Change-Id: I2a126b360d70d3626f8a3c5e73ac72dc980ac8b3
Reviewed-on: https://chromium-review.googlesource.com/946129
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51699}
Turbofan can only handle 64K control inputs for merges. Such large
can only be created by functions with 64K jumps, so we limit the
bytecode size to the minimum size of bytecode arrays with 64K jumps.
Bug: chromium:815392, v8:7438
Change-Id: I674705e87e19ce451b40d5827c9fe3e6ec17293a
Reviewed-on: https://chromium-review.googlesource.com/938421
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51598}
This fixes issues where the stack track contained 'Promise' but
not 'new'.
Bug: v8:7253
Change-Id: I840fcc0a76e2376aab0b64d321f5cf8ccc672956
Reviewed-on: https://chromium-review.googlesource.com/928762
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51516}
This adds a frame state for the call to the executor in inlined promise
constructors. We provide a continuation function in case of deopts which
just returns the created promise. This is not totally correct yet: if
the executor function also throws, we need to catch it and call the
reject function instead.
We also still need to add a frame state for the isCallable check on the
executor, so that the stack is correct for the thrown TypeError.
Bug: v8:7253
Change-Id: I3ee042ec82f1a9a35d59e576f6c8efe9bc98698c
Reviewed-on: https://chromium-review.googlesource.com/926523
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51417}
Without processing the input, a phi node can be flagged as unused and
replaced with {Dead}, although it is used by a {DeadValue} node.
Bug: chromium:808472
Change-Id: I7446883535b34770e31e4e26e1c242eb05673a91
Reviewed-on: https://chromium-review.googlesource.com/919362
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51395}
This makes the inlining of the default resolve/reject closures generated
by the Promise constructor effective. To be really useful we still need
to have the Promise constructor inlined (work-in-progress) and eventually
track SharedFunctionInfo feedback in the CALL_IC.
Bug: v8:2206, v8:7253
Change-Id: I08fa8ca72754f459ae36027a55377ef57d411cdc
Reviewed-on: https://chromium-review.googlesource.com/926103
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51390}
Inline the promise constructor when we have one argument and target
matches new_target.
This is not complete, and is sitting behind an experimental flag for
now. We need to fix deoptimization by providing proper frame states.
Create a unittest class for JSCallReducer - just assert whether there
was a change or not, rather than specify the exact graph that should be
produced.
Bug: v8:7253
Change-Id: Ib6886a8feb2799f47cd647853cabcf12a189bc25
Reviewed-on: https://chromium-review.googlesource.com/919282
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51389}
This CL introduces new operators JSFulfillPromise and JSPromiseResolve,
corresponding to the specification operations with the same name, and
uses that to lower calls to Promise.resolve() builtin to JSPromiseResolve.
We also optimize JSPromiseResolve and JSResolvePromise further based on
information found about the value/resolution in the graph. This applies
to both Promise.resolve() builtin calls and implicit resolve operations
in async functions and async generators.
On a very simple microbenchmark like
console.time('resolve');
for (let i = 0; i < 1e8; ++i) Promise.resolve({i});
console.timeEnd('resolve');
this CL reduces the execution time from around 3049ms to around 947ms,
which is a pretty significant 3x improvement. On the wikipedia benchmark
we observe an improvement around 2% with this CL.
Bug: v8:7253
Change-Id: Ic69086cdc1b724f35dbe83305795539c562ab817
Reviewed-on: https://chromium-review.googlesource.com/913488
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51387}
This introduces dedicated builtins
- FulfillPromise,
- RejectPromise, and
- ResolvePromise,
which perform the corresponding operations from the language
specification, and removes the redundant entry points and the
excessive inlining of these operations into other builtins. We
also add the same logic on the C++ side, so that we don't need
to go into JavaScript land when resolving/rejecting from the
API.
The C++ side has a complete implementation, including full support
for the debugger and the current PromiseHook machinery. This is to
avoid constantly crossing the boundary for those cases, and to also
simplify the CSA side (and soon the TurboFan side), where we only
do the fast-path and bail out to the runtime for the general handling.
On top of this we introduce %_RejectPromise and %_ResolvePromise,
which are entry points used by the bytecode and parser desugarings
for async functions, and also used by the V8 Extras API. Thanks to
this we can uniformly optimize these in TurboFan, where we have
corresponding operators JSRejectPromise and JSResolvePromise, which
currently just call into the builtins, but middle-term can be further
optimized, i.e. to skip the "then" lookup for JSResolvePromise when
we know something about the resolution.
In TurboFan we can also already inline the default PromiseCapability
[[Reject]] and [[Resolve]] functions, although this is not as effective
as it can be right now, until we have inlining support for the Promise
constructor (being worked on by petermarshall@ right now) and/or SFI
based CALL_IC feedback.
Overall this change is meant as a refactoring without significant
performance impact anywhere; it seems to improve performance of
simple async functions a bit, but otherwise is neutral.
Bug: v8:7253
Change-Id: Id0b979f9b2843560e38cd8df4b02627dad4b6d8c
Reviewed-on: https://chromium-review.googlesource.com/911632
Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Georg Neis <neis@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51260}
This reverts commit 14108f4c2e.
Reason for revert: Not the culprit for Canary microtask crashes
Original change's description:
> [builtins] Mega-revert to address the Dev blocker in crbug.com/808911.
>
> - Revert "[builtins] Save one word in contexts for Promise.all."
> This reverts commit 7632da067b.
> - Revert "[builtins] Also use the Promise#then protector for Promise#finally()."
> This reverts commit d4f072ced3.
> - Revert "[builtins] Don't mess with entered context for MicrotaskCallbacks."
> This reverts commit 6703dacdd6.
> - Revert "[debugger] Properly deal with settled promises in catch prediction."
> This reverts commit 40dd065823.
> - Revert "[builtins] Widen the fast-path for Promise builtins."
> This reverts commit db0556b7e8.
> - Revert "[builtins] Unify PerformPromiseThen and optimize it with TurboFan."
> This reverts commit a582199c5e.
> - Revert "[builtins] Remove obsolete PromiseBuiltinsAssembler::AppendPromiseCallback."
> This reverts commit 6bf8885290.
> - Revert "[builtins] Turn NewPromiseCapability into a proper builtin."
> This reverts commit 313b490ddd.
> - Revert "[builtins] Inline InternalPromiseThen into it's only caller"
> This reverts commit f7bd6a2fd6.
> - Revert "[builtins] Implement Promise#catch by really calling into Promise#then."
> This reverts commit b23b098fa0.
> - Revert "[promise] Remove incorrect fast path"
> This reverts commit 0f6eafe855.
> - Revert "[builtins] Squeeze JSPromise::result and JSPromise::reactions into a single field."
> This reverts commit 8a677a2831.
> - Revert "[builtins] Refactor promises to reduce GC overhead."
> This reverts commit 8e7737cb58.
>
> Tbr: hpayer@chromium.org
> Bug: chromium:800651, chromium:808911, v8:5691, v8:7253
> Change-Id: I8c8ea5ed32ed62f6cd8b0d027a3707ddd891e5f1
> Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
> Reviewed-on: https://chromium-review.googlesource.com/906991
> Commit-Queue: Yang Guo <yangguo@chromium.org>
> Commit-Queue: Adam Klein <adamk@chromium.org>
> Reviewed-by: Adam Klein <adamk@chromium.org>
> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#51158}
Change-Id: I09d958cbebd635a325809072a290f2f53df8c5d4
Tbr: adamk@chromium.org,yangguo@chromium.org,bmeurer@chromium.org
Bug: chromium:800651, chromium:808911, v8:5691, v8:7253
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Reviewed-on: https://chromium-review.googlesource.com/908988
Reviewed-by: Adam Klein <adamk@chromium.org>
Commit-Queue: Adam Klein <adamk@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51181}
- Revert "[builtins] Save one word in contexts for Promise.all."
This reverts commit 7632da067b.
- Revert "[builtins] Also use the Promise#then protector for Promise#finally()."
This reverts commit d4f072ced3.
- Revert "[builtins] Don't mess with entered context for MicrotaskCallbacks."
This reverts commit 6703dacdd6.
- Revert "[debugger] Properly deal with settled promises in catch prediction."
This reverts commit 40dd065823.
- Revert "[builtins] Widen the fast-path for Promise builtins."
This reverts commit db0556b7e8.
- Revert "[builtins] Unify PerformPromiseThen and optimize it with TurboFan."
This reverts commit a582199c5e.
- Revert "[builtins] Remove obsolete PromiseBuiltinsAssembler::AppendPromiseCallback."
This reverts commit 6bf8885290.
- Revert "[builtins] Turn NewPromiseCapability into a proper builtin."
This reverts commit 313b490ddd.
- Revert "[builtins] Inline InternalPromiseThen into it's only caller"
This reverts commit f7bd6a2fd6.
- Revert "[builtins] Implement Promise#catch by really calling into Promise#then."
This reverts commit b23b098fa0.
- Revert "[promise] Remove incorrect fast path"
This reverts commit 0f6eafe855.
- Revert "[builtins] Squeeze JSPromise::result and JSPromise::reactions into a single field."
This reverts commit 8a677a2831.
- Revert "[builtins] Refactor promises to reduce GC overhead."
This reverts commit 8e7737cb58.
Tbr: hpayer@chromium.org
Bug: chromium:800651, chromium:808911, v8:5691, v8:7253
Change-Id: I8c8ea5ed32ed62f6cd8b0d027a3707ddd891e5f1
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Reviewed-on: https://chromium-review.googlesource.com/906991
Commit-Queue: Yang Guo <yangguo@chromium.org>
Commit-Queue: Adam Klein <adamk@chromium.org>
Reviewed-by: Adam Klein <adamk@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51158}
Add a fast-path to Promise#finally, which skips the "then" lookup of the
Promise#then lookup chain is intact, similar to what we already do for
Promise#catch.
Drive-by-fix: Also use the @@species protector to speed up the lookup
of the SpeciesConstructor in Promise#finally.
Bug: v8:7253
Change-Id: If77e779a0188904effc4528beffc8f0bdd7c2efe
Reviewed-on: https://chromium-review.googlesource.com/902283
Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51116}
This adds a new isolate wide Promise#then protector, which guards the
"then" lookup for all JSPromise instances whose [[Prototype]] is the
initial %PromisePrototype%. Thus arbitrary mutations to the
Promise.prototype (i.e. monkey-patching other methods or installing
new functions) no longer sent you down the slow-path. Use this protector
in Promise.prototype.catch and in Promise.resolve.
Drive-by-fix: Restructure the resolve logic a bit and avoid the
expensive and large SameValue check, which can be turned into a simple
reference equal, as the promise in there is known to be a JSPromise
anyways.
Bug: v8:7253
Change-Id: If68b12c6bc6ca9c4d10552ae84854ebc3b5774f9
Reviewed-on: https://chromium-review.googlesource.com/899302
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Sathya Gunasekaran <gsathya@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51085}
This creates a uniform PerformPromiseThen builtin, which performs the
operation with the same name from the spec, except that it expects the
handlers to be either undefined or callable already, since this is only
relevant for a single callsite (namely Promise.prototype.then).
Introduce a matching operator JSPerformPromiseThen into TurboFan, which
represents this operation and removes the additional checks in case of
Promise.prototype.then based on the information we can derived from the
receiver maps.
This yields a nice 20-25% improvement on Promise.prototype.then, as
illustrated by the following micro-benchmark
```js
const N = 1e7;
function inc(x) { return x + 1; }
function chain(promise) {
return promise.then(inc).then(value => {
if (value < N) chain(Promise.resolve(value));
});
}
console.time('total');
chain(Promise.resolve(0));
setTimeout(console.timeEnd.bind(console, 'total'));
```
which goes from around 1230ms to 930ms with this patch.
Bug: v8:7253
Change-Id: I5712a863acdbe7da3bb8e621887c7b952148c51a
Reviewed-on: https://chromium-review.googlesource.com/899064
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51071}
We still avoid the "then" lookup using the current fast-path
mega-guard in the baseline case, but in TurboFan we simply
constant-fold the "then" lookup in the JSCallReducer. So all
further optimizations on Promise#then in TurboFan will automatically
apply to Promise#catch as well.
Bug: v8:7253
Change-Id: Idf7252157375a0ae3a91c7a3b42c30c5f367c0a8
Reviewed-on: https://chromium-review.googlesource.com/895446
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Cr-Commit-Position: refs/heads/master@{#51008}
This delays removing dead loop's loop exits after we iterate all uses of
the loop. That way, we avoid mutating the use collection while iterating
it.
Bug: chromium:803022
Change-Id: I17462dd82c3cb78f2f630e5db81d8ccdcc517d83
Reviewed-on: https://chromium-review.googlesource.com/878329
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50813}
This CL moves allocations in array-multiple-receiver-maps.js
to prevent gc fuzzing from cleaning out code objects, which
will mess with assertOptimized in the test.
Bug: v8:7338
Change-Id: I9ee88cf5518307ff12302df2fdaca5258c23b779
Reviewed-on: https://chromium-review.googlesource.com/880957
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50809}
DeadValue was a constant node of type None. This is unsound in the
presence of re-scheduling. This CL adds a value input to DeadValue,
which preserves the dependency on the original node of type None.
This reland addresses the bug that the EffectControlLinearizer could destroy dependencies of DeadValue by attaching DeadValue nodes to the effect chain in the EffectControlLinearizer.
Bug: chromium:796041 chromium:798938
Change-Id: If47b54a7986d257eb63b437f855769b503679ff5
Reviewed-on: https://chromium-review.googlesource.com/850392
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50360}
Add an early exit if the control op is Dead to prevent failing the
DCHECK.
BUG=chromium:797596, v8:5940, v8:3018
R=bmeurer@chromium.org, jarin@chromium.org
Change-Id: I6090380ea69c3205740b6c7a41d7c066d18d6a9f
Reviewed-on: https://chromium-review.googlesource.com/844978
Commit-Queue: Caitlin Potter <caitp@igalia.com>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50312}
This patch updates the instruction selector and code generator to pad arguments
for arm64 and drop an even number of slots when dropping the arguments. It also
updates the builtins that handle arguments. These changes need to be made at
the same time.
It also adds some tests for forwarding varargs, as this was affected by the
builtin changes and the existing tests did not catch all issues.
Bug: v8:6644
Change-Id: I81318d1d1c9ab2568f84f2bb868d2a2d4cb56053
Reviewed-on: https://chromium-review.googlesource.com/829933
Commit-Queue: Georgia Kouveli <georgia.kouveli@arm.com>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50259}
This CL passes feedback from the element kind deopt points
in Array.push to the deoptimizer. If the deopt points are
triggered, further speculation on Array.push is disallowed.
Bug: v8:7127, v8:7204
Change-Id: Ie91dee598bd8b8797110c8f468406327226893a4
Reviewed-on: https://chromium-review.googlesource.com/831523
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50171}
Add feedback to GrowFastElements operator and thread it
through to the deoptimize node it the lowering. The CL
uses the feedback to allow Array.push to disable speculation
if the grow operation deopts.
Bug: v8:7127, v8:7204
Change-Id: Ib5850a93759b9194c0fc2f191f6adf5d49cb7f55
Reviewed-on: https://chromium-review.googlesource.com/827128
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50145}
This reverts commit 917b9cb9fc.
In this CL, we canonicalize the fixed array when allocating storage for
empty fixed array. During initialization, we also make sure that we do
not write to the empty fixed array. This is quite hacky, but it
seems to be the least intrusive change.
Bug: chromium:793863
Change-Id: I1449ebac7c1e390467566a759bf70e7e2fabda31
Reviewed-on: https://chromium-review.googlesource.com/827013
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50119}
Add support for disallowing speculation upon deoptimize from
a CheckBound node, and use this in the case of array builtins
in js-call-reducer to prevent deoptimization loops.
Bug: v8:7127
Change-Id: I04cf655b10178d2938d2f0ee6b336601fab6463b
Reviewed-on: https://chromium-review.googlesource.com/822195
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50097}
This disallows speculation after deoptimization from any of
Array.{forEach,map,filter,find} due to CheckMap fails. Such
CheckMap fails happen if the builtins' function argument
causes the map of the array to change. The js-call-lowering
refrains from optimizing builtins for which speculation was
disallowed.
Bug: v8:6898, v8:7127
Change-Id: Ied6696f8fb023ee404fb82e9d37bfb061f293854
Reviewed-on: https://chromium-review.googlesource.com/819354
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Sigurd Schneider <sigurds@chromium.org>
Cr-Commit-Position: refs/heads/master@{#50069}
This relands commit e71b802279.
This can now back in as the fix for chromium:787301 had enough time to
be tested in Canary.
Original change's description:
> [deoptimizer] Staged materialization of objects.
>
> The existing object materialization in the deoptimizer has the following problems:
>
> - Objects do not necessarily verify during materialization (because during the
> depth first walk we might have inconsistent objects).
>
> - Stack can overflow (because we just materialize using recursive calls).
>
> - We generalize object fields.
>
>
> This CL re-implements the materialization algorithm to solve this problem. The
> new implementation creates the objects in two steps:
>
> 1. We allocate space for all the objects. In general, we allocate ByteArrays
> of the right size. For leaf objects that cannot participate in cycles,
> we build and initialize the materialized objects completely.
>
> For JS objects, we insert markers into the byte array at the positions
> where unboxed doubles are expected.
>
> 2. We initialize all the objects with the proper field values and change the
> map from the ByteArray map to the correct map. This requires some sync
> with the concurrent marker (Heap::NotifyObjectLayoutChange).
>
> When initializing the JS object fields, we make sure that we respect
> the unboxed double marker.
>
> Bug: chromium:770106, v8:3836
> Change-Id: I1ec466a9d19db9538df4ba915516d4c3ca825632
> Reviewed-on: https://chromium-review.googlesource.com/777559
> Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
> Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
> Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#49821}
Bug: chromium:770106, v8:3836
Change-Id: Ied6c4e0fbae52713e55ae6dc13794a7521dbb8a5
Reviewed-on: https://chromium-review.googlesource.com/817745
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49982}
We cannot remove a speculative operation when it's type relies on it to deopt.
Fix this by only relying on the lowering to remove operations.
Bug: chromium:786521
Change-Id: I2cf45e8d45b76cfeb06e6329f323cade74719124
Reviewed-on: https://chromium-review.googlesource.com/793043
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49882}
The proper fix would be to make TruncatingUseInfoFromRepresentation
respect tagged signed use representation, but requires extra work
to refine typing for all values that are stored into Smi fields.
Bug: chromium:791245
Change-Id: I83965bcc18a836d2c758a6a8b1477a4aa2c6133d
Reviewed-on: https://chromium-review.googlesource.com/808866
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49870}
Eventually, we want to fix this also for tagged pointers (tracking bug: https://crbug.com/v8/7162).
Bug: chromium:791245
Change-Id: I93d6deff36cedcc9a4665fab0abe6fffdae9b61b
Reviewed-on: https://chromium-review.googlesource.com/806457
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49850}
This reverts commit e71b802279.
Reason for revert: Need to have a back-mergeable fix.
Original change's description:
> [deoptimizer] Staged materialization of objects.
>
> The existing object materialization in the deoptimizer has the following problems:
>
> - Objects do not necessarily verify during materialization (because during the
> depth first walk we might have inconsistent objects).
>
> - Stack can overflow (because we just materialize using recursive calls).
>
> - We generalize object fields.
>
>
> This CL re-implements the materialization algorithm to solve this problem. The
> new implementation creates the objects in two steps:
>
> 1. We allocate space for all the objects. In general, we allocate ByteArrays
> of the right size. For leaf objects that cannot participate in cycles,
> we build and initialize the materialized objects completely.
>
> For JS objects, we insert markers into the byte array at the positions
> where unboxed doubles are expected.
>
> 2. We initialize all the objects with the proper field values and change the
> map from the ByteArray map to the correct map. This requires some sync
> with the concurrent marker (Heap::NotifyObjectLayoutChange).
>
> When initializing the JS object fields, we make sure that we respect
> the unboxed double marker.
>
> Bug: chromium:770106, v8:3836
> Change-Id: I1ec466a9d19db9538df4ba915516d4c3ca825632
> Reviewed-on: https://chromium-review.googlesource.com/777559
> Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
> Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
> Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#49821}
TBR=ulan@chromium.org,mstarzinger@chromium.org,jarin@chromium.org
Change-Id: I0657fb75330700dd7883c600dacb25676ebb47f9
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: chromium:770106, v8:3836
Reviewed-on: https://chromium-review.googlesource.com/806160
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49834}
The existing object materialization in the deoptimizer has the following problems:
- Objects do not necessarily verify during materialization (because during the
depth first walk we might have inconsistent objects).
- Stack can overflow (because we just materialize using recursive calls).
- We generalize object fields.
This CL re-implements the materialization algorithm to solve this problem. The
new implementation creates the objects in two steps:
1. We allocate space for all the objects. In general, we allocate ByteArrays
of the right size. For leaf objects that cannot participate in cycles,
we build and initialize the materialized objects completely.
For JS objects, we insert markers into the byte array at the positions
where unboxed doubles are expected.
2. We initialize all the objects with the proper field values and change the
map from the ByteArray map to the correct map. This requires some sync
with the concurrent marker (Heap::NotifyObjectLayoutChange).
When initializing the JS object fields, we make sure that we respect
the unboxed double marker.
Bug: chromium:770106, v8:3836
Change-Id: I1ec466a9d19db9538df4ba915516d4c3ca825632
Reviewed-on: https://chromium-review.googlesource.com/777559
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49821}
Finally address that long-standing TODO where ConsString allocation in
TurboFan would always go for the two byte map instead of choosing the
one byte map if the inputs are one byte strings.
Bug: v8:5269, v8:7109
Change-Id: Ibcfceaf499ceebef0ef928ebc5f204bcacf29bc0
Reviewed-on: https://chromium-review.googlesource.com/799700
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49744}
This addresses two TODOs in Ignition where the Construct and the
ConstructWithSpread bytecodes didn't collect JSBoundFunction
new.target feedback. This is fairly trivial to add now with the
existing machinery and the TurboFan side of this was already fixed
before, so we can leverage the new feedback.
Bug: v8:5267, v8:7109
Change-Id: Iae257836716c14f05f5d301326cbe8b2acaeb38b
Reviewed-on: https://chromium-review.googlesource.com/793048
Reviewed-by: Mythri Alle <mythria@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49712}
In TurboFan we can easily recognize calls to String.prototype.slice
where the start parameter is -1 and the end parameter is either
undefined or not present. These calls either return an empty string if
the input string is empty, or the last character of the input string
as a single character string. So we can just make use of the existing
StringCharAt operator.
This reduces the overhead of the String.prototype.slice calls from
optimized code in the chai test of the web-tooling-benchmark
significantly. We observe a 2-3% improvement on the test.
Bug: v8:6936, v8:7137
Change-Id: Iebe02667446880f5760e3e8c80f8b7cc712df663
Reviewed-on: https://chromium-review.googlesource.com/795726
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49704}
Properly handle known JSBoundFunction instances as targets to
JSConstruct by inlining the construction of the eventual target.
Also if the target is the result of a JSCreateBoundFunction call,
where we can also fold the construction and construct the bound
target function directly instead.
This addresses half of the TODO in the JSConstruct lowering in the
JSCallReducer where so far we didn't handle bound functions.
Bug: v8:5267, v8:7109
Change-Id: I022dc7d4fbbe2c9972472e78a6d64f51e3134c94
Reviewed-on: https://chromium-review.googlesource.com/792947
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49664}
This extends the typing rule for NumberTrunc to deal with general number
inputs properly, thus addressing a long-standing TODO. We also add test
cases to ensure that the typing rule gets the corner cases for NaN and
-0 right.
Bug: v8:5267, v8:7109
Change-Id: Iedc541a0f4619f37da37ea36940f92472034cdf2
Reviewed-on: https://chromium-review.googlesource.com/792932
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49652}
This extends the typing rule for NumberRound to deal with general number
inputs properly, thus addressing a long-standing TODO. We also add test
cases to ensure that the typing rule gets the corner cases for NaN and
-0 right.
Bug: v8:5267, v8:7109
Change-Id: Ia865ec1d6f8d96f20641bee96891740a9fc6e627
Reviewed-on: https://chromium-review.googlesource.com/792931
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49651}
This extends the typing rule for NumberCeil to deal with general number
inputs properly, thus addressing a long-standing TODO. We also add test
cases to ensure that the typing rule gets the corner cases for NaN and
-0 right.
Bug: v8:5267, v8:7109
Change-Id: I9154e47e58ad106791613db0030051f2a802a981
Reviewed-on: https://chromium-review.googlesource.com/792930
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49650}
The typer's ToNumber (and thus ToInteger etc.) returns type None when
the input type is BigInt, but we weren't quite ready for that in a few
places.
R=jarin@chromium.org
Bug: v8:7121
Change-Id: Ib12c726338f1ec3dfb9ba5cf54b00cc8d1351a89
Reviewed-on: https://chromium-review.googlesource.com/785130
Commit-Queue: Georg Neis <neis@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49604}
Reland of https://chromium-review.googlesource.com/c/v8/v8/+/727893
The crashes should be fixed by https://chromium-review.googlesource.com/c/v8/v8/+/763531
Original change's description:
> Revert "Reland^5 "[turbofan] eagerly prune None types and deadness from the graph""
>
> This reverts commit ac0661b358.
>
> Reason for revert: Clusterfuzz unhappy: chromium:783019 chromium:783035
>
> Original change's description:
> > Reland^5 "[turbofan] eagerly prune None types and deadness from the graph"
> >
> > This gives up on earlier attempts to interpret DeadValue as a signal of
> > unreachable code. This does not work because free-floating dead value
> > nodes, and even pure branch nodes that use them, can get scheduled so
> > early that they get reachable. Instead, we now eagerly remove branches
> > that use DeadValue in DeadCodeElimination and replace DeadValue inputs
> > to value phi nodes with dummy values.
> >
> > Reland of https://chromium-review.googlesource.com/715716
> >
> > Bug: chromium:741225 chromium:776256
> > Change-Id: I251efd507c967d4a8882ad8fd2fd96c4185781fe
> > Reviewed-on: https://chromium-review.googlesource.com/727893
> > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > Cr-Commit-Position: refs/heads/master@{#49188}
>
> TBR=jarin@chromium.org,tebbi@chromium.org
>
> Bug: chromium:741225 chromium:776256 chromium:783019 chromium:783035
> Change-Id: I6a8fa3a08ce2824a858ae01817688e63ed1f442e
> Reviewed-on: https://chromium-review.googlesource.com/758770
> Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
> Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#49262}
TBR=jarin@chromium.org,tebbi@chromium.org
# Not skipping CQ checks because original CL landed > 1 day ago.
Bug: chromium:741225 chromium:776256 chromium:783019 chromium:783035
Change-Id: I6c02b4beb02997ec34015ed2f6791a93c70f5e36
Reviewed-on: https://chromium-review.googlesource.com/772150
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49429}
This reverts commit ac0661b358.
Reason for revert: Clusterfuzz unhappy: chromium:783019 chromium:783035
Original change's description:
> Reland^5 "[turbofan] eagerly prune None types and deadness from the graph"
>
> This gives up on earlier attempts to interpret DeadValue as a signal of
> unreachable code. This does not work because free-floating dead value
> nodes, and even pure branch nodes that use them, can get scheduled so
> early that they get reachable. Instead, we now eagerly remove branches
> that use DeadValue in DeadCodeElimination and replace DeadValue inputs
> to value phi nodes with dummy values.
>
> Reland of https://chromium-review.googlesource.com/715716
>
> Bug: chromium:741225 chromium:776256
> Change-Id: I251efd507c967d4a8882ad8fd2fd96c4185781fe
> Reviewed-on: https://chromium-review.googlesource.com/727893
> Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#49188}
TBR=jarin@chromium.org,tebbi@chromium.org
Bug: chromium:741225 chromium:776256 chromium:783019 chromium:783035
Change-Id: I6a8fa3a08ce2824a858ae01817688e63ed1f442e
Reviewed-on: https://chromium-review.googlesource.com/758770
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49262}
This gives up on earlier attempts to interpret DeadValue as a signal of
unreachable code. This does not work because free-floating dead value
nodes, and even pure branch nodes that use them, can get scheduled so
early that they get reachable. Instead, we now eagerly remove branches
that use DeadValue in DeadCodeElimination and replace DeadValue inputs
to value phi nodes with dummy values.
Reland of https://chromium-review.googlesource.com/715716
Bug: chromium:741225 chromium:776256
Change-Id: I251efd507c967d4a8882ad8fd2fd96c4185781fe
Reviewed-on: https://chromium-review.googlesource.com/727893
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49188}
This adds support to the KeyedLoadIC to ignore out of bounds accesses
for Strings and return undefined instead. We add a dedicated bit to the
Smi handler to encode the OOB state and have TurboFan generate appropriate
code for that case as well. This is mostly useful when programs
accidentially access past the length of a string, which was observed and
fixed for example in Babel recently, see
https://github.com/babel/babel/pull/6589
for details. The idea is to also extend this mechanism to Arrays and
maybe other receivers, as reading beyond the length is also often used
in jQuery and other popular libraries.
Note that this is considered a mitigation for a performance cliff and
not a general optimization of OOB accesses. These should still be
avoided and handled properly instead.
This seems to further improve the babel test on the web-tooling-benchmark
by around 1%, because the OOB access no longer turns the otherwise
MONOMORPHIC access into MEGAMORPHIC state.
Bug: v8:6936, v8:7014
Change-Id: I9df03304e056d7001a65da8e9621119f8e9bb55b
Reviewed-on: https://chromium-review.googlesource.com/744022
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Camillo Bruni <cbruni@chromium.org>
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49049}
Following up on adding n-ary nodes, this extends the parser and
interpreter to support n-ary logical operations.
Bug: v8:6964
Bug: chromium:731861
Change-Id: Ife2141c389b9abccd917ab2aaddf399c436ef777
Reviewed-on: https://chromium-review.googlesource.com/735497
Reviewed-by: Adam Klein <adamk@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/master@{#49029}
We now represent the SameValue operation explicitly in TurboFan and the
operation can thus participate in all kinds of optimizations. Especially
we get rid of the JSCall node in the general case, which blocks several
optimizations across the call. The general, baseline performance is now
always on par with StrictEqual.
Once the StrictEqual operator is also a simplified operator, we should
start unifying the type based optimizations in SimplifiedLowering.
In the micro-benchmark we go from
testStrictEqual: 1422 ms.
testObjectIs: 1520 ms.
testManualSameValue: 1759 ms.
to
testStrictEqual: 1426 ms.
testObjectIs: 1357 ms.
testManualSameValue: 1766 ms.
which gives the expected result.
Bug: v8:7007
Change-Id: I0de3ff6ff6209ab4c3edb69de6a16e387295a9c8
Reviewed-on: https://chromium-review.googlesource.com/741228
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48994}
Expressions of the form
a_0 + a_1 + a_2 + a_3 + ... + a_n
seem to be reasonably common for cases such as building templates.
However, parsing these expressions results in a n-deep expression tree:
...
/
+
/ \
+ a_2
/ \
a_0 a_1
Traversing this tree during compilation can cause a stack overflow when n is
large.
Instead, for left-associate operations such as add, we now build up an
n-ary node in the parse tree, of the form
n-ary +
/ | \
/ | ... \
a_0 a_1 a_n
The bytecode compiler can now iterate through the child expressions
rather than recursing.
This patch only supports arithmetic operations -- subsequent patches
will enable the same optimization for logical tests and comma
expressions.
Bug: v8:6964
Bug: chromium:724961
Bug: chromium:731861
Bug: chromium:752081
Bug: chromium:771653
Bug: chromium:777302
Change-Id: Ie97e4ce42506fe62a7bc4ffbdaa90a9f698352cb
Reviewed-on: https://chromium-review.googlesource.com/733120
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48920}
When TurboFan sees a call to Reflect.get with exactly two parameters,
we can lower that to a direct call to the GetPropertyStub, which is
certainly faster than the general C++ builtin. This gives a nice
7-8% improvement on the chai test in the web-tooling-benchmark.
The micro-benchmark on the issue goes from
reflectGetPresent: 461 ms.
reflectGetAbsent: 470 ms.
to
reflectGetPresent: 141 ms.
reflectGetAbsent: 245 ms.
which is an up to 3.2x improvement.
Bug: v8:5996, v8:6936, v8:6937
Change-Id: Ic439fccb13f1a2f84386bf9fc31b4283d101afc4
Reviewed-on: https://chromium-review.googlesource.com/732988
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48841}
This adds a new InstanceOfIC where the TestInstanceOf bytecode collects
constant feedback about the right-hand side of instanceof operators,
including both JSFunction and JSBoundFunction instances. TurboFan then
uses the feedback to optimize instanceof in places where the right-hand
side is not a known constant (known to TurboFan).
This addresses the odd performance cliff that we see with instanceof in
functions with multiple closures. It was discovered as one of the main
bottlenecks on the uglify-es test in the web-tooling-benchmark. The
uglify-es test (run in separation) is ~18% faster with this change.
On the micro-benchmark in the tracking bug we go from
instanceofSingleClosure_Const: 69 ms.
instanceofSingleClosure_Class: 246 ms.
instanceofMultiClosure: 246 ms.
instanceofParameter: 246 ms.
to
instanceofSingleClosure_Const: 70 ms.
instanceofSingleClosure_Class: 75 ms.
instanceofMultiClosure: 76 ms.
instanceofParameter: 73 ms.
boosting performance by roughly 3.6x and thus effectively removing the
performance cliff around instanceof.
Bug: v8:6936, v8:6971
Change-Id: Ib88dbb9eaef9cafa4a0e260fbbde73427a54046e
Reviewed-on: https://chromium-review.googlesource.com/730686
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48820}
This revert is manual, but almost completely automatic.
It was just blocked by a single-line irrelevant refactoring change.
This reverts commit 1cee0e012e.
Reason for revert: chromium:776256
Original change's description:
> Reland^4 "[turbofan] eagerly prune None types and deadness from the graph"
>
> This fixes https://bugs.chromium.org/p/chromium/issues/detail?id=773954.
> The issue was that in the EffectControlLinearizer, the effect input of an
> {Unreachable} node was not updated, leaving a {Checkpoint} behind.
>
> This is a reland of 4cf476458f
> Original change's description:
> > Reland^3 "[turbofan] eagerly prune None types and deadness from the graph"
> >
> > This fixes the issues
> > https://bugs.chromium.org/p/chromium/issues/detail?id=772873
> > and https://bugs.chromium.org/p/chromium/issues/detail?id=772872.
> >
> > One problem was that mutating an effect node into Unreachable confused
> > the LoadElimination sidetables, so I just always create a new node now.
> >
> > The other problem was that UpdateBlockControl() was executed after
> > UpdateEffectPhi() in the lazy case. This reverted the update to the Merge input.
> > So now I make sure that UpdateEffectPhi() is always executed last.
> >
> > This is a reland of 6ddb5e7da7
> > Original change's description:
> > > Reland^2 "[turbofan] eagerly prune None types and deadness from the graph"
> > >
> > > Now, the EffectControlLinearizer connects all occurrences of Unreachable to the
> > > graph end. This fixes issues with later phases running DeadCodeElimination and
> > > introducing new DeadValue nodes when processing uses of Unreachable.
> > >
> > > This is a reland of 3c4bc27f13
> > > Original change's description:
> > > > Reland "[turbofan] eagerly prune None types and deadness from the graph"
> > > >
> > > > This is a reland of e1cdda2512
> > > > Original change's description:
> > > > > [turbofan] eagerly prune None types and deadness from the graph
> > > > >
> > > > > In addition to using the {Dead} node to prune dead control nodes and nodes that
> > > > > depend on them, we introduce a {DeadValue} node representing an impossible value
> > > > > that can occur at any position in the graph. The extended {DeadCodeElimination}
> > > > > prunes {DeadValue} and its uses, inserting a crashing {Unreachable} node into
> > > > > the effect chain when possible. The remaining uses of {DeadValue} are handled
> > > > > in {EffectControlLinearizer}, where we always have access to the effect chain.
> > > > > In addition to explicitly introduced {DeadValue} nodes, we consider any value use
> > > > > of a node with type {None} as dead.
> > > > >
> > > > > Bug: chromium:741225
> > > > > Change-Id: Icc4b636d1d018c452ba1a2fa7cd3e00e522f1655
> > > > > Reviewed-on: https://chromium-review.googlesource.com/641250
> > > > > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > > > > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > > > > Cr-Commit-Position: refs/heads/master@{#48208}
> > > >
> > > > Bug: chromium:741225
> > > > Change-Id: I21316913dae02864f7a6d7c9269405a79f054138
> > > > Reviewed-on: https://chromium-review.googlesource.com/692034
> > > > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > > > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > > > Cr-Commit-Position: refs/heads/master@{#48232}
> > >
> > > Bug: chromium:741225
> > > Change-Id: I5702ec34856c075717162153adc765774453c45f
> > > Reviewed-on: https://chromium-review.googlesource.com/702264
> > > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > > Cr-Commit-Position: refs/heads/master@{#48366}
> >
> > Bug: chromium:741225
> > Change-Id: I4054a694d2521c2e1f0c4a3ad0f3cf100b5c536f
> > Reviewed-on: https://chromium-review.googlesource.com/709214
> > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > Cr-Commit-Position: refs/heads/master@{#48469}
>
> Bug: chromium:741225
> Change-Id: Id9d4f3a3ae36cb3e38f80edcdba88efa7922ca24
> Reviewed-on: https://chromium-review.googlesource.com/715716
> Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#48660}
TBR=jarin@chromium.org,tebbi@chromium.org,bmeurer@chromium.org
Bug: chromium:741225 chromium:776256
Change-Id: Iaf2af3cb6dea5fdece43297cb9d987e7decc726d
Reviewed-on: https://chromium-review.googlesource.com/727804
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48749}
This addresses the odd performance cliff, where the CallIC tracks known
JSFunction targets, but goes MEGAMORPHIC when it sees a JSBoundFunction
target. With this fix in place the micro-benchmark on the bug goes from
arrowCall: 82 ms.
boundCall: 234 ms.
to
arrowCall: 81 ms.
boundCall: 80 ms.
so Function#bind doesn't cause any additional overhead anymore.
Bug: v8:5267, v8:6962
Change-Id: Iaceaf89fd3e99e2afe2ae45e96a6813a3ef8b1d2
Reviewed-on: https://chromium-review.googlesource.com/727879
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48722}
So far the JSCallReducer was only able to unfold constant
JSBoundFunction targets for JSCall nodes, which is not the
common case. With the introduction of JSCreateBoundFunction
operator earlier, we can now also recognize calls to bound
functions where the bind happens earlier in the function,
i.e. as the example of
a.map(f.bind(self))
in https://twitter.com/BenLesh/status/920700003974123520, which
is a handy way to use Function#bind. So this transformation
takes a node like
JSCall(JSCreateBoundFunction(bound_target_function,
bound_this,
a1,...,aN),
receiver, p1,...,pM)
and turns that into
JSCall(bound_target_function, bound_this, a1,...,aN,p1,...,pM)
allowing TurboFan to further inline the bound_target_function
at this call site if that's also inlinable (i.e. it's a known
constant JSFunction or the result of a JSCreateClosure call).
This improves the micro-benchmark from
arrowCall: 55 ms.
boundCall: 221 ms.
arrowMap: 181 ms.
boundMap: 806 ms.
to
arrowCall: 71 ms.
boundCall: 76 ms.
arrowMap: 188 ms.
boundMap: 186 ms.
so that Function#bind in this case is as fast as using closures,
which is an up to 4.3x improvement in the Array#map example.
Bug: v8:5257, v8:6961
Change-Id: Ibca650faad912bf9db1db6fbc48772e7551289a6
Reviewed-on: https://chromium-review.googlesource.com/727799
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48713}
This fixes https://bugs.chromium.org/p/chromium/issues/detail?id=773954.
The issue was that in the EffectControlLinearizer, the effect input of an
{Unreachable} node was not updated, leaving a {Checkpoint} behind.
This is a reland of 4cf476458f
Original change's description:
> Reland^3 "[turbofan] eagerly prune None types and deadness from the graph"
>
> This fixes the issues
> https://bugs.chromium.org/p/chromium/issues/detail?id=772873
> and https://bugs.chromium.org/p/chromium/issues/detail?id=772872.
>
> One problem was that mutating an effect node into Unreachable confused
> the LoadElimination sidetables, so I just always create a new node now.
>
> The other problem was that UpdateBlockControl() was executed after
> UpdateEffectPhi() in the lazy case. This reverted the update to the Merge input.
> So now I make sure that UpdateEffectPhi() is always executed last.
>
> This is a reland of 6ddb5e7da7
> Original change's description:
> > Reland^2 "[turbofan] eagerly prune None types and deadness from the graph"
> >
> > Now, the EffectControlLinearizer connects all occurrences of Unreachable to the
> > graph end. This fixes issues with later phases running DeadCodeElimination and
> > introducing new DeadValue nodes when processing uses of Unreachable.
> >
> > This is a reland of 3c4bc27f13
> > Original change's description:
> > > Reland "[turbofan] eagerly prune None types and deadness from the graph"
> > >
> > > This is a reland of e1cdda2512
> > > Original change's description:
> > > > [turbofan] eagerly prune None types and deadness from the graph
> > > >
> > > > In addition to using the {Dead} node to prune dead control nodes and nodes that
> > > > depend on them, we introduce a {DeadValue} node representing an impossible value
> > > > that can occur at any position in the graph. The extended {DeadCodeElimination}
> > > > prunes {DeadValue} and its uses, inserting a crashing {Unreachable} node into
> > > > the effect chain when possible. The remaining uses of {DeadValue} are handled
> > > > in {EffectControlLinearizer}, where we always have access to the effect chain.
> > > > In addition to explicitly introduced {DeadValue} nodes, we consider any value use
> > > > of a node with type {None} as dead.
> > > >
> > > > Bug: chromium:741225
> > > > Change-Id: Icc4b636d1d018c452ba1a2fa7cd3e00e522f1655
> > > > Reviewed-on: https://chromium-review.googlesource.com/641250
> > > > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > > > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > > > Cr-Commit-Position: refs/heads/master@{#48208}
> > >
> > > Bug: chromium:741225
> > > Change-Id: I21316913dae02864f7a6d7c9269405a79f054138
> > > Reviewed-on: https://chromium-review.googlesource.com/692034
> > > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > > Cr-Commit-Position: refs/heads/master@{#48232}
> >
> > Bug: chromium:741225
> > Change-Id: I5702ec34856c075717162153adc765774453c45f
> > Reviewed-on: https://chromium-review.googlesource.com/702264
> > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > Cr-Commit-Position: refs/heads/master@{#48366}
>
> Bug: chromium:741225
> Change-Id: I4054a694d2521c2e1f0c4a3ad0f3cf100b5c536f
> Reviewed-on: https://chromium-review.googlesource.com/709214
> Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#48469}
Bug: chromium:741225
Change-Id: Id9d4f3a3ae36cb3e38f80edcdba88efa7922ca24
Reviewed-on: https://chromium-review.googlesource.com/715716
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48660}
Ensure we only lower SpeculativeNumberBinops to a pure operator for
non-string plain primitives. Previously we could lower if a value might be
the-hole, however this would fail a CHECK in ConvertInputsToNumber which
expects a plain primitive.
BUG=chromium:772420
Change-Id: I0c755d10db7afd9cabfb638eca5662d70dfc8d51
Reviewed-on: https://chromium-review.googlesource.com/715717
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48649}
OSR for functions which use arguments no longer needs to be disabled, since
TurboFan handles the case.
Bug:
Change-Id: I121f1190a142c18f113bd5f875e258812645c43f
Reviewed-on: https://chromium-review.googlesource.com/721661
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Toon Verwaest <verwaest@chromium.org>
Commit-Queue: Marja Hölttä <marja@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48631}
Port the baseline version of Reflect.has to the CodeStubAssembler and
reuse the existing logic for HasProperty (i.e. the HasProperty builtin).
Also inline the Reflect.has builtin into TurboFan, by adding a check
on the target in front of a use of the JSHasProperty operator.
Technically this additional check is not necessary, because the
JSHasProperty operator already throws if the target is not a JSReceiver,
but the exception message is confusing then.
This improves the performance of the micro-benchmark from
reflectHasPresent: 337 ms.
reflectHasAbsent: 472 ms.
to
reflectHasPresent: 121 ms.
reflectHasAbsent: 216 ms.
which is a nice 2.8x improvement in the best case. It also improves the
chai test on the web-tooling-benchmark by around 1-2%, which is roughly
the expected win (since Reflect.has overall accounts for around 3-4%).
Bug: v8:5996, v8:6936, v8:6937
Change-Id: I856183229677a71c19936f06f2a4fc7a794a9a4a
Reviewed-on: https://chromium-review.googlesource.com/720959
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48608}
This reverts commit 4cf476458f.
Reason for revert: Broken effect chains detected by Clusterfuzz. Playing it safe for the 63 branch.
Original change's description:
> Reland^3 "[turbofan] eagerly prune None types and deadness from the graph"
>
> This fixes the issues
> https://bugs.chromium.org/p/chromium/issues/detail?id=772873
> and https://bugs.chromium.org/p/chromium/issues/detail?id=772872.
>
> One problem was that mutating an effect node into Unreachable confused
> the LoadElimination sidetables, so I just always create a new node now.
>
> The other problem was that UpdateBlockControl() was executed after
> UpdateEffectPhi() in the lazy case. This reverted the update to the Merge input.
> So now I make sure that UpdateEffectPhi() is always executed last.
>
> This is a reland of 6ddb5e7da7
> Original change's description:
> > Reland^2 "[turbofan] eagerly prune None types and deadness from the graph"
> >
> > Now, the EffectControlLinearizer connects all occurrences of Unreachable to the
> > graph end. This fixes issues with later phases running DeadCodeElimination and
> > introducing new DeadValue nodes when processing uses of Unreachable.
> >
> > This is a reland of 3c4bc27f13
> > Original change's description:
> > > Reland "[turbofan] eagerly prune None types and deadness from the graph"
> > >
> > > This is a reland of e1cdda2512
> > > Original change's description:
> > > > [turbofan] eagerly prune None types and deadness from the graph
> > > >
> > > > In addition to using the {Dead} node to prune dead control nodes and nodes that
> > > > depend on them, we introduce a {DeadValue} node representing an impossible value
> > > > that can occur at any position in the graph. The extended {DeadCodeElimination}
> > > > prunes {DeadValue} and its uses, inserting a crashing {Unreachable} node into
> > > > the effect chain when possible. The remaining uses of {DeadValue} are handled
> > > > in {EffectControlLinearizer}, where we always have access to the effect chain.
> > > > In addition to explicitly introduced {DeadValue} nodes, we consider any value use
> > > > of a node with type {None} as dead.
> > > >
> > > > Bug: chromium:741225
> > > > Change-Id: Icc4b636d1d018c452ba1a2fa7cd3e00e522f1655
> > > > Reviewed-on: https://chromium-review.googlesource.com/641250
> > > > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > > > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > > > Cr-Commit-Position: refs/heads/master@{#48208}
> > >
> > > Bug: chromium:741225
> > > Change-Id: I21316913dae02864f7a6d7c9269405a79f054138
> > > Reviewed-on: https://chromium-review.googlesource.com/692034
> > > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > > Cr-Commit-Position: refs/heads/master@{#48232}
> >
> > Bug: chromium:741225
> > Change-Id: I5702ec34856c075717162153adc765774453c45f
> > Reviewed-on: https://chromium-review.googlesource.com/702264
> > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > Cr-Commit-Position: refs/heads/master@{#48366}
>
> Bug: chromium:741225
> Change-Id: I4054a694d2521c2e1f0c4a3ad0f3cf100b5c536f
> Reviewed-on: https://chromium-review.googlesource.com/709214
> Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#48469}
TBR=jarin@chromium.org,tebbi@chromium.org
Change-Id: Icf6a6af4feaafd4bde28cb7b996735ff91bb3810
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: chromium:741225
Reviewed-on: https://chromium-review.googlesource.com/715096
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48482}
This fixes the issues
https://bugs.chromium.org/p/chromium/issues/detail?id=772873
and https://bugs.chromium.org/p/chromium/issues/detail?id=772872.
One problem was that mutating an effect node into Unreachable confused
the LoadElimination sidetables, so I just always create a new node now.
The other problem was that UpdateBlockControl() was executed after
UpdateEffectPhi() in the lazy case. This reverted the update to the Merge input.
So now I make sure that UpdateEffectPhi() is always executed last.
This is a reland of 6ddb5e7da7
Original change's description:
> Reland^2 "[turbofan] eagerly prune None types and deadness from the graph"
>
> Now, the EffectControlLinearizer connects all occurrences of Unreachable to the
> graph end. This fixes issues with later phases running DeadCodeElimination and
> introducing new DeadValue nodes when processing uses of Unreachable.
>
> This is a reland of 3c4bc27f13
> Original change's description:
> > Reland "[turbofan] eagerly prune None types and deadness from the graph"
> >
> > This is a reland of e1cdda2512
> > Original change's description:
> > > [turbofan] eagerly prune None types and deadness from the graph
> > >
> > > In addition to using the {Dead} node to prune dead control nodes and nodes that
> > > depend on them, we introduce a {DeadValue} node representing an impossible value
> > > that can occur at any position in the graph. The extended {DeadCodeElimination}
> > > prunes {DeadValue} and its uses, inserting a crashing {Unreachable} node into
> > > the effect chain when possible. The remaining uses of {DeadValue} are handled
> > > in {EffectControlLinearizer}, where we always have access to the effect chain.
> > > In addition to explicitly introduced {DeadValue} nodes, we consider any value use
> > > of a node with type {None} as dead.
> > >
> > > Bug: chromium:741225
> > > Change-Id: Icc4b636d1d018c452ba1a2fa7cd3e00e522f1655
> > > Reviewed-on: https://chromium-review.googlesource.com/641250
> > > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > > Cr-Commit-Position: refs/heads/master@{#48208}
> >
> > Bug: chromium:741225
> > Change-Id: I21316913dae02864f7a6d7c9269405a79f054138
> > Reviewed-on: https://chromium-review.googlesource.com/692034
> > Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> > Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> > Cr-Commit-Position: refs/heads/master@{#48232}
>
> Bug: chromium:741225
> Change-Id: I5702ec34856c075717162153adc765774453c45f
> Reviewed-on: https://chromium-review.googlesource.com/702264
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#48366}
Bug: chromium:741225
Change-Id: I4054a694d2521c2e1f0c4a3ad0f3cf100b5c536f
Reviewed-on: https://chromium-review.googlesource.com/709214
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48469}
The Object.is builtin provides an entry point to the abstract operation
SameValue, which properly distinguishes -0 and 0, and also identifies
NaNs. Most of the time you don't need these, but rather just regular
strict equality, but when you do, Object.is(o, -0) is the most readable
way to check for minus zero.
This is for example used in Node.js by formatNumber to properly print -0
for negative zero. However since the builtin thus far implemented as C++
builtin and TurboFan didn't know anything about it, Node.js considering
to go with a more performant, less readable version (which also makes
assumptions about the input value) in
https://github.com/nodejs/node/pull/15726
until the performance of Object.is will be on par (so hopefully we can
go back to Object.is in Node 9).
This CL ports the baseline implementation of Object.is to CSA, which
is pretty straight-forward since SameValue is already available in
CodeStubAssembler, and inlines a few interesting cases into TurboFan,
i.e. comparing same SSA node, and checking for -0 and NaN explicitly.
On the micro-benchmarks we go from
testNumberIsMinusZero: 1000 ms.
testObjectIsMinusZero: 929 ms.
testObjectIsNaN: 954 ms.
testObjectIsSame: 793 ms.
testStrictEqualSame: 104 ms.
to
testNumberIsMinusZero: 89 ms.
testObjectIsMinusZero: 88 ms.
testObjectIsNaN: 88 ms.
testObjectIsSame: 86 ms.
testStrictEqualSame: 105 ms.
which is a nice 10x to 11x improvement and brings Object.is on par with
strict equality for most cases.
Drive-by-fix: Also refactor and optimize the SameValue check in the
CodeStubAssembler to avoid code bloat (by not inlining StrictEqual
into every user of SameValue, and also avoiding useless checks).
Bug: v8:6882
Change-Id: Ibffd8c36511f219fcce0d89ed4e1073f5d6c6344
Reviewed-on: https://chromium-review.googlesource.com/700254
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48275}
Until now keyed accesses to properties with string or symbol keys were
only optimized properly while the IC was monomorphic and would go
megamorphic as soon as there's another receiver map, even if the name
was still the same (i.e. the same symbol or internalized string). This
was a weird performance-cliff, that'll hurt modern code especially
because for symbols you can only access them via keyed loads and stores.
This CL fixes the state machine inside the ICs to properly transition to
POLYMORPHIC state (and stay there) as long as the new name matches the
previously recorded name. The FeedbackVector and TurboFan were already
able to deal with this and didn't need any updates.
On the micro-benchmark from the tracking bug we go from
testStringMonomorphic: 429 ms.
testSymbolMonomorphic: 431 ms.
testStringPolymorphic: 429 ms.
testSymbolPolymorphic: 5621 ms.
to
testStringMonomorphic: 429 ms.
testSymbolMonomorphic: 429 ms.
testStringPolymorphic: 429 ms.
testSymbolPolymorphic: 430 ms.
effectively eliminating the overhead for symbols completely, and
yielding a 13.5x performance boost.
This also seems to yield a 1% improvement on the ARES6 ML benchmark,
because it eliminates the KEYED_LOAD_ICs for the Symbol.species lookups.
Bug: v8:6367, v8:6278, v8:6344
Change-Id: I879fe56387b4c56203c1ad8ef8cafb6cc4c32897
Reviewed-on: https://chromium-review.googlesource.com/695108
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48261}
The TypedArray.prototype[Symbol.toStringTag] getter is currently the best (and
as far as I can tell only definitely side-effect free) way to check whether an
arbitrary object is a TypedArray - either generally TypedArray or a specific
one like Uint8Array. Using the getter is thus emerging as the general pattern
to detect TypedArrays, even Node.js now adapted it starting with
https://github.com/nodejs/node/pull/15663
for the isTypedArray and isUint8Array type checks in lib/internal/util/types.js
now.
The getter returns either the string with the TypedArray subclass name
(i.e. "Uint8Array") or undefined if the receiver is not a TypedArray.
This can be implemented with a simple elements kind dispatch, instead of
checking the instance type and then loading the class name from the
constructor, which requires a loop walking up the transition tree. This
CL ports the builtin to CSA and TurboFan, and changes the logic to a
simple elements kind check. On the micro-benchmark mentioned in the
referenced bug, the time goes from
testIsArrayBufferView: 565 ms.
testIsTypedArray: 2403 ms.
testIsUint8Array: 3847 ms.
to
testIsArrayBufferView: 566 ms.
testIsTypedArray: 965 ms.
testIsUint8Array: 965 ms.
which presents an up to 4x improvement.
Bug: v8:6874
Change-Id: I9c330b4529d9631df2f052acf023c6a4fae69611
Reviewed-on: https://chromium-review.googlesource.com/695021
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48254}
This is a reland of 9d3c4b4b91
Original change's description:
> [turbofan] Implement lowering of {JSCreateClosure}.
>
> This adds support for inline allocation of {JSFunction} objects as part
> of closures instantiation for {JSCreateClosure} nodes. The lowering is
> limited to instantiation sites which have already seen more than one
> previous instantiation, this avoids the need to increment the respective
> counter.
>
> R=jarin@chromium.org
>
> Change-Id: I462c557453fe58bc5f09020a3d5ebdf11c2ea68b
> Reviewed-on: https://chromium-review.googlesource.com/594287
> Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#48176}
Change-Id: I3ec3880bea89798a34a3878e6122b95db1014151
Reviewed-on: https://chromium-review.googlesource.com/686834
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48198}
This reverts commit 9d3c4b4b91.
Reason for revert: Breaks cctest/test-debug/NoBreakWhenBootstrapping in no-snap mode.
Original change's description:
> [turbofan] Implement lowering of {JSCreateClosure}.
>
> This adds support for inline allocation of {JSFunction} objects as part
> of closures instantiation for {JSCreateClosure} nodes. The lowering is
> limited to instantiation sites which have already seen more than one
> previous instantiation, this avoids the need to increment the respective
> counter.
>
> R=jarin@chromium.org
>
> Change-Id: I462c557453fe58bc5f09020a3d5ebdf11c2ea68b
> Reviewed-on: https://chromium-review.googlesource.com/594287
> Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
> Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#48176}
TBR=mstarzinger@chromium.org,jarin@chromium.org
Change-Id: Id52281f6a3c0b7c2603053ecf002777d5b0d6f1f
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Reviewed-on: https://chromium-review.googlesource.com/686534
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48178}
This adds support for inline allocation of {JSFunction} objects as part
of closures instantiation for {JSCreateClosure} nodes. The lowering is
limited to instantiation sites which have already seen more than one
previous instantiation, this avoids the need to increment the respective
counter.
R=jarin@chromium.org
Change-Id: I462c557453fe58bc5f09020a3d5ebdf11c2ea68b
Reviewed-on: https://chromium-review.googlesource.com/594287
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#48176}
SetForceInlineFlag is now only used in tests. Earlier, it was also used
in js builtins, because unless this flag was specified the js builtins
were not inlined. All the performance critical js builtins are moved
to turbofan builtins and SetForceInlineFlag is no longer used. We would
like to remove this flag completely to simplify inlining heuristics.
Also, this uses a bit on the SharedFuntionInfo.
Bug: v8:6682
Change-Id: I19afd27381afc212f29179f2c5477095c8174f39
Reviewed-on: https://chromium-review.googlesource.com/660739
Commit-Queue: Mythri Alle <mythria@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47997}
So far we didn't properly constant-fold JSToString operators in
JSTypedLowering where the input was a known number constant.
Bug: v8:6815
Change-Id: Iac87346b7d38f0f75461f285ea7daa2d5a5e1524
Reviewed-on: https://chromium-review.googlesource.com/663358
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47972}
When accessing elements of a global (constant) JSArray, whose backing
store is copy-on-write, we can just constant-fold the value and insert
a check that the backing store stays the same.
Bug: v8:6816, v8:6815
Change-Id: I090bcec7b1ce72a1f9ed8625680ed91e8c67f27f
Reviewed-on: https://chromium-review.googlesource.com/662757
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47963}
TurboFan wasn't able to inline calls to Array.prototype.push which
didn't have exactly one parameter. This was a rather artifical
limitation and was mostly due to the way the MaybeGrowFastElements
operator was implemented (which was not ideal by itself). Refactoring
this a bit, allows us to inline the operation in general, independent
of the number of values to push.
Array#push with multiple parameters is used quite a lot inside Ember (as
discovered by Apple, i.e. https://bugs.webkit.org/show_bug.cgi?id=175823)
and is also dominating the Six-Speed/SpreadLiterals/ES5 benchmark (see
https://twitter.com/SpiderMonkeyJS/status/906528938452832257 from the
SpiderMonkey folks). The micro-benchmark mentioned in the tracking bug
(v8:6808) improves from
arrayPush0: 2422 ms.
arrayPush1: 2567 ms.
arrayPush2: 4092 ms.
arrayPush3: 4308 ms.
to
arrayPush0: 798 ms.
arrayPush1: 2563 ms.
arrayPush2: 2623 ms.
arrayPush3: 2773 ms.
with this change, effectively removing the odd 50-60% performance
cliff that was associated with going from one parameter to two or
more.
Bug: v8:2229, v8:6808
Change-Id: Iffe4c1233903c04c3dc2062aad39d99769c8ab57
Reviewed-on: https://chromium-review.googlesource.com/657582
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47940}
The previous %StringCharCodeAt runtime entry (and the inlined intrinsic)
are obsolete and not used anymore (except in dedicated tests for this
runtime function), so remove it. And rename the %StringCharCodeAtRT
function, which is actually used to %StringCharCodeAt instead to have
a consistent naming scheme for runtime fallbacks.
Bug: v8:5049
Change-Id: I619429ef54f6efea61fc51ab9ed1d5cfe4417f99
Reviewed-on: https://chromium-review.googlesource.com/657719
Commit-Queue: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Yang Guo <yangguo@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47928}
The advantage of an explicit Abort that the interpreter and the compiler know
that aborting cannot continue or throw or deopt. As a result we generate less
code and we do not confuse the compiler if the environment is not set up for
throwing (as in the generator dispatch that fails validation in
crbug.com/762057).
Bug: chromium:762057
Change-Id: I3e88f78be32f31ac49b1845595255f802c405ed7
Reviewed-on: https://chromium-review.googlesource.com/657025
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47922}
Add support to the JSCallReducer to recognize JSConstruct nodes where
the target is the Object constructor, and reduce them to JSCreate
nodes if either
(a) no value is passed to the Object constructor, or
(b) the target and new.target are definitely not identical, by checking
whether both target and new.target are different HeapConstants
(if they are not, then the JSCreateLowering will not be able to
do a lot with the JSCreate anyways).
This should cover the relevant cases for subclassing appropriately. It
fixes the 3-4x slowdown on the micro-benchmark mentioned in the linked
bug,
baseNoExtends: 752 ms.
baseExtendsObject: 752 ms.
baseExtendsViaFactory: 751 ms.
and thus removes the performance cliff.
R=jarin@chromium.org
Bug: v8:6801
Change-Id: Id265fd1399302a67b5790a6d0156679920c58bdd
Reviewed-on: https://chromium-review.googlesource.com/657019
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47913}
We emitted rotation by 24 bits with bitwise and, but that is wrong
because the low 8 bits can wrap around and "leak" into the result.
Bug: chromium:739902
Change-Id: Id49251e89405afb1581b8c60cde808c2d8bf693d
Reviewed-on: https://chromium-review.googlesource.com/645848
Reviewed-by: Martyn Capewell <martyn.capewell@arm.com>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47746}
When calling
Object(value)
where the value is known to be a JSReceiver, we can just replace it with
value, as the Object constructor call is a no-op in that case. Otherwise
when value is known to be not null or undefined then we can replace the
Object constructor call with an invocation of ToObject.
This covers the common pattern found in bundles generated by Webpack,
where the Object constructor is used to call imported functions, i.e.
Object(module.foo)(1, 2, 3)
There's a lot of detail in https://github.com/webpack/webpack/issues/5600
on this matter and why this pattern was chosen.
Bug: v8:6772
Change-Id: I2b4f0b4542b68b97b337ce571d6d79946c73d8bb
Reviewed-on: https://chromium-review.googlesource.com/643868
Reviewed-by: Yang Guo <yangguo@chromium.org>
Commit-Queue: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47728}
This change prevents constant folding of uninhabited RefenceEqual node
because that could widen a type (from None type to the type of the
boolean constant).
Hopefully, this is a temporary workaround that will be replaced
by a better dead code elimination.
Bug: v8:6631
Change-Id: Ie25e7d710aaf1d37c9adba60f92438570843dd5d
Reviewed-on: https://chromium-review.googlesource.com/627916
Reviewed-by: Tobias Tebbi <tebbi@chromium.org>
Commit-Queue: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47545}
This is in preparation to the removal of the FullCodeGenerator, we no
longer need the ability to stress the underlying implementation.
R=rmcilroy@chromium.org
BUG=v8:6409
Cq-Include-Trybots: master.tryserver.v8:v8_linux_noi18n_rel_ng
Change-Id: Iad3177d6de4a68b57c12a770b6e85ed7a9710254
Reviewed-on: https://chromium-review.googlesource.com/584747
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#47276}
With TurboFan, there should no longer be any deopt loops (aside from
bugs). So, the "too many deopts" bailout is no longer needed, at least
in its current form.
This fixes an issue where deopt counts are leaked between native
contexts, resulting in optimization being disabled unnecessarily.
Bug: v8:6402
Change-Id: Ia06374ae6b5c2d473bcdd8eef1284bf02766c2fb
Reviewed-on: https://chromium-review.googlesource.com/588894
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46961}
Adding (very) small tests for deoptimization.
Some of these tests were failing when the safepoints were not found,
after setting the return address.
BUG=V8:6563
Change-Id: I3af36b193a5982cd73414cc1884c5f0a7a727f5a
Reviewed-on: https://chromium-review.googlesource.com/584751
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Juliana Patricia Vicente Franco <jupvfranco@google.com>
Cr-Commit-Position: refs/heads/master@{#46922}
Reland of https://chromium-review.googlesource.com/c/543042/.
Now the OSR phase is only used when OSRing from the ast graph builder.
When OSRing from Turbofan, the implementation is now in the graph
building phase, at the beginning of the VisitBytecode function.
We are no longer generating any OSRLoopEntry or OSRNormalEntry nodes,
nor nodes for the possible code of the OSRed function which is before
the OSRed loops.
The trimming and reducing of the OSR phase is not done either. This
change in the way the way the OSR is done enabled to remove the
workaround to the bug mentioned below.
Bug: v8:6112
Bug: v8:6518
Change-Id: Ia02f2138f54fc79cab2f02fed68d9bb522d6ce14
Reviewed-on: https://chromium-review.googlesource.com/584756
Commit-Queue: Alexandre Talon <alexandret@google.com>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46899}
This reverts commit 69c8f16da7.
Reason for revert: Causing crashes on Clusterfuzz - http://crbug.com/747154
BUG=chromium:747154
Original change's description:
> [Turbofan] Merged the OSR phase into the graph building phase.
>
> Now the OSR phase is only used when OSRing from the ast graph builder.
> When OSRing from Turbofan, the implementation is now in the graph
> building phase, at the beginning of the VisitBytecode function.
> We are no longer generating any OSRLoopEntry or OSRNormalEntry nodes,
> nor nodes for the possible code of the OSRed function which is before
> the OSRed loops.
>
> The trimming and reducing of the OSR phase is not done either. This
> change in the way the way the OSR is done enabled to remove the
> workaround to the bug mentioned below.
>
> Bug: v8:6112
> Bug: v8:6518
> Change-Id: I1c9231810b923486d55ea618d550d981d695d797
> Reviewed-on: https://chromium-review.googlesource.com/543042
> Commit-Queue: Alexandre Talon <alexandret@google.com>
> Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
> Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#46801}
TBR=rmcilroy@chromium.org,mstarzinger@chromium.org,leszeks@chromium.org,alexandret@google.com
Change-Id: Ifa9bf5d86e888a47cad7fb10446b36fda5029604
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Bug: v8:6112, v8:6518
Reviewed-on: https://chromium-review.googlesource.com/581288
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46817}
Now the OSR phase is only used when OSRing from the ast graph builder.
When OSRing from Turbofan, the implementation is now in the graph
building phase, at the beginning of the VisitBytecode function.
We are no longer generating any OSRLoopEntry or OSRNormalEntry nodes,
nor nodes for the possible code of the OSRed function which is before
the OSRed loops.
The trimming and reducing of the OSR phase is not done either. This
change in the way the way the OSR is done enabled to remove the
workaround to the bug mentioned below.
Bug: v8:6112
Bug: v8:6518
Change-Id: I1c9231810b923486d55ea618d550d981d695d797
Reviewed-on: https://chromium-review.googlesource.com/543042
Commit-Queue: Alexandre Talon <alexandret@google.com>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46801}
Inlining heuristics in Turbofan used ast node count. Bytecode size
is a better approximation of the size of the graph than the
ast node count. This cl changes the heuristics to use the bytecode
size instead. Also removing the ast_node_count filed in the shared
function info. It was used only for the inlining heuristics.
Also removed the max_inlined_source_size flag which is no longer used.
Bug:
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng
Change-Id: I8a2d2509c8e8d2779b33b817bb217de203d54ec3
Reviewed-on: https://chromium-review.googlesource.com/570055
Commit-Queue: Mythri Alle <mythria@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46771}
There remained a few of regressions and we didn't see any significant
improvement in the real world with this turned on. This CL reverts all the
StringConcat bytecode work which landed.
BUG=v8:6243
Change-Id: I832eb72e880ad41411dbec8fe29f71ef0f2025c8
Reviewed-on: https://chromium-review.googlesource.com/575130
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46769}
The tail call implementation is hidden behind the --harmony-tailcalls
flag, which is off-by-default (and has been unstaged since February).
It is known to be broken in a variety of cases, including clusterfuzz
security issues (see sample Chromium issues below). To avoid letting
the implementation bitrot further on trunk, this patch removes it.
Bug: v8:4698, chromium:636914, chromium:724746
Cq-Include-Trybots: master.tryserver.chromium.linux:linux_chromium_rel_ng;master.tryserver.v8:v8_linux_noi18n_rel_ng
Change-Id: I9cb547101456a582374fdf7b1a3f044a9ef33e5c
Reviewed-on: https://chromium-review.googlesource.com/569069
Commit-Queue: Adam Klein <adamk@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46651}
These are no longer necessary since we only have one optimizing compiler.
Also avoid changing --turbo-filter when --no-opt is set, and instead
explicitly check both the FLAG_opt and FLAG_turbo_filter in
GetOptimizedCode to check whether optimization is disabled.
BUG=v8:6408
Change-Id: I0948f788e8ff111c08022270d86c22f848da300a
Reviewed-on: https://chromium-review.googlesource.com/568484
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46648}
Removes the --ignition flag which is now on by default. Adds a
--stress-fullcodegen flag which enables running all functions supported
by fullcodegen to be compiled by fullcodegen.
This will enable moving parser internalization later when we are not
stressing fullcodegen or compiling asm.js functions.
BUG=v8:5203, v8:6409, v8:6589
Change-Id: I7fa68016d4e734755434ec0b4e749ef65ffa7f4e
Reviewed-on: https://chromium-review.googlesource.com/565569
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46635}
By creating the boilerplate only on the second instantiation we cannot
propagate back the elements transitions early enough. The resulting literals
would change the initial ElementsKind one step too late and already pollute
ICs that went to monomorphic state.
- Disable lazy AllocationSites for literals containing arrays
- Introduce new ComplexLiteral class to share code between ObjectLiteral
and ArrayLiteral
- RegexpLiteral now no longer needs a depth_ field
Bug: v8:6517, v8:6519, v8:6211
Change-Id: Ia88d1878954e8895c3d00a7dda8d71e95bba005c
Reviewed-on: https://chromium-review.googlesource.com/563305
Reviewed-by: Adam Klein <adamk@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Camillo Bruni <cbruni@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46603}
Adds constant folding for the StringConcat bytecode to
NativeContextSpecialization. Can reduce operator to either a fully folded
constant string, or a JSAdd or a StringConcat with a reduced number of
operators.
BUG=v8:6243, chromium:738312
Change-Id: I6b2be6a3d95230a23f3c7390a4f7be5181c49a2a
Reviewed-on: https://chromium-review.googlesource.com/559146
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46461}
This mostly reverts commit c503b80595 but fixes
an issue where literals would always be pretenured on first instantiation.
As a cleanup we pass in a PretenureFlag instead of using the FeedbackVector as
indicator.
Bug: v8:6211
Change-Id: Id328552620e33f5083519bcba1e24396d162d516
Reviewed-on: https://chromium-review.googlesource.com/555670
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Camillo Bruni <cbruni@chromium.org>
Commit-Queue: Camillo Bruni <cbruni@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46342}
This adds support for lowering of nodes having the {JSToObject} operator
even if they have exceptional control projections (e.g. are inside of a
try-block).
R=bmeurer@chromium.org
TEST=mjsunit/compiler/optimized-with
Change-Id: I711ff4935db68c43243a971a8b21989487c86317
Reviewed-on: https://chromium-review.googlesource.com/554628
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46318}
This adds support for lowering of nodes having the {JSCreateArray}
operator even if they have exceptional control projections (e.g. are
placed inside a try-block).
R=mvstanton@chromium.org
TEST=mjsunit/compiler/array-constructor
Change-Id: I2fe34dbb3729b4763471f2638a960b01c531c038
Reviewed-on: https://chromium-review.googlesource.com/554732
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46315}
Adds a CheckString to all operand inputs of JSStringConcat. The operands are
already known to be strings, so this will get eliminated in almost all cases,
however, if there is a yield within the concatenation then we lose the
knowledge that the previous operands are strings since the values are loaded
from the generator object. Adds a test for this case.
BUG=v8:6243
Change-Id: I1601a316e6efbed1c53486f1027cb0ea023ff030
Reviewed-on: https://chromium-review.googlesource.com/549301
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46243}
This removes the --turbo flag and solely relies on the filter pattern
provided via --turbo-filter when deciding whether to use TurboFan. Note
that disabling optimization wholesale can still be done with --no-opt,
which should be used in favor of --no-turbo everywhere.
Also note that this contains semantic changes to the TurboFan activation
criteria. We respect the filter pattern more stringently and no longer
activate TurboFan just because the source contains patterns forcing use
of Ignition via {AstNumberingVisitor::DisableFullCodegenAndCrankshaft}.
R=rmcilroy@chromium.org
BUG=v8:6408
Change-Id: I0c855f6a62350eb62283a3431c8cc1baa750950e
Reviewed-on: https://chromium-review.googlesource.com/528121
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Michael Stanton <mvstanton@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46167}
Adds typed lowering of JSStringConcat to ConsString allocation if the
following conditions hold:
- All concatinations will result in a ConsString of >= ConString::kMinLength
- No concatinations will result in a empty string in the RHS unless there is
a sequential string in the LHS.
This also means JSStringConcat needs an eager checkpoint since it can
deopt if throwing a RangeError when the string length protector is valid.
BUG=v8:6243
Change-Id: I01ca79f884df467c10f2c032c72d51b5199c1a3c
Reviewed-on: https://chromium-review.googlesource.com/526636
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46093}
Add a new JSConstructWithArrayLike operator that is backed by the
ConstructWithArrayLike builtin (similar to what was done before
for the JSCallWithArrayLike operator), and use that operator to
optimize Reflect.construct inlining in TurboFan. This is handled
uniformly with JSConstructWithSpread in the JSCallReducer.
Also add missing test coverage for Reflect.construct in optimized
code, especially for some interesting corner cases.
R=petermarshall@chromium.org
BUG=v8:4587,v8:5269
Review-Url: https://codereview.chromium.org/2949813002
Cr-Commit-Position: refs/heads/master@{#46087}
Add a new JSCallWithArrayLike operator that is backed by the
CallWithArrayLike builtin, and use that operator for both
Function.prototype.apply and Reflect.apply inlining. Also unify
the handling of JSCallWithArrayLike and JSCallWithSpread in
the JSCallReducer to reduce the copy&paste overhead.
Drive-by-fix: Add a lot of test coverage for Reflect.apply and
Function.prototype.apply in optimized code, especially for some
corner cases, which was missing so far.
BUG=v8:4587,v8:5269
R=petermarshall@chromium.org
Review-Url: https://codereview.chromium.org/2950773002
Cr-Commit-Position: refs/heads/master@{#46041}
We can remove a lot of native code and rely on CallOrConstructVarargs
to do the stack manipulation for us.
This will also take advantage of the fast-path for double arrays in
CallOrConstructDoubleVarargs.
We can also remove Runtime_SpreadIterableFixed because it isn't used
anymore. We just call directly into spread_iterable from CSA.
Bug: v8:6488, chromium:704966
Change-Id: I81a18281f062619851134fff7ce88471566ee3b5
Reviewed-on: https://chromium-review.googlesource.com/535615
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Commit-Queue: Peter Marshall <petermarshall@chromium.org>
Cr-Commit-Position: refs/heads/master@{#46038}
Storing the boilerplate on the first run leads to memory ovehead for code
that is run only once. Hence we directly return the creating literal on the
first run and only start creating copies from the second run on.
Bug: v8:6211
Change-Id: I69b96d124a5b594b991fdbcc76dbf935d973ffad
Reviewed-on: https://chromium-review.googlesource.com/530688
Commit-Queue: Camillo Bruni <cbruni@chromium.org>
Reviewed-by: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45975}
Port the baseline implementation of Object.prototype.isPrototypeOf to
the CodeStubAssembler, sharing the existing prototype chain lookup logic
with the instanceof / OrdinaryHasInstance implementation. Based on that,
do the same in TurboFan, introducing a new JSHasInPrototypeChain
operator, which encapsulates the central prototype chain walk logic.
This speeds up Object.prototype.isPrototypeOf by more than a factor of
four, so that the code
A.prototype.isPrototypeOf(a)
is now performance-wise on par with
a instanceof A
for the case where A is a regular constructor function and a is an
instance of A.
Since instanceof does more than just the fundamental prototype chain
lookup, it was discovered in Node core that O.p.isPrototypeOf would
be a more appropriate alternative for certain sanity checks, since
it's less vulnerable to monkey-patching. In addition, the Object
builtin would also avoid the performance-cliff associated with
instanceof (due to the Symbol.hasInstance hook), as for example hit
by https://github.com/nodejs/node/pull/13403#issuecomment-305915874.
The main blocker was the missing performance of isPrototypeOf, since
it was still a JS builtin backed by a runtime call.
This CL also adds more test coverage for the
Object.prototype.isPrototypeOf builtin, especially when called from
optimized code.
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_chromium_rel_ng
BUG=v8:5269,v8:5989,v8:6483
R=jgruber@chromium.org
Review-Url: https://codereview.chromium.org/2934893002
Cr-Commit-Position: refs/heads/master@{#45925}
This takes into account the type of the type guard when choosing
representation for a node. To make the representation changes
unambiguous, we pass the restricted type to the changer.
BUG=chromium:726554
Review-Url: https://codereview.chromium.org/2920193004
Cr-Commit-Position: refs/heads/master@{#45734}
Previously Ignition would collect precise Number feedback for binary
operators, but TurboFan would just ignore that and treat it the same as
NumberOrOddball. That however generates a lot of unnecessary code, plus
it defeats redundancy elimination if the same input is also used by
compare operations, which do properly distinguish feedback Number and
NumberOrOddball.
This CL adds the missing bits to connect the existing functionality
properly, i.e. adding the missing BinaryOperationHint and using the
NumberOperationHint::kNumber in the representation selection for tagged
inputs.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2923543003
Cr-Commit-Position: refs/heads/master@{#45732}
Introduces ThrowReferenceErrorIfHole / ThrowSuperNotCalledIfHole
/ ThrowSuperAlreadyCalledIfNotHole bytecodes to handle hole checks.
In the bytecode-graph builder they are handled by introducing a deopt point
instead of adding explicit control flow. JumpIfNotHole / JumpIfNotHoleConstant
bytecodes are removed since they are no longer required.
Bug: v8:4280, v8:6383
Change-Id: I58b70c556b0ffa30e41a0cd44016874c3e9c5fe1
Reviewed-on: https://chromium-review.googlesource.com/509613
Commit-Queue: Mythri Alle <mythria@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45720}
When the input to Function.prototype.bind is a known function, we can
inline the allocation of the JSBoundFunction into TurboFan, which
provides a 2x speed-up for several hot functions in Node streams (as
discovered by Matteo Collina). One of example of this can be found in
https://github.com/nodejs/node/pull/13322, which can be optimized and
made more readable using bind instead of closures.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2916063002
Cr-Commit-Position: refs/heads/master@{#45679}
This fixes an issue with ful-codegen where code target entries for the OSR
check were being incorrectly shared. We now explicitly disable sharing of code
target constant pool entries for full-codegen and for calls to builtins from
WASM code, using a scope.
BUG=chromium:725743
Review-Url: https://codereview.chromium.org/2922433002
Cr-Commit-Position: refs/heads/master@{#45661}
Previously the inlining of accessors into try-blocks (i.e. try/catch,
try/finally, for-of, etc.) was disabled in JSNativeContextSpecialization,
which prevented a couple of interesting optimizations, i.e. we end up
with a LOAD_IC in optimized code for this simple example:
class A { get x() { return 1; } }
function foo(a) {
try {
return a.x;
} catch (e) {
return 0;
}
}
foo(new A)
This is now fixed and the accessors are properly rewired into the
handler chain.
BUG=v8:6278,v8:6344,v8:6424
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2902533003
Cr-Commit-Position: refs/heads/master@{#45485}
When a virtual object passes by a store node that updates a field to the existing value, then the object and its state were not copied, which lead to the original object being passed on.
If then later the store actually modifies and copies the virtual object, this new copy is not passed down the effect chain, so subsequent nodes still refer to the original virtual object and try to update it once new information flows in.
This conflicts with updates on the node that originally created the virtual object, leading to divergence.
Bug: v8:6345
Change-Id: Iab1ce98a60b48478b343eae765c80bdfcb8ba390
Reviewed-on: https://chromium-review.googlesource.com/496267
Commit-Queue: Tobias Tebbi <tebbi@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#45120}
1. Replaces --crankshaft with --opt in tests.
2. Also fixes presubmit to check for --opt flag when
assertOptimized is used.
3. Updates testrunner/local/variants.py and
v8_foozie.py to use --opt flag.
This would mean, nooptimize variant means there are
no optimizations. Not even with %OptimizeFunctionOnNextCall.
Bug:v8:6325
Change-Id: I638e743d0773a6729c6b9749e2ca1e2537f12ce6
Reviewed-on: https://chromium-review.googlesource.com/490206
Commit-Queue: Mythri Alle <mythria@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44985}
This also fixes incorrect type for fixed array accesses.
BUG=chromium:715651,v8:6309,chromium:715204
Review-Url: https://codereview.chromium.org/2848583002
Cr-Commit-Position: refs/heads/master@{#44926}
This still doesn't cover all the paths yet, since some paths are
impossible to trigger at this point due to the way the CanInlineCall
predicate works on the AllocationSite, which says multiple things:
- In case of Array(len), the len was always a Smi so far.
- In case of Array(...args), storing the args didn't change the
elements kind.
- In case of Array(len), the len was always less than the initial
maximum fast element array size.
These conditions are tailored towards Crankshaft and don't really
make a lot of sense in the TurboFan world. We'd need more fine
grained protections, which we will achieve by refactoring the Array
constructor.
BUG=chromium:715404,v8:6262
TBR=machenbach@chromium.org
Review-Url: https://codereview.chromium.org/2843033002
Cr-Commit-Position: refs/heads/master@{#44901}
Also add more local variables to regress-v8-6077 to force
register spill on platform with 32 float registers.
BUG=
Review-Url: https://codereview.chromium.org/2822073003
Cr-Commit-Position: refs/heads/master@{#44865}
We already have an optimization to constant-fold access to an object's
prototype via the special __proto__ accessor (specified in appendix B).
We can use the same optimization to also constant-fold accesses to an
object's prototype via the official Object.getPrototypeOf function.
Also add the optimization for Reflect.getPrototypeOf, which is
equivalent for object inputs.
This is commonly used by Babel to implement various new language
features, for example subclassing and certain property lookups.
R=yangguo@chromium.org
BUG=v8:6292
Review-Url: https://codereview.chromium.org/2841463002
Cr-Commit-Position: refs/heads/master@{#44788}
Move JSOrdinaryHasInstance lowering to JSNativeContextSpecialization,
which was previously mostly done in JSTypedLowering (for no reason).
Add new logic to the lowering to constant-fold OrdinaryHasInstance
checks when the map of the left-hand side and the "prototype" of the
right-hand side is known. This address the performance issue with the
(base) class constructors generated by Babel, i.e.:
function _classCallCheck(instance, Constructor) {
if (!(instance instanceof Constructor)) {
throw new TypeError("Cannot call a class as a function");
}
}
var C = function C() { _classCallCheck(this, C); };
for
class C {}
Also ensure that a known constructor being used inside an instanceof
get's a proper initial map on-demand.
BUG=v8:6275
R=mstarzinger@chromium.org
Review-Url: https://codereview.chromium.org/2827013002
Cr-Commit-Position: refs/heads/master@{#44727}
Since we no longer support the ignition-staging configuration
any longer, we can retire the three tier pipeline and the
CompileBaseline functionallity.
We still need support for JSFunction self healing due to
liveedit (which for --no-turbo might end up replacing a
forced Ignition function with a FCG function) - we can
remove this once we remove --no-turbo support.
BUG=v8:4280
Change-Id: I5482abd17785324654e022affd6bdb555b19b181
Reviewed-on: https://chromium-review.googlesource.com/452620
Commit-Queue: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44141}
Test regress-694088.js is adapted for execution on big endian platforms.
TEST=test/mjsunit/compiler/regress-694088.js
BUG=
Review-Url: https://codereview.chromium.org/2739403002
Cr-Commit-Position: refs/heads/master@{#43746}
Now that the --turbo flag is on by default, this implication makes it
impossible to disable the runtime profiler to prevent functions from
being considered hot. Essentially the --nocrankshaft flag was borked.
R=machenbach@chromium.org
Change-Id: I698008b4b69038e8bfab6d8ec3ccf57e2fd71701
Reviewed-on: https://chromium-review.googlesource.com/452619
Reviewed-by: Michael Achenbach <machenbach@chromium.org>
Reviewed-by: Igor Sheludko <ishell@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#43720}
The new NewUnmappedArgumentsElements node now takes two inputs:
- the frame holding the arguments (current frame or arguments adaptor frame)
- the length of the suffix of passed arguments to be copied into the backing store
These inputs are computed with two new node types:
ArgumentsFrame()
ArgumentsLength[formal_parameter_count,is_rest_length](Node* arguments_frame)
The node type NewRestParameterElements can now be expressed with NewUnmappedArgumentsElements and an appropriate length and is thus not needed anymore.
In escape analysis, we lower loads from the length field of NewUnmappedArgumentsElements with its length input and if we find out that no write access to the arguments elements exists, we replace element loads with direct stack access and replace the NewUnmappedArgumentsElements node with a node of the new node type ArgumentsElementsState. This corresponds to an ObjectState node and gets translated into a deoptimizer instruction to allocate the backing store. Together with the already existing deoptimizer support for the actual arguments object/rest parameters, this allows to remove all allocations for arguments objects/rest parameters in this case.
In the deoptimizer, we read the actual parameters from the stack while transforming the static deopt info into TranslatedValue objects.
If escape analysis cannot remove the backing store allocation, NewUnmappedArgumentsElements gets lo
BUG=v8:5726
Review-Url: https://codereview.chromium.org/2692753004
Cr-Commit-Position: refs/heads/master@{#43475}
If the RHS is 0 and we have Smi feedback, speculate that the result (the LHS)
will continue to be in the Unsigned31 range. This helps us avoid converting
the result to double when merging with Signed32.
R=jarin@chromium.org
BUG=
Review-Url: https://codereview.chromium.org/2709423002
Cr-Commit-Position: refs/heads/master@{#43415}
This implements the proposal at
https://github.com/tc39/proposal-template-literal-revision
staged behind a flag --harmony-template-escapes. The proposal allows
invalid octal, unicode, and hexadecimal escape sequences to appear in
tagged template literals, instead of being a syntax error. These have
a 'cooked' value of 'undefined', but are still accessible through the
'raw' property.
BUG=v8:5546
Review-Url: https://codereview.chromium.org/2665513002
Cr-Commit-Position: refs/heads/master@{#43384}
