Use an unordered_map<Module, unordered_set<String>> to keep track
of visited Module/ExportName pairs during ResolveExport.
This required adding a Hash() method to Module, which is accomplished
by allocating a Symbol and storing it in the SharedFunctionInfo::name
slot, then delegating the hash to that Symbol.
Also added a helper method Module::shared() to easily get ahold of
the SharedFunctionInfo and call it in the appropriate places instead
of re-doing the ternary operator.
BUG=v8:1569
Review-Url: https://codereview.chromium.org/2367623004
Cr-Commit-Position: refs/heads/master@{#39743}
Reason for revert:
Tanks EarleyBoyer.
Original issue's description:
> [compiler] Properly guard the speculative optimizations for instanceof.
>
> Add a general feedback slot for instanceof similar to what we already have
> for for-in, which basically has a fast (indicated by the uninitialized
> sentinel) and a slow (indicated by the megamorphic sentinel) mode. Now
> we can only take the fast path when the feedback slot says it hasn't
> seen any funky inputs and nothing funky appeared in the prototype chain.
> In the TurboFan code we also deoptimize whenever we see a funky object
> (i.e. a proxy or an object that requires access checks) in the prototype
> chain (similar to what Crankshaft already did).
>
> Drive-by-fix: Also make Crankshaft respect the mode and therefore
> address the deopt loop in Crankshaft around instanceof.
>
> We might want to introduce an InstanceOfIC mechanism at some point and
> track the map of the right-hand side.
>
> BUG=v8:5267
> R=mvstanton@chromium.org
>
> Committed: https://crrev.com/a0484bc6116ebc2b855de87d862945e2ae07169b
> Cr-Commit-Position: refs/heads/master@{#39718}
TBR=mvstanton@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:5267
Review-Url: https://codereview.chromium.org/2365223003
Cr-Commit-Position: refs/heads/master@{#39736}
Reason for revert:
We currently have some stability issues on Canary. Let's reland this after we verified that we "fixed" Canary again.
Original issue's description:
> Preparse inner functions (new try)
>
> This is an overly pessimistic approach where PreParser only keeps
> track of unresolved variables, but doesn't declare anything. This
> will result in context-allocating variables in the outer function
> unnecessarily, if the variable names clash with variable names
> used by the inner function (even if the variables are not the
> same). However, we have been unable to prove that this approach
> wouldn't be good enough for the practical purposes.
>
> Fixes after the previous try ( https://codereview.chromium.org/2322243002/ ):
> Keep the context-allocation decision stable when compiling fully eagerly.
>
> Tests which exercise this functionality:
> mjsunit/fixed-context-shapes-when-recompiling.js
>
> Design document (chromium):
>
> https://docs.google.com/a/chromium.org/document/d/1rRv5JJZ0JpOZAZN2CSUwZPFJiBAdRnTiSYhazseNHFg/edit?usp=sharing
>
> BUG=
>
> Committed: https://crrev.com/7c73cf32c60484cdf37c84f1d61b4640e87068d7
> Cr-Commit-Position: refs/heads/master@{#39719}
TBR=verwaest@chromium.org,adamk@chromium.org,marja@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=
Review-Url: https://codereview.chromium.org/2373443003
Cr-Commit-Position: refs/heads/master@{#39730}
The WebAssembly spec requires a HasProperty() check for the maximum
property of the descriptor object which is used to set up a
WebAssembly.Memory object or a WebAssembly.Table object.
The original implementation only approximated the HasProperty() check.
It used Get() to get the value of the maximum property of the descriptor
object and compared the resulting value to {undefined}. However, this
approximation is incorrect if the property exists but its value is
{undefined}.
R=titzer@chromium.org, franzih@chromium.org
BUG=chromium:649461
TEST=mjsunit/wasm/memory
Review-Url: https://codereview.chromium.org/2367673003
Cr-Commit-Position: refs/heads/master@{#39722}
This is an overly pessimistic approach where PreParser only keeps
track of unresolved variables, but doesn't declare anything. This
will result in context-allocating variables in the outer function
unnecessarily, if the variable names clash with variable names
used by the inner function (even if the variables are not the
same). However, we have been unable to prove that this approach
wouldn't be good enough for the practical purposes.
Fixes after the previous try ( https://codereview.chromium.org/2322243002/ ):
Keep the context-allocation decision stable when compiling fully eagerly.
Tests which exercise this functionality:
mjsunit/fixed-context-shapes-when-recompiling.js
Design document (chromium):
https://docs.google.com/a/chromium.org/document/d/1rRv5JJZ0JpOZAZN2CSUwZPFJiBAdRnTiSYhazseNHFg/edit?usp=sharing
BUG=
Review-Url: https://codereview.chromium.org/2352593002
Cr-Commit-Position: refs/heads/master@{#39719}
Add a general feedback slot for instanceof similar to what we already have
for for-in, which basically has a fast (indicated by the uninitialized
sentinel) and a slow (indicated by the megamorphic sentinel) mode. Now
we can only take the fast path when the feedback slot says it hasn't
seen any funky inputs and nothing funky appeared in the prototype chain.
In the TurboFan code we also deoptimize whenever we see a funky object
(i.e. a proxy or an object that requires access checks) in the prototype
chain (similar to what Crankshaft already did).
Drive-by-fix: Also make Crankshaft respect the mode and therefore
address the deopt loop in Crankshaft around instanceof.
We might want to introduce an InstanceOfIC mechanism at some point and
track the map of the right-hand side.
BUG=v8:5267
R=mvstanton@chromium.org
Review-Url: https://codereview.chromium.org/2370693002
Cr-Commit-Position: refs/heads/master@{#39718}
V8 is collecting a growing amount of fuzzers, all of which take substantial
space on the bots and in chromium build archives. This CL improves that
situation by allowing component (shared library) builds for almost all fuzzers.
The parser fuzzer is handled as an exception since it would require exporting a
large number of additional functions.
A component build results in about a 50-100x improvement in file size for each
fuzzer (~50M-100M to around 1.1M).
BUG=chromium:648864
CQ_INCLUDE_TRYBOTS=master.tryserver.chromium.linux:linux_chromium_compile_dbg_ng;master.tryserver.chromium.android:android_clang_dbg_recipe
Review-Url: https://codereview.chromium.org/2360983002
Cr-Commit-Position: refs/heads/master@{#39709}
If DevTools is turned on in the middle of doing some things with async
functions, then more items may be popped from the Promise stack than were
pushed to it. In this sort of case, it's OK to have a catch misprediction,
but a crash is unacceptable. This patch defensively handles this edge
case where the Promise stack is unexpectedly empty for that reason.
BUG=v8:5167
Review-Url: https://codereview.chromium.org/2361333003
Cr-Commit-Position: refs/heads/master@{#39705}
This patch tracks the stack of async functions differently from other
Promise async stack tracking. With this patch, the stack trace of a
callstack of async functions should look similarly to the call stack
if all of the functions were synchronous. An example can be found in
the updated test expectations: https://codereview.chromium.org/2362923002 .
The new stack traces are implemented using existing mechanisms in the
inspector. The inspector has two ways to save async stack traces: recurring
and non-recurring stacks. An example of a non-recurring stack is setTimeout,
and a recurring one is saved for setInterval. Recurring stacks are deleted
only when a special "cancel" function is called, rather than being deleted
after being used the first time. Previous Promise async stack tracking always
used non-recurring stacks.
For async functions, this patch saves a recurring stack. The top frame of
the stack is duplicated, as the resuming function contains a similar frame;
the devtools frontend is responsible for removing or marking this frame,
which it can do based on seeing the [async function] line which follows it.
The second frame will instead be provided by the resuming execution
context. The recurring stack is saved when the async function is entered, and
it is deleted from a finally block. The id of the stack is saved in the outer
Promise being constructed by the async function. When an intermediate
throwaway Promise will be triggered as a reaction, it will be identified as
such based on its debugging metadata, and the corresponding async function's
recurring stack will be used.
BUG=v8:4483
Review-Url: https://codereview.chromium.org/2357423002
Cr-Commit-Position: refs/heads/master@{#39695}
There's no reason (anymore) to have empty imports in special_imports. Remove
them from there and rename special_imports to namespace_imports to be more
precise.
R=adamk@chromium.org
BUG=v8:1569
Review-Url: https://codereview.chromium.org/2368613002
Cr-Commit-Position: refs/heads/master@{#39693}
This matches current Crankshaft/fullcodegen behavior more closely and
thus reduces the chances that we run into unnecessary polymorphism due
to the field representation tracking in our object model.
R=jarin@chromium.org
BUG=v8:5267
Committed: 6a939714e9
Review-Url: https://codereview.chromium.org/2367593003
Cr-Commit-Position: refs/heads/master@{#39692}
Resolve imports and indirect exports at instantiation time.
With this CL we have some basic functionality for modules working. Not yet
supported: star exports, namespace imports, cycle detection, proper variable
initialisation in mutually recursive modules.
BUG=v8:1569
Review-Url: https://codereview.chromium.org/2362083002
Cr-Commit-Position: refs/heads/master@{#39689}
DevTools wants to be able to detect async functions in order to print
their synopsis better in stack traces and tooltips. This patch provides a
simple method to do the check.
BUG=v8:4483
Review-Url: https://codereview.chromium.org/2365833002
Cr-Commit-Position: refs/heads/master@{#39687}
Reason for revert:
Main suspect for tsan:
https://build.chromium.org/p/client.v8/builders/V8%20Linux64%20TSAN/builds/11893
Also changes layout tests:
https://build.chromium.org/p/client.v8.fyi/builders/V8-Blink%20Linux%2064/builds/10036
+mips builder:
https://build.chromium.org/p/client.v8.ports/builders/V8%20Mips%20-%20builder/builds/4032
Original issue's description:
> [wasm] Master CL for Binary 0xC changes.
>
> [0xC] Convert to stack machine semantics.
> [0xC] Use section codes instead of names.
> [0xC] Add elements section decoding.
> [0xC] Decoding of globals section.
> [0xC] Decoding of memory section.
> [0xC] Decoding of imports section.
> [0xC] Decoding of exports section.
> [0xC] Decoding of data section.
> [0xC] Remove CallImport bytecode.
> [0xC] Function bodies have an implicit block.
> [0xC] Remove the bottom label from loops.
> [0xC] Add signatures to blocks.
> [0xC] Remove arities from branches.
> Add tests for init expression decoding.
> Rework compilation of import wrappers and how they are patched.
> Rework function indices in debugging.
> Fix ASM->WASM builder for stack machine.
> Reorganize asm.js foreign functions due to import indices change.
>
> R=ahaas@chromium.org,rossberg@chromium.org,bradnelson@chromium.org
> BUG=chromium:575167
> LOG=Y
>
> Committed: https://crrev.com/76eb976a67273b8c03c744f64ad850b0432554b9
> Cr-Commit-Position: refs/heads/master@{#39678}
TBR=ahaas@chromium.org,bradnelson@chromium.org,mtrofin@chromium.org,rossberg@chromium.org,bradnelson@google.com,titzer@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:575167
Review-Url: https://codereview.chromium.org/2361053004
Cr-Commit-Position: refs/heads/master@{#39685}
[0xC] Convert to stack machine semantics.
[0xC] Use section codes instead of names.
[0xC] Add elements section decoding.
[0xC] Decoding of globals section.
[0xC] Decoding of memory section.
[0xC] Decoding of imports section.
[0xC] Decoding of exports section.
[0xC] Decoding of data section.
[0xC] Remove CallImport bytecode.
[0xC] Function bodies have an implicit block.
[0xC] Remove the bottom label from loops.
[0xC] Add signatures to blocks.
[0xC] Remove arities from branches.
Add tests for init expression decoding.
Rework compilation of import wrappers and how they are patched.
Rework function indices in debugging.
Fix ASM->WASM builder for stack machine.
Reorganize asm.js foreign functions due to import indices change.
R=ahaas@chromium.org,rossberg@chromium.org,bradnelson@chromium.org
BUG=chromium:575167
LOG=Y
Review-Url: https://codereview.chromium.org/2345593003
Cr-Commit-Position: refs/heads/master@{#39678}
CpuFeatures::IsSupported(feature) indicates that the feature is
available on the target. AssemblerBase::IsEnabled(feature) indicates
that we've checked for support (using CpuFeatureScope). The main benefit
is that we can test on (for example) ARMv8, but have some assurance that
we won't generate ARMv8 instructions on ARMv7 targets.
This patch simply cleans up the usage, which had become inconsistent.
The instruction emission functions now check not only that their
dependent features are supported, but also that we've verified that
using CpuFeatureScope.
BUG=
Review-Url: https://codereview.chromium.org/2360243002
Cr-Commit-Position: refs/heads/master@{#39676}
It looks like it tried to trigger lazy inner function parsing by
inserting a comment into an inner function.
1) We don't have lazy inner functions yet.
2) Even if we had, there's no way this heuristic could trigger laziness:
we need to do the laziness decision upfront, without looking at the
contents / size of the function.
3) Some of the combinations were weird: lazy_outer but non-lazy inner?
In the current heuristics, only the total script size affects laziness;
in particular, it doesn't matter where the long comment is.
R=mstarzinger@chromium.org
BUG=
Review-Url: https://codereview.chromium.org/2364003002
Cr-Commit-Position: refs/heads/master@{#39673}
This makes sure the {EscapeAnalysisReducer} inserts proper {TypeGuard}
nodes if the replacement node is not a subtype of the original node.
This happens predominantly for code that has been made unreachable by
type checks.
R=jarin@chromium.org
TEST=mjsunit/regress/regress-crbug-640497
BUG=chromium:640497
Review URL: https://codereview.chromium.org/2363573003 .
Cr-Commit-Position: refs/heads/master@{#39656}
This matches current Crankshaft/fullcodegen behavior more closely and
thus reduces the chances that we run into unnecessary polymorphism due
to the field representation tracking in our object model.
R=jarin@chromium.org
BUG=v8:5267
Review URL: https://codereview.chromium.org/2367593003 .
Cr-Commit-Position: refs/heads/master@{#39655}
Rename the high-level operators CheckTaggedSigned to CheckSmi and
CheckTaggedPointer to CheckHeapObject, to better match the naming
convention (i.e. ObjectIsSmi and CheckSmi, ObjectIsString and
CheckString, etc.).
For lowering CheckSmi, always report TaggedSigned representation
and let the RepresentationChanger come up with a reasonable conversion
from whatever input representation to TaggedSigned. This way we no
longer insert the useless ChangeSomethingToTagged and then Smi check
the result sequences, i.e. mostly reduces the amount of useless code
being generated. But we also observe a few performance improvements
on some crypto benchmarks.
This would enable us to avoid the Smi canonicalization when going from
Float64 to Tagged completely and thus match the representation selection
of Crankshaft in many areas (which might reduce the amount of
polymorphism until we fix our object model).
A follow-up CL will do the same for CheckHeapObject.
BUG=v8:5267
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/2362173003 .
Cr-Commit-Position: refs/heads/master@{#39654}
This CL optimizes the code in BytecodeArrayBuilder and
BytecodeArrayWriter by making the following main changes:
- Move operand scale calculation out of BytecodeArrayWriter to the
BytecodeNode constructor, where the decision on which operands are
scalable can generally be statically decided by the compiler.
- Move the maximum register calculation out of BytecodeArrayWriter
and into BytecodeRegisterOptimizer (which is the only place outside
BytecodeGenerator which updates which registers are used). This
avoids the BytecodeArrayWriter needing to know the operand types
of a node as it writes it.
- Modify EmitBytecodes to use individual push_backs rather than
building a buffer and calling insert, since this turns out to be faster.
- Initialize BytecodeArrayWriter's bytecode vector by reserving 512
bytes,
- Make common functions in Bytecodes constexpr so that they
can be statically calculated by the compiler.
- Move common functions and constructors in Bytecodes and
BytecodeNode to the header so that they can be inlined.
- Change large static switch statements in Bytecodes to const array
lookups, and move to the header to allow inlining.
I also took the opportunity to remove a number of unused helper
functions, and rework some others for consistency.
This reduces the percentage of time spent in making BytecodeArrays
in CodeLoad from ~15% to ~11% according to perf. The
CoadLoad score increase by around 2%.
BUG=v8:4280
Committed: https://crrev.com/b11a8b4d41bf09d6b3d6cf214fe3fb61faf01a64
Review-Url: https://codereview.chromium.org/2351763002
Cr-Original-Commit-Position: refs/heads/master@{#39599}
Cr-Commit-Position: refs/heads/master@{#39637}
