This change introduces a pipeline for the final stages of
bytecode generation.
The peephole optimizer is made distinct from the BytecodeArrayBuilder.
A new BytecodeArrayWriter is responsible for writing bytecode. It
also keeps track of the maximum register seen and offers a potentially
smaller frame size.
R=rmcilroy@chromium.org
LOG=N
BUG=v8:4280
Review-Url: https://codereview.chromium.org/1947403002
Cr-Commit-Position: refs/heads/master@{#36220}
We eagerly inserted Int32Mul for Math.imul during builtin lowering and
messed up with the types, which confused the representation selection.
This adds a proper NumberImul operator, and fixes the builtin reducer to
do the right thing according to the spec.
R=mstarzinger@chromium.org
BUG=v8:5006
LOG=n
Review-Url: https://codereview.chromium.org/1971163002
Cr-Commit-Position: refs/heads/master@{#36219}
The usat instruction is available from ARMv6, so there's no need to
check for the ARMv7 feature before using it. ARMv6 is the oldest
supported architecture in V8.
Correcting this allows the removal of a special case for predictable
code size.
BUG=
Review-Url: https://codereview.chromium.org/1974903002
Cr-Commit-Position: refs/heads/master@{#36218}
V8 tracing controller uses 2 clocks: wall clock and cpu clock. This patch
implements CPU time for OS X and POSIX to provide more accurate
accounting of CPU time used by each thread.
BUG=v8:4984
LOG=n
Review-Url: https://codereview.chromium.org/1966183003
Cr-Commit-Position: refs/heads/master@{#36213}
Names passed for imports and exports are checked during decoding,
leading to errors if they are no valid UTF-8. Function names are not
checked during decode, but rather lead to undefined being returned at
runtime if they are not UTF-8.
We need to do these checks on the Wasm side, since the factory
methods assume to get valid UTF-8 strings.
R=titzer@chromium.org, yangguo@chromium.org
Review-Url: https://codereview.chromium.org/1967023004
Cr-Commit-Position: refs/heads/master@{#36208}
With this CL it is possible to compile a wasm module with multiple
threads in parallel. Parallel compilation works as follows:
1) The main thread allocates a compilation unit for each wasm function.
2) The main thread spawns WasmCompilationTasks which run on the
background threads.
3.a) The background threads and the main thread pick one compilation unit
at a time and execute the parallel phase of the compilation unit.
After finishing the execution of the parallel phase, the compilation
unit is stored in a result queue.
3.b) If the result queue contains a compilation unit, the main thread
dequeues it and finishes its compilation.
4) After the execution of the parallel phase of all compilation units has
started, the main thread waits for all WasmCompilationTasks to finish.
5) The main thread finalizes the compilation of the module.
I'm going to add some additional tests before committing this CL.
R=titzer@chromium.org, bmeurer@chromium.org, mlippautz@chromium.org, mstarzinger@chromium.org
Committed: https://crrev.com/17215438659d8ff2d7d55f95226bf8a1477ccd79
Cr-Commit-Position: refs/heads/master@{#36178}
Review-Url: https://codereview.chromium.org/1961973002
Cr-Commit-Position: refs/heads/master@{#36207}
Currently we do not check for @@isConcatSpreadable properly. If the Symbol is
set on the Array.prototype or Object.prototype the current fast paths fail.
This CL adds a fix to globally invalidate a isConcatSpreadable_protector.
Drive-by-fix: use named accessors for context variables
LOG=N
BUG=chromium:542504, v8:903
Review-Url: https://codereview.chromium.org/1409123003
Cr-Commit-Position: refs/heads/master@{#36201}
We got the condition wrong and actually deoptimized when the typed array
was not neutered. This fixes the deopt loop in Math.random and actually
many programs that use typed arrays.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/1970123002
Cr-Commit-Position: refs/heads/master@{#36194}
This patch removes the following properties, as their use count is
very low, they are V8-only, and not on a standards track.
- v8Parse
- resolved
- pattern
v8BreakIterator is left in as it has significantly more usage.
BUG=v8:3785
R=adamk,jshin@chromium.org
Review-Url: https://codereview.chromium.org/1968893002
Cr-Commit-Position: refs/heads/master@{#36190}
Reason for revert:
Buildbot is failing on Mac release build.
Original issue's description:
> Implement CPU time for OS X and POSIX.
>
> V8 tracing controller uses 2 clocks: wall clock and cpu clock. This patch
> implements CPU time for OS X and POSIX to provide more accurate
> accounting of CPU time used by each thread.
>
> BUG=v8:4984
> LOG=n
>
> Committed: https://crrev.com/025f3d262bab2748362374f1b90ac723a9655ee4
> Cr-Commit-Position: refs/heads/master@{#36188}
TBR=jochen@chromium.org,bmeurer@chromium.org,fmeawad@chromium.org,rsesek@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:4984
Review-Url: https://codereview.chromium.org/1966173003
Cr-Commit-Position: refs/heads/master@{#36189}
V8 tracing controller uses 2 clocks: wall clock and cpu clock. This patch
implements CPU time for OS X and POSIX to provide more accurate
accounting of CPU time used by each thread.
BUG=v8:4984
LOG=n
Review-Url: https://codereview.chromium.org/1959103004
Cr-Commit-Position: refs/heads/master@{#36188}
When I18N is enabled, use ICU's case conversion API and transliteration
API [1] to implement String.prototype.to{Upper,Lower}Case and
String.prototype.toLocale{Upper,Lower}Case.
* ICU-based case conversion was implemented in runtime-i18n.cc/i18n.js
* The above 4 functions are overridden with those in i18n.js when
--icu_case_mapping flag is turned on. To control the override by the flag,
they're overriden in icu-case-mapping.js
Previously, toLocale{U,L}Case just called to{U,L}Case so that they didn't
support locale-sensitive case conversion for Turkic languages (az, tr),
Greek (el) and Lithuanian (lt).
Before ICU APIs for the most general case are called, a fast-path for Latin-1
is tried. It's taken from Blink and adopted as necessary. This fast path
is always tried for to{U,L}Case. For toLocale{U,L}Case, it's only taken
when a locale (explicitly specified or default) is not in {az, el, lt, tr}.
With these changes, a build with --icu_case_mapping=true passes a bunch
of tests in test262/intl402/Strings/* and intl/* that failed before.
Handling of pure ASCII strings (aligned at word boundary) are not as fast
as Unibrow's implementation that uses word-by-word case conversion. OTOH,
Latin-1 input handling is faster than Unibrow. General Unicode input
handling is slower but more accurate.
See https://docs.google.com/spreadsheets/d/1KJCJxKc1FxFXjwmYqABS0_2cNdPetvnd8gY8_HGSbrg/edit?usp=sharing for the benchmark.
This CL started with http://crrev.com/1544023002#ps200001 by littledan@,
but has changed significantly since.
[1] See why transliteration API is needed for uppercasing in Greek.
http://bugs.icu-project.org/trac/ticket/10582
R=yangguo
BUG=v8:4476,v8:4477
LOG=Y
TEST=test262/{built-ins,intl402}/Strings/*, webkit/fast/js/*, mjsunit/string-case,
intl/general/case*
Review-Url: https://codereview.chromium.org/1812673005
Cr-Commit-Position: refs/heads/master@{#36187}
Reason for revert:
The ThreadSanitizer finds data races.
Original issue's description:
> [wasm] Implement parallel compilation.
>
> With this CL it is possible to compile a wasm module with multiple
> threads in parallel. Parallel compilation works as follows:
>
> 1) The main thread allocates a compilation unit for each wasm function.
> 2) The main thread spawns WasmCompilationTasks which run on the
> background threads.
> 3.a) The background threads and the main thread pick one compilation unit
> at a time and execute the parallel phase of the compilation unit.
> After finishing the execution of the parallel phase, the compilation
> unit is stored in a result queue.
> 3.b) If the result queue contains a compilation unit, the main thread
> dequeues it and finishes its compilation.
> 4) After the execution of the parallel phase of all compilation units has
> started, the main thread waits for all WasmCompilationTasks to finish.
> 5) The main thread finalizes the compilation of the module.
>
> I'm going to add some additional tests before committing this CL.
>
> R=titzer@chromium.org, bmeurer@chromium.org, mlippautz@chromium.org, mstarzinger@chromium.org
>
> Committed: https://crrev.com/17215438659d8ff2d7d55f95226bf8a1477ccd79
> Cr-Commit-Position: refs/heads/master@{#36178}
TBR=bmeurer@chromium.org,mlippautz@chromium.org,mstarzinger@chromium.org,titzer@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review-Url: https://codereview.chromium.org/1965243003
Cr-Commit-Position: refs/heads/master@{#36182}
The previous approach taken by FastNew[Sloppy,Strict,Rest]ArgumentsStub
looked at the function slot in order to skip stub frames
and find the JS frame. However, stub frames do not have a
function slot (in fact their fixed frame ends one slot
before the JS frame's function slot). Therefore, if this
location in the stub frame happens to have the function
object the create arguments stubs won't skip this frame
correctly.
Replace this approach with one where the stub is
specialized to either skip a frame if required (since
there will only ever be one extra frame on Ignition
the loop approach isn't necessary).
BUG=v8:4928
LOG=N
CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux_nosnap_dbg
Review-Url: https://codereview.chromium.org/1949023003
Cr-Commit-Position: refs/heads/master@{#36181}
This code was used exclusively by its own unit test and can be safely removed.
BUG=v8:4947
LOG=N
Review-Url: https://codereview.chromium.org/1965373002
Cr-Commit-Position: refs/heads/master@{#36180}
When instantiating a subclassed API function, the instance cache is avoided. There is currently no direct API yet to instantiate a Template while passing in a new.target. It probably makes sense to extend ObjectTemplate::NewInstance to accept a new.target, in line with Reflect.construct.
BUG=v8:3330, v8:5001
Review-Url: https://codereview.chromium.org/1972613002
Cr-Commit-Position: refs/heads/master@{#36179}
With this CL it is possible to compile a wasm module with multiple
threads in parallel. Parallel compilation works as follows:
1) The main thread allocates a compilation unit for each wasm function.
2) The main thread spawns WasmCompilationTasks which run on the
background threads.
3.a) The background threads and the main thread pick one compilation unit
at a time and execute the parallel phase of the compilation unit.
After finishing the execution of the parallel phase, the compilation
unit is stored in a result queue.
3.b) If the result queue contains a compilation unit, the main thread
dequeues it and finishes its compilation.
4) After the execution of the parallel phase of all compilation units has
started, the main thread waits for all WasmCompilationTasks to finish.
5) The main thread finalizes the compilation of the module.
I'm going to add some additional tests before committing this CL.
R=titzer@chromium.org, bmeurer@chromium.org, mlippautz@chromium.org, mstarzinger@chromium.org
Review-Url: https://codereview.chromium.org/1961973002
Cr-Commit-Position: refs/heads/master@{#36178}
Prints source position information alongside bytecode.
BUG=v8:4280
LOG=N
Review-Url: https://codereview.chromium.org/1963663002
Cr-Commit-Position: refs/heads/master@{#36171}
By now the runtime entry function in question is a duplicate of the
existing Runtime_ToFastProperties function. This just gets rid of the
duplication.
R=bmeurer@chromium.org
Review-Url: https://codereview.chromium.org/1963973003
Cr-Commit-Position: refs/heads/master@{#36161}
Up until now we had two places where we did the function prototype
folding, once in the Typer and once in JSTypedLowering. Put this logic
into JSNativeContextSpecialization instead.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/1965293002
Cr-Commit-Position: refs/heads/master@{#36157}
Make JSCreateArguments eliminatable, and remove the need for frame
states on JSCreateArguments nodes being lowered to (optimized) stub
calls. Only the runtime fallback needs a frame state, because in that
case we need to ask the deoptimizer for arguments to inlined functions.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/1965013005
Cr-Commit-Position: refs/heads/master@{#36154}
The data from UseCounters are in--V8's custom Promise methods are not
in wide use on the web (<.002%). Therefore, this patch removes them.
That includes:
- Promise.prototype.chain
- Promise.defer -- the most widely used of the bunch
- Promise.accept
For now, those methods are still available by checking the "disable
latest stable JavaScript features" flag, or --promise-extra at the
command line, but I expect them to be fully removable.
R=adamk
CC=rossberg
BUG=v8:3238,v8:4633
Review-Url: https://codereview.chromium.org/1965183002
Cr-Commit-Position: refs/heads/master@{#36152}
Old code failed to walk over deleted elements, instead treating
deleted elements as "undefined" in the output array.
This is the Map equivalent of commit 2d9bfe9ad5.
Also micro-optimized the loops to avoid an extra call to KeyAt()
and used a direct hole comparison instead of calling IsTheHole().
R=cbruni@chromium.org
BUG=v8:4946
LOG=y
Review-Url: https://codereview.chromium.org/1965593002
Cr-Commit-Position: refs/heads/master@{#36149}
Renames IsDouble* predicates to IsFP*.
Adds specific IsFloat*, IsDouble*, and IsSimd128* predicates.
Adds specific GetFloatRegister, GetDoubleRegister, and
GetSimd128Register methods.
This is mostly a mechanical renaming of IsDouble* to IsFP* methods.
This shouldn't change code generation at all. All fp registers are still
treated as double registers.
LOG=N
BUG=v8:4124
Review-Url: https://codereview.chromium.org/1959763002
Cr-Commit-Position: refs/heads/master@{#36146}
Rather than pushing zero for the initial bytecode offset, we should push the
offset of the first bytecode handler, Smi tagged. This fixes the line number
for the top stack frame on overflow errors.
BUG=v8:4981
LOG=N
Review-Url: https://codereview.chromium.org/1950913004
Cr-Commit-Position: refs/heads/master@{#36137}
Trying to re-land http://crrev.com/1948453002 after fixing assembler-x64.cc in http://crrev.com/1962563003.
Before this patch, we would emit a cmp or test with a memory operand only if both of the operands in the IR were loads. Now if either of them is a load and the other one is an immediate, we can use a memory operand if the load representation machine size is wide enough to represent the latter.
Review-Url: https://codereview.chromium.org/1968453002
Cr-Commit-Position: refs/heads/master@{#36136}
In the bytecode graphbuilder, translate the two generator-specific
bytecodes as a couple of runtime calls for now.
BUG=v8:4907
LOG=n
Review-Url: https://codereview.chromium.org/1957393004
Cr-Commit-Position: refs/heads/master@{#36134}
This adds a new pass MemoryOptimizer that walks over the effect chain
from Start and lowers all Allocate, LoadField, StoreField, LoadElement,
and StoreElement nodes, trying to fold allocations into allocation
groups and eliminate write barriers on StoreField and StoreElement if
possible (i.e. if the object belongs to the current allocation group and
that group allocates in new space).
R=hpayer@chromium.org, jarin@chromium.org
BUG=v8:4931, chromium:580959
LOG=n
Review-Url: https://codereview.chromium.org/1963583004
Cr-Commit-Position: refs/heads/master@{#36128}
This implements declaration of lookup slots for variables and functions
within optimized code. Such a declaration only appears with top-level
eval code, which we only recently started handling in TurboFan.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/1962723002
Cr-Commit-Position: refs/heads/master@{#36125}
This patch installs %TypedArray% and its prototype on the native
context, and wires them up to each TypedArray subclass. This is later
used to check the holder of length, byteLength and byteOffset is
%Typedarray% and apply the appropriate optimizations.
BUG=chromium:593634
LOG=Y
Review-Url: https://codereview.chromium.org/1949863002
Cr-Commit-Position: refs/heads/master@{#36116}
Fixes 4 compile errors on Mac due to specifiers not matching type casts.
LOG=N
BUG=
Review-Url: https://codereview.chromium.org/1960113002
Cr-Commit-Position: refs/heads/master@{#36115}
atomic-utils.h only uses functionality from base/, and also by moving it into
base/, code outside of v8_base can benefit from it.
BUG=v8:4991
LOG=n
Review-Url: https://codereview.chromium.org/1954603002
Cr-Commit-Position: refs/heads/master@{#36114}
