Currently we count optimizations to decide to disable optimization, and
count deopts to detect this decision and allow re-enabling optimizations
after a while.
However, throwing out TurboFan OSR code and GC optimized code evictions
do not count as deopts, which means that the optimization count
increases without increasing the deopt count. This increased optimization
count disables further optimization -- which is bad, because these are
not "true" deopts -- and can stop the optimization from being re-enabled,
because the deopt count can't go high enough.
Instead, we now only ever look at deopts to disable/re-enable
optimization, and opt counts are only used for naming log files and in
tests.
Change-Id: I0c7d6be497545449a38cf952cd2f007ee51982ba
Reviewed-on: https://chromium-review.googlesource.com/468811
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44647}
The hole NaN should also have proper Type::Hole, and not silently hide
in the Type::Number. This way we can remove all the special casing for
the hole NaN, and we also finally get the CheckNumber right.
This also allows us to remove some ducktape from the Deoptimizer, as for
escape analyzed FixedDoubleArrays we always pass the hole value now to
represent the actual holes.
Also-By: jarin@chromium.org
BUG=chromium:684208,chromium:709753,v8:5267
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2814013003
Cr-Commit-Position: refs/heads/master@{#44603}
JSObject is slow: creating strings for keys and storing values by these keys after takes significant amount of time.
With this CL console methods (most of them collect top stack frame to calculate source location) are ~33% faster.
V8Debugger::captureStackTrace is ~50% faster.
BUG=v8:6189
R=yangguo@chromium.orgTBR=bmeurer@chromium.org
Review-Url: https://codereview.chromium.org/2789073002
Cr-Commit-Position: refs/heads/master@{#44344}
- Introduce new struct AsyncGeneratorRequest, which holds
information pertinent to resuming execution of an
AsyncGenerator, such as the Promise associated with the async
generator request. It is intended to be used as a singly
linked list, and holds a pointer to the next item in te queue.
- Introduce JSAsyncGeneratorObject (subclass of
JSGeneratorObject), which includes several new internal fields
(`queue` which contains a singly linked list of
AsyncGeneratorRequest objects, and `await_input` which
contains the sent value from an Await expression (This is
necessary to prevent function.sent (used by yield*) from
having the sent value observably overwritten during
execution).
- Modify SuspendGenerator to accept a set of Flags, which
indicate whether the suspend is for a Yield or Await, and
whether it takes place on an async generator or ES6
generator.
- Introduce interpreter intrinsics and TF intrinsic lowering for
accessing the await input of an async generator
- Modify the JSGeneratorStore operator to understand whether or
not it's suspending for a normal yield, or an AsyncGenerator
Await. This ensures appropriate registers are stored.
- Add versions of ResumeGeneratorTrampoline which store the
input value in a different field depending on wether it's an
AsyncGenerator Await resume, or an ordinary resume. Also modifies
whether debug code will assert that the generator object is a
JSGeneratorObject or a JSAsyncGeneratorObject depending on the
resume type.
BUG=v8:5855
R=bmeurer@chromium.org, rmcilroy@chromium.org, jgruber@chromium.org,
littledan@chromium.org, neis@chromium.orgTBR=marja@chromium.org
Change-Id: I9d58df1d344465fc937fe7eed322424204497187
Reviewed-on: https://chromium-review.googlesource.com/446961
Commit-Queue: Caitlin Potter <caitp@igalia.com>
Reviewed-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Hannes Payer <hpayer@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Jakob Gruber <jgruber@chromium.org>
Cr-Commit-Position: refs/heads/master@{#44240}
This adds optimization and deoptimization counts to the Web UI. Also, the function timeline
now shows optimization and deoptimization marks.
Review-Url: https://codereview.chromium.org/2753543006
Cr-Commit-Position: refs/heads/master@{#44033}
This is a first step towards moving Turbofan code generation off the main thread.
Summary of the changes:
- AssemblerBase no longer has a pointer to the isolate. Instead, its
constructor receives the few things that it needs from the isolate (on most
architectures this is just the serializer_enabled flag).
- RelocInfo no longer has a pointer to the isolate. Instead, the functions
that need it take it as an argument. (There are currently still a few that
implicitly access the isolate through a HeapObject.)
- The MacroAssembler now explicitly holds a pointer to the isolate (before, it
used to get it from the Assembler).
- The jit_cookie also moved from AssemblerBase to the MacroAssemblers, since
it's not used at all in the Assemblers.
- A few architectures implemented parts of the Assembler with the help
of a Codepatcher that is based on MacroAssembler. Since the Assembler no
longer has the isolate, but the MacroAssembler still needs it, this doesn't
work anymore. Instead, these Assemblers now use a new PatchingAssembler.
BUG=v8:6048
Review-Url: https://codereview.chromium.org/2732273003
Cr-Commit-Position: refs/heads/master@{#43890}
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}
Use an opaque format for the frame type marker on the stack, where the
marker is simply shifted left by 1 instead of being a Smi. This allows
us to generate simpler code for frame initialisation, as we can push a
smaller value, decreasing the prologue by 4 bytes and one instruction.
Drive-by: Use the same format for JsFrameMarker.
Change-Id: I812dde9b37869fe20de4148a665d06cf23ce7372
Reviewed-on: https://chromium-review.googlesource.com/443426
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Reviewed-by: Benedikt Meurer <bmeurer@chromium.org>
Reviewed-by: Daniel Clifford <danno@chromium.org>
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Cr-Commit-Position: refs/heads/master@{#43347}
This adds support for deoptimizing into the JSConstructStub after the
receiver instantiation but before the actual constructor invocation.
Such a deoptimization point is needed for cases where instantiation
might be observed (e.g. when new.target is a proxy) and hence might
trigger a deopt.
We use this new deoptimization point for the "after" frame-state the
inliner attaches to {JSCreate} nodes being inserted when constructor
calls are being inlined.
R=jarin@chromium.org
TEST=mjsunit/regress/regress-5638b
BUG=v8:5638
Change-Id: I7c72c807ee8fb76d12e0e9ccab86d970ab1a0efd
Reviewed-on: https://chromium-review.googlesource.com/440125
Reviewed-by: Hannes Payer <hpayer@chromium.org>
Reviewed-by: Jaroslav Sevcik <jarin@chromium.org>
Commit-Queue: Michael Starzinger <mstarzinger@chromium.org>
Cr-Commit-Position: refs/heads/master@{#43149}
TypeFeedbackVectors are strongly rooted by a closure. However, in modern
JavaScript closures are created and abandoned more freely. An important
closure may not be present in the root-set at time of garbage collection,
even though we've cached optimized code and use it regularly. For
example, consider leaf functions in an event dispatching system. They may
well be "hot," but tragically non-present when we collect the heap.
Until now, we've relied on a weak root to cache the feedback vector in
this case. Since there is no way to signal intent or relative importance,
this weak root is as susceptible to clearing as any other weak root at
garbage collection time.
Meanwhile, the feedback vector has become more important. All of our
ICs store their data there. Literal and regex boilerplates are stored there.
If we lose the vector, then we not only lose optimized code built from
it, we also lose the very feedback which allowed us to create that optimized
code. Therefore it's vital to express that dependency through the root
set.
This CL does this by creating a strong link to a feedback
vector at the instantiation site of the function closure.
This instantiation site is in the code and feedback vector
of the outer closure.
BUG=v8:5456
Review-Url: https://codereview.chromium.org/2674593003
Cr-Commit-Position: refs/heads/master@{#42953}
They have the same lifetime. It's a match!
Both structures are native context dependent and dealt with (creation,
clearing, gathering feedback) at the same time. By treating the spaces used
for literal boilerplates as feedback vector slots, we no longer have to keep
track of the materialized literal count elsewhere.
A follow-on CL removes even more parser infrastructure related to this count.
BUG=v8:5456
Review-Url: https://codereview.chromium.org/2655853010
Cr-Commit-Position: refs/heads/master@{#42771}
This makes sure that the deoptimizer preserves the exact bit pattern of
floating-point values (both 32-bit and 64-bit) up to the point where a
potential {HeapNumber} is allocated. It in turn allows us to correctly
recognize the {hole_nan_value} when stored into a {FixedDouleArray}.
R=jarin@chromium.org
TEST=mjsunit/regress/regress-crbug-684208
BUG=chromium:684208
Review-Url: https://codereview.chromium.org/2652303002
Cr-Commit-Position: refs/heads/master@{#42679}
This also introduces exhaustive switch-cases for instance types.
BUG=chromium:681383
Review-Url: https://codereview.chromium.org/2646433002
Cr-Commit-Position: refs/heads/master@{#42447}
This is mainly to catch a crash that we see in Canary with escape
analysis on.
Review-Url: https://codereview.chromium.org/2625893003
Cr-Commit-Position: refs/heads/master@{#42223}
This fixes the corner-case where the method in question failed to lookup
the very last deoptimization bailout without subsequent entries within
the relocation info. Also enable a test covering this.
R=tebbi@chromium.org
TEST=cctest/test-cpu-profiler/CollectDeoptEvents
Review-Url: https://codereview.chromium.org/2565733002
Cr-Commit-Position: refs/heads/master@{#41623}
This ensures the deoptimization triggered due to materialization of
objects by the {TranslatedState} works in conjunction with OSR. The
optimized code used for OSR is not installed on the function, hence
needs to be specified explicitly when requesting deoptimization for
specific stack frames.
R=jarin@chromium.org
TEST=mjsunit/regress/regress-crbug-668795
BUG=chromium:668795
Review-Url: https://codereview.chromium.org/2534143002
Cr-Commit-Position: refs/heads/master@{#41348}
The range-based exception handler table is by now only used for bytecode
arrays. The semantics of the interpreter are that bytecode offsets point
to the beginning of the currently executing bytecode instruction. Uses
hence need to compensate for lookups based on a "retrun address". This
change removes the need for such off-by-one compensations by changing
lookup semantics to be based on "current instruction" offsets.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2534893002
Cr-Commit-Position: refs/heads/master@{#41339}
This removes the supporting function to perform a range-lookup in the
exception handler table for unoptimized code. Such tables are by now
guaranteed to be empty, the deoptimizer cannot encounter this case.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2529343003
Cr-Commit-Position: refs/heads/master@{#41318}
When generating snapshot on a machine with a different page size than
the target machine, we can run into problems as the v8 page area size
changes. This is because v8 has page guards which depend on os page
size, so if the target has larger os page, v8 page area is smaller and
may not fit the contents.
The solution proposed here is adding a flag, v8_os_page_size, that
would, if used, override local os page size and use the one specified
during snapshot generation.
BUG=
Review-Url: https://codereview.chromium.org/2488403003
Cr-Commit-Position: refs/heads/master@{#40997}
SourcePosition::InliningId() refers to a the new table DeoptimizationInputData::InliningPositions(), which provides the following data for every inlining id:
- The inlined SharedFunctionInfo as an offset into DeoptimizationInfo::LiteralArray
- The SourcePosition of the inlining. Recursively, this yields the full inlining stack.
Before the Code object is created, the same information can be found in CompilationInfo::inlined_functions().
If SourcePosition::InliningId() is SourcePosition::kNotInlined, it refers to the outer (non-inlined) function.
So every SourcePosition has full information about its inlining stack, as long as the corresponding Code object is known. The internal represenation of a source position is a positive 64bit integer.
All compilers create now appropriate source positions for inlined functions. In the case of Turbofan, this required using AstGraphBuilderWithPositions for inlined functions too. So this class is now moved to a header file.
At the moment, the additional information in source positions is only used in --trace-deopt and --code-comments. The profiler needs to be updated, at the moment it gets the correct script offsets from the deopt info, but the wrong script id from the reconstructed deopt stack, which can lead to wrong outputs. This should be resolved by making the profiler use the new inlining information for deopts.
I activated the inlined deoptimization tests in test-cpu-profiler.cc for Turbofan, changing them to a case where the deopt stack and the inlining position agree. It is currently still broken for other cases.
The following additional changes were necessary:
- The source position table (internal::SourcePositionTableBuilder etc.) supports now 64bit source positions. Encoding source positions in a single 64bit int together with the difference encoding in the source position table results in very little overhead for the inlining id, since only 12% of the source positions in Octane have a changed inlining id.
- The class HPositionInfo was effectively dead code and is now removed.
- SourcePosition has new printing and information facilities, including computing a full inlining stack.
- I had to rename compiler/source-position.{h,cc} to compiler/compiler-source-position-table.{h,cc} to avoid clashes with the new src/source-position.cc file.
- I wrote the new wrapper PodArray for ByteArray. It is a template working with any POD-type. This is used in DeoptimizationInputData::InliningPositions().
- I removed HInlinedFunctionInfo and HGraph::inlined_function_infos, because they were only used for the now obsolete Crankshaft inlining ids.
- Crankshaft managed a list of inlined functions in Lithium: LChunk::inlined_functions. This is an analog structure to CompilationInfo::inlined_functions. So I removed LChunk::inlined_functions and made Crankshaft use CompilationInfo::inlined_functions instead, because this was necessary to register the offsets into the literal array in a uniform way. This is a safe change because LChunk::inlined_functions has no other uses and the functions in CompilationInfo::inlined_functions have a strictly longer lifespan, being created earlier (in Hydrogen already).
BUG=v8:5432
Review-Url: https://codereview.chromium.org/2451853002
Cr-Commit-Position: refs/heads/master@{#40975}
This changes {FrameState} nodes modeling "after" states to use bytecode
offsets pointing to the deoptimizing bytecode. This is in sync with the
normal execution, as the bytecode offset is advanced after operations
complete in regular bytecode handlers.
The change is necessary to ensure lazy deoptimized frames contain an
accurate bytecode offset while they are on the stack. Such frames can be
inspected by various stack walks. The continuation builtin will advance
the bytecode offset upon return.
R=jarin@chromium.org
TEST=mjsunit/regress/regress-crbug-660379
BUG=chromium:660379
Review-Url: https://codereview.chromium.org/2487173002
Cr-Commit-Position: refs/heads/master@{#40887}
Exchanged the ZoneList for a ZoneChunkList to avoid
unnecessary growing.
Review-Url: https://codereview.chromium.org/2468183004
Cr-Commit-Position: refs/heads/master@{#40736}
This slot is completely unused and always undefined anyways, so there's
no need to maintain the slot during object construction.
R=yangguo@chromium.org
BUG=v8:5049
Review-Url: https://codereview.chromium.org/2423323002
Cr-Commit-Position: refs/heads/master@{#40377}
This adds more useful information to the v8-heap-stats tool.
BUG=v8:5489
Review-Url: https://codereview.chromium.org/2394213003
Cr-Commit-Position: refs/heads/master@{#40361}
This reverts commit 7db0ecdec3.
Manual revert since automatic revert is too large for the web interface.
BUG=
TBR=bmeurer@chromium.org,mstarzinger@chromium.org,yangguo@chromium.org,ahaas@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
Review-Url: https://codereview.chromium.org/2396353002
Cr-Commit-Position: refs/heads/master@{#40082}
This add support for ConsString objects allocated inline to the escape
analysis pass. The raw hash field in such strings needs special handling
similar to existing raw fields. This also contains materialization code
within the deoptimizer as usual.
R=bmeurer@chromium.org
TEST=mjsunit/regress/regress-crbug-648737
BUG=chromium:648737
Review-Url: https://codereview.chromium.org/2357153002
Cr-Commit-Position: refs/heads/master@{#39594}
This fixes the materialization of JSFunction objects to not rely on a
context being available. The context has been cleared because it might
be de-materiallized itself.
R=bmeurer@chromium.org
TEST=mjsunit/compiler/escape-analysis-materialize
BUG=chromium:644245
Review-Url: https://codereview.chromium.org/2320983002
Cr-Commit-Position: refs/heads/master@{#39277}
This fixes the materialization of JSArray objects to not rely on a
context being available. The context has been cleared because it might
be de-materiallized itself.
R=bmeurer@chromium.org
BUG=chromium:644245
Review-Url: https://codereview.chromium.org/2323713002
Cr-Commit-Position: refs/heads/master@{#39274}
This adds support to the deoptimizer to materialize ContextExtension
objects that have been de-materialized by escape analysis. This is
follow-up to the inline allocation of such objects during the create
lowering phase (i.e. JSCreateWithContext and JSCreateCatchContext).
R=bmeurer@chromium.org
TEST=mjsunit/regress/regress-crbug-644245
BUG=chromium:644245
Review-Url: https://codereview.chromium.org/2317353003
Cr-Commit-Position: refs/heads/master@{#39270}
This clears the context register by setting it to Smi(0) before calling
the Runtime::kNotifyDeoptimized helper. The deoptimizer must be able to
materialize all heap objects without any context available. The context
itself might be dematerialized.
With this change we make sure that invariant is maintained even without
escape analysis kicking in. We also satisfy the check that the context
register is either Smi(0) or a valid context. It might have been the
special {arguments_marker} in this particular case.
R=bmeurer@chromium.org
BUG=chromium:644245
Review-Url: https://codereview.chromium.org/2320673002
Cr-Commit-Position: refs/heads/master@{#39267}
This adds support for dematerialized context values as part of an
interpreted frame (similar to an FCG frame). Both frame translations
should be kept in sync as much as possible.
R=rmcilroy@chromium.org
BUG=chromium:644728
Review-Url: https://codereview.chromium.org/2313343002
Cr-Commit-Position: refs/heads/master@{#39256}
After we landed the tracing runtime call stats, which gave
us a lot of V8 insight in tracing, we noticed that there is
some arising issues and discrepancies.
Issues include:
Missing trace events, that happened due to
transforming those trace events into runtime calls
Discrepancies include:
Missing categories in Runtime call stats like GC,
because we are not handling the Scoped runtime calls
properly in the tracing version.
To reduce/eliminate those issue, we are taking a small
step back. We are unifying the RuntimeStats code and
using the original one. That would allow us to use all
the original probes but emit trace events from them.
We are also putting back the trace-events in their place.
The output from both system should be intact (Except of
the addition of the missing trace-events).
Also as a byproduct, we are reducing the number of context
scopes by half since we are using the same scope as
runtime call stats.
As a follow up to this CL, we will address the non-scoped
Runtime Call Stats (mainly in GC).
BUG=642373
Review-Url: https://codereview.chromium.org/2296243002
Cr-Commit-Position: refs/heads/master@{#39180}
We removed some TRACE_EVENT0 instrumentation when we landed runtime call stats
tracing, however some of them are used in metrics, this patch adds these
TRACE_EVENT0 back.
BUG=642373
LOG=N
Review-Url: https://codereview.chromium.org/2301743002
Cr-Commit-Position: refs/heads/master@{#39104}
The accumulator is always part of the translation for every interpreted
frame. The assumption is that all frames are in {TOS_REGISTER} state.
This however is not supported for non-topmost frames and we need to
avoid pushing the accumulator onto the machine stack.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2271153003
Cr-Commit-Position: refs/heads/master@{#38945}
This makes sure the deoptimizer picks bytecode prepared for debugging
when materializing an interpreted frame if one is available. This is
normally done by the interpreter entry trampoline and hence needs to be
replicated by the deoptimizer.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2271443003
Cr-Commit-Position: refs/heads/master@{#38815}
With --ignition-preserve-bytecode, we don't have the guarantee that
SharedFunctionInfo::abstract_code() returns the code we deopt to.
R=mstarzinger@chromium.org
BUG=v8:5265
Review-Url: https://codereview.chromium.org/2239773003
Cr-Commit-Position: refs/heads/master@{#38614}
Currently we have V8 RuntimeCallStats that is independently from tracing when
running d8 with flag --runtime_call_stats. This patch embeds V8 runtime call
stats into tracing, by having a global table of runtime call counters each
isolate, resetting the table each time we enter a top level trace event, and
dumping the table for each top level trace event. This will make trace file more
compat, as well as enable runtime call stats in tracing system.
This patch adds ~5% overhead to V8 when the category is enabled, we measure the
overhead by running a script when category is enabled.
BUG=v8:5089
Committed: https://crrev.com/d014866173eaa2b548c566217b2c94b1d49385fa
Committed: https://crrev.com/1ca3b73bba4a7253ca8eeef39321d70e7d414331
Committed: https://crrev.com/3f936a5b17754783e92d2146eaf66c88a78ee45b
Review-Url: https://codereview.chromium.org/2187693002
Cr-Original-Original-Original-Commit-Position: refs/heads/master@{#38270}
Cr-Original-Original-Commit-Position: refs/heads/master@{#38314}
Cr-Original-Commit-Position: refs/heads/master@{#38403}
Cr-Commit-Position: refs/heads/master@{#38510}
Reason for revert:
Leaks block chromium roll:
https://codereview.chromium.org/2219083003/
Example build:
https://build.chromium.org/p/tryserver.chromium.linux/builders/linux_chromium_asan_rel_ng/builds/205435
You can add the trybot from tryserver.chromium.linux, linux_chromium_asan_rel_ng, on reland.
Original issue's description:
> [Reland][Tracing] Embed V8 runtime call stats into tracing.
>
> Currently we have V8 RuntimeCallStats that is independently from tracing when
> running d8 with flag --runtime_call_stats. This patch embeds V8 runtime call
> stats into tracing, by having a global table of runtime call counters each
> isolate, resetting the table each time we enter a top level trace event, and
> dumping the table for each top level trace event. This will make trace file more
> compat, as well as enable runtime call stats in tracing system.
>
> This patch adds ~5% overhead to V8 when the category is enabled, we measure the
> overhead by running a script when category is enabled.
>
> BUG=v8:5089
>
> Committed: https://crrev.com/d014866173eaa2b548c566217b2c94b1d49385fa
> Committed: https://crrev.com/1ca3b73bba4a7253ca8eeef39321d70e7d414331
> Committed: https://crrev.com/3f936a5b17754783e92d2146eaf66c88a78ee45b
> Cr-Original-Original-Commit-Position: refs/heads/master@{#38270}
> Cr-Original-Commit-Position: refs/heads/master@{#38314}
> Cr-Commit-Position: refs/heads/master@{#38403}
TBR=cbruni@chromium.org,fmeawad@chromium.org,bmeurer@chromium.org,adamk@chromium.org,rmcilroy@chromium.org,lpy@chromium.org
# Not skipping CQ checks because original CL landed more than 1 days ago.
BUG=v8:5089
NOTRY=true
NOPRESUBMIT=true
Review-Url: https://codereview.chromium.org/2221853002
Cr-Commit-Position: refs/heads/master@{#38412}
Currently we have V8 RuntimeCallStats that is independently from tracing when
running d8 with flag --runtime_call_stats. This patch embeds V8 runtime call
stats into tracing, by having a global table of runtime call counters each
isolate, resetting the table each time we enter a top level trace event, and
dumping the table for each top level trace event. This will make trace file more
compat, as well as enable runtime call stats in tracing system.
This patch adds ~5% overhead to V8 when the category is enabled, we measure the
overhead by running a script when category is enabled.
BUG=v8:5089
Committed: https://crrev.com/d014866173eaa2b548c566217b2c94b1d49385fa
Committed: https://crrev.com/1ca3b73bba4a7253ca8eeef39321d70e7d414331
Review-Url: https://codereview.chromium.org/2187693002
Cr-Original-Original-Commit-Position: refs/heads/master@{#38270}
Cr-Original-Commit-Position: refs/heads/master@{#38314}
Cr-Commit-Position: refs/heads/master@{#38403}
Reason for revert:
Mac64 ASAN failure. https://build.chromium.org/p/client.v8/builders/V8%20Mac64%20ASAN/builds/7810/steps/Check/logs/Threading4
Original issue's description:
> [Reland][Tracing] Embed V8 runtime call stats into tracing.
>
> Currently we have V8 RuntimeCallStats that is independently from tracing when
> running d8 with flag --runtime_call_stats. This patch embeds V8 runtime call
> stats into tracing, by having a global table of runtime call counters each
> isolate, resetting the table each time we enter a top level trace event, and
> dumping the table for each top level trace event. This will make trace file more
> compat, as well as enable runtime call stats in tracing system.
>
> This patch adds ~5% overhead to V8 when the category is enabled, we measure the
> overhead by running a script when category is enabled.
>
> BUG=v8:5089
>
> Committed: https://crrev.com/d014866173eaa2b548c566217b2c94b1d49385fa
> Committed: https://crrev.com/1ca3b73bba4a7253ca8eeef39321d70e7d414331
> Cr-Original-Commit-Position: refs/heads/master@{#38270}
> Cr-Commit-Position: refs/heads/master@{#38314}
TBR=cbruni@chromium.org,fmeawad@chromium.org,machenbach@chromium.org,bmeurer@chromium.org,adamk@chromium.org,rmcilroy@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:5089
Review-Url: https://codereview.chromium.org/2200373003
Cr-Commit-Position: refs/heads/master@{#38315}
Currently we have V8 RuntimeCallStats that is independently from tracing when
running d8 with flag --runtime_call_stats. This patch embeds V8 runtime call
stats into tracing, by having a global table of runtime call counters each
isolate, resetting the table each time we enter a top level trace event, and
dumping the table for each top level trace event. This will make trace file more
compat, as well as enable runtime call stats in tracing system.
This patch adds ~5% overhead to V8 when the category is enabled, we measure the
overhead by running a script when category is enabled.
BUG=v8:5089
Committed: https://crrev.com/d014866173eaa2b548c566217b2c94b1d49385fa
Review-Url: https://codereview.chromium.org/2187693002
Cr-Original-Commit-Position: refs/heads/master@{#38270}
Cr-Commit-Position: refs/heads/master@{#38314}
Reason for revert:
Sanitizer failures:
https://build.chromium.org/p/client.v8/builders/V8%20Linux64%20TSAN/builds/10989/https://build.chromium.org/p/client.v8/builders/V8%20Mac64%20ASAN/builds/7786/
Original issue's description:
> [Tracing] Embed V8 runtime call stats into tracing.
>
> Currently we have V8 RuntimeCallStats that is independently from tracing when
> running d8 with flag --runtime_call_stats. This patch embeds V8 runtime call
> stats into tracing, by having a global table of runtime call counters each
> isolate, resetting the table each time we enter a top level trace event, and
> dumping the table for each top level trace event. This will make trace file more
> compat, as well as enable runtime call stats in tracing system.
>
> This patch adds ~5% overhead to V8 when the category is enabled, we measure the
> overhead by running a script when category is enabled.
>
> BUG=v8:5089
>
> Committed: https://crrev.com/d014866173eaa2b548c566217b2c94b1d49385fa
> Cr-Commit-Position: refs/heads/master@{#38270}
TBR=cbruni@chromium.org,fmeawad@chromium.org,machenbach@chromium.org,bmeurer@chromium.org,adamk@chromium.org,rmcilroy@chromium.org,lpy@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:5089
Review-Url: https://codereview.chromium.org/2203913004
Cr-Commit-Position: refs/heads/master@{#38275}
Currently we have V8 RuntimeCallStats that is independently from tracing when
running d8 with flag --runtime_call_stats. This patch embeds V8 runtime call
stats into tracing, by having a global table of runtime call counters each
isolate, resetting the table each time we enter a top level trace event, and
dumping the table for each top level trace event. This will make trace file more
compat, as well as enable runtime call stats in tracing system.
This patch adds ~5% overhead to V8 when the category is enabled, we measure the
overhead by running a script when category is enabled.
BUG=5089
Review-Url: https://codereview.chromium.org/2187693002
Cr-Commit-Position: refs/heads/master@{#38270}
This allows to pass deoptimization reasons to the profiler without the
requirement of always providing a source position. The absence of deopt
reasons is now communicated via a sentinel as the deopt id value. The
deoptimization reasons recently added to TurboFan are now passed to the
profiler.
R=bmeurer@chromium.org
TEST=cctest/test-cpu-profiler
Review-Url: https://codereview.chromium.org/2159793002
Cr-Commit-Position: refs/heads/master@{#37852}
So far TurboFan wasn't adding the deoptimization reasons for eager/soft
deoptimization exits that can be used by either the DevTools profiler or
the --trace-deopt flag. This adds basic support for deopt reasons on
Deoptimize, DeoptimizeIf and DeoptimizeUnless nodes and threads through
the reasons to the code generation.
Also moves the DeoptReason to it's own file (to resolve include cycles)
and drops unused reasons.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/2161543002
Cr-Commit-Position: refs/heads/master@{#37823}
Now we can turn it into a turbofan stub.
Create new instance types JS_ARGUMENTS_TYPE and JS_ERROR_TYPE.
Review-Url: https://codereview.chromium.org/2080243003
Cr-Commit-Position: refs/heads/master@{#37219}
The patch introduces a dedicated dispatching class for JIT code events. It is
set as a helper on the isolate.
This allows classes across v8 to break their dependency on Logger and CpuProfiler.
These two became just regular clients of the dispatcher.
BUG=v8:4789
Review-Url: https://codereview.chromium.org/2061623002
Cr-Commit-Position: refs/heads/master@{#37005}
Passing in the isolate and pointer compare the instnance against the
corresponding constant is always faster than decoding the instance types.
BUG=
Review-Url: https://codereview.chromium.org/2028983002
Cr-Commit-Position: refs/heads/master@{#36744}
This allows sloppy arguments objects to be allocated by the generic
Heap::AllocateJSObjectFromMap allocation function. An empty elements
backing store for such objects is provided. This is needed in order to
materialize such objects without access to a specific native context.
R=bmeurer@chromium.org
TEST=mjsunit/regress/regress-crbug-613919
BUG=chromium:613919
Review-Url: https://codereview.chromium.org/2014343004
Cr-Commit-Position: refs/heads/master@{#36547}
This makes sure the deoptimizer can materialize sloppy arguments objects
with the FAST_SLOPPY_ARGUMENTS_ELEMENTS elements kind. TurboFan's escape
analysis treates those as normal JSObject types and hence materializes
them differently than Crankshaft does.
R=verwaest@chromium.org
TEST=mjsunit/regress/regress-crbug-613919
BUG=chromium:613919
Review-Url: https://codereview.chromium.org/2001133002
Cr-Commit-Position: refs/heads/master@{#36440}
This adds support to materialize JSFunction objects from deoptimization
information. By now we need to support this because TurboFan's escape
analysis can produce virtual (i.e. non-escaping) closures.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/1998143002
Cr-Commit-Position: refs/heads/master@{#36400}
In order to support compiling to baseline on return we need to be able to
return to the actual return address. With this change this is what the
Return bytecode now does, removing the need for the
InterpreterExitTrampoline.
This change also removes the InterpreterNotifyDeoptXXX builtins and
unifies FCG and Igntion to both use NotifyDeoptXXX. As part of this
change, FullCodegenerator::State is moved to Deoptimize::BailoutState.
BUG=v8:4280
LOG=N
Committed: https://crrev.com/34c9626e2ee56fe805de549697ca5323aed7cb66
Cr-Commit-Position: refs/heads/master@{#36288}
Review-Url: https://codereview.chromium.org/1969423002
Cr-Commit-Position: refs/heads/master@{#36310}
Reason for revert:
Breaks
https://build.chromium.org/p/client.v8.ports/builders/V8%20Linux%20-%20arm%20-%20sim/builds/619
Might only affect pure release builds?
Original issue's description:
> [Interpreter] Remove InterpreterExitTrampoline and replace with returning to the entry trampoline.
>
> In order to support compiling to baseline on return we need to be able to
> return to the actual return address. With this change this is what the
> Return bytecode now does, removing the need for the
> InterpreterExitTrampoline.
>
> This change also removes the InterpreterNotifyDeoptXXX builtins and
> unifies FCG and Igntion to both use NotifyDeoptXXX. As part of this
> change, FullCodegenerator::State is moved to Deoptimize::BailoutState.
>
> BUG=v8:4280
> LOG=N
>
> Committed: https://crrev.com/34c9626e2ee56fe805de549697ca5323aed7cb66
> Cr-Commit-Position: refs/heads/master@{#36288}
TBR=mstarzinger@chromium.org,oth@chromium.org,rmcilroy@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:4280
Review-Url: https://codereview.chromium.org/1986353002
Cr-Commit-Position: refs/heads/master@{#36294}
In order to support compiling to baseline on return we need to be able to
return to the actual return address. With this change this is what the
Return bytecode now does, removing the need for the
InterpreterExitTrampoline.
This change also removes the InterpreterNotifyDeoptXXX builtins and
unifies FCG and Igntion to both use NotifyDeoptXXX. As part of this
change, FullCodegenerator::State is moved to Deoptimize::BailoutState.
BUG=v8:4280
LOG=N
Review-Url: https://codereview.chromium.org/1969423002
Cr-Commit-Position: refs/heads/master@{#36288}
By fully annotating the API with runtime counters we can properly measure
how much time we spend in total in v8. When --runtime-call-stats is specified
we now disable the fast-paths for callbacks to properly measure them.
As a drive-by-fix this CL unifies the LOG messages in api.cc.
Additionally we added missing timers to gain better resolution in the parser
and callbacks.
BUG=
Review-Url: https://codereview.chromium.org/1923893002
Cr-Commit-Position: refs/heads/master@{#36248}
This makes the profiler reconstruct inlined function frames at eager
deopt points from the deoptimization data. The main goal of this is to
remove the last side-channel where Crankshaft communicates directly to
the profiler. This is the last preparatory step towards deprecating the
side-channel in question.
R=yangguo@chromium.org
Review-Url: https://codereview.chromium.org/1973993002
Cr-Commit-Position: refs/heads/master@{#36229}
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}
This passes the inlining_id of deoptimization points via the relocation
info instead of via a side-channel to the CPU profiler. This is one step
towards deprecating the side-channel in question and avoid the need for
performing a lookup of the return address of the deopt point.
R=jarin@chromium.org
Review-Url: https://codereview.chromium.org/1956693002
Cr-Commit-Position: refs/heads/master@{#36177}
- Move the concurrent unmapping to MemoryAllocator
- Hide (private) members where possible
- MemoryAllocator:Free is now the bottleneck for freeing
- Pooled pages are either allocated from a set of pooled pages are obtained
through work stealing from the concurrent unmapper
BUG=chromium:605866, chromium:581412
LOG=N
This reverts commit 25ff296c4c.
Review-Url: https://codereview.chromium.org/1929503002
Cr-Commit-Position: refs/heads/master@{#35910}
Reason for revert:
Creates a spike of OOM(v8) crashers on Win32.
Original issue's description:
> [heap] Uncommit pooled pages concurrently
>
> - Move the concurrent unmapping to MemoryAllocator
> - Hide (private) members where possible
> - MemoryAllocator:Free is now the bottleneck for freeing
> - Pooled pages are either allocated from a set of pooled pages are obtained
> through work stealing from the concurrent unmapper
>
> BUG=chromium:605866, chromium:581412
> LOG=N
>
> Committed: https://crrev.com/2158df87116906160cebc3ad20c97f454822da03
> Cr-Commit-Position: refs/heads/master@{#35797}
TBR=hpayer@chromium.org,mlippautz@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:605866, chromium:581412
Review URL: https://codereview.chromium.org/1925563003
Cr-Commit-Position: refs/heads/master@{#35819}
- Move the concurrent unmapping to MemoryAllocator
- Hide (private) members where possible
- MemoryAllocator:Free is now the bottleneck for freeing
- Pooled pages are either allocated from a set of pooled pages are obtained
through work stealing from the concurrent unmapper
BUG=chromium:605866, chromium:581412
LOG=N
Review URL: https://codereview.chromium.org/1913083002
Cr-Commit-Position: refs/heads/master@{#35797}
The approximate mode enables taking an approximate stack trace from GC,
where the top frames might be missing if inlined. Note that in that case,
the frame summary will refer to optimized code, so it will not be possible
to take source position. (The user of the summary will have to handle the
case frame_summary.abstract_code()->kind() == AbstractCode::OPTIMIZED_CODE
specially.)
Review URL: https://codereview.chromium.org/1907443002
Cr-Commit-Position: refs/heads/master@{#35728}
Introduce a ResumeGeneratorTrampoline, which does the actual stack state
reconstruction (currently always restores a fullcodegen frame), and
introduce appropriate TurboFan builtins for %GeneratorPrototype%.next,
%GeneratorPrototype%.return and %GeneratorPrototype%.throw based on
this native builtin.
Also unify the flooding in case of step-in to always work based on
JSFunction and remove the special casing for JSGeneratorObject.
R=mstarzinger@chromium.org, neis@chromium.orgTBR=rossberg@chromium.org
BUG=chromium:513471
LOG=n
Review URL: https://codereview.chromium.org/1865833002
Cr-Commit-Position: refs/heads/master@{#35283}
We expect that the majority of malloc'd memory held by V8 is allocated
in Zone objects. Introduce an Allocator class that is used by Zones to
manage memory, and allows for querying the current usage.
BUG=none
R=titzer@chromium.org,bmeurer@chromium.org,jarin@chromium.org
LOG=n
TBR=rossberg@chromium.org
Review URL: https://codereview.chromium.org/1847543002
Cr-Commit-Position: refs/heads/master@{#35196}
These checks can fail if there bottommost function is a tail caller and the next function has different number of arguments than the bottommost one.
BUG=chromium:593697,v8:4698
LOG=N
Review URL: https://codereview.chromium.org/1785253003
Cr-Commit-Position: refs/heads/master@{#34718}
In case when F was called with incompatible number of arguments (and therefore
the arguments adator frame was created), F inlines a tail call of G which then
deopts the deoptimizer should also remove the arguments adaptor frame for F.
This CL adds required machinery to the deoptimizer.
BUG=v8:4698
LOG=N
Review URL: https://codereview.chromium.org/1768263004
Cr-Commit-Position: refs/heads/master@{#34610}
BailoutId points to the next bytecode in the bytecode array. Code offset
is set to one less than the bail out id. This would point to the end of the
current instruction. Since we use it only for summarizing the frame and to
compute the source position, it should be safe to set it to the end of current
instruction.
BUG=v8:4280, v8:4689
LOG=N
Review URL: https://codereview.chromium.org/1763783003
Cr-Commit-Position: refs/heads/master@{#34580}
Before this CL, various code stubs used different techniques
for marking their frames to enable stack-crawling and other
access to data in the frame. All of them were based on a abuse
of the "standard" frame representation, e.g. storing the a
context pointer immediately below the frame's fp, and a
function pointer after that. Although functional, this approach
tends to make stubs and builtins do an awkward, unnecessary
dance to appear like standard frames, even if they have
nothing to do with JavaScript execution.
This CL attempts to improve this by:
* Ensuring that there are only two fundamentally different
types of frames, a "standard" frame and a "typed" frame.
Standard frames, as before, contain both a context and
function pointer. Typed frames contain only a minimum
of a smi marker in the position immediately below the fp
where the context is in standard frames.
* Only interpreted, full codegen, and optimized Crankshaft and
TurboFan JavaScript frames use the "standard" format. All
other frames use the type frame format with an explicit
marker.
* Typed frames can contain one or more values below the
type marker. There is new magic macro machinery in
frames.h that simplifies defining the offsets of these fields
in typed frames.
* A new flag in the CallDescriptor enables specifying whether
a frame is a standard frame or a typed frame. Secondary
register location spilling is now only enabled for standard
frames.
* A zillion places in the code have been updated to deal with
the fact that most code stubs and internal frames use the
typed frame format. This includes changes in the
deoptimizer, debugger, and liveedit.
* StandardFrameConstants::kMarkerOffset is deprecated,
(CommonFrameConstants::kContextOrFrameTypeOffset
and StandardFrameConstants::kFrameOffset are now used
in its stead).
LOG=N
Review URL: https://codereview.chromium.org/1696043002
Cr-Commit-Position: refs/heads/master@{#34571}
This is a prerequisite for teaching deoptimizer to drop possible arguments adapter frame below current input frame which is needed to support tail call inlining.
Review URL: https://codereview.chromium.org/1761303002
Cr-Commit-Position: refs/heads/master@{#34533}
The support was already removed from the Crankshaft and this CL removes the last piece.
Review URL: https://codereview.chromium.org/1764593002
Cr-Commit-Position: refs/heads/master@{#34456}
Frame slots indexes numbers are used more consistently for
computation in both TurboFan and Crankshaft. Specifically,
Crankshaft now uses frame slot indexes in LChunk, removing
the need for some special-case maths when building the
deoptimization translation table.
LOG=N
R=mstarzinger@chromium.org
Committed: https://crrev.com/81423b84dbb2eaf7e1a57b0f6029fc8e643b4755
Cr-Commit-Position: refs/heads/master@{#34078}
Review URL: https://codereview.chromium.org/1702593002
Cr-Commit-Position: refs/heads/master@{#34114}
Moves the accumulator value on-heap to be restored in the
InterpreterNotifyDeopt handler rather than explicitly
setting the accumulator register. This allows it to be
materialized correctly if required.
BUG=v8:4678
LOG=N
Review URL: https://codereview.chromium.org/1707133003
Cr-Commit-Position: refs/heads/master@{#34113}
This CL adds a TRACE_EVENT where there is an isolated LOG, a HistogramTimer
or a TimerEvent.
Once we have a d8 tracing controller, all TimerEvents will be removed since
they do not provide an added value over TRACE_EVENTs. HistogramTimers will
remain, but their functionality will be limited to Histograms only.
BUG=v8:4562
LOG=N
Review URL: https://codereview.chromium.org/1707563002
Cr-Commit-Position: refs/heads/master@{#34099}
Replaces the push of the dispatch table on the interpreted stack frame with a
push of the bytecode array. This enables the debugger to replace the bytecode
array with a patched version containing breakpoints.
BUG=v8:4690
LOG=N
Review URL: https://codereview.chromium.org/1699013002
Cr-Commit-Position: refs/heads/master@{#34032}
Before this CL, the context of the parent frame was used when deoptimizing a
stub failure rather than the context value passed to the stub itself. In order
to guarantee that the right context is passed to the runtime upon stub failure,
this CL adds the context explicitly to the stub's environment that's used to
compute the failure deoptimizing translations. The context can then be extracted
during deoptimization translation to ensure that the precise context that was
passed to the stub is also passed to the runtime.
R=jarin@chromium.org
LOG=N
Review URL: https://codereview.chromium.org/1694183003
Cr-Commit-Position: refs/heads/master@{#34030}
V8 tracks already most useful information, but lacks proper tracing scopes
that make it possible to distinguish certain events from each other.
- add trace-scope to track lazy-parsing due to optimization
- add trace-scope to track code optimization
BUG=
Review URL: https://codereview.chromium.org/1661883003
Cr-Commit-Position: refs/heads/master@{#34002}
This is mostly preparation for allowing the function closure to be materialized.
As a drive-by fix, I have added ignition source position support to the frame inspector (this fixed some ignition test failures).
Review URL: https://codereview.chromium.org/1698743002
Cr-Commit-Position: refs/heads/master@{#33975}
The idea here is to perform the handler lookup in the deoptimizer, and then take the information from the handler table to build the catch handler frame in the deoptimizer. Specifically, we use the pc offset, context location and stack height (in full-code) to tweak the output frame.
Sadly, this still requires nasty voodoo for the liveness analyzer so that it keeps variables alive if they are used in the catch handler.
Review URL: https://codereview.chromium.org/1416543006
Cr-Commit-Position: refs/heads/master@{#33936}
Saves and restores the dispatch pointer during calls to enable the debugger to
switch the dispatch table used by a function during it's execution.
Also moves the accumulator and context nodes to be Variables so that they will
be properly merged across branches.
BUG=v8:4280,v8:4690
LOG=N
Review URL: https://codereview.chromium.org/1684073002
Cr-Commit-Position: refs/heads/master@{#33894}
This removes uses of JSFunction by the (proper) deoptimizer. This will be useful
when we escape analyze JSFunction away. Unfortunately, the debugger still needs
JSFunction, so escape analysis would not work yet.
Review URL: https://codereview.chromium.org/1686183003
Cr-Commit-Position: refs/heads/master@{#33891}
JS_FRAME_FUNCTION can be expressed using the STACK_SLOT translation.
Review URL: https://codereview.chromium.org/1688023003
Cr-Commit-Position: refs/heads/master@{#33882}
There are a bunch of places in our builtins where we use %_Arguments and
%_ArgumentsLength for no good reason, as arguments object and/or rest
parameter is as good and performant in these cases. Now the only uses
of %_Arguments and %_ArgumentsLength left are in string.js, which
requires dedicated investigation.
CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux_nosnap_dbg
R=yangguo@chromium.org
Committed: https://crrev.com/2160429fd458e3c095475e718c97f77ac90d906f
Cr-Commit-Position: refs/heads/master@{#33834}
Review URL: https://codereview.chromium.org/1678953004
Cr-Commit-Position: refs/heads/master@{#33881}
This change adds AbstractCode, which can be either Code or
BytecodeArray, and adds methods to calculate source position based
on that. Also cleans up to use code offsets instead of raw PC
where possible, and consistently uses the offset from instruction
start (as opposed to code object start).
R=rmcilroy@chromium.org, vogelheim@chromium.org
BUG=v8:4690
LOG=N
Review URL: https://codereview.chromium.org/1618343002
Cr-Commit-Position: refs/heads/master@{#33579}
Rename IntepreterExceptionEntryHandler builtin to InterpreterEnterBytecodeDispatch
and use it as the return address when building interpreter frames during deopt.
This ensures that we restart execution of the outer frame at the correct
bytecode.
BUG=v8:4280,v8:4678
LOG=N
Review URL: https://codereview.chromium.org/1633633002
Cr-Commit-Position: refs/heads/master@{#33512}
Escape analysis will eliminate the context, if possible. Materialization
must then ensure that the materialized object has the right map, and that
the context register is set accordingly.
R=jarin@chromium.org
BUG=569530,v8:4586
LOG=n
Review URL: https://codereview.chromium.org/1572293002
Cr-Commit-Position: refs/heads/master@{#33201}
Deopt support is added on two levels. On the IR level,
a new ObjectState node is added, which represenents an
object to be materialized. ObjectState nodes appear as
inputs of FrameState and StateValues nodes. On the
instruction select/code-generation level, the
FrameStateDescriptor class handles the nesting
introduced by ObjectState, and ensures that deopt code
with CAPTURED_OBJECT/DUPLICATED_OBJECT entries are
generated similarly to what crankshaft's escape
analysis does.
Two unittests test correctness of the IR level implementation.
Correctness for instruction selection / code generation
is tested by mjsunit tests.
R=jarin@chromium.org,mstarzinger@chromium.org
BUG=v8:4586
LOG=n
Review URL: https://codereview.chromium.org/1485183002
Cr-Commit-Position: refs/heads/master@{#33115}
Adds support for generating deoptimization translations for interpreter
stack frames, and building interpreter frames for these translations
when a function deopts. Also adds builtins for
InterpreterNotifyDeoptimized which resume the function's continuation at
the correct point in the interpreter after deopt.
MIPS patch contributed by balazs.kilvady@igmtec.com
BUG=v8:4280
LOG=N
TEST=test-deoptimization.cc with --ignition and --turbo
Review URL: https://codereview.chromium.org/1528913003
Cr-Commit-Position: refs/heads/master@{#32971}
This drops the specific slot containing the new.target value from our
construct stub frames. This side-channel has been deprecated and will
no longer be accessed by any consumers.
R=verwaest@chromium.org
Review URL: https://codereview.chromium.org/1489353004
Cr-Commit-Position: refs/heads/master@{#32550}
Moves all files related to AST and scopes into ast/,
and all files related to scanner & parser to parsing/.
Also eliminates a couple of spurious dependencies.
R=mstarzinger@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1481613002
Cr-Commit-Position: refs/heads/master@{#32351}
Currently we still (mis)used some machine operators in typed lowering
(namely Word32Or, Word32Xor and Word32And). But these operators are
"polymorphic" in the signedness of their inputs and output, hence the
representation selection (and thereby simplified lowering) was unable to
figure out whether a bitwise operation that was seen would produce an
unsigned or a signed result. If such nodes also have frame state uses,
the only safe choice was float64, which was not only a lot less ideal,
but also the main cause of the for-in related deoptimizer loops.
Adding dedicated NumberBitwiseOr, NumberBitwiseAnd and NumberBitwiseXor
simplified operators not only gives us precise (and correct) typing for
the bitwise operations, but also allows us to actually verify the graph
properly after typed lowering.
Drive-by-fix: Remove the double-to-smi magic from the Deoptimizer, which
is responsible for various deopt-loops in TurboFan, and is no longer
needed with the addition of the NumberBitwise operators.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1422213002
Cr-Commit-Position: refs/heads/master@{#31594}
Previous to this patch, both the lithium and TurboFan register
allocators tracked allocated registers by "indices", rather than
the register codes used elsewhere in the runtime. This patch
ensures that codes are used everywhere, and in the process cleans
up a bunch of redundant code and adds more structure to how the
set of allocatable registers is defined.
Some highlights of changes:
* TurboFan's RegisterConfiguration class moved to V8's top level
so that it can be shared with Crankshaft.
* Various "ToAllocationIndex" and related methods removed.
* Code that can be easily shared between Register classes on
different platforms is now shared.
* The list of allocatable registers on each platform is declared
as a list rather than implicitly via the register index <->
code mapping.
Committed: https://crrev.com/80bc6f6e11f79524e3f1ad05579583adfd5f18b2
Cr-Commit-Position: refs/heads/master@{#30913}
Committed: https://crrev.com/7b7a8205d9a00c678fb7a6e032a55fecbc1509cf
Cr-Commit-Position: refs/heads/master@{#31075}
Review URL: https://codereview.chromium.org/1287383003
Cr-Commit-Position: refs/heads/master@{#31087}
Reason for revert:
Failures on MIPS
Original issue's description:
> Remove register index/code indirection
>
> Previous to this patch, both the lithium and TurboFan register
> allocators tracked allocated registers by "indices", rather than
> the register codes used elsewhere in the runtime. This patch
> ensures that codes are used everywhere, and in the process cleans
> up a bunch of redundant code and adds more structure to how the
> set of allocatable registers is defined.
>
> Some highlights of changes:
>
> * TurboFan's RegisterConfiguration class moved to V8's top level
> so that it can be shared with Crankshaft.
> * Various "ToAllocationIndex" and related methods removed.
> * Code that can be easily shared between Register classes on
> different platforms is now shared.
> * The list of allocatable registers on each platform is declared
> as a list rather than implicitly via the register index <->
> code mapping.
>
> Committed: https://crrev.com/80bc6f6e11f79524e3f1ad05579583adfd5f18b2
> Cr-Commit-Position: refs/heads/master@{#30913}
>
> Committed: https://crrev.com/7b7a8205d9a00c678fb7a6e032a55fecbc1509cf
> Cr-Commit-Position: refs/heads/master@{#31075}
TBR=akos.palfi@imgtec.com,bmeurer@chromium.org,jarin@chromium.org,paul.lind@imgtec.com,titzer@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1380863004
Cr-Commit-Position: refs/heads/master@{#31083}
Previous to this patch, both the lithium and TurboFan register
allocators tracked allocated registers by "indices", rather than
the register codes used elsewhere in the runtime. This patch
ensures that codes are used everywhere, and in the process cleans
up a bunch of redundant code and adds more structure to how the
set of allocatable registers is defined.
Some highlights of changes:
* TurboFan's RegisterConfiguration class moved to V8's top level
so that it can be shared with Crankshaft.
* Various "ToAllocationIndex" and related methods removed.
* Code that can be easily shared between Register classes on
different platforms is now shared.
* The list of allocatable registers on each platform is declared
as a list rather than implicitly via the register index <->
code mapping.
Committed: https://crrev.com/80bc6f6e11f79524e3f1ad05579583adfd5f18b2
Cr-Commit-Position: refs/heads/master@{#30913}
Review URL: https://codereview.chromium.org/1287383003
Cr-Commit-Position: refs/heads/master@{#31075}
Reason for revert:
Failures on greedy RegAlloc, Fuzzer
Original issue's description:
> Remove register index/code indirection
>
> Previous to this patch, both the lithium and TurboFan register
> allocators tracked allocated registers by "indices", rather than
> the register codes used elsewhere in the runtime. This patch
> ensures that codes are used everywhere, and in the process cleans
> up a bunch of redundant code and adds more structure to how the
> set of allocatable registers is defined.
>
> Some highlights of changes:
>
> * TurboFan's RegisterConfiguration class moved to V8's top level
> so that it can be shared with Crankshaft.
> * Various "ToAllocationIndex" and related methods removed.
> * Code that can be easily shared between Register classes on
> different platforms is now shared.
> * The list of allocatable registers on each platform is declared
> as a list rather than implicitly via the register index <->
> code mapping.
>
> Committed: https://crrev.com/80bc6f6e11f79524e3f1ad05579583adfd5f18b2
> Cr-Commit-Position: refs/heads/master@{#30913}
TBR=akos.palfi@imgtec.com,bmeurer@chromium.org,jarin@chromium.org,paul.lind@imgtec.com,titzer@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1365073002
Cr-Commit-Position: refs/heads/master@{#30914}
Previous to this patch, both the lithium and TurboFan register
allocators tracked allocated registers by "indices", rather than
the register codes used elsewhere in the runtime. This patch
ensures that codes are used everywhere, and in the process cleans
up a bunch of redundant code and adds more structure to how the
set of allocatable registers is defined.
Some highlights of changes:
* TurboFan's RegisterConfiguration class moved to V8's top level
so that it can be shared with Crankshaft.
* Various "ToAllocationIndex" and related methods removed.
* Code that can be easily shared between Register classes on
different platforms is now shared.
* The list of allocatable registers on each platform is declared
as a list rather than implicitly via the register index <->
code mapping.
Review URL: https://codereview.chromium.org/1287383003
Cr-Commit-Position: refs/heads/master@{#30913}
This makes it clear that only components within the "heap" directory
should be friends with the Heap class. The two notable exceptions are
Factory and Isolate which represent external interfaces into the heap.
R=mlippautz@chromium.org
Review URL: https://codereview.chromium.org/1320843002
Cr-Commit-Position: refs/heads/master@{#30408}
Replaces all instances of the code which computed the debug
name of a stub or function with calls to CompileInfo::GetDebugName instead.
Also:
- Removes useless parameter on CodeStub::GetMajorName
- Removes FakeStubForTesting since it is no longer required
- Adds CompileInfo::ShouldEnsureSpaceForLazyDeopt() to replace unclear calls to IsStub().
Review URL: https://codereview.chromium.org/1297203002
Cr-Commit-Position: refs/heads/master@{#30324}
Previously, it was not possible to specify StackSlotOperands for all
slots in both the caller and callee stacks. Specifically, the region
of the callee's stack including the saved return address, frame
pointer, function pointer and context pointer could not be addressed
by the register allocator/gap resolver.
In preparation for better tail call support, which will use the gap
resolver to reconcile outgoing parameters, this change makes it
possible to address all slots on the stack, because slots in the
previously inaccessible dead zone may become parameter slots for
outgoing tail calls. All caller stack slots are accessible as they
were before, with slot -1 corresponding to the last stack
parameter. Stack slot indices >= 0 access the callee stack, with slot
0 corresponding to the callee's saved return address, 1 corresponding
to the saved frame pointer, 2 corresponding to the current function
context, 3 corresponding to the frame marker/JSFunction, and slots 4
and above corresponding to spill slots.
The following changes were specifically needed:
* Frame has been changed to explicitly manage three areas of the
callee frame, the fixed header, the spill slot area, and the
callee-saved register area.
* Conversions from stack slot indices to fp offsets all now go through
a common bottleneck: OptimizedFrame::StackSlotOffsetRelativeToFp
* The generation of deoptimization translation tables has been changed
to support the new stack slot indexing scheme. Crankshaft, which
doesn't support the new slot numbering in its register allocator,
must adapt the indexes when creating translation tables.
* Callee-saved parameters are now kept below spill slots, not above,
to support saving only the optimal set of used registers, which is
only known after register allocation is finished and spill slots
have been allocated.
Review URL: https://codereview.chromium.org/1261923007
Cr-Commit-Position: refs/heads/master@{#30224}
The calculation now takes into account the size of the arguments object
if it is present in the optimized frame.
(Yang, many thanks for the awesome repro!)
BUG=chromium:514362
LOG=N
R=yangguo@chromium.org
Review URL: https://codereview.chromium.org/1264483008
Cr-Commit-Position: refs/heads/master@{#29973}
This fixes a bug introduced by r28826 (Unify decoding of deoptimization
translations, https://codereview.chromium.org/1136223004), where we
started leaking arguments marker sentinel to the debugger, which would
then cause crashes. This change replaces the sentinel with the undefined
value in the debugger-inspectable frame.
BUG=chromium:514362
LOG=n
R=yangguo@chromium.org
Review URL: https://codereview.chromium.org/1263333002
Cr-Commit-Position: refs/heads/master@{#29971}
All of this is controlled by the CallDescriptor. It's simply the case
that if you specify less registers than the function arity calls for,
the rest are assumed to be on the stack.
Bailout handlers accept these constant stack arguments too.
BUG=
Review URL: https://codereview.chromium.org/1250563004
Cr-Commit-Position: refs/heads/master@{#29813}
This reverts commit 80b3f16951.
Revert "Record code slots that may point to evacuation candidate objects after deoptimizing them."
This reverts commit 4621210cfe.
BUG=chromium:507840
LOG=n
Review URL: https://codereview.chromium.org/1239383004
Cr-Commit-Position: refs/heads/master@{#29778}
Additionally, push the allocation site or undefined independently of creating a memento to preserve a fixed size for the construct frames.
BUG=
Review URL: https://codereview.chromium.org/1239593003
Cr-Commit-Position: refs/heads/master@{#29719}
Note that there are currently no objects that require a pre-allocated
properties backing store, all such slots are in-object properties from
the begining. Hence {unused + pre_allocated - inobject == 0} holds.
R=verwaest@chromium.org
Review URL: https://codereview.chromium.org/1226203011
Cr-Commit-Position: refs/heads/master@{#29590}
The only right way to enable access checks is to install access check callbacks on an object template via v8::ObjectTemplate::SetAccessCheckCallbacks(). It does not make sense to enable access checks on an arbitrary object.
Review URL: https://codereview.chromium.org/1217893012
Cr-Commit-Position: refs/heads/master@{#29439}
Up until now the context register was listed explicitly in each stub's
CallInterfaceDescriptor. This was problematic, because it was listed
first in the list of register parameters--which is fine for Crankshaft,
which is more or less built to handle the context as the first
parameter-- but not ideal for TurboFan, which adds the context at
the end of all function parameters. Now the context register is no
longer in the register list and can be handled appropriately by both
compilers. Specifically, this allows the FunctionType specified for
each CallInterfaceDescriptor to exactly match the parameter register
list.
Review URL: https://codereview.chromium.org/1211333003
Cr-Commit-Position: refs/heads/master@{#29402}
On MIPS32 we can't read a 8 bytes long data from a not 8 bytes aligned memory address.
BUG=
TEST=mjsunit/debug-backtrace
Review URL: https://codereview.chromium.org/1193433002
Cr-Commit-Position: refs/heads/master@{#29100}
Up until now we can only inline based on JSFunction, because of the way
the deoptimization works. With this change we will be able to inline
based on the SharedFunctionInfo and materialize the JSFunction from a
literal or a stack slot when necessary.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1169103004
Cr-Commit-Position: refs/heads/master@{#28906}
Use the new TranslatedState and friends, which work at a higher level
than the TranslationIterator, which will make it easier to change the
deoptimization commands in subsequent CLs.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1166353004
Cr-Commit-Position: refs/heads/master@{#28862}
This unifies methods Deoptimizer::DoTranslateCommand, Deotpimizer::DoTranslateObject and the arguments object materializer.
To unify these, we have to separate reading of the input frame from writing to the output frame because the argument materializer does not write to output frames.
Instead, we now deoptimize in following stages:
1. Read out the input frame/registers, decode them using the translations from the deoptimizer and store them in the deoptimizer (Deoptimizer::translated_state_). This is done in TranslatedState::Init.
2. Write out into the output frame buffer all the values that do not require allocation. We also remember references to the values that require materialization. As before, this is done in Deoptimizer::DoCompute*Frame method, but instead calling to DoTranslateCommand, we use the translated frame to obtain the values and write them to the output frames.
3. The platform specific code then sets up the output frames and calls into the deoptimization notification. This has not been changed at all.
4. Once the stack is setup, we handlify all the references in the saved translated values (TranslatedState::Prepare).
5. Finally, we materialize all the values we remembered in step (1) and write them to their frames on the stack (using the TranslatedValue::GetValue method).
BUG=
Review URL: https://codereview.chromium.org/1136223004
Cr-Commit-Position: refs/heads/master@{#28826}
Embed constant pools within their corresponding Code
objects.
This removes support for out-of-line constant pools in favor
of the new approach -- the main advantage being that it
eliminates the need to allocate and manage separate constant
pool array objects.
Currently supported on PPC and ARM. Enabled by default on
PPC only.
This yields a 6% improvment in Octane on PPC64.
R=bmeurer@chromium.org, rmcilroy@chromium.org, michael_dawson@ca.ibm.com
BUG=chromium:478811
LOG=Y
Review URL: https://codereview.chromium.org/1162993006
Cr-Commit-Position: refs/heads/master@{#28801}
When compiling on a laptop I like to concatenate the small test files.
This makes a big difference to compile times. These changes make that
easier.
R=ulan@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1163803002
Cr-Commit-Position: refs/heads/master@{#28742}
Replace the --turbo-deoptimization flag with --turbo-asm-deoptimization
and enable deoptimization for non-asm.js TurboFan code unconditionally.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1153483002
Cr-Commit-Position: refs/heads/master@{#28543}
TurboFan compilation was committing dependencies long before the
optimized function made it's way into the optimized code list for
the native context. The problem is that once the code pointer is out
there in dependency arrays, it is eligible for deopt. But the deopt
logic needs the code to be in the optimized code list to fully do it's
job.
BUG=
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1136693003
Cr-Commit-Position: refs/heads/master@{#28312}
AFAICT none of these can actually be triggered currently; but it's still good to harden the code a little.
Review URL: https://codereview.chromium.org/1058533007
Cr-Commit-Position: refs/heads/master@{#27848}
Deoptimization infrastructure already handles it correctly.
This change fixes repetitive deoptimizations in the code like this:
var u32 = new Uint32Array(1);
u32[0] = -1;
function tr(x) { return x|0; }
function ld() { return tr(u32[0]); }
while (true) ld();
Currently inlined tr will contain HArgumentsObject that is considered uint32-unsafe use and prevents u32[0] from becoming uint32 load - instead a speculative int32 load is generated which just deopts.
BUG=
Review URL: https://codereview.chromium.org/1077113002
Cr-Commit-Position: refs/heads/master@{#27781}
The original code always returned the first entry from RelocInfo that matched with
bailout_id. But we may have a few different deopt reasons for one bailout_id.
So we need to get the one which matches with a particular call from JumpTable.
We can do this by checking not 'target_address' (it maps to bailout_id)
but 'from' address which maps to a particular JumpTable entry.
The test was reworked so it tests identical functions against different reasons.
BUG=chromium:452067
LOG=n
Review URL: https://codereview.chromium.org/984773003
Cr-Commit-Position: refs/heads/master@{#27076}
Save Unknown position as zero in RelocInfo.
Remove copy constructor of SourcePosition because it is trivial.
Mechanical replace int raw_position with SourcePosition position.
BUG=452067
LOG=n
Review URL: https://codereview.chromium.org/959203002
Cr-Commit-Position: refs/heads/master@{#26916}
We accessed to cpu_profiler for tracking SharedFunctionInfo objects movements and used their addresses for generating function_id. Actually we could replace the manually generated shared_id by the pair script_id + position. In this case we can drop SharedFunctionInfo events support from cpu_profiler and remove the dependency.
BTW GetCallUid was used as an unique identifier of the function on the front-end side. Actually it is a hash which might not be unique. So I renamed GetCallUid with GetHash and implemented GetFunctionId method.
BUG=452067
LOG=n
Review URL: https://codereview.chromium.org/941973002
Cr-Commit-Position: refs/heads/master@{#26775}
1) Deoptimizer::Reason was replaced with Deoptimizer::DeoptInfo
because it also has raw position. Also the old name clashes with DeoptReason enum.
2) c_entry_fp assignment call was added to EntryGenerator::Generate
So we can calculate sp and have a chance to record the stack for the deopting function.
btw it makes the test stable.
3) new kind of CodeEvents was added to cpu-profiler
4) GetDeoptInfo method was extracted from PrintDeoptLocation.
So it could be reused in cpu profiler.
BUG=452067
LOG=n
Review URL: https://codereview.chromium.org/910773002
Cr-Commit-Position: refs/heads/master@{#26545}
1) The hardcoded strings were converted into DeoptReason enum.
2) Deopt comment were converted into a pair location and deopt reason entries so
the deopt reason tracking mode would less affect the size of the RelocInfo table and heap.
3) DeoptReason entry in RelocInfo reuses kCommentTag value and generates short entry in RelocInfo table.
BUG=452067
LOG=n
Committed: https://crrev.com/c49820e45b57f128a98690940875c049f612dde6
Cr-Commit-Position: refs/heads/master@{#26434}
Committed: https://crrev.com/ec42e002da03adb2db968dd5b7453341ddc59a5c
Cr-Commit-Position: refs/heads/master@{#26448}
Review URL: https://codereview.chromium.org/874323003
Cr-Commit-Position: refs/heads/master@{#26463}
Reason for revert:
This CL breaks RelocInfo::INTERNAL_REFERENCE (and maybe others that come after DEOPT_REASON), which is currently (mostly) unused on tip-of-tree, but will be used for jump tables soon(ish) and should therefore work. It seems to be a problem with implicitly assumptions about the number of reloc info modes. Needs further investigation.
Original issue's description:
> Externalize deoptimization reasons.
>
> 1) The hardcoded strings were converted into DeoptReason enum.
>
> 2) Deopt comment were converted into a pair location and deopt reason entries so
> the deopt reason tracking mode would less affect the size of the RelocInfo table and heap.
>
> 3) DeoptReason entry in RelocInfo reuses kCommentTag value and generates short entry in RelocInfo table.
>
> BUG=452067
> LOG=n
>
> Committed: https://crrev.com/c49820e45b57f128a98690940875c049f612dde6
> Cr-Commit-Position: refs/heads/master@{#26434}
>
> Committed: https://crrev.com/ec42e002da03adb2db968dd5b7453341ddc59a5c
> Cr-Commit-Position: refs/heads/master@{#26448}
TBR=alph@chromium.org,mstarzinger@chromium.org,svenpanne@chromium.org,yurys@chromium.org,loislo@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=452067
Review URL: https://codereview.chromium.org/900223002
Cr-Commit-Position: refs/heads/master@{#26457}
1) The hardcoded strings were converted into DeoptReason enum.
2) Deopt comment were converted into a pair location and deopt reason entries so
the deopt reason tracking mode would less affect the size of the RelocInfo table and heap.
3) DeoptReason entry in RelocInfo reuses kCommentTag value and generates short entry in RelocInfo table.
BUG=452067
LOG=n
Committed: https://crrev.com/c49820e45b57f128a98690940875c049f612dde6
Cr-Commit-Position: refs/heads/master@{#26434}
Review URL: https://codereview.chromium.org/874323003
Cr-Commit-Position: refs/heads/master@{#26448}
Reason for revert:
it broke the build
Original issue's description:
> Externalize deoptimization reasons.
>
> 1) The hardcoded strings were converted into DeoptReason enum.
>
> 2) Deopt comment were converted into a pair location and deopt reason entries so
> the deopt reason tracking mode would less affect the size of the RelocInfo table and heap.
>
> 3) DeoptReason entry in RelocInfo reuses kCommentTag value and generates short entry in RelocInfo table.
>
> BUG=452067
> LOG=n
>
> Committed: https://crrev.com/c49820e45b57f128a98690940875c049f612dde6
> Cr-Commit-Position: refs/heads/master@{#26434}
TBR=alph@chromium.org,mstarzinger@chromium.org,svenpanne@chromium.org,yurys@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=452067
Review URL: https://codereview.chromium.org/892843007
Cr-Commit-Position: refs/heads/master@{#26435}
1) The hardcoded strings were converted into DeoptReason enum.
2) Deopt comment were converted into a pair location and deopt reason entries so
the deopt reason tracking mode would less affect the size of the RelocInfo table and heap.
3) DeoptReason entry in RelocInfo reuses kCommentTag value and generates short entry in RelocInfo table.
BUG=452067
LOG=n
Review URL: https://codereview.chromium.org/874323003
Cr-Commit-Position: refs/heads/master@{#26434}
