Add a notion of "invocation count" to the baseline compilers, which
increment a special slot in the TypeFeedbackVector for each invocation
of a given function (the optimized code doesn't currently collect this
information).
Use this invocation count to relativize the call counts on the call
sites within the function, so that the inlining heuristic has a view
of relative importance of a call site rather than some absolute numbers
with unclear meaning for the current function. Also apply the call site
frequency as a factor to all frequencies in the inlinee by passing this
to the graph builders so that the importance of a call site in an
inlinee is relative to the topmost optimized function.
Note that all functions that neither have literals nor need type
feedback slots will share a single invocation count cell in the
canonical empty type feedback vector, so their invocation count is
meaningless, but that doesn't matter since we only use the invocation
count to relativize call counts within the function, which we only have
if we have at least one type feedback vector (the CallIC slot).
See the design document for additional details on this change:
https://docs.google.com/document/d/1VoYBhpDhJC4VlqMXCKvae-8IGuheBGxy32EOgC2LnT8
BUG=v8:5267,v8:5372
R=mvstanton@chromium.org,rmcilroy@chromium.org,mstarzinger@chromium.org
Review-Url: https://codereview.chromium.org/2337123003
Cr-Commit-Position: refs/heads/master@{#39410}
This makes for slightly faster rebuilds when touching parser-base.h
(which changes frequently!). Also takes care of an old TODO,
moving CompileTimeValue into its own file under ast/, where it
properly belongs.
BUG=v8:5294
Review-Url: https://codereview.chromium.org/2305883002
Cr-Commit-Position: refs/heads/master@{#39141}
This changes the compilation pipeline so that mixed stacks are allowed
when bytecode is preserved. This means there can be activations of both,
"baseline" as well as "unoptimized" code active on the stack at the same
time for any single given function.
R=rmcilroy@chromium.org
BUG=v8:4280
Review-Url: https://codereview.chromium.org/2267693002
Cr-Commit-Position: refs/heads/master@{#38809}
This fixes the self-healing mechanism for closures in the interpreter
entry trampoline not that bytecode can be preserved even when baseline
code is already available.
R=rmcilroy@chromium.org
TEST=cctest/test-compiler/IgnitionEntryTrampolineSelfHealing
BUG=chromium:638225
Review-Url: https://codereview.chromium.org/2257143002
Cr-Commit-Position: refs/heads/master@{#38747}
Collect type feedback in the call bytecode handler. The current
implementation only collects feedback for JS function objects. The other
objects and Array functions do not collect any feedback. They will be
marked Megamorphic.
BUG=v8:4280, v8:4780
LOG=N
Review-Url: https://codereview.chromium.org/2122183002
Cr-Commit-Position: refs/heads/master@{#37700}
We get less "pollution" of type feedback if we have one vector per native
context, rather than one for the whole system. This CL moves the vector
appropriately.
BUG=
Review-Url: https://codereview.chromium.org/1906823002
Cr-Commit-Position: refs/heads/master@{#36539}
We cannot tier up from interpreted to baseline code when there is an
activation of the function on the stack. This significantly regresses
the performance of recursive functions since they are unlikely to get
tiered up.
This CL adds the ability for a function to be marked for baseline
compilation when it returns. To do this we patch the
InterpreterEntryTrampoline return address to point to
InterpreterMarkBaselineOnReturn, which leaves the
interpreted frame and recompile the function for
baseline.
This improves the score of EarlyBoyer by ~8x for Ignition.
BUG=v8:4280
LOG=N
Review-Url: https://codereview.chromium.org/1965343002
Cr-Commit-Position: refs/heads/master@{#36360}
Visit the Optimized Code Map on first call rather than closure creation.
This is useful for escape analysis, and helps upcoming changes to
type feedback gathering.
Adding notry due to crashed builders:
NOTRY=true
BUG=
Committed: https://crrev.com/9336f4cc6d25d39a128176679a70dbd13a6d946e
Cr-Commit-Position: refs/heads/master@{#35395}
Review URL: https://codereview.chromium.org/1670143002
Cr-Commit-Position: refs/heads/master@{#35440}
Reason for revert:
Blocks roll. See https://codereview.chromium.org/1877003002/ for detailed messages.
You should be able to repro this with Linux ASAN.
Original issue's description:
> Visit the Optimized Code Map on first call rather than closure creation.
>
> This is useful for escape analysis, and helps upcoming changes to
> type feedback gathering.
>
> BUG=
>
> Committed: https://crrev.com/9336f4cc6d25d39a128176679a70dbd13a6d946e
> Cr-Commit-Position: refs/heads/master@{#35395}
TBR=mstarzinger@chromium.org,bmeurer@chromium.org,mvstanton@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=
Review URL: https://codereview.chromium.org/1878063004
Cr-Commit-Position: refs/heads/master@{#35404}
This is useful for escape analysis, and helps upcoming changes to
type feedback gathering.
BUG=
Review URL: https://codereview.chromium.org/1670143002
Cr-Commit-Position: refs/heads/master@{#35395}
Makes --ignition cause eager compilation if we aren't building the startup
snapshot.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1811553003
Cr-Commit-Position: refs/heads/master@{#35066}
This is a pure refactoring and renaming of methods in the compiler API
with the goal to increase readability. Also the compiler API is moved to
the top of the file, as it is the central piece in that file.
R=yangguo@chromium.org
Review URL: https://codereview.chromium.org/1766623004
Cr-Commit-Position: refs/heads/master@{#34579}
Reason for revert:
Must revert for now due to chromium api natives issues.
Original issue's description:
> Type Feedback Vector lives in the closure
>
> (RELAND: the problem before was a missing write barrier for adding the code
> entry to the new closure. It's been addressed with a new macro instruction
> and test. The only change to this CL is the addition of two calls to
> __ RecordWriteCodeEntryField() in the platform CompileLazy builtin.)
>
> We get less "pollution" of type feedback if we have one vector per native
> context, rather than one for the whole system. This CL moves the vector
> appropriately.
>
> We rely more heavily on the Optimized Code Map in the SharedFunctionInfo. The
> vector actually lives in the first slot of the literals array (indeed there is
> great commonality between those arrays, they can be thought of as the same
> thing). So we make greater effort to ensure there is a valid literals array
> after compilation.
>
> This meant, for performance reasons, that we needed to extend
> FastNewClosureStub to support creating closures with literals. And ultimately,
> it drove us to move the optimized code map lookup out of FastNewClosureStub
> and into the compile lazy builtin.
>
> The heap change is trivial so I TBR Hannes for it...
> Also, Yang has had a look at the debugger changes already and approved 'em. So he is TBR style too.
> And Benedikt reviewed it as well.
>
> TBR=hpayer@chromium.org, yangguo@chromium.org, bmeurer@chromium.org
>
> BUG=
>
> Committed: https://crrev.com/bb31db3ad6de16f86a61f6c7bbfd3274e3d957b5
> Cr-Commit-Position: refs/heads/master@{#33741}
TBR=bmeurer@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=
Review URL: https://codereview.chromium.org/1670813005
Cr-Commit-Position: refs/heads/master@{#33766}
(RELAND: the problem before was a missing write barrier for adding the code
entry to the new closure. It's been addressed with a new macro instruction
and test. The only change to this CL is the addition of two calls to
__ RecordWriteCodeEntryField() in the platform CompileLazy builtin.)
We get less "pollution" of type feedback if we have one vector per native
context, rather than one for the whole system. This CL moves the vector
appropriately.
We rely more heavily on the Optimized Code Map in the SharedFunctionInfo. The
vector actually lives in the first slot of the literals array (indeed there is
great commonality between those arrays, they can be thought of as the same
thing). So we make greater effort to ensure there is a valid literals array
after compilation.
This meant, for performance reasons, that we needed to extend
FastNewClosureStub to support creating closures with literals. And ultimately,
it drove us to move the optimized code map lookup out of FastNewClosureStub
and into the compile lazy builtin.
The heap change is trivial so I TBR Hannes for it...
Also, Yang has had a look at the debugger changes already and approved 'em. So he is TBR style too.
And Benedikt reviewed it as well.
TBR=hpayer@chromium.org, yangguo@chromium.org, bmeurer@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1668103002
Cr-Commit-Position: refs/heads/master@{#33741}
Reason for revert:
Bug: failing to use write barrier when writing code entry into closure.
Original issue's description:
> Reland of Type Feedback Vector lives in the closure
>
> (Fixed a bug found by nosnap builds.)
>
> We get less "pollution" of type feedback if we have one vector per native
> context, rather than one for the whole system. This CL moves the vector
> appropriately.
>
> We rely more heavily on the Optimized Code Map in the SharedFunctionInfo. The
> vector actually lives in the first slot of the literals array (indeed there is
> great commonality between those arrays, they can be thought of as the same
> thing). So we make greater effort to ensure there is a valid literals array
> after compilation.
>
> This meant, for performance reasons, that we needed to extend
> FastNewClosureStub to support creating closures with literals. And ultimately,
> it drove us to move the optimized code map lookup out of FastNewClosureStub
> and into the compile lazy builtin.
>
> The heap change is trivial so I TBR Hannes for it...
>
> TBR=hpayer@chromium.org
> BUG=
>
> Committed: https://crrev.com/d984b3b0ce91e55800f5323b4bb32a06f8a5aab1
> Cr-Commit-Position: refs/heads/master@{#33548}
TBR=bmeurer@chromium.org,yangguo@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=
Review URL: https://codereview.chromium.org/1643533003
Cr-Commit-Position: refs/heads/master@{#33556}
(Fixed a bug found by nosnap builds.)
We get less "pollution" of type feedback if we have one vector per native
context, rather than one for the whole system. This CL moves the vector
appropriately.
We rely more heavily on the Optimized Code Map in the SharedFunctionInfo. The
vector actually lives in the first slot of the literals array (indeed there is
great commonality between those arrays, they can be thought of as the same
thing). So we make greater effort to ensure there is a valid literals array
after compilation.
This meant, for performance reasons, that we needed to extend
FastNewClosureStub to support creating closures with literals. And ultimately,
it drove us to move the optimized code map lookup out of FastNewClosureStub
and into the compile lazy builtin.
The heap change is trivial so I TBR Hannes for it...
TBR=hpayer@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1642613002
Cr-Commit-Position: refs/heads/master@{#33548}
Reason for revert:
FAilure on win32 bot, need to investigate webkit failures.
Original issue's description:
> Type Feedback Vector lives in the closure
>
> We get less "pollution" of type feedback if we have one vector per native
> context, rather than one for the whole system. This CL moves the vector
> appropriately.
>
> We rely more heavily on the Optimized Code Map in the SharedFunctionInfo. The
> vector actually lives in the first slot of the literals array (indeed there is
> great commonality between those arrays, they can be thought of as the same
> thing). So we make greater effort to ensure there is a valid literals array
> after compilation.
>
> This meant, for performance reasons, that we needed to extend
> FastNewClosureStub to support creating closures with literals. And ultimately,
> it drove us to move the optimized code map lookup out of FastNewClosureStub
> and into the compile lazy builtin.
>
> The heap change is trivial so I TBR Hannes for it...
>
> TBR=hpayer@chromium.org
>
> BUG=
>
> Committed: https://crrev.com/a5200f7ed4d11c6b882fa667da7a1864226544b4
> Cr-Commit-Position: refs/heads/master@{#33518}
TBR=bmeurer@chromium.org,akos.palfi@imgtec.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=
Review URL: https://codereview.chromium.org/1632993003
Cr-Commit-Position: refs/heads/master@{#33520}
We get less "pollution" of type feedback if we have one vector per native
context, rather than one for the whole system. This CL moves the vector
appropriately.
We rely more heavily on the Optimized Code Map in the SharedFunctionInfo. The
vector actually lives in the first slot of the literals array (indeed there is
great commonality between those arrays, they can be thought of as the same
thing). So we make greater effort to ensure there is a valid literals array
after compilation.
This meant, for performance reasons, that we needed to extend
FastNewClosureStub to support creating closures with literals. And ultimately,
it drove us to move the optimized code map lookup out of FastNewClosureStub
and into the compile lazy builtin.
The heap change is trivial so I TBR Hannes for it...
TBR=hpayer@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1563213002
Cr-Commit-Position: refs/heads/master@{#33518}
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}
Introduce a new JSGlobalSpecialization advanced reducer that runs
during the initial inlining and context specialization, and specializes
the graph to the globals of the native context. Currently we assume
that we do not inline cross native context, but long-term we will grab
the global object from the JSLoadGlobal/JSStoreGlobal feedback (with the
new global load/store ICs that are currently in the workings), and then
this whole specialization will be fully compositional even across
cross-context inlining.
Note that we cannot really handle most of the stores to global object
property cells because TurboFan doesn't have a mechanism to enforce
certain representations. Also note that we cannot yet fully benefit
from the type feedback collected on the global object property cells,
because the type system cannot deal with maps in a reasonable way.
CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux_nosnap_rel
R=jarin@chromium.org
BUG=v8:4470
LOG=n
Committed: https://crrev.com/6fbf7903f94924ea066af481719898bd9667b6eb
Cr-Commit-Position: refs/heads/master@{#31139}
Review URL: https://codereview.chromium.org/1387393002
Cr-Commit-Position: refs/heads/master@{#31148}
Reason for revert:
Breaks GC stress: http://build.chromium.org/p/client.v8/builders/V8%20Linux64%20GC%20Stress%20-%20custom%20snapshot/builds/1984/steps/Bisect%20c5528ac1.Retry/logs/regress-crbug-450960
Original issue's description:
> [turbofan] Add initial support for global specialization.
>
> Introduce a new JSGlobalSpecialization advanced reducer that runs
> during the initial inlining and context specialization, and specializes
> the graph to the globals of the native context. Currently we assume
> that we do not inline cross native context, but long-term we will grab
> the global object from the JSLoadGlobal/JSStoreGlobal feedback (with the
> new global load/store ICs that are currently in the workings), and then
> this whole specialization will be fully compositional even across
> cross-context inlining.
>
> Note that we cannot really handle most of the stores to global object
> property cells because TurboFan doesn't have a mechanism to enforce
> certain representations. Also note that we cannot yet fully benefit
> from the type feedback collected on the global object property cells,
> because the type system cannot deal with maps in a reasonable way.
>
> CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux_nosnap_rel
> R=jarin@chromium.org
> BUG=v8:4470
> LOG=n
>
> Committed: https://crrev.com/6fbf7903f94924ea066af481719898bd9667b6eb
> Cr-Commit-Position: refs/heads/master@{#31139}
TBR=jarin@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:4470
Review URL: https://codereview.chromium.org/1390073004
Cr-Commit-Position: refs/heads/master@{#31144}
Introduce a new JSGlobalSpecialization advanced reducer that runs
during the initial inlining and context specialization, and specializes
the graph to the globals of the native context. Currently we assume
that we do not inline cross native context, but long-term we will grab
the global object from the JSLoadGlobal/JSStoreGlobal feedback (with the
new global load/store ICs that are currently in the workings), and then
this whole specialization will be fully compositional even across
cross-context inlining.
Note that we cannot really handle most of the stores to global object
property cells because TurboFan doesn't have a mechanism to enforce
certain representations. Also note that we cannot yet fully benefit
from the type feedback collected on the global object property cells,
because the type system cannot deal with maps in a reasonable way.
CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux_nosnap_rel
R=jarin@chromium.org
BUG=v8:4470
LOG=n
Review URL: https://codereview.chromium.org/1387393002
Cr-Commit-Position: refs/heads/master@{#31139}
This CL also allows to use arbitrary number of feedback vector elements for particular slot kind.
Review URL: https://codereview.chromium.org/1370303004
Cr-Commit-Position: refs/heads/master@{#31050}
Note that this tests performed unnecessary many iterations which led to
long runtimes in debug mode and also caused flaky GCs during that would
cause the optimized code map to be flushed and violated assumptions.
R=yangguo@chromium.org
BUG=v8:4363
LOG=N
Review URL: https://codereview.chromium.org/1280973002
Cr-Commit-Position: refs/heads/master@{#30070}
This allows the optimized code map to contain no context-dependent
entries, but still hold one context-independent entry. This is a
precursor to extending the lifetime of the context-independent entry.
R=mvstanton@chromium.org
Review URL: https://codereview.chromium.org/1249543005
Cr-Commit-Position: refs/heads/master@{#29788}
This allows context-independent code generated by TurboFan to be cached
in the optimized code map and reused across native contexts. Note that
currently this cache is still flushed at GC time.
R=bmeurer@chromium.org,mvstanton@chromium.org
TEST=cctest/test-compiler/OptimizedCodeSharing
Review URL: https://codereview.chromium.org/1208013002
Cr-Commit-Position: refs/heads/master@{#29313}
Before this we had 3 super related lexical bindings that got injected
into method bodies: .home_object, .this_function, and new.target.
With this change we get rid of the .home_object one in favor of using
.this_function[home_object_symbol] which allows some simplifications
throughout the code base.
BUG=v8:3768
LOG=N
R=adamk@chromium.org, wingo@igalia.com
Review URL: https://codereview.chromium.org/1154103005
Cr-Commit-Position: refs/heads/master@{#28802}
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 we enter a method that needs access to the [[HomeObject]]
we allocate a local variable `.home_object` and assign it the
value from the [[HomeObject]] private symbol. Something along
the lines of:
method() {
var .home_object = %ThisFunction()[home_object_symbol];
...
}
BUG=v8:3867, v8:4031
LOG=N
Review URL: https://codereview.chromium.org/1135243004
Cr-Commit-Position: refs/heads/master@{#28644}
This just delegates to SharedFunctionInfo::optimization_disabled and
was primarily used for assertions. Removing it due to misleading name
because already optimized functions reported being "non-optimizable".
This relands commit 181d7b8597.
R=titzer@chromium.org
Review URL: https://codereview.chromium.org/1146423002
Cr-Commit-Position: refs/heads/master@{#28577}
Reason for revert:
Causes assertions to fire when serializing optimized code.
Original issue's description:
> Remove obsolete JSFunction::IsOptimizable predicate.
>
> This just delegates to SharedFunctionInfo::optimization_disabled and
> was primarily used for assertions. Removing it due to misleading name
> because already optimized functions reported being "non-optimizable".
>
> R=titzer@chromium.org
>
> Committed: https://crrev.com/181d7b85977eb752b19e1de902093783e31330ef
> Cr-Commit-Position: refs/heads/master@{#28551}
TBR=titzer@chromium.org,bmeurer@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1148973005
Cr-Commit-Position: refs/heads/master@{#28554}
This just delegates to SharedFunctionInfo::optimization_disabled and
was primarily used for assertions. Removing it due to misleading name
because already optimized functions reported being "non-optimizable".
R=titzer@chromium.org
Review URL: https://codereview.chromium.org/1150683002
Cr-Commit-Position: refs/heads/master@{#28551}
When the page is controlled by a ServiceWorker, the ServiceWorker can return an opaque (non-CORS cross origin) resource response.
We need to treat the messages from such script resource as opaque.
Committed: https://crrev.com/7a599c5e1242d3c5ab7515ee149623da90ae69ec
Cr-Commit-Position: refs/heads/master@{#28445}
Review URL: https://codereview.chromium.org/1140673002
Cr-Commit-Position: refs/heads/master@{#28459}
When the page is controlled by a ServiceWorker, the ServiceWorker can return an opaque (non-CORS cross origin) resource response.
We need to treat the messages from such script resource as opaque.
Review URL: https://codereview.chromium.org/1140673002
Cr-Commit-Position: refs/heads/master@{#28445}
According to Source Map specification [1] source map url can be passed either as a magic comment at the end of script or as SourceMap http header. We already parse the former value and expose it on Script object. This change allows to unify the way we deal with source map urls received in http header by providing api for passing that url into the script being compiled.
source_map_url is intentionally not passed into CompilationCacheScript::Lookup. The cache is anyways disabled when debugger is on.
[1] https://docs.google.com/document/d/1U1RGAehQwRypUTovF1KRlpiOFze0b-_2gc6fAH0KY0k/edit
LOG=Y
BUG=chromium:462572
Review URL: https://codereview.chromium.org/983603003
Cr-Commit-Position: refs/heads/master@{#27017}
This adds an "experimental" API hook (v8::ScriptCompiler::CompileModule)
allowing compilation of modules. The code gen is incredibly basic: the
module body is represented by a Block in the AST. But this at least gets
more of the pipeline working, and opens the door to writing mjsunit tests
(once d8 is modified to support module compilation).
BUG=v8:1569
LOG=n
Review URL: https://codereview.chromium.org/902093002
Cr-Commit-Position: refs/heads/master@{#26496}
The first try failed because I needed to make a better distinction
between clearing ICs according to policy at GC time or unconditional
clearing (say, via %ClearFunctionTypeFeedback).
It was also blocked by an issue in super constructor calls.
This fix (https://codereview.chromium.org/892113002/) needs to land
before checking in this CL.
R=ulan@chromium.org
Review URL: https://codereview.chromium.org/866493003
Cr-Commit-Position: refs/heads/master@{#26420}
