Now that it is no longer needed, this also removes the invalid inclusion
of "object-inl.h" within the "unique.h" header file.
Note that this change still leaves 2 violations of that rule in the
code, checked with the "tools/check-inline-includes.sh" tool.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1321223002
Cr-Commit-Position: refs/heads/master@{#30503}
The usage of Unique<T> throughout the TurboFan IR does not have any
advantage. There is no single point in time when they are initialized
and most use-sites looked through to the underlying Handle<T> anyways.
Also there already was a mixture of Handle<T> versus Unique<T> in the
graph and this unifies the situation to use Handle<T> everywhere.
R=bmeurer@chromium.org,titzer@chromium.org
Review URL: https://codereview.chromium.org/1314473007
Cr-Commit-Position: refs/heads/master@{#30458}
In many cases, the context that TurboFan's ASTGraphBuilder or subsequent
reduction operations attaches to nodes does not need to be that exact
context, but rather only needs to be one with the same native context,
because it is used internally only to fetch the native context, e.g. for
creating and throwing exceptions.
This reducer recognizes common cases where the context that is specified
for a node can be relaxed to a canonical, less specific one. This
relaxed context can either be the enclosing function's context or a specific
Module or Script context that is explicitly created within the function.
This optimization is especially important for TurboFan-generated code stubs
which use context specialization and inlining to generate optimal code.
Without context relaxation, many extraneous moves are generated to pass
exactly the right context to internal functions like ToNumber and
AllocateHeapNumber, which only need the native context. By turning context
relaxation on, these moves disappear because all these common internal
context uses are unified to the context passed into the stub function, which
is typically already in the correct context register and remains there for
short stubs. It also eliminates the explicit use of a specialized context
constant in the code stub in these cases, which could cause memory leaks.
Review URL: https://codereview.chromium.org/1244583003
Cr-Commit-Position: refs/heads/master@{#29763}
Prior to this patch, we enter a global debug mode whenever a break point
is set. By entering this mode, all code is deoptimized and activated
frames are recompiled and redirected to newly compiled debug code.
After this patch, we only deoptimize/redirect for functions we want to
debug. Trigger for this is Debug::EnsureDebugInfo, and having DebugInfo
object attached to the SFI prevents optimization/inlining.
The result is that we can have optimized code for functions without break
points alongside functions that do have break points, which are not
optimized.
R=mstarzinger@chromium.org, ulan@chromium.org
BUG=v8:4132
LOG=Y
Review URL: https://codereview.chromium.org/1233073005
Cr-Commit-Position: refs/heads/master@{#29758}
Remove the context specialization hack from the AstGraphBuilder, and
properly specialize to the function context in the context specialization.
And replace the correct context in the JSInliner.
R=mstarzinger@chromium.org
BUG=v8:4273
LOG=n
Review URL: https://codereview.chromium.org/1218873005
Cr-Commit-Position: refs/heads/master@{#29493}
The deoptimizer (and probably various other places) cannot deal properly
with recursive function inlining, so we disallow it in TurboFan as well.
We might want to reconsider that decision at some point in the future.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1211243007
Cr-Commit-Position: refs/heads/master@{#29374}
Ideally inliner itself should not deal with context specialization at
all, since this is all handled in the pipeline instead (actually
inlining already runs together with context specialization), and the
inlining logic should not care about the specialization mode.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1217973003
Cr-Commit-Position: refs/heads/master@{#29366}
This also threads through the parameter count and local count to the instruction selector. This will be later used to allow merging of various StateValues vector (and prepare for differential encoding which will not distinguish between parameters, locals and expression stack).
BUG=
Review URL: https://codereview.chromium.org/1191243003
Cr-Commit-Position: refs/heads/master@{#29214}
The three different concerns that the ControlReducer used to deal with
are now properly separated into
a.) DeadCodeElimination, which is a regular AdvancedReducer, that
propagates Dead via control edges,
b.) CommonOperatorReducer, which does strength reduction on common
operators (i.e. Branch, Phi, and friends), and
c.) GraphTrimming, which removes dead->live edges from the graph.
This will make it possible to run the DeadCodeElimination together with
other passes that actually introduce Dead nodes, i.e. typed lowering;
and it opens the door for general inlining without two stage fix point
iteration.
To make the DeadCodeElimination easier and more uniform, we basically
reverted the introduction of DeadValue and DeadEffect, and changed the
Dead operator to produce control, value and effect. Note however that
this is not a requirement, but merely a way to make dead propagation
easier and more uniform. We could always go back and decide to have
different Dead operators if some other change requires that.
Note that there are several additional opportunities for cleanup now,
i.e. OSR deconstruction could be a regular reducer now, and we don't
need to use TheHole as dead value marker in the GraphReducer. And we can
actually run the dead code elimination together with the other passes
instead of using separate passes over the graph. We will do this in
follow up CLs.
R=jarin@chromium.org, mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1193833002
Cr-Commit-Position: refs/heads/master@{#29146}
Reason for revert:
Breaks Windows debug.
Original issue's description:
> [turbofan] Record the SharedFunctionInfo of ALL inlined functions.
>
> Previously we only recorded the SharedFunctionInfo of inlined functions
> that had at least one (lazy) deopt point left at code generation time.
>
> R=mstarzinger@chromium.org
>
> Committed: ffa0b4007cTBR=mstarzinger@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1178683004
Cr-Commit-Position: refs/heads/master@{#28920}
Previously we only recorded the SharedFunctionInfo of inlined functions
that had at least one (lazy) deopt point left at code generation time.
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1175953002.
Cr-Commit-Position: refs/heads/master@{#28919}
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}
This allows any AdvancedReducer to remove exception projections from
graphs. This is the common case when JS-operators are being replaced
with pure values. The old NodeProperties::ReplaceWithValue is being
deprecated in favor of AdvancedReducer::ReplaceWithValue.
R=titzer@chromium.org
TEST=unittests/AdvancedReducerTest
Review URL: https://codereview.chromium.org/1168693002
Cr-Commit-Position: refs/heads/master@{#28810}
This simplifies inlining, in that we only need to update uses of Start
and inputs of End instead of walking the whole inlinee to update all
outer frame states.
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1146403008
Cr-Commit-Position: refs/heads/master@{#28649}
This way we don't need to connect (potentially) non-terminating loops
later during control reduction, which saves one forward pass over the
control graph. Long term we will move the trimming functionality of
the control reducer to the GraphReducer, and get rid of the Finish
method again.
As a bonus, this change also properly rewires Terminate, Throw and
Deoptimize during inlining.
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1155683004
Cr-Commit-Position: refs/heads/master@{#28625}
This simplifies the handling of the End node. Based on this CL we will
finally fix terminating every loop from the beginning (via Terminate
nodes) and fix inlining of Throw, Deoptimize and Terminate.
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1157023002
Cr-Commit-Position: refs/heads/master@{#28620}
The inliner previously assumed that there will only be returns reaching
the end node, but that's not true. This refactoring will make it
possible to also hook up Deoptimize, Throw and Terminate nodes reaching
end properly.
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1146393002
Cr-Commit-Position: refs/heads/master@{#28550}
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}
During inlining we should pay attention to only rewire the outer frame
states of the inlinee, but leave any inner frame states of the inlinee
untouched. Otherwise we might run into trouble once we start caching
graphs, or do getter/setter inlining.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1134973006
Cr-Commit-Position: refs/heads/master@{#28466}
Note that this is just a duplication for now. We'll want to get rid of the
NodeProperties::ReplaceWithValue() method in the long run.
R=bmeurer@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1135483004
Cr-Commit-Position: refs/heads/master@{#28363}
Now all nodes that care about deoptimization always take frame state
inputs no matter whether deoptimization is enabled for a particular
function. In case that deoptimization is off, the AstGraphBuilder just
inserts the empty frame state. This greatly simplifies the logic in
various places and makes testing easier as well, and is probably the
first step towards enabling --turbo-deoptimization by default.
There seems to be no noticable performance impact on asm.js programs.
Also fix the graph replay in order to regenerate the scheduler unittests.
Review URL: https://codereview.chromium.org/1106613003
Cr-Commit-Position: refs/heads/master@{#28026}
This removes the CompilationInfoWithZone class from the header file
because it is more than a pure convenience class and shouldn't be used
outside of the compiler at all.
R=titzer@chromium.org
Review URL: https://codereview.chromium.org/1000353004
Cr-Commit-Position: refs/heads/master@{#27411}
Context specialization enables inlining (at least currently it is the
only enabler for inlining), but inlining enables more possibilities for
context specialization. So we really need to run them together.
This is especially important with the "module based builtins" that we're
working towards.
BUG=v8:3952
LOG=n
Review URL: https://codereview.chromium.org/988423004
Cr-Commit-Position: refs/heads/master@{#27085}
Rationale: separate the inputs and outputs of parsing + analysis from the business of compiling (i.e. generating machine code).
BUG=
Review URL: https://codereview.chromium.org/974213002
Cr-Commit-Position: refs/heads/master@{#27078}
The JSInliner used to load the context from the JSFunction node at
runtime, which introduced a HeapConstant (because we had to materialize
the JSFunction after context specialization) and a LoadField operation,
independent whether the inlinee actually uses the context. This is
rather cumbersome currently, and therefore this is now changed to just
embed the context constant instead. Once we do inlining based on
SharedFunctionInfo rather than JSFunction, we should reconsider this
decision and come up with a proper heuristic.
BUG=v8:3952
LOG=n
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/994523002
Cr-Commit-Position: refs/heads/master@{#27069}
We mark certain builtins for inlining, and those should always be
inlined into optimized code (CrankShaft already handles it this way), so
we should support that in TurboFan as well. Currently this mainly
affects a certain set of Math functions, but once have the basics in
place we can extend this to any kind of builtin/code stub/accessor.
This adds a new flag --turbo_builtin_inlining (enabled by default), that
forces the inliner to always inline builtins marked for inlining, but
does not affect inlining of other functions (this is still controlled by
the --turbo-inlining flag).
BUG=v8:3952
LOG=n
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/993473002
Cr-Commit-Position: refs/heads/master@{#27059}
This is currently the cleanest approach to avoid the useless stack check
during inlining. We might be able to just remove the useless stack
checks later when we have a phase that also takes care of removing
redundant stack checks on loop back edges (which we do not generate
currently).
On the other hand, the flag introduced here might be useful when
building code stubs/builtins/dom stubs using JS based DSL, because you
certainly don't want a JS-level stack check in a code stub.
R=jarin@chromium.org
BUG=v8:3952
LOG=n
Review URL: https://codereview.chromium.org/994433002
Cr-Commit-Position: refs/heads/master@{#27058}
This enables adding more language modes in the future.
For maximum flexibility, LanguageMode is a bitmask, so we're not restricted to
use a sequence of language modes which are progressively stricter, but we can
express the language mode as combination of features.
For now, LanguageMode can only be "sloppy" or "strict", and there are
STATIC_ASSERTS in places which need to change when more modes are added.
LanguageMode is a bit like the old LanguageMode when "extended" mode was still
around (see https://codereview.chromium.org/8417035 and
https://codereview.chromium.org/181543002 ) except that it's transmitted through
all the layers (there's no StrictModeFlag).
BUG=
Review URL: https://codereview.chromium.org/894683003
Cr-Commit-Position: refs/heads/master@{#26419}
