Reason for revert:
Tanks benchmarks (e.g., Octane box2d TF).
Original issue's description:
> [turbofan] Connect ObjectIsNumber to effect and control chains.
>
> In theory, we could connect the nodes when doing
> the schedule-in-the-middle pass, but that would require creating two
> versions of the operator (effectful and pure). I believe we do not
> lose anything by wiring the node up eagerly.
>
> Committed: https://crrev.com/2894e80a0a4a51a0d72e72aa48fcd01968f7949f
> Cr-Commit-Position: refs/heads/master@{#34141}
TBR=bmeurer@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1718483002
Cr-Commit-Position: refs/heads/master@{#34147}
In theory, we could connect the nodes when doing
the schedule-in-the-middle pass, but that would require creating two
versions of the operator (effectful and pure). I believe we do not
lose anything by wiring the node up eagerly.
Review URL: https://codereview.chromium.org/1709093002
Cr-Commit-Position: refs/heads/master@{#34141}
This reducer doesn't really add value, because:
(a) it is only concerned with JSCallFunction and JSToNumber, but when
we get to it, all JSCallFunction nodes will have been replaced by
Call nodes, and in the not so far future, we will also have
replaced almost all JSToNumber nodes with better code,
(b) and the reducer tries to be smart and use one of the outermost
contexts, but that might not be beneficial always; actually it
might even create longer live ranges and lead to more spilling
in some cases.
But most importantly, the JSContextRelaxation currently blocks inlining
based on SharedFunctionInfo, because it requires the inliner to check
the native context, which in turn requires JSFunction knowledge. So I'm
removing this reducer for now to unblock the more important inliner
changes.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1715633002
Cr-Commit-Position: refs/heads/master@{#34139}
This removes the language mode parameter from all JSCall operators. The
information is no longer used anywhere and is not threaded through the
interpreter bytecode. We should only thread it through the bytecode if
it has a semantic impact on the compilation.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1709493002
Cr-Commit-Position: refs/heads/master@{#34073}
Drive-by-fix: Remove the (now) unused %_SetValueOf and %_JSValueGetValue
intrinsics from the various compilers and the runtime.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1698343002
Cr-Commit-Position: refs/heads/master@{#34037}
Support SBFX in the instruction selector for sign-extension patterns like
Sar(Shl(x, a), b), where a and b are immediate values.
BUG=
Review URL: https://codereview.chromium.org/1695293002
Cr-Commit-Position: refs/heads/master@{#34029}
Passing floating point params to/from C has never quite worked correctly,
but we've never enforced the restriction early in the CallDescriptor
creation process because of unittests. Fix unittests to make their own
simple call descriptors and not rely on the C ones.
R=bmeurer@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1701593003
Cr-Commit-Position: refs/heads/master@{#33993}
Improve instruction selector for mask and shift operations by using cheaper
instructions where possible, in preference to UBFX.
Reverted because it was suspected of causing a couple of flaky tests to fail,
but investigation suggests this is unlikely.
Original review: https://codereview.chromium.org/1677023002
BUG=
Review URL: https://codereview.chromium.org/1684073006
Cr-Commit-Position: refs/heads/master@{#33988}
This adds initial support for inline allocation of object and array
literals to the JSCreateLowering pass. It's basically identical to
what Crankshaft does.
This also unstages the TurboFan escape analysis, as the lowering seems
to trigger a bunch of bugs in it; those bugs will be fixed separately,
and we will re-enable escape analysis afterwards.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1698783002
Cr-Commit-Position: refs/heads/master@{#33972}
This was causing code like:
REX.W cmpq r9,r8
setzl r8l
movzxbl r8,r8
REX.W cmpq r8,0x0
jz 185
(note the cmpq instead of cmpl above) on x64 instead of:
REX.W cmpq r9,r8
jnz 149
http://crrev.com/1677503002 is now obsolete and has been reverted.
Review URL: https://codereview.chromium.org/1685183003
Cr-Commit-Position: refs/heads/master@{#33934}
The FastNewStrictArgumentsStub is very similar to the recently added
FastNewRestParameterStub, it's actually almost a copy of it, except that
it doesn't have the fast case we have for the empty rest parameter. This
patch improves strict arguments in TurboFan and fullcodegen by up to 10x
compared to the previous version.
Also introduce proper JSSloppyArgumentsObject and JSStrictArgumentsObject
for the in-object properties instead of having them as constants in the
Heap class.
Drive-by-fix: Use this stub and the FastNewRestParameterStub in the
interpreter to avoid the runtime call overhead for strict arguments
and rest parameter creation.
R=jarin@chromium.orgTBR=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1693513002
Cr-Commit-Position: refs/heads/master@{#33925}
Moves InterpreterAssembler out of the compiler directory and into the
interpreter directory. Makes InterpreterAssembler as subclass of
CodeStubAssembler.
As part of this change, the special bytecode dispatch linkage type
is removed and instead we use a InterfaceDispatchDescriptor and
a normal CodeStub linkage type.
Removes a bunch of duplicated logic in InterpreterAssembler and
instead uses the CodeStubAssembler logic. Refactors Interpreter
with these changes.
Modifies CodeStubAssembler to add the extra operations required
by the Interpreter (extra call types, raw memory access and some extra
binary ops). Also adds the ability for subclasses to add extra
prologue and epilogue operations around calls, which is required
for the Interpreter.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1673333004
Cr-Commit-Position: refs/heads/master@{#33873}
Reason for revert:
Possibly causing Mozilla test failures - will investigate.
Original issue's description:
> [turbofan] ARM: Improve AND instruction selection
>
> Improve instruction selector for mask and shift operations by using cheaper
> instructions where possible, in preference to UBFX.
>
> BUG=
>
> Committed: https://crrev.com/53d9c12977f07f55b6f2a72128b8d02c4c857845
> Cr-Commit-Position: refs/heads/master@{#33843}
TBR=bmeurer@chromium.org,jarin@chromium.org,danno@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/1681953003
Cr-Commit-Position: refs/heads/master@{#33850}
Improve instruction selector for mask and shift operations by using cheaper
instructions where possible, in preference to UBFX.
BUG=
Review URL: https://codereview.chromium.org/1677023002
Cr-Commit-Position: refs/heads/master@{#33843}
This moves the JSCreate related functionality from JSTypedLowering into
a dedicated JSCreateLowering reducer. This is in preparation of landing
the support for optimized literals in TurboFan, which would blow up
JSTypedLowering quite seriously otherwise.
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1678833002
Cr-Commit-Position: refs/heads/master@{#33813}
Replace the somewhat awkward RestParamAccessStub, which would always
call into the runtime anyway with a proper FastNewRestParameterStub,
which is basically based on the code that was already there for strict
arguments object materialization. But for rest parameters we could
optimize even further (leading to 8-10x improvements for functions with
rest parameters), by fixing the internal formal parameter count:
Every SharedFunctionInfo has a formal_parameter_count field, which
specifies the number of formal parameters, and is used to decide whether
we need to create an arguments adaptor frame when calling a function
(i.e. if there's a mismatch between the actual and expected parameters).
Previously the formal_parameter_count included the rest parameter, which
was sort of unfortunate, as that meant that calling a function with only
the non-rest parameters still required an arguments adaptor (plus some
other oddities). Now with this CL we fix, so that we do no longer
include the rest parameter in that count. Thereby checking for rest
parameters is very efficient, as we only need to check whether there is
an arguments adaptor frame, and if not create an empty array, otherwise
check whether the arguments adaptor frame has more parameters than
specified by the formal_parameter_count.
The FastNewRestParameterStub is written in a way that it can be directly
used by Ignition as well, and with some tweaks to the TurboFan backends
and the CodeStubAssembler, we should be able to rewrite it as
TurboFanCodeStub in the near future.
Drive-by-fix: Refactor and unify the CreateArgumentsType which was
different in TurboFan and Ignition; now we have a single enum class
which is used in both TurboFan and Ignition.
R=jarin@chromium.org, rmcilroy@chromium.orgTBR=rossberg@chromium.org
BUG=v8:2159
LOG=n
Review URL: https://codereview.chromium.org/1676883002
Cr-Commit-Position: refs/heads/master@{#33809}
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}
So far, we've been moving down gaps wholesale. This change moves
individual move operations instead. This improves some benchmarks,
and should overall reduce code size, because it improves the chance of
reducing the number of moves.
For example, there are improvements on x64 in Emscripten (Bullet, in
particular) , JetStream geomean, Embenchen (zlib).
In the process of making this change, I noticed we can separate the
tasks performed by the move optimizer, as follows:
- group gaps into 1
- push gaps down, jumping instructions (these 2 were together before)
- merge blocks (and then push gaps down)
- finalize
We can do without a finalization list. This avoids duplicating storage -
we already have the list of instructions; it also simplifies the logic, since,
with this change, we may process an instruction's gap twice.
Compile time doesn't regress much (see pathological cases), but we
may want to avoid the allocations of the few sets used in the new code.
I'll do that in a subsequent change.
BUG=
Review URL: https://codereview.chromium.org/1634093002
Cr-Commit-Position: refs/heads/master@{#33715}
Move all the code that deals with falling back to object creation via
stubs to JSGenericLowering, where we can already deal well with stub
calls. This includes JSCreateLiteralArray, JSCreateLiteralObject,
JSCreateClosure, JSCreateFunctionContext and JSCreateArray.
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1653353002
Cr-Commit-Position: refs/heads/master@{#33682}
Avoid the hacking in JSIntrinsicLowering and provide a proper simplified
operator ObjectIsReceiver instead that is used to implement %_IsJSReceiver
which is used by our JavaScript builtins and the JSInliner.
R=jarin@chromium.org
BUG=v8:4544
LOG=n
Review URL: https://codereview.chromium.org/1657863004
Cr-Commit-Position: refs/heads/master@{#33675}
There's no point in having %_IsFunction as inline intrinsic, as it
is only used in non performance critical code, which is already full
of runtime calls anyway, so %IsFunction will do the trick as well.
R=yangguo@chromium.org
Review URL: https://codereview.chromium.org/1658123002
Cr-Commit-Position: refs/heads/master@{#33660}
This CL removes the Config templatization from the types. It is not
necessary anymore, after the HeapTypes have been removed.
The CL also changes the type hierarchy - the specific type kinds are
not inner classes of the Type class and they do not inherit from Type.
This is partly because it seems impossible to make this work without
templates. Instead, a new TypeBase class is introduced and all the
structural (i.e., non-bitset) types inherit from it.
The bitset type still requires the bit-munging hack and some nasty
reinterpret-casts to pretend bitsets are of type Type*. Additionally,
there is now the same hack for TypeBase - all pointers to the sub-types
of TypeBase are reinterpret-casted to Type*. This is to keep the type
constructors in inline method definitions (although it is unclear how
much that actually buys us).
In future, we would like to move to a model where we encapsulate Type*
into a class (or possibly use Type where we used to use Type*). This
would loosen the coupling between bitset size and pointer size, and
eventually we would be able to have more bits.
TBR=bradnelson@chromium.org
Review URL: https://codereview.chromium.org/1655833002
Cr-Commit-Position: refs/heads/master@{#33656}
Reason for revert:
problems on Mac64
Original issue's description:
> [turbofan] Add the StackSlot operator to turbofan.
>
> The StackSlot operator allows to allocate a spill slot on the stack. We
> are going to use this operator to pass floats through pointers to c
> functions, which we need for floating point rounding in the case where
> the architecture does not provide rounding instructions.
>
> R=titzer@chromium.org, v8-arm-ports@googlegroups.com, v8-ppc-ports@googlegroups.com, v8-mips-ports@googlegroups.com
>
> Committed: https://crrev.com/7a693437787090d62d937b862e29521debcc5223
> Cr-Commit-Position: refs/heads/master@{#33600}
TBR=titzer@chromium.org,v8-arm-ports@googlegroups.com,v8-mips-ports@googlegroups.com,v8-ppc-ports@googlegroups.com
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1644283002
Cr-Commit-Position: refs/heads/master@{#33601}
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}
Introduces the concept of transfer direction to register operands. This
enables the register translator to emit exactly the moves that a
bytecode having it's register operands translated needs.
BUG=v8:4280,v8:4675
LOG=N
Review URL: https://codereview.chromium.org/1633153002
Cr-Commit-Position: refs/heads/master@{#33544}
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}
* Add caching to handling of dangling loads
* Add two unittests for load elimination on escaped objects
BUG=v8:4586
LOG=n
Review URL: https://codereview.chromium.org/1619103004
Cr-Commit-Position: refs/heads/master@{#33498}
moves, we move those to the node, and remove them from the
predecessors ("merge" them to the common node).
If only some of the moves are common, we don't do anything. This is
what this change addresses.
The bug linked below should be addressed by this change. The only
difference in codegen before/after the change that introduced the bug
was un-merged moves.
BUG=chromium:549262
LOG=N
Review URL: https://codereview.chromium.org/1527203002
Cr-Commit-Position: refs/heads/master@{#33481}
Change the interpreter to always store the current context in the frame's
context slot instead of the function context. This makes it possible to
restore the correct context during deopt.
BUG=v8:4678,v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1604923002
Cr-Commit-Position: refs/heads/master@{#33477}
* Treat Select nodes as escaping
* Correctly void virtual field information
after a store to a non-const index
* Add a shortcut if all allocates escape
* Add a shortcut if no allocates are discovered
* Only reduce FrameState/StateValues nodes if they
have virtual allocates as input (transitively)
* Fix bug in FrameState/StateValues duplication
* Add check to verifier: First 3 inputs of FrameState
must be StateValues
R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n
Review URL: https://codereview.chromium.org/1583213003
Cr-Commit-Position: refs/heads/master@{#33406}
o Adds wide variants of bytecodes that have operands describing ranges
of registers. The upcoming wide register support does not suppport
re-mapping ranges.
o Adds kRegPair16 and kRegTriple16 operands required for new wide
bytecodes and renames Count8/Count16 operands to RegCount8/RegCount16.
o Removes Exchange bytecodes
BUG=v8:4675
LOG=NO
Review URL: https://codereview.chromium.org/1595103006
Cr-Commit-Position: refs/heads/master@{#33389}
Make ForInPrepare take a kRegTriple8 operand and ForInNext take kRegPair8
operand for cache state. This is to ensure that the cache state output of
ForInPrepare is in consecutive registers to allow us to deopt the
ForInPrepare node from TF->Ignition (to be done in a followup CL).
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1584813002
Cr-Commit-Position: refs/heads/master@{#33357}
This bug improves performance of escape analysis.
* A allocation discovery phase (EscapeAnalysis::AssignAliases)
ensures compact representation of virtual state
* Node revisiting in EscapeStatusAnalysis has been improved
* Escape analysis no longer requires a trimmed graph
BUG=v8:4586
LOG=n
Review URL: https://codereview.chromium.org/1559123003
Cr-Commit-Position: refs/heads/master@{#33267}
Adds support for calling runtime functions which return a pair of
values. Adds the bytecode CallRuntimePair. Also adds support to TurboFan
for calling stubs which return multiple values.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1568493002
Cr-Commit-Position: refs/heads/master@{#33181}
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}
New bytecodes for making registers with indicies wider than 1-byte
accessible.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1555713002
Cr-Commit-Position: refs/heads/master@{#33091}
According to the ES2015 specification, bound functions are exotic
objects, and thus don't need to be implemented as JSFunctions. So
we introduce a new JSBoundFunction type to represent bound functions
and make them optimizable. This already improves the performance of
calling or constructing bound functions by 10-100x depending on the
use case because we avoid the crazy dance between JavaScript and C++
that was implemented in v8natives.js previously.
There's still room for improvement in the performance of actually
creating bound functions, which is also relevant in practice, but
we already have a plan how to accomplish that later.
The mips/mips64 ports were contributed by akos.palfi@imgtec.com.
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_chromium_rel_ng;tryserver.blink:linux_blink_rel
BUG=chromium:535408, chromium:571299, v8:4629
LOG=n
Committed: https://crrev.com/ca8623eaa468cba65a5adafcdfb4615966f43ce2
Cr-Commit-Position: refs/heads/master@{#33042}
Review URL: https://codereview.chromium.org/1542963002
Cr-Commit-Position: refs/heads/master@{#33044}
Reason for revert:
Breaks arm64 sim nosnap: https://build.chromium.org/p/client.v8/builders/V8%20Linux%20-%20arm64%20-%20sim%20-%20nosnap%20-%20debug/builds/805/steps/Check/logs/function-bind
Original issue's description:
> [runtime] Introduce dedicated JSBoundFunction to represent bound functions.
>
> According to the ES2015 specification, bound functions are exotic
> objects, and thus don't need to be implemented as JSFunctions. So
> we introduce a new JSBoundFunction type to represent bound functions
> and make them optimizable. This already improves the performance of
> calling or constructing bound functions by 10-100x depending on the
> use case because we avoid the crazy dance between JavaScript and C++
> that was implemented in v8natives.js previously.
>
> There's still room for improvement in the performance of actually
> creating bound functions, which is also relevant in practice, but
> we already have a plan how to accomplish that later.
>
> The mips/mips64 ports were contributed by akos.palfi@imgtec.com.
>
> CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_chromium_rel_ng;tryserver.blink:linux_blink_rel
> BUG=chromium:535408, chromium:571299, v8:4629
> LOG=n
>
> Committed: https://crrev.com/ca8623eaa468cba65a5adafcdfb4615966f43ce2
> Cr-Commit-Position: refs/heads/master@{#33042}
TBR=cbruni@chromium.org,hpayer@chromium.org,yangguo@chromium.org,akos.palfi@imgtec.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=chromium:535408, chromium:571299, v8:4629
Review URL: https://codereview.chromium.org/1552473002
Cr-Commit-Position: refs/heads/master@{#33043}
According to the ES2015 specification, bound functions are exotic
objects, and thus don't need to be implemented as JSFunctions. So
we introduce a new JSBoundFunction type to represent bound functions
and make them optimizable. This already improves the performance of
calling or constructing bound functions by 10-100x depending on the
use case because we avoid the crazy dance between JavaScript and C++
that was implemented in v8natives.js previously.
There's still room for improvement in the performance of actually
creating bound functions, which is also relevant in practice, but
we already have a plan how to accomplish that later.
The mips/mips64 ports were contributed by akos.palfi@imgtec.com.
CQ_INCLUDE_TRYBOTS=tryserver.chromium.linux:linux_chromium_rel_ng;tryserver.blink:linux_blink_rel
BUG=chromium:535408, chromium:571299, v8:4629
LOG=n
Review URL: https://codereview.chromium.org/1542963002
Cr-Commit-Position: refs/heads/master@{#33042}
Move replacements out of virtual object. Replacements are
global to the graph and are not dependent on the virtual
state (after they are discovered).
R=mstarzinger@chromium.org
BUG=v8:4586
LOG=n
Review URL: https://codereview.chromium.org/1527533002
Cr-Commit-Position: refs/heads/master@{#32838}
Removes the dummy control and effect edges from the RMA Call nodes. This
requires a change to the node matchers to allow them to cope with nodes
which don't have control or effect matchers.
Review URL: https://codereview.chromium.org/1518673002
Cr-Commit-Position: refs/heads/master@{#32752}
Nowadays, representation inference and simplified lowering can insert the
right truncations based on the use.
Review URL: https://codereview.chromium.org/1512243002
Cr-Commit-Position: refs/heads/master@{#32747}
MachineType is now a class with two enum fields:
- MachineRepresentation
- MachineSemantic
Both enums are usable on their own, and this change switches some places from using MachineType to use just MachineRepresentation. Most notably:
- register allocator now uses just the representation.
- Phi and Select nodes only refer to representations.
Review URL: https://codereview.chromium.org/1513543003
Cr-Commit-Position: refs/heads/master@{#32738}
Extract ToBoolean hints from the fullcodegen code object and put them
into the ToBoolean nodes created by the AstGraphBuilder. We currently
do not yet consume this feedback, that will be done in a followup CL.
R=mstarzinger@chromium.org
BUG=v8:4583
LOG=n
Review URL: https://codereview.chromium.org/1494973002
Cr-Commit-Position: refs/heads/master@{#32576}
This moves the proper handling for the end node withing the constructed
graph into the RawMachineAssembler. This simplifies all assemblers and
makes the handling of {Start} and {End} symmetrical.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1493963003
Cr-Commit-Position: refs/heads/master@{#32563}
Also remove the ResultMode from ToBooleanStub and always return true or
false and use the same mechanism in fullcodegen. This is in preparation
for adding ToBoolean hints to TurboFan.
Drive-by-fix: We can use the power of the ToBooleanIC in TurboFan now
that the ResultMode is gone (and the runtime always returns true or
false from the miss handler).
R=mstarzinger@chromium.org
BUG=v8:4583
LOG=n
Review URL: https://codereview.chromium.org/1491223002
Cr-Commit-Position: refs/heads/master@{#32524}
We can constant fold %_IsJSReceiver(x) based on whether x is always a
receiver or can never be a receiver. This is important as
%_IsJSReceiver is inserted by the JSInliner.
R=jarin@chromium.org
BUG=v8:4544
LOG=n
Review URL: https://codereview.chromium.org/1486383003
Cr-Commit-Position: refs/heads/master@{#32519}
This is the first part of escape analysis for turbofan.
At the moment, there is no deopt support, and support
for loops is partial (only binary Phis are handled).
The CL includes 4 unittests.
There are also 8 new mjsunit tests, some of which are
skiped as they require features not yet implemented.
BUG=v8:4586
LOG=n
Review URL: https://codereview.chromium.org/1457683003
Cr-Commit-Position: refs/heads/master@{#32498}
Xori instruction can only have unisgned 16-bit immediates for right input,
as such it is not suitable for bit negation on mips.
TEST=unittests/InstructionSecetorTest.Word(32|64)XorMinusOneWithParameter
BUG=
Review URL: https://codereview.chromium.org/1485833003
Cr-Commit-Position: refs/heads/master@{#32478}
This is the initial support for binary operation hints on javascript
binary operators, i.e. JSAdd, JSSubtract and so on. The hints are
extracted from the fullcodegen code object before graph building and the
AstGraphBuilder puts those hints on the operators if available.
R=jarin@chromium.org
BUG=v8:4583
LOG=n
Review URL: https://codereview.chromium.org/1487973002
Cr-Commit-Position: refs/heads/master@{#32443}
an optomization to remove redundant cast operations.
1. Adds an optimization to remove redundant ToBoolean and ToName operations.
2. Adds implementation and tests for cast operatorts to bytecode graph builder.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1468003002
Cr-Commit-Position: refs/heads/master@{#32408}
Add initial support to optimize certain "prototype" loads from known
JSFunctions which have a prototype. This includes an appropriate typing
rule plus a matching rule for typed lowering.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1482213002
Cr-Commit-Position: refs/heads/master@{#32390}
Previously all contexts had a link to the global object, but what is
required in most cases (except for the global load, store and delete
case) is the native context.
This also removes the second dummy global object that was still linked
to every native context. We will add a different mechanism to ensure
that builtins do not pollute the actual global object during
bootstrapping.
Drive-by-fix: Unify some MacroAssembler magic and drop obsolete stuff.
CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux_nosnap_rel
R=yangguo@chromium.org,mstarzinger@chromium.org
Committed: https://crrev.com/d290f204938295bfecc5c8e645ccfcff6e80ddb8
Cr-Commit-Position: refs/heads/master@{#32375}
Review URL: https://codereview.chromium.org/1480003002
Cr-Commit-Position: refs/heads/master@{#32381}
Reason for revert:
[Sheriff] Breaks:
https://build.chromium.org/p/client.v8/builders/V8%20Linux%20-%20nosnap/builds/5472
Original issue's description:
> [runtime] Replace global object link with native context link in all contexts.
>
> Previously all contexts had a link to the global object, but what is
> required in most cases (except for the global load, store and delete
> case) is the native context.
>
> This also removes the second dummy global object that was still linked
> to every native context. We will add a different mechanism to ensure
> that builtins do not pollute the actual global object during
> bootstrapping.
>
> Drive-by-fix: Unify some MacroAssembler magic and drop obsolete stuff.
>
> R=yangguo@chromium.org
>
> Committed: https://crrev.com/d290f204938295bfecc5c8e645ccfcff6e80ddb8
> Cr-Commit-Position: refs/heads/master@{#32375}
TBR=yangguo@chromium.org,mstarzinger@chromium.org,bmeurer@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1478303002
Cr-Commit-Position: refs/heads/master@{#32377}
Previously all contexts had a link to the global object, but what is
required in most cases (except for the global load, store and delete
case) is the native context.
This also removes the second dummy global object that was still linked
to every native context. We will add a different mechanism to ensure
that builtins do not pollute the actual global object during
bootstrapping.
Drive-by-fix: Unify some MacroAssembler magic and drop obsolete stuff.
R=yangguo@chromium.org
Review URL: https://codereview.chromium.org/1480003002
Cr-Commit-Position: refs/heads/master@{#32375}
Decouple the implicit ToObject for with statements from the actual
creation of the with context. This way we can handle/optimize those
constructs separately.
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1481753003
Cr-Commit-Position: refs/heads/master@{#32341}
Removing bytecode graph builder tests as they are high maintenance and
have limited use, ie they track changes in the implementation rather
than behaviour.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1477783003
Cr-Commit-Position: refs/heads/master@{#32333}
Reason for revert:
This is also unsound for the reasons outlined in
https://codereview.chromium.org/1473073004/
Will reland the mjsunit test separately and help Fedor to implement a solution based on simplified operators.
Original issue's description:
> [machine-operator-reducer] fix float truncation
>
> Don't replace `TruncateFloat64ToInt32(RoundInt64ToFloat64(value))` with
> `value`. Generally, `value` may have a range bigger than the one that
> could fit into Int32. Replace it with `TruncateInt64ToInt32(value)`
> instead, and only if the `value` fits into Float64 without precision
> loss.
>
> Add missing mjsunit test for 52bit multiplication/division optimization
> that has landed in refs/heads/master@{#31899}.
>
> BUG=
> R=titzer@google.com
>
> Committed: https://crrev.com/64efa2a904773816968992628f0bf0f1b7ae82be
> Cr-Commit-Position: refs/heads/master@{#32227}
TBR=titzer@chromium.org,fedor@indutny.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=
Review URL: https://codereview.chromium.org/1468313009
Cr-Commit-Position: refs/heads/master@{#32312}
Reason for revert:
Unsound use of types in the MachineOperatorReducer. Will work on a sound solution with Fedor.
Original issue's description:
> [compiler] merge binary-operator-reducer
>
> Merge BinaryOperatorReducer into the MachineOperatorReducer class.
> It does not need `Revisit()` calls, because the newly inserted nodes are
> visited anyway, and there are no other methods that need AdvancedReducer
> there.
>
> BUG=
> R=titzer@chromium.org
>
> Committed: https://crrev.com/993ba9d2529a6401b3040b9263f8d06db7dbb4f1
> Cr-Commit-Position: refs/heads/master@{#32298}
TBR=titzer@chromium.org,fedor@indutny.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=
Review URL: https://codereview.chromium.org/1476763006
Cr-Commit-Position: refs/heads/master@{#32310}
Merge BinaryOperatorReducer into the MachineOperatorReducer class.
It does not need `Revisit()` calls, because the newly inserted nodes are
visited anyway, and there are no other methods that need AdvancedReducer
there.
BUG=
R=titzer@chromium.org
Review URL: https://codereview.chromium.org/1473073004
Cr-Commit-Position: refs/heads/master@{#32298}
Don't replace `TruncateFloat64ToInt32(RoundInt64ToFloat64(value))` with
`value`. Generally, `value` may have a range bigger than the one that
could fit into Int32. Replace it with `TruncateInt64ToInt32(value)`
instead, and only if the `value` fits into Float64 without precision
loss.
Add missing mjsunit test for 52bit multiplication/division optimization
that has landed in refs/heads/master@{#31899}.
BUG=
R=titzer@google.com
Review URL: https://codereview.chromium.org/1433353006
Cr-Commit-Position: refs/heads/master@{#32227}
Adds code and tests to support CreateClosure bytecode when building
graphs.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1458603012
Cr-Commit-Position: refs/heads/master@{#32224}
Change the runtime entries and their associated code stubs for object
and array literal creation to take the closure instead of the raw
literals pointer. This is way easier to deal with (and cleaner) in
TurboFan.
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1469833005
Cr-Commit-Position: refs/heads/master@{#32220}
Put the constant parts of the CreateLiteralArray and CreateLiteralObject
operators into CreateLiteralParameters and properly use them everywhere.
R=mstarzinger@chromium.org
Review URL: https://codereview.chromium.org/1475613002
Cr-Commit-Position: refs/heads/master@{#32207}
* Adds a PrepareForTailCall instruction that bumps the stack in the case that
the number of parameters passed to the callee causes the stack to exceed the
calleer's frame size.
* Uses the gap resolver to move the saved caller return address and frame
pointer to the approprate location in the tail-called frame.
BUG=v8:4076
LOG=n
Review URL: https://codereview.chromium.org/1455833004
Cr-Commit-Position: refs/heads/master@{#32151}
Adds support for the New, CallRuntime and CallJSRuntime bytecodes in
BytecodeGraphBuilder. Also adds BuildLoadObjectField,
BuildLoadGlobalObject and BuildLoadNativeContextField helpers.
Landed on behalf of rmcilroy.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1456483002
Cr-Commit-Position: refs/heads/master@{#32136}
This adds an explicit parameter to the call descriptor having kind
kJSCallFunction representing the new.target value. Note that for now
this parameter is not yet passed in and hence cannot be used yet. Also
contains some refactoring of how parameter index value are calculated,
establishing Linkage as the central point for such index computations.
This is a preparatory CL to allows us passing new.target in a register
instead of via a side-channel through the construct stub frame.
R=bmeurer@chromium.org
BUG=v8:4544
LOG=n
Review URL: https://codereview.chromium.org/1461973002
Cr-Commit-Position: refs/heads/master@{#32112}
This fixes undefined behavior with the OpParameter helper in some of our
node matchers. There was a constness mismatch of the template parameter.
R=oth@chromium.org,mythria@chromium.org
Review URL: https://codereview.chromium.org/1453973004
Cr-Commit-Position: refs/heads/master@{#32087}
Adds support for the LdaGlobal and StaGlobal bytecodes to the
BytecodeGraphBuilder. Also fixes a bug in the context node's parameter
index and start node inputs.
Landed on behalf of rmcilroy.
TBR=bmeuer@chromium.org,mythria@chromium.org
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1449373002
Cr-Commit-Position: refs/heads/master@{#32049}
Adds support for visiting the Call bytecode to the bytecode graph builder.
This change also adds the call type feedback slot to the Call bytecode.
This is not currently used by the interpreter, but is used by the
graph builder.
Also adds a CallWide varient of the Call bytecode, and adds the kCount16
operand type.
Landed on behalf of rmcilroy.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1456453002
Cr-Commit-Position: refs/heads/master@{#32033}
Adds implementation and tests for following operators in bytecode graph builder:
-VisitLoadICSloppy
-VisitLoadICStrict
-VisitLoadICSloppyWide
-VisitLoadICStrictWide
The current implementation introduces empty frame states for frame state inputs expected by these operations.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1419373007
Cr-Commit-Position: refs/heads/master@{#32026}
* Limit triggering of tail calls to explicit use of a new inline runtime
function %_TailCall. %_TailCall works just like %_Call except for using
tail-calling mechanics (currently only in TF).
* Remove hack that recognized some specific usages of %_Call and converted them
into tail calls.
* Support tail calls for all calls where the number of callee stack parameters
is less than or equal to the number of caller stack parameters.
* Use the gap resolver to swizzle parameters and registers to tail calls.
BUG=v8:4076
LOG=n
Review URL: https://codereview.chromium.org/1439613003
Cr-Commit-Position: refs/heads/master@{#31987}
This makes sure that inlining a constructor call to a function which
cannot be used as a constructor (e.g. strong mode function) still does
throw correctly when the implicit receiver is created.
R=bmeurer@chromium.org
TEST=mjsunit/regress/regress-inline-strong-as-construct
BUG=v8:4544
LOG=n
Review URL: https://codereview.chromium.org/1447443002
Cr-Commit-Position: refs/heads/master@{#31982}
This is necessary to allow more optimizations to take place between
the representation inference and change lowering. Perhaps we want
to rename SimplifiedLowering -> RepresentationInference and
ChangeLowering -> SimplifiedLowering.
Review URL: https://codereview.chromium.org/1439473003
Cr-Commit-Position: refs/heads/master@{#31976}
This adds initial support for fast inline allocations of JSObject
instances. It currently has exactly the same limitations as Crankshaft.
R=mstarzinger@chromium.org
BUG=v8:4544
LOG=n
Review URL: https://codereview.chromium.org/1441573004
Cr-Commit-Position: refs/heads/master@{#31957}
This passes both, the actual constructor and the original constructor,
to nodes having the {JSCreate} operator. This is required for allocating
properly subclassed implicit receiver objects.
R=verwaest@chromium.org
BUG=v8:4544
LOG=n
Review URL: https://codereview.chromium.org/1434873004
Cr-Commit-Position: refs/heads/master@{#31955}
This implements a first version of support for constructor call inlining
in the inlining machinery. For now we can only inline calls where the
actual constructor and the original constructor coincide (i.e. no super
constructor calls). Note that the target of a super constructor call is
loaded with a runtime call, so there is no way for it to be constant
promoted at the moment.
R=bmeurer@chromium.org
BUG=v8:4544
LOG=n
Review URL: https://codereview.chromium.org/1435873002
Cr-Commit-Position: refs/heads/master@{#31954}
This patch extends the typed lowering with a specialized version of 'instanceof' that is used if the "class", i.e. the constructor function, is a known constant.
Unittests check that replacement occurs as intended. Functional correctness is ensured by extensive unit tests covering instanceof already in the testsuite.
TESTS=unittests/JSTypedLoweringTest.{JSInstanceOfSpecializationWithSmiCheck,JSInstanceOfSpecializationWithoutSmiCheck,JSInstanceOfNoSpecialization}
Review URL: https://codereview.chromium.org/1407413014
Cr-Commit-Position: refs/heads/master@{#31916}
Use compare-negate instruction if the right-hand input to a compare is a
negate operation.
BUG=
Review URL: https://codereview.chromium.org/1410123009
Cr-Commit-Position: refs/heads/master@{#31866}
TurboFan didn't fully support the relevant ES6 type conversion
intrinsics like %_ToNumber, %_ToLength, %_ToName, %_ToString and
%_ToInteger until now, we always went to the runtime instead. These
intrinsics are now well supported in TurboFan, and we are even able to
generate quite decent code in some cases.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1428243003
Cr-Commit-Position: refs/heads/master@{#31820}
This deprecates the ability of the raw machine assembler to utilize the
CallFunctionStub in preparation of the stub itself being deprecated. We
only used this to test instruction selection of calls to stubs that can
deoptimize, the test has been adapted.
R=verwaest@chromium.org
TEST=unittests/InstructionSelectorTest
Review URL: https://codereview.chromium.org/1408193006
Cr-Commit-Position: refs/heads/master@{#31799}
Use the Call builtin instead, which does the right thing(TM)
always, especially since the CallFunctionStub is going away.
R=jarin@chromium.org
BUG=v8:4413
LOG=n
Review URL: https://codereview.chromium.org/1410853007
Cr-Commit-Position: refs/heads/master@{#31794}
The callees are expected to properly set the number of actual
arguments passed to the callee, which is now represented correctly
in the TurboFan graphs by a new Parameter right before the context
Parameter. Currently this is only being used for outgoing calls.
Note that this requires disabling two of the TF code stub tests,
because of the JavaScript graphs are not automagically compatible
with abitrary (incoming) code stub interface descriptors. If we
want to support JS code stubs at all, then we need to find a sane
way to feed in this information.
Drive-by-fix: Don't insert a direct call to a classConstructor.
R=mstarzinger@chromium.org
BUG=v8:4413, v8:4428
LOG=n
Review URL: https://codereview.chromium.org/1410633006
Cr-Commit-Position: refs/heads/master@{#31789}
Adds support for calling JS runtime functions. Also changes the bytecode
array builder to allow calling functions with an invalid argument
register if the call takes no arguments.
Adds the bytecode CallJSRuntime.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1410003003
Cr-Commit-Position: refs/heads/master@{#31774}
This moves the optimization for variables loads targeting lookup slots
in DYNAMIC_GLOBAL and DYNAMIC_LOCAL mode into the AstGraphBuilder. This
way we implicitly get all optimizations that target global loads and
context loads for free.
R=bmeurer@chromium.org
BUG=v8:4513
LOG=n
Review URL: https://codereview.chromium.org/1424943008
Cr-Commit-Position: refs/heads/master@{#31713}
This adds optimized lowering for JSConvertReceiver (in the general case)
and JSToObject in typed lowering. It also uses JSConvertReceiver for
direct calls in typed lowering.
R=mstarzinger@chromium.org
BUG=v8:4493
LOG=n
Review URL: https://codereview.chromium.org/1431543002
Cr-Commit-Position: refs/heads/master@{#31676}
In order to properly (lazy) bailout when converting the receiver for
sloppy mode functions (using the newly added JSConvertReceiver
operator), we need to have a bailout location right before every call
(also right before every %_Call and %_CallFunction), otherwise if the
JSConvertReceiver just reuses the lazy bailout frame state from the
JSCallFunction node, it will skip the whole function in case of lazy
bailout.
Note it should be impossible to trigger this currently because we do not
yet support AllocationSite code dependencies in TurboFan, which can
trigger this kind of lazy bailout; therefore it's not possible to write
a regression test (yet).
R=yangguo@chromium.org
BUG=v8:4493
LOG=n
Review URL: https://codereview.chromium.org/1425883004
Cr-Commit-Position: refs/heads/master@{#31668}
we may introduce moves that are redundant in the context of
moves on subsequent instructions. Currently, we only detect such
redundancies by allowing moves to skip over Nop instructions (true
nops, with no input/output). We can also skip over other cases, for
example over constant definitions (nop with an output), since whatever
moves happen above it do not influence the instruction's outcome.
We may be able to handle other cases, too - in subsequent CLs.
BUG=
Review URL: https://codereview.chromium.org/1422333003
Cr-Commit-Position: refs/heads/master@{#31662}
For..in introduces 3 new bytecodes ForInPrepare, ForInNext, and
ForInDone to start a for..in loop, get the next element, and check if
the loop is done.
For..in builds upon new LoopBuilder constructs for conditionally
breaking and continuing during iteration: BreakIf{Null|Undefined}
and ContinueIf{Null|Undefined}. New conditional jump bytecodes
support this succinctly: JumpIfNull and JumpIfUndefined.
Add missing check to BytecodeLabel that could allow multiple
forward referencess to the same label which is not supported.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1422033002
Cr-Commit-Position: refs/heads/master@{#31651}
Rename ZoneTypeCache to TypeCache and use a single shared (immutable)
instance consistently to cache the most commonly used types. Also serves
as a chokepoint for defining those types, so we don't repeat the
definition (and possible bugs) in various places.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1409763004
Cr-Commit-Position: refs/heads/master@{#31631}
Float(32|64)Min:
// (a < b) ? a : b
fcmp da, db
fcsel dd, da, db, lo
Float(32|64)Max:
// (b < a) ? a : b
fcmp db, da
fcsel dd, da, db, lo
BUG=
Review URL: https://codereview.chromium.org/1360603003
Cr-Commit-Position: refs/heads/master@{#31621}
Previously ChangeLowering would always box float64 values when going to
tagged representation, but that introduces a lot of deoptimizer loops
and polymorphism into TurboFan, which is unfortunate and unnecessary.
This adds some logic to ChangeFloat64ToTagged to try harder to create a
Smi when going from Float64 to Tagged, instead of always allocating a
HeapNumber. This might need some additional tweaking, but at least it
makes it possible to start comparing TurboFan and Crankshaft for some
regular JavaScript.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1420913003
Cr-Commit-Position: refs/heads/master@{#31609}
Up until now, if one wanted to specify an explicit stack location or register as an operand for an instruction, it had to also be
explicitly associated with a virtual register as a so-called
FixedRegister or FixedStackSlot.
For the implementation of tail calls, the plan is to use the gap
resolver needs to shuffle stack locations from the caller to the
tail-called callee. In order to do this, it must be possible to
explicitly address operand locations on the stack that are not
associated with virtual registers.
This CL introduces ExplictOperands, which can specify a specific
register or stack location that is not associated with virtual
register. This will allow tail calls to specify the target
locations for the necessary stack moves in the gap for the tail
call without the core register allocation having to know about
the target of the stack moves at all.
In the process this CL:
* creates a new Operand kind, ExplicitOperand, with which
instructions can specify register and stack slots without an
associated virtual register.
* creates a LocationOperand class from which AllocatedOperand and
ExplicitOperand are derived and provides a common interface to
get Register, DoubleRegister and spill slot information.
* removes RegisterOperand, DoubleRegisterOperand,
StackSlotOperand and DoubleStackSlotOperand, they are subsumed
by LocationOperand.
* addresses a cleanup TODO in AllocatedOperand to reduce the
redundancy of AllocatedOperand::Kind by using machine_type() to
determine if an operand corresponds to a general purpose or
double register.
BUG=v8:4076
LOG=n
Review URL: https://codereview.chromium.org/1389373002
Cr-Commit-Position: refs/heads/master@{#31603}
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}
Adds support for loading from and storing to outer context
variables. Also adds support for declaring functions on contexts and
locals. Finally, fixes a couple of issues with StaContextSlot where
we weren't emitting the write barrier and therefore would crash in the
GC.
Also added code so that --print-bytecode will output the
function name before the bytecodes, and replaces MachineType with StoreRepresentation in RawMachineAssembler::Store and updates tests.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1425633002
Cr-Commit-Position: refs/heads/master@{#31584}
From the Google C++ style guide: "You may not use a using-directive to
make all names from a namespace available". This would be covered by
presubmit linter checks if build/namespaces were not blacklisted.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1410073004
Cr-Commit-Position: refs/heads/master@{#31565}
register configurations currently. This CL provides a mechanism so that
optimizing compilers can select different Register Configuration.
BUG=
Review URL: https://codereview.chromium.org/1405673003
Cr-Commit-Position: refs/heads/master@{#31476}
The plan is to implement the same idea using vector IC machinery.
Stubs implementations and scopes modifications are left untouched for now.
Review URL: https://codereview.chromium.org/1419823003
Cr-Commit-Position: refs/heads/master@{#31458}
Use a unified NamedAccess operator parameter for both JSLoadNamed and
JSStoreNamed, and similar use PropertyAccess for both JSLoadProperty and
JSStoreProperty.
Review URL: https://codereview.chromium.org/1418993002
Cr-Commit-Position: refs/heads/master@{#31456}
Revert "Revert of [turbofan] Splinter into one range.
(patchset #2 id:80001 of https://codereview.chromium.org/1391023007/ )"
This reverts commit 23a8837fcc.
Also added a CHECK in Merge to validate that splitting yields a different
range and thus advances the algorithm. Ran stress bots successfully. Likely my earlier change in Splintering addressed the stress test scenario
that was looping infinitely.
BUG=
Review URL: https://codereview.chromium.org/1406983004
Cr-Commit-Position: refs/heads/master@{#31430}
This removes all locally constructed SimplifiedOperatorBuilder instances
and uses the one passed along the JSGraph. It ensures that the correct
zone is used to allocate operators, no matter where the reducer is used.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1410003002
Cr-Commit-Position: refs/heads/master@{#31355}
Removes a branch that checks for a condition that has been checked on dominators of the branch.
This introduces a new reducer that propagates the list of checked conditions (and their boolean values) through the control flow graph. If it encounters a branch checking a condition with a known value, the branch is eliminated.
The analysis relies on loops being reducible: if a condition has been checked on all paths to loop entry, then it is checked in the loop (regardless what of the conditions checked inside the loop).
The implementation is fairly naive and could be improved:
- all the operation on the condition lists could be made allocation-free when revisited.
- we could try to use a map structure rather than a linked list (to make
lookups faster).
- the merging of control flow could be changed to take into account
conditions from non-dominating paths (as long as all paths check
the condition).
Review URL: https://codereview.chromium.org/1376293005
Cr-Commit-Position: refs/heads/master@{#31347}
Adds support for local context loads and stores. Also adds support for
creation of new block contexts (e.g., for let variables) and initializing
const / let variables with the hole appropriately.
Also adds some checks to ensure BytecodeArrayBuilder::context_count is set
appropriately and fixes tests to do so.
Adds the bytecode StaContextSlot.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1403943004
Cr-Commit-Position: refs/heads/master@{#31343}
This fixes the lifetime of nodes created by JSGlobalSpecialization that
contain a simplified operator. In the case where this reducer runs as
part of the inliner, the SimplifiedOperatorBuilder was instantiated with
the wrong zone. This led to use-after-free of simplified operators.
To avoid such situations in the future, we decided to move this operator
builder into the JSGraph and make the situation uniform with all other
operator builders.
R=bmeurer@chromium.org
BUG=chromium:543528
LOG=n
Review URL: https://codereview.chromium.org/1409993002
Cr-Commit-Position: refs/heads/master@{#31334}
Reason for revert:
Weird endless loop in TopLevelLiveRange::Merge() due to always splitting first and not making progress. See comments, unfortunately no useable repro.
Original issue's description:
> [turbofan] Splinter into one range.
>
> Before this CL, we created one live range per successive set of
> deferred blocks. For scenarios with many such blocks, this creates
> an upfront pressure for the register allocator to deal with many ranges.
> Linear sorts ranges, which is a super-linear operation.
>
> The change places all deferred intervals into one range, meaning that,
> at most, there will be twice as many live ranges as the original set. In
> pathological cases (benchmarks/Compile/slow_nbody1.js), this change
> halves the compilation time. We see some improvements elsewhere,
> notably SQLite at ~4-5%.
>
> We may be able to avoid the subsequent merge. Its cost is the
> additional ranges it may need to create. The sole reason for the merge
> phase is to provide an unchanged view of the world to the subsequent
> phases. With the at-most-one splinter model, we may be able to teach
> the other phases about splintering - should we find perf hindrances
> due to merging.
>
> Committed: https://crrev.com/efdcd20267870276c5824f1ccf4e171ac378f7ae
> Cr-Commit-Position: refs/heads/master@{#31224}
TBR=jarin@chromium.org,mtrofin@google.com,mtrofin@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1403163003
Cr-Commit-Position: refs/heads/master@{#31300}
This CL re-purposes ValueEffect and Finish as delimiters for regions
that are scheduled atomically (renamed to BeginRegion, FinishRegion).
The BeginRegion node takes and produces an effect. For the uses that do
not care about the placement in the effect chain, it is ok to feed
graph->start() as an effect input.
The FinishRegion takes a value and an effect and produces a value and
an effect. It is important that any value or effect produced inside the
region is not used outside the region. The FinishRegion node is the only
way to smuggle an effect and a value out.
At the moment, this does not support control flow inside the region. Control flow would be hard.
During scheduling we do some sanity check, but the checks are not exhaustive. Here is what we check:
- the effect chain between begin and finish is linear (no splitting,
single effect input and output).
- any value produced is consumed by the FinishRegion node.
- no control flow outputs.
Review URL: https://codereview.chromium.org/1399423002
Cr-Commit-Position: refs/heads/master@{#31265}
Support negate with shifted input on ARM64 by supporting lhs zero registers for
binary operations, and removing explicit Neg instruction support.
Review URL: https://codereview.chromium.org/1404093003
Cr-Commit-Position: refs/heads/master@{#31263}
Before this CL, we created one live range per successive set of
deferred blocks. For scenarios with many such blocks, this creates
an upfront pressure for the register allocator to deal with many ranges.
Linear sorts ranges, which is a super-linear operation.
The change places all deferred intervals into one range, meaning that,
at most, there will be twice as many live ranges as the original set. In
pathological cases (benchmarks/Compile/slow_nbody1.js), this change
halves the compilation time. We see some improvements elsewhere,
notably SQLite at ~4-5%.
We may be able to avoid the subsequent merge. Its cost is the
additional ranges it may need to create. The sole reason for the merge
phase is to provide an unchanged view of the world to the subsequent
phases. With the at-most-one splinter model, we may be able to teach
the other phases about splintering - should we find perf hindrances
due to merging.
Review URL: https://codereview.chromium.org/1391023007
Cr-Commit-Position: refs/heads/master@{#31224}
The test had an effect phi with one effect input connected to a loop with two control inputs. Also, the Terminate node was used by the effect phi.
Review URL: https://codereview.chromium.org/1398763002
Cr-Commit-Position: refs/heads/master@{#31193}
This will allow exploration of possibilities like passing around buffer base and length.
BUG=None
TEST=test-multiple-return
LOG=N
R=mtrofin@chromium.org,titzer@chromium.org
Review URL: https://codereview.chromium.org/1391333003
Cr-Commit-Position: refs/heads/master@{#31184}
Implements support for declaring global variables. Also adds support for loading
from and storing to both global and unallocated global variables. Adds the
following bytecodes:
- StoreGlobal
- LoadContextSlot
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1378523005
Cr-Commit-Position: refs/heads/master@{#31166}
Optimizing global constants such as "NaN", "Infinity" and "undefined" is
best performed during graph building. Then the optimization and lowering
passes only need to deal with real loads in case of JSLoadGlobal.
R=mstarzinger@chromium.org
BUG=v8:4470
LOG=n
Review URL: https://codereview.chromium.org/1384953002
Cr-Commit-Position: refs/heads/master@{#31135}
Adds support for strict mode load / store ICs and cleans up BinaryOp and
CompareOp to only trigger an UNIMPLEMENTED abort if called with STRONG
mode (which is the only language mode which has different compare/binary ops.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1385623002
Cr-Commit-Position: refs/heads/master@{#31134}
A subset of the LiveRangeUnitTests (SplitInvalidPreStart,
InvalidSplitEnd, SplitInvalidPostEnd) fail or hang on AIX in release
mode.
These tests fork a child which is expected to crash in
register-allocator code after feeding in bad inputs.
In debug mode, they behave as expected due to hitting a debug assert.
In release mode, however, the tests rely only on the fact that
dereferencing a null pointer will cause a SEGFAULT. This is true on
most platforms, but not AIX. An AIX process has valid low memory
pages mapped for reading and will not fault. Thus, these tests fail
or hang because the child process survives the load from address zero
and either completes (with undefined results) or goes into an infinite
loop.
R=bmeurer@chromium.org, danno@chromium.org, michael_dawson@ca.ibm.com
BUG=
Review URL: https://codereview.chromium.org/1384733002
Cr-Commit-Position: refs/heads/master@{#31090}
Adds support for calling runtime functions from the interpreter. Adds the
CallRuntime bytecode which takes a Runtime::FunctionId of the function to call
and the arguments in sequential registers. Adds a InterpreterCEntry builtin
to enable the interpreter to enter C++ code based on the functionId.
Also renames Builtin::PushArgsAndCall to Builtin::InterpreterPushArgsAndCall
and groups all the interpreter builtins together.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1362383002
Cr-Commit-Position: refs/heads/master@{#31089}
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}
Reason for revert:
Now breaking arm32 debug bot (worked locally even with --debug-code, so I'll need to figure out what's different on the bot)
Original issue's description:
> [Interpreter] Add CallRuntime support to the interpreter.
>
> Adds support for calling runtime functions from the interpreter. Adds the
> CallRuntime bytecode which takes a Runtime::FunctionId of the function to call
> and the arguments in sequential registers. Adds a InterpreterCEntry builtin
> to enable the interpreter to enter C++ code based on the functionId.
>
> Also renames Builtin::PushArgsAndCall to Builtin::InterpreterPushArgsAndCall
> and groups all the interpreter builtins together.
>
> BUG=v8:4280
> LOG=N
>
TBR=bmeurer@chromium.org,oth@chromium.org,mstarzinger@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:4280
Review URL: https://codereview.chromium.org/1379933003
Cr-Commit-Position: refs/heads/master@{#31078}
Adds support for calling runtime functions from the interpreter. Adds the
CallRuntime bytecode which takes a Runtime::FunctionId of the function to call
and the arguments in sequential registers. Adds a InterpreterCEntry builtin
to enable the interpreter to enter C++ code based on the functionId.
Also renames Builtin::PushArgsAndCall to Builtin::InterpreterPushArgsAndCall
and groups all the interpreter builtins together.
BUG=v8:4280
LOG=N
Committed: https://crrev.com/40e8424b744f8b6e3e1d93e20f23487419911dfc
Cr-Commit-Position: refs/heads/master@{#31064}
Review URL: https://codereview.chromium.org/1362383002
Cr-Commit-Position: refs/heads/master@{#31076}
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}
This lowers JSCreateFunctionContext nodes to call the above stub for
help with allocating function contexts when possible. It also contains
an implementation for inlined allocations of such contexts, which is
still behind a flag until inlined allocations are ready for prime time.
TEST=unittests/JSTypedLoweringTest.JSCreateFunctionContext
R=mvstanton@chromium.org
Review URL: https://codereview.chromium.org/1380113002
Cr-Commit-Position: refs/heads/master@{#31068}
Reason for revert:
Broke Arm64 bot (CEntry stub is trying to pop arguments off stack when argv_in_reg, so I need to fix this).
Original issue's description:
> [Interpreter] Add CallRuntime support to the interpreter.
>
> Adds support for calling runtime functions from the interpreter. Adds the
> CallRuntime bytecode which takes a Runtime::FunctionId of the function to call
> and the arguments in sequential registers. Adds a InterpreterCEntry builtin
> to enable the interpreter to enter C++ code based on the functionId.
>
> Also renames Builtin::PushArgsAndCall to Builtin::InterpreterPushArgsAndCall
> and groups all the interpreter builtins together.
>
> BUG=v8:4280
> LOG=N
>
> Committed: https://crrev.com/40e8424b744f8b6e3e1d93e20f23487419911dfc
> Cr-Commit-Position: refs/heads/master@{#31064}
TBR=bmeurer@chromium.org,oth@chromium.org,mstarzinger@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:4280
Review URL: https://codereview.chromium.org/1387543002
Cr-Commit-Position: refs/heads/master@{#31066}
Adds support for calling runtime functions from the interpreter. Adds the
CallRuntime bytecode which takes a Runtime::FunctionId of the function to call
and the arguments in sequential registers. Adds a InterpreterCEntry builtin
to enable the interpreter to enter C++ code based on the functionId.
Also renames Builtin::PushArgsAndCall to Builtin::InterpreterPushArgsAndCall
and groups all the interpreter builtins together.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1362383002
Cr-Commit-Position: refs/heads/master@{#31064}
Adds support for short operands, starting with kIdx16. Introduces
BytecodeTraits to enable compile time determination of various traits for a
bytecode, such as size, operands, etc. Reworks BytecodeIterator,
BytecodeArrayBuilder and Bytecodes::Decode to support 16 bit operands. Adds
support to Interpreter to load 16 bit operands.
Also fixes a bug with ToBoolean where it wouldn't get emitted at the start
of a block, and added a test.
BytecodeTraits template magic inspired by oth@chromium.org.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1370893002
Cr-Commit-Position: refs/heads/master@{#31058}
This enables linter checking for "readability/namespace" violations
during presubmit and instead marks the few known exceptions that we
allow explicitly.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1371083003
Cr-Commit-Position: refs/heads/master@{#31019}
This changes the operators for JSCreate[Block|Script]Context to take
their ScopeInfo as a static parameter as opposed to a value input and
in turn allows for easier access to that parameter during lowerings.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1379593002
Cr-Commit-Position: refs/heads/master@{#31009}
This is necessary because these operators can read heap (equality can actually write heap when flattening strings).
BUG=v8:4446
LOG=n
Review URL: https://codereview.chromium.org/1374683002
Cr-Commit-Position: refs/heads/master@{#31005}
Verifies consistency of node inputs and uses:
- node inputs should agree with the input count computed from the node's operator.
- effect inputs should have effect outputs (or be a sentinel).
- control inputs should have control outputs (or be a sentinel).
- frame state inputs should be frame states (or be a sentinel).
- if the node has control uses, it should produce control.
- if the node has effect uses, it should produce effect.
- if the node has frame state uses, it must be a frame state.
I also removed some tests, either because they did not seem to be useful (scheduler) or they tested dead functionality (diamond effect phi).
Review URL: https://codereview.chromium.org/1368913002
Cr-Commit-Position: refs/heads/master@{#30927}
+ Add bytecodes for conditional and unconditional jumps.
+ Add bytecodes for test/compare operations.
+ Expose jumps in bytecode-array-builder and add BytecodeLabel class for
identifying jump targets.
+ Add support for if..then...else in the bytecode-generator.
+ Implement jump bytecodes in the interpreter. Test/compare operations
dependent on runtime call for comparisons.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1343363002
Cr-Commit-Position: refs/heads/master@{#30918}
This introduces the NodeProperties::ChangeOp helper which guards node
operator changes so that additional checking can be done without any
additional dependencies being pulled into the Node class. For now only
the input count is checked, but additional checking might follow.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1366753003
Cr-Commit-Position: refs/heads/master@{#30916}
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 patch checks the type of the lhs operand of a floating point
comparison for ARM, and commutes the operands if it is #0.0. It allows
us to optimize a comparison with zero, as the vcmp instruction
accepts #0.0 as rhs operand.
Code before for "0.0 < 0.123":
------------------------------
movw ip, #29360
movt ip, #37224
movw r9, #31981
movt r9, #16319
vmov d0, ip, r9
mov ip, #0
vmov d1, ip, ip
vcmp.f64 d1, d0
vmrs APSR, FPSCR
bcc +12
Code after:
-----------
movw ip, #29360
movt ip, #37224
movw r9, #31981
movt r9, #16319
vmov d0, ip, r9
vcmp.f64 d0, #0.0
vmrs APSR, FPSCR
bgt +12
BUG=
Review URL: https://codereview.chromium.org/1361913003
Cr-Commit-Position: refs/heads/master@{#30911}
Reduce operations of the form f64cmp(fp32to64(x), k) to f32cmp(x, k) when k
can be encoded as a 32-bit float.
Review URL: https://codereview.chromium.org/1365623002
Cr-Commit-Position: refs/heads/master@{#30909}
This patch explicitly names commuted conditions for floating point
comparisons, instead of relying on CommuteFlagsCondition. Otherwise, a
bug in this function would not be caught.
BUG=
Review URL: https://codereview.chromium.org/1364773002
Cr-Commit-Position: refs/heads/master@{#30905}
This patch checks the type of the lhs operand of a floating point
comparison, and commutes the operands if it is #0.0. It allows us to
optimize a comparison with zero, as the fcmp instruction accepts #0.0 as
rhs operand.
Code before for "0.0 < 0.123":
------------------------------
fmov d1, xzr
ldr d0, pc+96
fcmp d1, d0
b.lo #+0xc
Code after:
-----------
ldr d0, pc+92
fcmp d0, #0.0
b.gt #+0xc
Before this patch, we used unsigned condition codes for floating point
comparisons, but the unordered case was not correctly commuted.
Review URL: https://codereview.chromium.org/1356283003
Cr-Commit-Position: refs/heads/master@{#30881}
Adds support for JS calls to the interpreter. In order to support
calls from the interpreter, the PushArgsAndCall builtin is added
which pushes a sequence of arguments onto the stack and calls
builtin::Call.
Adds the Call bytecode.
MIPS port contributed by akos.palfi@imgtec.com in https://codereview.chromium.org/1334873002/
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1323463005
Cr-Commit-Position: refs/heads/master@{#30710}
Replace the ADD, SUB, etc. builtins with proper runtime implementations,
and expose them as runtime calls that can be used by the code stubs and
the interpreter (for now).
Also remove all the support runtime functions for ADD, SUB and friends,
namely %NumberAdd, %NumberSub, and so on.
R=mstarzinger@chromium.org
CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux_layout_dbg,v8_linux_nosnap_dbg
Review URL: https://codereview.chromium.org/1333843002
Cr-Commit-Position: refs/heads/master@{#30680}
There are now two runtime entries %NewClosure and %NewClosure_Tenured,
with the same signature (one parameter, the SharedFunctionInfo, and the
context of the caller).
Also remove the HFunctionLiteral special case instruction from Crankshaft,
as HCallWithDescriptor with FastNewClosureStub or HCallRuntime with
either %NewClosure or %NewClosure_Tenured can easily do that for you.
Also remove the redundant context parameter from the JSCreateClosure
operator, because every JS operator already takes a context input.
CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_linux_nosnap_dbg
Review URL: https://codereview.chromium.org/1329293003
Cr-Commit-Position: refs/heads/master@{#30671}
A LiveRange is identified by 2 integers: the vreg() of its TopLevel,
which is the virtual register (operand) ID; and a relative_id(), which has
no meaning in the program, but is valuable in debugging or tracing
scenarios.
This change ensures that relative_id is unique even in cases of splinter
ranges and their children.
Review URL: https://codereview.chromium.org/1318493005
Cr-Commit-Position: refs/heads/master@{#30665}
Similar to DELETE, the IN builtin is just a thin wrapper for %HasElement
and %HasProperty anyway, and cannot be optimized, plus it had a weird
special fast case (which also involved at least one LOAD_IC plus some
intrinsic magic).
R=yangguo@chromium.org,jarin@chromium.org
CQ_INCLUDE_TRYBOTS=tryserver.v8:v8_win_nosnap_shared_rel
Committed: https://crrev.com/72d60a1e80e81e2e68ca402665e2acbc46c5e471
Cr-Commit-Position: refs/heads/master@{#30154}
Review URL: https://codereview.chromium.org/1295433002
Cr-Commit-Position: refs/heads/master@{#30582}
Once a range is found to have a conflict, split around all the calls it
crosses over, since it will anyway have conflicts there, too.
Incrementally, from the last change to greedy, this change brings
overall improvement in benchmarks. In fact, except for 2 regressions
in Jetstream (splay-latency and date-format-xparb, at 6 and 7%
respectivelly), everything else is in the green or noise. Quite a few
benchmarks are over 3%, with a few (zlib, for example) in the double
digits.
Review URL: https://codereview.chromium.org/1328783002
Cr-Commit-Position: refs/heads/master@{#30579}
Adds support for property load operations via Load/KeyedLoad ICs. Adds the
following bytecodes:
- LoadIC
- KeyedLoadIC
Also adds support to the interpreter assembler for loading the type feedback
vector from the function on the stack, and calling ICs.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1309843007
Cr-Commit-Position: refs/heads/master@{#30543}
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}
This CL us a pure refactoring that makes an empty compilation unit
including just "isolate.h" or "contexts.h" but not "objects-inl.h"
compile without warnings or errors. This is needed to further reduce
the header dependency tangle.
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1322883002
Cr-Commit-Position: refs/heads/master@{#30500}
I plan to reuse this for live range splitting/splintering/merging tests.
BUG=
Review URL: https://codereview.chromium.org/1305313008
Cr-Commit-Position: refs/heads/master@{#30461}
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}
Adds support to the interpreter for loading literals from the constant pool.
Adds the LoadConstant bytecode and makes use of it for loading large Smis and
HeapObject literals.
Also removes unused HandleVector from utils.h.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1321663003
Cr-Commit-Position: refs/heads/master@{#30450}
A TopLevelLiveRange is the live range of a virtual register. Through
register allocation, it may end up being split in a succession of child
live ranges, where data flow is handled through moves from
predecessor to successor child.
Today, the concepts of "top level" and "child" live ranges are conflated
under the LiveRange class. However, a good few APIs pertain solely
to TopLevelLiveRanges. This was communicated through comments or
DCHECKs - but this makes for poor code comprehensibility and maintainability.
For example, the worklist of the register allocator (live_ranges()) needs
to only contain TopLevelLiveRanges; spill range concerns are associated
only with the top range; phi-ness; certain phases in the allocation pipeline;
APIs on LiveRange used for initial construction - before splitting;
splintering - these are all responsibilities associated to TopLevelLiveRanges,
and not child live ranges.
This change separates the concepts.
An effect of this change is that child live range allocation need not involve
RegisterAllocationData. That's "a good thing" (lower coupling), but it has
the side-effect of not having a good way to construct unique identifiers for
child live ranges, relative to a given InstructionSequence.
LiveRange Id are used primarily for tracing/output-ing, and debugging.
I propose a 2-component identifier: a virtual register (vreg) number,
uniquely identifying TopLevelLiveRanges; and a relative identifier, which
uniquely identifies children of a given TopLevelLiveRange. "0" is reserved
for the TopLevel range. The relative identifier does not necessarily
indicate order in the child chain, which is no worse than the current state
of affairs.
I believe this change should make it easier to understand a trace output
(because the virtual register number is readily available). I plan to formalize
with a small structure the notion of live range id, and consolidate tracing
around that, as part of a separate CL. (there are seemingly disparate ways
to trace - printf or stream-based APIs - so this seems like an opportune
change to consolidate that)
Review URL: https://codereview.chromium.org/1311983002
Cr-Commit-Position: refs/heads/master@{#30370}
This change encompasses what is necessary to enable stack checks in loops without suffering large regressions.
Primarily, it consists of a new mechanism for dealing with deferred blocks by "splintering", rather than splitting, inside deferred blocks.
My initial change was splitting along deferred block boundaries, but the regression introduced by stackchecks wasn't resolved conclusively. After investigation, it appears that just splitting ranges along cold block boundaries leads to a greater opportunity for moves on the hot path, hence the suboptimal outcome.
The alternative "splinters" ranges rather than splitting them. While splitting creates 2 ranges and links them (parent-child), in contrast, splintering creates a new independent range with no parent-child relation to the original. The original range appears as if it has a liveness hole in the place of the splintered one. All thus obtained ranges are then register allocated with no change to the register allocator.
The splinters (cold blocks) do not conflict with the hot path ranges, by construction. The hot path ones have less pressure to split, because we remove a source of conflicts. After allocation, we merge the splinters back to their original ranges and continue the pipeline. We leverage the previous changes made for deferred blocks (determining where to spill, for example).
Review URL: https://codereview.chromium.org/1305393003
Cr-Commit-Position: refs/heads/master@{#30357}
Adds implementations and tests for the following bytecodes:
- Add
- Sub
- Mul
- Div
- Mod
Also adds the Mod bytecode and adds support to BytecodeGenerator and
BytecodeArrayBuilder to enable it's use.
The current bytecodes always call through to the JS builtins. This also adds
LoadObjectField and CallJSBuiltin operators to the InterpreterAssembler.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1300813005
Cr-Commit-Position: refs/heads/master@{#30352}
Passes the current context to bytecode interpreter handlers. This is held in the
context register on all architectures except for ia32 where there are too few
registers and it is instead spilled to the stack.
Also changes Load/StoreRegister to use kMachAnyTagged representation since they
should only ever hold tagged values.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1294133004
Cr-Commit-Position: refs/heads/master@{#30325}
Makes the following modifications to the interpreter builtins and
InterpreterAssembler:
- Adds an accumulator register and initializes it to undefined()
- Adds a register file pointer register and use it instead of FramePointer to
access registers
- Modifies builtin to support functions with 0 regiters in the register file
- Modifies builtin to Call rather than TailCall to first bytecode handler.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1289863003
Cr-Commit-Position: refs/heads/master@{#30219}
Reason for revert:
Breaks win32 nosnap
Original issue's description:
> [runtime] Remove useless IN builtin.
>
> Similar to DELETE, the IN builtin is just a thin wrapper for %HasElement
> and %HasProperty anyway, and cannot be optimized, plus it had a weird
> special fast case (which also involved at least one LOAD_IC plus some
> intrinsic magic).
>
> R=yangguo@chromium.org,jarin@chromium.org
>
> Committed: https://crrev.com/72d60a1e80e81e2e68ca402665e2acbc46c5e471
> Cr-Commit-Position: refs/heads/master@{#30154}
TBR=yangguo@chromium.org,jarin@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
Review URL: https://codereview.chromium.org/1288923002
Cr-Commit-Position: refs/heads/master@{#30155}
Similar to DELETE, the IN builtin is just a thin wrapper for %HasElement
and %HasProperty anyway, and cannot be optimized, plus it had a weird
special fast case (which also involved at least one LOAD_IC plus some
intrinsic magic).
R=yangguo@chromium.org,jarin@chromium.org
Review URL: https://codereview.chromium.org/1295433002
Cr-Commit-Position: refs/heads/master@{#30154}
TurboFan is now a requirement and supported by all backends, so we don't
need those macros (plus all the machinery on top) anymore.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1282763002
Cr-Commit-Position: refs/heads/master@{#30082}
Rationale: The {kind} of a call descriptor describes what the {target} being
called is--i.e. a JSFunction, code object, or address. That kind materially
dictates the instruction(s) generated for an outgoing call.
The other flags on a call descriptor should describe specific properties
(like whether a roots register is valid or not) so that backend logic doesn't
have to switch over the kind, but is informed directly of what it wants to
know.
R=mstarzinger@chromium.org
BUG=
Review URL: https://codereview.chromium.org/1268273003
Cr-Commit-Position: refs/heads/master@{#30065}
Add factory methods for different types of LinkageLocations, and ensure that
accesses to the underlying data in the location are classified by type and
funneled through explicit accessors.
Also change the representation of LinkageLocation to use a BitField rather
than using a reserved section of the integer range.
Review URL: https://codereview.chromium.org/1262343002
Cr-Commit-Position: refs/heads/master@{#29938}
Adds interpreter entry and exit trampoline builtins. Also implements the
Return bytecode handler and fixes a few bugs in InterpreterAssembler
highlighted by running on other architectures.
MIPS and MIPS64 port contributed by Paul Lind (paul.lind@imgtec.com)
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1245133002
Cr-Commit-Position: refs/heads/master@{#29929}
Changes the interpreter to use a BytecodeArray pointer and an offset to avoid
having an inner pointer to a BytecodeArray object in registers during dispatch.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1254293006
Cr-Commit-Position: refs/heads/master@{#29910}
Adds basic support for generation of interpreter bytecode handler code
snippets. The InterpreterAssembler class exposes a set of low level,
interpreter specific operations which can be used to build a Turbofan
graph. The Interpreter class generates a bytecode handler snippet for
each bytecode by assembling operations using an InterpreterAssembler.
Currently only two simple bytecodes are supported: LoadLiteral0 and Return.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1239793002
Cr-Commit-Position: refs/heads/master@{#29814}
Consolidated conflict detection and traversal logic in CoalescedLiveRanges to avoid duplication in both code and testing. In addition, this change achieves better separation between CoalescedLiveRanges and other register allocator components, improving testability and maintainability.
BUG=
Review URL: https://codereview.chromium.org/1219063017
Cr-Commit-Position: refs/heads/master@{#29783}
The InterpreterAssembler needs to specify a specific CallDescriptor type
instead of using the SimplifiedCDescriptor type. This CL makes it possible
to specify the CallDescriptor used by the RawMachineAssembler instead of
specifying a MachineSignature.
Also removes instruction-selector-tester.h which was erroneously resurrected
at some point.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1235313002
Cr-Commit-Position: refs/heads/master@{#29777}
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}
This CL also adds hydrogen stubs for global loads and global stores, full-codegen and TurboFan now uses this machinery.
Review URL: https://codereview.chromium.org/1224793002
Cr-Commit-Position: refs/heads/master@{#29592}
Reduce Float64 comparison to Float32 when both inputs are conversions from
Float32.
Review URL: https://codereview.chromium.org/1235663002
Cr-Commit-Position: refs/heads/master@{#29586}
Optimize string "length" property access based on static type
information if possible, but also optimistically optimize the access
based on type feedback from the LoadIC.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1216593003
Cr-Commit-Position: refs/heads/master@{#29543}
We actually need round to zero truncation to implement the counterpart
of LDoubleToI in TurboFan, which tries to convert a double to an integer
as required for keyed load/store optimizations.
Drive-by-cleanup: Reduce some code duplication in the InstructionSelector
implementations.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1225993002
Cr-Commit-Position: refs/heads/master@{#29527}
The RawMachineAssembler will be used to build the interpreter, so it needs
to move back to src/compiler.
This reverts commit b5b00cc031.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1221303014
Cr-Commit-Position: refs/heads/master@{#29519}
Currently we lower shifts directly to machine operators, and add an
appropriate Word32And to implement the & 0x1F operation on the right
hand side required by the specification. However for Word32And we assume
Int32 in simplified lowering, which is basically changes the right hand
side bit interpretation for the shifts from Uint32 to Int32, which is
obviously wrong. So now we represent that explicitly by proper
simplified operators for the shifts, which are lowered to machine in
simplified lowering.
R=jarin@chromium.org
Review URL: https://codereview.chromium.org/1213803008
Cr-Commit-Position: refs/heads/master@{#29465}
Support 32-bit cmp with shift/extend by reusing the existing add/sub shift and
extend code.
Review URL: https://codereview.chromium.org/1218103005
Cr-Commit-Position: refs/heads/master@{#29435}
Revert "Revert relanded strong property access CL"
Regression issues should be solved. Initial patchset is the original, subsequent patchsets are the fixing modifications.
This reverts commit 4ac7be5656.
BUG=v8:3956
LOG=N
Review URL: https://codereview.chromium.org/1199983002
Cr-Commit-Position: refs/heads/master@{#29384}
