Make Ignition collect BinaryOperationFeedback on ToNumber, using the
shared type feedback slot with the following Inc/Dec bytecode, and use
this feedback in TurboFan to turn the ToNumber(x) operation into a
SpeculativeNumberMultiply(x,1) with the feedback hint.
R=jarin@chromium.org, mstarzinger@chromium.org, rmcilroy@chromium.org
BUG=v8:6214,v8:5267
Review-Url: https://codereview.chromium.org/2804813003
Cr-Commit-Position: refs/heads/master@{#44440}
They have the same lifetime. It's a match!
Both structures are native context dependent and dealt with (creation,
clearing, gathering feedback) at the same time. By treating the spaces used
for literal boilerplates as feedback vector slots, we no longer have to keep
track of the materialized literal count elsewhere.
A follow-on CL removes even more parser infrastructure related to this count.
BUG=v8:5456
Review-Url: https://codereview.chromium.org/2655853010
Cr-Commit-Position: refs/heads/master@{#42771}
This changes the NewClosure interface descriptor, but ignores
the additional vector/slot arguments for now. The feedback vector
gets larger, as it holds a space for each literal array. A follow-on
CL will constructively use this space.
BUG=v8:5456
Review-Url: https://codereview.chromium.org/2614373002
Cr-Commit-Position: refs/heads/master@{#42146}
Reason for revert:
Speculative revert because of blocked roll: https://codereview.chromium.org/2596013002/
Original issue's description:
> [TypeFeedbackVector] Root literal arrays in function literals slots
>
> Literal arrays and feedback vectors for a function can be garbage
> collected if we don't have a rooted closure for the function, which
> happens often. It's expensive to come back from this (recreating
> boilerplates and gathering feedback again), and the cost is
> disproportionate if the function was inlined into optimized code.
>
> To guard against losing these arrays when we need them, we'll now
> create literal arrays when creating the feedback vector for the outer
> closure, and root them strongly in that vector.
>
> BUG=v8:5456
>
> Review-Url: https://codereview.chromium.org/2504153002
> Cr-Commit-Position: refs/heads/master@{#41893}
> Committed: 93df094081TBR=bmeurer@chromium.org,mlippautz@chromium.org,mvstanton@chromium.org
# Skipping CQ checks because original CL landed less than 1 days ago.
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=v8:5456
Review-Url: https://codereview.chromium.org/2597163002
Cr-Commit-Position: refs/heads/master@{#41917}
Literal arrays and feedback vectors for a function can be garbage
collected if we don't have a rooted closure for the function, which
happens often. It's expensive to come back from this (recreating
boilerplates and gathering feedback again), and the cost is
disproportionate if the function was inlined into optimized code.
To guard against losing these arrays when we need them, we'll now
create literal arrays when creating the feedback vector for the outer
closure, and root them strongly in that vector.
BUG=v8:5456
Review-Url: https://codereview.chromium.org/2504153002
Cr-Commit-Position: refs/heads/master@{#41893}
This introduces an explicit struct for the communication channel between
the {ArrayLiteral} AST node and the corresponding runtime methods. Those
methods take a pair of {ElementsKind} as well as an array (can either be
a FixedArray or a FixedDoubleArray) of constant values.
For bonus points it also reduces the size of the involved heap object by
one word (i.e. length field of FixedArray not needed anymore).
R=mvstanton@chromium.org
Review-Url: https://codereview.chromium.org/2581683003
Cr-Commit-Position: refs/heads/master@{#41752}
The majority of context slot accesses are to the local context (current context
register and depth 0), so this adds bytecodes to optimise for that case.
This cuts down bytecode size by roughly 1% (measured on Octane and Top25).
Review-Url: https://codereview.chromium.org/2459513002
Cr-Commit-Position: refs/heads/master@{#40641}
The CreateArrayLiteral bytecode handler now directly inlines the FastCloneShallowArrayStub.
BUG=v8:4280
Review-Url: https://codereview.chromium.org/2341743003
Cr-Commit-Position: refs/heads/master@{#39562}
Add a notion of "invocation count" to the baseline compilers, which
increment a special slot in the TypeFeedbackVector for each invocation
of a given function (the optimized code doesn't currently collect this
information).
Use this invocation count to relativize the call counts on the call
sites within the function, so that the inlining heuristic has a view
of relative importance of a call site rather than some absolute numbers
with unclear meaning for the current function. Also apply the call site
frequency as a factor to all frequencies in the inlinee by passing this
to the graph builders so that the importance of a call site in an
inlinee is relative to the topmost optimized function.
Note that all functions that neither have literals nor need type
feedback slots will share a single invocation count cell in the
canonical empty type feedback vector, so their invocation count is
meaningless, but that doesn't matter since we only use the invocation
count to relativize call counts within the function, which we only have
if we have at least one type feedback vector (the CallIC slot).
See the design document for additional details on this change:
https://docs.google.com/document/d/1VoYBhpDhJC4VlqMXCKvae-8IGuheBGxy32EOgC2LnT8
BUG=v8:5267,v8:5372
R=mvstanton@chromium.org,rmcilroy@chromium.org,mstarzinger@chromium.org
Review-Url: https://codereview.chromium.org/2337123003
Cr-Commit-Position: refs/heads/master@{#39410}
Moves the context chain search loop out of generated bytecode, and into
the (Lda|Ldr|Sda)ContextSlot handler, by passing the context depth in as
an additional operand. This should decrease the bytecode size and
increase performance for deep context chain searches, at the cost of
slightly increasing bytecode size for shallow context access.
Review-Url: https://codereview.chromium.org/2336643002
Cr-Commit-Position: refs/heads/master@{#39378}
In ignition, allocation site mementos were disabled when creating array
literals. Enabled them in this cl.
BUG=v8:4280
LOG=N
Review-Url: https://codereview.chromium.org/2294913006
Cr-Commit-Position: refs/heads/master@{#39234}
For historical reasons, the interpreter's bytecode expectations tests
required a type for the constant pool. This had two disadvantages:
1. Strings and numbers were not visible in mixed pools, and
2. Mismatches of pool types (e.g. when rebaselining) would cause parser
errors
This removes the pool types, making everything 'mixed', but appending
the values to string and number valued constants. Specifying a pool type
in the *.golden header now prints a warning (for backwards compatibility).
BUG=v8:5350
Review-Url: https://codereview.chromium.org/2310103002
Cr-Commit-Position: refs/heads/master@{#39216}
Avoids the always generated Star bytecodes after ObjectLiteral.
BUG=v4:4820
LOG=n
Review-Url: https://codereview.chromium.org/2216023003
Cr-Commit-Position: refs/heads/master@{#38480}
Assign feedback slots in the type feedback vector for binary operations.
Update bytecode-generator to use these slots and add them as an operand
to binary operations.
BUG=v8:4280
LOG=N
Review-Url: https://codereview.chromium.org/2209633002
Cr-Commit-Position: refs/heads/master@{#38408}
Add a new bytecode to create a function context. The handler inlines
FastNewFunctionContextStub.
BUG=v8:4280
LOG=n
Review-Url: https://codereview.chromium.org/2187523002
Cr-Commit-Position: refs/heads/master@{#38301}
ToNumber's result is always directly stored to a register using a Star
bytecode. Fuse it into ToNumber.
BUG=v8:4280
LOG=n
Review-Url: https://codereview.chromium.org/2165953002
Cr-Commit-Position: refs/heads/master@{#37976}
With this change the bytecode array builder only emits expression
positions for bytecodes that can throw. This allows more peephole
optimization opportunities and results in smaller code.
BUG=v8:4280,chromium:615979
LOG=N
Review-Url: https://codereview.chromium.org/2038323002
Cr-Commit-Position: refs/heads/master@{#36863}
The original peephole optimizer logic in the BytecodeArrayBuilder did
not respect source positions as it was written before there were
bytecode source positions. This led to some minor differences to
FCG and was problematic when combined with pending bytecode
optimizations. This change makes the new peephole optimizer fully
respect source positions.
BUG=v8:4280
LOG=N
Review-Url: https://codereview.chromium.org/1998203002
Cr-Commit-Position: refs/heads/master@{#36439}
Prints source position information alongside bytecode.
BUG=v8:4280
LOG=N
Review-Url: https://codereview.chromium.org/1963663002
Cr-Commit-Position: refs/heads/master@{#36171}
Adapts FastCloneShallowObjectStub to enable it to be used by the
CreateObjectLiteral bytecode.
BUG=v8:4280
LOG=N
Review-Url: https://codereview.chromium.org/1922523002
Cr-Commit-Position: refs/heads/master@{#35909}
Adds IncStub and DecStub TurboFan code stubs and hooks them up to the
interpreter's Inc and Dec bytecodes (which are used for count
operations, e.g. i++).
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1901083002
Cr-Commit-Position: refs/heads/master@{#35720}
This change introduces wide prefix bytecodes to support wide (16-bit)
and extra-wide (32-bit) operands. It retires the previous
wide-bytecodes and reduces the number of operand types.
Operands are now either scalable or fixed size. Scalable operands
increase in width when a bytecode is prefixed with wide or extra-wide.
The bytecode handler table is extended to 256*3 entries. The
first 256 entries are used for bytecodes with 8-bit operands,
the second 256 entries are used for bytecodes with operands that
scale to 16-bits, and the third group of 256 entries are used for
bytecodes with operands that scale to 32-bits.
LOG=N
BUG=v8:4747,v8:4280
Review URL: https://codereview.chromium.org/1783483002
Cr-Commit-Position: refs/heads/master@{#34955}
This mechanism was used to ensure that functions ended up as constants on the map of prototypes defined using object literals, e.g.,:
function.prototype = {
method: function() { ... }
}
Nowadays we treat prototypes specially, and make all their functions constants when an object turns prototype. Hence this special custom code isn't necessary anymore.
This also affects boilerplates that do not become prototypes. Their functions will not be constants but fields instead. Calling their methods will slow down. However, multiple instances of the same boilerplate will stay monomorphic. We'll have to see what the impact is for such objects, but preliminary benchmarks do not show this as an important regression.
BUG=chromium:593008
LOG=n
Review URL: https://codereview.chromium.org/1772423002
Cr-Commit-Position: refs/heads/master@{#34602}
Bytecode expectations have been moved to external (.golden) files,
one per test. Each test in the suite builds a representation of the
the compiled bytecode using BytecodeExpectationsPrinter. The output is
then compared to the golden file. If the comparision fails, a textual
diff can be used to identify the discrepancies.
Only the test snippets are left in the cc file, which also allows to
make it more compact and meaningful. Leaving the snippets in the cc
file was a deliberate choice to allow keeping the "truth" about the
tests in the cc file, which will rarely change, as opposed to golden
files.
Golden files can be generated and kept up to date using
generate-bytecode-expectations, which also means that the test suite
can be batch updated whenever the bytecode or golden format changes.
The golden format has been slightly amended (no more comments about
`void*`, add size of the bytecode array) following the consideration
made while converting the tests.
There is also a fix: BytecodeExpectationsPrinter::top_level_ was left
uninitialized, leading to undefined behaviour.
BUG=v8:4280
LOG=N
Review URL: https://codereview.chromium.org/1717293002
Cr-Commit-Position: refs/heads/master@{#34285}
