This addresses an issue where the at-start splitting used
in the splintering mechanism was in conflict with the mechanics
used in linear allocator, in particular in the initial split/spill of
ranges for memory operands. We are already doing a split-at-start in
Greedy, so this change centralizes that to the base RegisterAllocator.
Verified locally that v8:4508 is addressed by this. Also, this fixes
the failures that required the revert
5308a999eb. See trybots at
issue 1425533002.
R=bmeurer@chromium.org,jarin@chromium.org
BUG= v8:4508
LOG=n
Review URL: https://codereview.chromium.org/1426583002
Cr-Commit-Position: refs/heads/master@{#31544}
Reason for revert:
because of merge mistake, "regress/" is missed when skipping one test case for X87.
"regress/" will be added when relanding it.
Original issue's description:
> X87: disable the regress-undefined-nan test case for x87.
>
> The reason is when native_context_specialization flag is ture, X87 turbofan
> will hit the known issue that X87 will change a sNaN to qNaN by default. And
> then it will fail when bit-comparing the source (sNaN) and the result (qNaN).
>
> BUG=
>
> Committed: https://crrev.com/b3c719ebbad6c87afefa33a7d0b3f412b2e304db
> Cr-Commit-Position: refs/heads/master@{#31530}
TBR=bmeurer@chromium.org,weiliang.lin@intel.com
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=
Review URL: https://codereview.chromium.org/1417303005
Cr-Commit-Position: refs/heads/master@{#31543}
port e678a0f9a9 (r31358)
original commit message:
Use %_ToLength for TO_LENGTH, implemented via a ToLengthStub
that supports a fast path for small integers. Everything else is still
handled in the runtime.
BUG=
Review URL: https://codereview.chromium.org/1421803002
Cr-Commit-Position: refs/heads/master@{#31542}
For some reason, the DisableCrankshaft() in ast-numbering.cc does not always
prevent crankshaft from happening. Bailout here rather than asserting an
unreachable condition.
BUG=546967, v8:4488
LOG=N
R=bmeurer@chromium.org
Review URL: https://codereview.chromium.org/1414713004
Cr-Commit-Position: refs/heads/master@{#31537}
Reason for revert:
[sheriff] breaks benchmarks:
http://build.chromium.org/p/client.v8/builders/V8%20Linux%20-%20debug/builds/4998
Original issue's description:
> [Turbofan] Fix perf regression introduced by per-range change.
>
> When the range ends just at the gap of a non-deferred block, the last
> instruction the range covers is in the predecessor. If that predecessor is
> a deferred block, before this CL, we would splinter the remainder of the
> range all the way to the end. That leads to inefficient codegen, because
> we still want a split inside the deferred block.
>
> Also, opportunistically added a trace before we splinter, for better
> diagnostics.
>
> BUG= chromium:546416
> LOG=N
>
> Committed: https://crrev.com/32b6e085e74a8fcf94a01d20740fe4fdede07a86
> Cr-Commit-Position: refs/heads/master@{#31529}
TBR=bmeurer@chromium.org,jarin@chromium.org,mtrofin@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG= chromium:546416
Review URL: https://codereview.chromium.org/1412893007
Cr-Commit-Position: refs/heads/master@{#31531}
The reason is when native_context_specialization flag is ture, X87 turbofan
will hit the known issue that X87 will change a sNaN to qNaN by default. And
then it will fail when bit-comparing the source (sNaN) and the result (qNaN).
BUG=
Review URL: https://codereview.chromium.org/1414733004
Cr-Commit-Position: refs/heads/master@{#31530}
When the range ends just at the gap of a non-deferred block, the last
instruction the range covers is in the predecessor. If that predecessor is
a deferred block, before this CL, we would splinter the remainder of the
range all the way to the end. That leads to inefficient codegen, because
we still want a split inside the deferred block.
Also, opportunistically added a trace before we splinter, for better
diagnostics.
BUG= chromium:546416
LOG=N
Review URL: https://codereview.chromium.org/1412123009
Cr-Commit-Position: refs/heads/master@{#31529}
Fix aborting compaction for pages by doing two separate passes, one that scans
all live objects, and one that later on sweeps the page.
Detailed description see bug report.
BUG=chromium:539356,chromium:524425
LOG=N
Review URL: https://codereview.chromium.org/1413763011
Cr-Commit-Position: refs/heads/master@{#31526}
This patch switches sloppy-mode code from legacy const semantics
to ES2015 semantics. It is unknown how much of the web will be
broken by this; likely the patch will have to be reverted before
a branch happens.
BUG=v8:3739
LOG=Y
R=rossberg,adamk
Review URL: https://codereview.chromium.org/1420223003
Cr-Commit-Position: refs/heads/master@{#31525}
We don't need to have an (untested) fallback for the case that a
prototype map is not stable when specializing a named property,
because prototype maps are always stable (which is now guarded
by a DCHECK in CompilationDependencies). Less (dead) code is
better code.
R=verwaest@chromium.org
BUG=v8:4470
LOG=n
Review URL: https://codereview.chromium.org/1417973006
Cr-Commit-Position: refs/heads/master@{#31524}
port 7557dc5a70 (r31312).
original commit message:
This change add a new bytecode for operator new and implements it using
the Construct() builtin.
BUG=
Review URL: https://codereview.chromium.org/1423733002
Cr-Commit-Position: refs/heads/master@{#31518}
This removes the requirement for handles as arguments, but also removes
concurrency support, which is not being used at the moment.
Supporting concurrency could be done by introducing a sibling class to
IdentityMap that includes RelocationLock on method calls.
R=bmeurer@chromium.org, ulan@chromium.org
Review URL: https://codereview.chromium.org/1419563004
Cr-Commit-Position: refs/heads/master@{#31510}
Fix lookup for storing to properties, and also make sure we don't embed
deprecated map (using Map::TryUpdate).
R=jarin@chromium.org
BUG=v8:4470
LOG=n
Review URL: https://codereview.chromium.org/1424523002
Cr-Commit-Position: refs/heads/master@{#31509}
port f2f46aff8b (r31242).
original commit message:
The lack of a vector slot for the keyed store operation in filling in
non-constant array literal properties led to undesirable contortions in
compilers downwind of full-codegen. The use of a single slot to initialize all
the array elements is sufficient.
BUG=
Review URL: https://codereview.chromium.org/1422443004
Cr-Commit-Position: refs/heads/master@{#31505}
port 2d4aeaad2f (r31204).
original commit message:
The stack manipulation was expensive. Two virtual registers are better.
BUG=
Review URL: https://codereview.chromium.org/1410573003
Cr-Commit-Position: refs/heads/master@{#31504}
port 9c8262f11e (r31120).
original commit message:
When calling into C++ builtins, we need to make sure that the argument
count register contains the correct number of arguments, otherwise the
CEntryStub will not be able to leave the stack in the correct state.
BUG=
Review URL: https://codereview.chromium.org/1418533009
Cr-Commit-Position: refs/heads/master@{#31503}
port 75f6ad74b2 (r31089).
original commit message:
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/1418213002
Cr-Commit-Position: refs/heads/master@{#31502}
This is in preparation of implementing Reflect.set.
R=rossberg
BUG=
Review URL: https://codereview.chromium.org/1394983005
Cr-Commit-Position: refs/heads/master@{#31501}
