Legacy const is changed so that a declaration declares a configurable, but non-writable, slot, and the initializer reconfigures it (when possible) to non-configurable non-writable. This avoids the need for "the hole" as marker value in JSContextExtensionObjects and GlobalObjects. Undefined is used instead.
BUG=
R=rossberg@chromium.org
Review URL: https://codereview.chromium.org/379893002
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@22379 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Calling OffsetOfElementAt becomes expensive when compiling functions with many
constant pool entries. This was causing a regression in MandreelLatency due
to the time spent populating the constant pool array for large compiled
functions.
This change avoids calling OffsetOfElementAt for each entry, and instead keeps
track of the current offsets in ConstantPoolBuilder::Populate. This gives the
following improvements on a Nexus 5:
Inline CP | OOL CP (before CL) | OOL CP (after CL)
Mandreel: 4305 | 3961 | 4120
MandreelLatency: 2298 | 1198 | 1994
Octane Score: 5197 | 4982 | 5152
R=ulan@chromium.org
Review URL: https://codereview.chromium.org/376973002
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@22293 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
- Adds support to the Arm assembler to use extended constant pools.
- Update (set_)target_address_at to support extended constant pool load
updates.
- Replace Operand::is_single_instruction with Operand::instructions_required
Due to the fact that different constant pool load types require different
numbers of instructions.
- Various cleanups of ConstantPoolBuilder to cleaner integration of the
extended constant pool building.
- Update GetRelocatedValue such that offsets to both map_check and bool_load
are explicitly provided, rather than location of bool_load being inferred
based on map_check, since the code inbetween is no longer of a predictable
size.
- Update MacroAssembler::GetRelocatedValueLocation() to add support for
getting a value from an extended constant pool entry.
- Update Debug::SetAfterBreakTarget() to use target_address_from_return_address
when checking for debug breaks at constant pool load points.
- Change ConstantPoolIterateBody to iterate over both heap and code pointer in
the small section before moving onto the extended section, to work around the
requirement of the serializer that pointers are iterated in-order.
- Increase old_pointer_space SizeOfFirstPage() to offset the fact that constant
pools are now in the old pointer space (rather than code).
R=ulan@chromium.org
Review URL: https://codereview.chromium.org/356393003
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@22209 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Centralize a register definition in an IC that provides:
1) symbolic names for the register (like, edx == receiver)
2) defines ordering when passed on the stack
Code that implements or uses the IC should use this definition instead of "knowing" what the registers are. Or at least have the definition to validate it's assumptions.
As a side effect of avoiding runtime static initializers (enforced by tools/check-static-initializers.sh, neat), I gave ownership of the registers array to CodeStubInterfaceDescriptor. This prompted a cleanup of that struct.
R=jkummerow@chromium.org
Review URL: https://codereview.chromium.org/352583002
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@22011 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
In theory, the Thumb version of FlushICache doesn't have to be any
different from the ARM version: There is no need to switch to ARM mode
for the svc, and r7 can be used just like any other register.
In practice, GCC uses r7 as a frame pointer in Thumb code, and it
refuses to allow it be used in inline assembly. To avoid complicated
checks to determine whether or not it is possible to use it directly,
the FlushICache function explicitly preserves r7.
R=ulan@chromium.org
Review URL: https://codereview.chromium.org/335133002
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21949 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
1) symbolic names for the register (like, edx == receiver)
2) can return an array of registers
3) defines ordering when passed on the stack
Code that implements or uses the IC should use this RegisterSpec instead of "knowing" what the registers are. Or at least have the RegisterSpec to validate it's assumptions.
R=jkummerow@chromium.org
Review URL: https://codereview.chromium.org/340363002
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21939 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Simplify the compile time feature detection on Arm:
- Define CAN_USE_XXX definitions unconditionally for all target/host
toolchain combinations
- Rename arm_test / ARM_TEST to arm_test_noprob / ARM_TEST_NO_FEATURE_PROBE
- Don't set ARM_TEST_NO_FEATURE_PROBE implicitly on the simulator to make
make simulator / native more consistent
- Unify CpuFeatures::PrintTarget for simulator and native builds
- Remove unecessary CAN_USE_VFP_INSTRUCTIONS definition for android (this is
the default for arm_fpu=default)
- Add a CpuFeatures::Probe() before calling CpuFeatures::PrintFeatures() in PrintHelp
to ensure we have probed features before printing them.
BUG=384474
LOG=N
R=jkummerow@chromium.org
Review URL: https://codereview.chromium.org/340373002
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21935 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
A couple of cleanups to the Arm backend to enable support of extended
OOL constant pools in a following CL.
- Remove instruction pattern extern const's and replace their use with IsXXX()
functions.
- Do calculation of the target address of a load from constant pool in one
place.
- A couple of other small cleanups.
R=ulan@chromium.org
Review URL: https://codereview.chromium.org/317653003
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21932 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Avoid generating duplicate vmla instructions for vmul/vadd sequences where the
vmul has more than one use.
For example: function f(a, b, c) { return (a * b) + c + (a * b); }
Previously, this would produce a vmul for the subexpression (a * b), then vmla
for (a * b) + c, then vmla for (a * b) + [(a * b) + c].
Now it produces vmul, vadd, vadd, as expected.
BUG=
R=ulan@chromium.org
Review URL: https://codereview.chromium.org/323423003
git-svn-id: https://v8.googlecode.com/svn/branches/bleeding_edge@21899 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
