The old code used a separate HToInt32 instruction which had a wrong register
constraint for the input register which caused wrong result when the stored value
is used after a typed array store. (UseRegister instead of UseTempRegister) when no
SSE3 is available.
This change fixes it by replacing HToInt32 with the corresponding HChange
instruction which has correct register contraints.
TEST=mjsunit/compiler/regress-toint32.js
Review URL: https://chromiumcodereview.appspot.com/9565007
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10891 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Port r10881 (0d25c61e).
Original commit message:
Generates inlined code for object allocation specific to the initial map
of the given constructor function. Also forces completion of inobject
slack tracking while crankshafting to finalize instance size of these
objects.
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/9569008
Patch from Daniel Kalmar <kalmard@homejinni.com>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10890 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Port r10794 (654fe910).
Original commit message:
Only JSObject enumerables with enum cache (fast case properties, no interceptors, no enumerable properties on the prototype) are supported.
HLoadKeyedGeneric with keys produced by for-in enumeration are recognized and rewritten into direct property load by index. For this enum-cache was extended to store property indices in a separate array (see handles.cc).
New hydrogen instructions:
- HForInPrepareMap: checks for-in fast case preconditions and returns map that contains enum-cache;
- HForInCacheArray: extracts enum-cache array from the map;
- HCheckMapValue: map check with HValue map instead of immediate;
- HLoadFieldByIndex: load fast property by it's index, positive indexes denote in-object properties, negative - out of object properties;
Changed hydrogen instructions:
- HLoadKeyedFastElement: added hole check suppression for loads from internal FixedArrays that are knows to have no holes inside.
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/9453009
Patch from Daniel Kalmar <kalmard@homejinni.com>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10865 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Port r10849 (b0fe79c).
Also included: Fixed a bug in GenerateRecordCallTarget.
This bug prevented certain functions from being registered as Monomorphic
and thus prevented them from being inlined using the new system (b0fe79c).
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/9511002
Patch from Daniel Kalmar <kalmard@homejinni.com>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10862 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Port r10700 (c976fbd5).
Original commit message:
This change enables optimization of top-level and eval-code. For this to work, it adds
support for declaring global variables in optimized code.
At the same time it disables the eager generation of deoptimization support data
in the full code generator (originally introduced in
r10040). This speeds up initial compilation and saves
memory for functions that won't be optimized. It requires
recompiling the function with deoptimization
support when we decide to optimize it.
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/9418006
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10773 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This extends the current support for nested object literals we already
have in Crankshaft, to also support nested array literals and mixed
nested literals containing arrays and objects. All three types are
generated by the unified HFastLiteral instruction.
All previous upper bounds on nested literal graphs remain unchanged,
keeping the size of generated code in check.
The main intention is to boost performance of two-dimensional array
literals containing constant elements (aka. matrices).
R=danno@chromium.org
TEST=mjsunit/compiler/literals-optimized
Review URL: https://chromiumcodereview.appspot.com/9403018
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10734 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
This method works around the Branch offset and relocinfo issues by emulating a pc-relative jump.
This allows us to generate larger entry tables. The theoretical limit is 2^16 (number of entries)
but even that can be extended by allowing a larger instruction count.
Also reverted the mips-specific constant (kNumberOfEntries) in deoptimizer.h
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/9347016
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10717 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Port r10673 (18d3af5).
Original commit message:
Supports growing non-COW JSArray by a single element if the backing store has room, and initial allocation of a backing store for the store to index zero of an empty array to kPreallocatedArrayElements elements (e.g. the [] array literal).
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/9378005
Patch from Daniel Kalmar <kalmard@homejinni.com>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10707 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Some instructions can use >16 bit immediates if they represent a <=16 bit signed value.
However some logical instructions (andi, xori, ori, lui) should always treat the immediate value as unsigned.
This patch adds an ASSERT to these places and a minor change to MacroAssembler::li to satisfy this.
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/9309077
Patch from Daniel Kalmar <kalmard@homejinni.com>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10644 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
1. Instead of checking upfront and estimating a limit for the number, we
now are able to stop register allocation and bailout when we don't
have enough virtual registers.
2. GCed some out-dated flags from flag-definition.h
3. Simplified the interface from the Lithium builder to the
register allocator in lithium-*.cc: For uses and definitions, we
just record the virtual register number given by the Hydrogen value id.
For temporaries, we request a new virtual register from the allocator.
For fixed temps, we don't need to do anything.
4. Increased number of deoptimization entries to 16K. Eventually we
probably want to make this array grow dynamically.
Review URL: https://chromiumcodereview.appspot.com/9325019
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10597 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Port r10538 (21c5dc1).
MIPS-specific changes:
-moved MacroAssembler::SmiTagCheckOverflow definition to macro-assembler-mips.cc from the header
-added optimized 3-argument version of said function
-removed the related, completely unused TrySmiTag function
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/9296045
Patch from Daniel Kalmar <kalmard@homejinni.com>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10567 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Port r10531 (d61db240).
Original commit message:
This caches call targets of constructor calls by associating one element
caches with call sites. The type feedback oracle can use the recorded
valued to gather type information for monomorphic constructor call sites.
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/9298014
Patch from Daniel Kalmar <kalmard@homejinni.com>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10546 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Yak shaving for map sharing with accessor properties contd.: When CALLBACKS can
have map transitions, simply looking at the property type is not sufficient
anymore to decide if a property is there or not. One has to look at the actual
contents of the descriptor entry then, but this breaks down sometimes when the
lookup is being done with a NULL holder. Luckily enough, we can oftren replace
IsProperty by the simpler IsFound, because we inspect the type immediately
afterwards, anyway.
Review URL: https://chromiumcodereview.appspot.com/9280007
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10474 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
Port r10356 (69da81).
Original commit message:
Loosen the requirement for Map equivalency on several map checks, including checks up the prototype chain, that are not sensitive to ElementsKinds. These selected map checks should also match against FAST_DOUBLE_ELEMENT and FAST_ELEMENT transitions of the original map. This specifically helps all variants of transitioned JSArrays to still efficiently call builtins like push, pop and sort.
BUG=
TEST=
Review URL: https://chromiumcodereview.appspot.com/9265007
Patch from Daniel Kalmar <kalmard@homejinni.com>.
git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10461 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
