v8/src/arm/ic-arm.cc

2265 lines
78 KiB
C++
Raw Normal View History

// Copyright 2006-2008 the V8 project authors. All rights reserved.
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
// * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
// * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following
// disclaimer in the documentation and/or other materials provided
// with the distribution.
// * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived
// from this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#include "v8.h"
#if defined(V8_TARGET_ARCH_ARM)
#include "assembler-arm.h"
#include "code-stubs.h"
#include "codegen-inl.h"
#include "disasm.h"
#include "ic-inl.h"
#include "runtime.h"
#include "stub-cache.h"
namespace v8 {
namespace internal {
// ----------------------------------------------------------------------------
// Static IC stub generators.
//
#define __ ACCESS_MASM(masm)
static void GenerateGlobalInstanceTypeCheck(MacroAssembler* masm,
Register type,
Label* global_object) {
// Register usage:
// type: holds the receiver instance type on entry.
__ cmp(type, Operand(JS_GLOBAL_OBJECT_TYPE));
__ b(eq, global_object);
__ cmp(type, Operand(JS_BUILTINS_OBJECT_TYPE));
__ b(eq, global_object);
__ cmp(type, Operand(JS_GLOBAL_PROXY_TYPE));
__ b(eq, global_object);
}
// Generated code falls through if the receiver is a regular non-global
// JS object with slow properties and no interceptors.
static void GenerateStringDictionaryReceiverCheck(MacroAssembler* masm,
Register receiver,
Register elements,
Register t0,
Register t1,
Label* miss) {
// Register usage:
// receiver: holds the receiver on entry and is unchanged.
// elements: holds the property dictionary on fall through.
// Scratch registers:
// t0: used to holds the receiver map.
// t1: used to holds the receiver instance type, receiver bit mask and
// elements map.
// Check that the receiver isn't a smi.
__ tst(receiver, Operand(kSmiTagMask));
__ b(eq, miss);
// Check that the receiver is a valid JS object.
__ CompareObjectType(receiver, t0, t1, FIRST_JS_OBJECT_TYPE);
__ b(lt, miss);
// If this assert fails, we have to check upper bound too.
ASSERT(LAST_TYPE == JS_FUNCTION_TYPE);
GenerateGlobalInstanceTypeCheck(masm, t1, miss);
// Check that the global object does not require access checks.
__ ldrb(t1, FieldMemOperand(t0, Map::kBitFieldOffset));
__ tst(t1, Operand((1 << Map::kIsAccessCheckNeeded) |
(1 << Map::kHasNamedInterceptor)));
__ b(nz, miss);
__ ldr(elements, FieldMemOperand(receiver, JSObject::kPropertiesOffset));
__ ldr(t1, FieldMemOperand(elements, HeapObject::kMapOffset));
__ LoadRoot(ip, Heap::kHashTableMapRootIndex);
__ cmp(t1, ip);
__ b(nz, miss);
}
// Probe the string dictionary in the |elements| register. Jump to the
// |done| label if a property with the given name is found. Jump to
// the |miss| label otherwise.
static void GenerateStringDictionaryProbes(MacroAssembler* masm,
Label* miss,
Label* done,
Register elements,
Register name,
Register scratch1,
Register scratch2) {
// Compute the capacity mask.
const int kCapacityOffset = StringDictionary::kHeaderSize +
StringDictionary::kCapacityIndex * kPointerSize;
__ ldr(scratch1, FieldMemOperand(elements, kCapacityOffset));
__ mov(scratch1, Operand(scratch1, ASR, kSmiTagSize)); // convert smi to int
__ sub(scratch1, scratch1, Operand(1));
const int kElementsStartOffset = StringDictionary::kHeaderSize +
StringDictionary::kElementsStartIndex * kPointerSize;
// Generate an unrolled loop that performs a few probes before
// giving up. Measurements done on Gmail indicate that 2 probes
// cover ~93% of loads from dictionaries.
static const int kProbes = 4;
for (int i = 0; i < kProbes; i++) {
// Compute the masked index: (hash + i + i * i) & mask.
__ ldr(scratch2, FieldMemOperand(name, String::kHashFieldOffset));
if (i > 0) {
// Add the probe offset (i + i * i) left shifted to avoid right shifting
// the hash in a separate instruction. The value hash + i + i * i is right
// shifted in the following and instruction.
ASSERT(StringDictionary::GetProbeOffset(i) <
1 << (32 - String::kHashFieldOffset));
__ add(scratch2, scratch2, Operand(
StringDictionary::GetProbeOffset(i) << String::kHashShift));
}
__ and_(scratch2, scratch1, Operand(scratch2, LSR, String::kHashShift));
// Scale the index by multiplying by the element size.
ASSERT(StringDictionary::kEntrySize == 3);
// scratch2 = scratch2 * 3.
__ add(scratch2, scratch2, Operand(scratch2, LSL, 1));
// Check if the key is identical to the name.
__ add(scratch2, elements, Operand(scratch2, LSL, 2));
__ ldr(ip, FieldMemOperand(scratch2, kElementsStartOffset));
__ cmp(name, Operand(ip));
if (i != kProbes - 1) {
__ b(eq, done);
} else {
__ b(ne, miss);
}
}
}
// Helper function used from LoadIC/CallIC GenerateNormal.
//
// elements: Property dictionary. It is not clobbered if a jump to the miss
// label is done.
// name: Property name. It is not clobbered if a jump to the miss label is
// done
// result: Register for the result. It is only updated if a jump to the miss
// label is not done. Can be the same as elements or name clobbering
// one of these in the case of not jumping to the miss label.
// The two scratch registers need to be different from elements, name and
// result.
// The generated code assumes that the receiver has slow properties,
// is not a global object and does not have interceptors.
static void GenerateDictionaryLoad(MacroAssembler* masm,
Label* miss,
Register elements,
Register name,
Register result,
Register scratch1,
Register scratch2) {
// Main use of the scratch registers.
// scratch1: Used as temporary and to hold the capacity of the property
// dictionary.
// scratch2: Used as temporary.
Label done;
// Probe the dictionary.
GenerateStringDictionaryProbes(masm,
miss,
&done,
elements,
name,
scratch1,
scratch2);
// If probing finds an entry check that the value is a normal
// property.
__ bind(&done); // scratch2 == elements + 4 * index
const int kElementsStartOffset = StringDictionary::kHeaderSize +
StringDictionary::kElementsStartIndex * kPointerSize;
const int kDetailsOffset = kElementsStartOffset + 2 * kPointerSize;
__ ldr(scratch1, FieldMemOperand(scratch2, kDetailsOffset));
__ tst(scratch1, Operand(PropertyDetails::TypeField::mask() << kSmiTagSize));
__ b(ne, miss);
// Get the value at the masked, scaled index and return.
__ ldr(result,
FieldMemOperand(scratch2, kElementsStartOffset + 1 * kPointerSize));
}
// Helper function used from StoreIC::GenerateNormal.
//
// elements: Property dictionary. It is not clobbered if a jump to the miss
// label is done.
// name: Property name. It is not clobbered if a jump to the miss label is
// done
// value: The value to store.
// The two scratch registers need to be different from elements, name and
// result.
// The generated code assumes that the receiver has slow properties,
// is not a global object and does not have interceptors.
static void GenerateDictionaryStore(MacroAssembler* masm,
Label* miss,
Register elements,
Register name,
Register value,
Register scratch1,
Register scratch2) {
// Main use of the scratch registers.
// scratch1: Used as temporary and to hold the capacity of the property
// dictionary.
// scratch2: Used as temporary.
Label done;
// Probe the dictionary.
GenerateStringDictionaryProbes(masm,
miss,
&done,
elements,
name,
scratch1,
scratch2);
// If probing finds an entry in the dictionary check that the value
// is a normal property that is not read only.
__ bind(&done); // scratch2 == elements + 4 * index
const int kElementsStartOffset = StringDictionary::kHeaderSize +
StringDictionary::kElementsStartIndex * kPointerSize;
const int kDetailsOffset = kElementsStartOffset + 2 * kPointerSize;
const int kTypeAndReadOnlyMask
= (PropertyDetails::TypeField::mask() |
PropertyDetails::AttributesField::encode(READ_ONLY)) << kSmiTagSize;
__ ldr(scratch1, FieldMemOperand(scratch2, kDetailsOffset));
__ tst(scratch1, Operand(kTypeAndReadOnlyMask));
__ b(ne, miss);
// Store the value at the masked, scaled index and return.
const int kValueOffset = kElementsStartOffset + kPointerSize;
__ add(scratch2, scratch2, Operand(kValueOffset - kHeapObjectTag));
__ str(value, MemOperand(scratch2));
// Update the write barrier. Make sure not to clobber the value.
__ mov(scratch1, value);
__ RecordWrite(elements, scratch2, scratch1);
}
static void GenerateNumberDictionaryLoad(MacroAssembler* masm,
Label* miss,
Register elements,
Register key,
Register result,
Register t0,
Register t1,
Register t2) {
// Register use:
//
// elements - holds the slow-case elements of the receiver on entry.
// Unchanged unless 'result' is the same register.
//
// key - holds the smi key on entry.
// Unchanged unless 'result' is the same register.
//
// result - holds the result on exit if the load succeeded.
// Allowed to be the same as 'key' or 'result'.
// Unchanged on bailout so 'key' or 'result' can be used
// in further computation.
//
// Scratch registers:
//
// t0 - holds the untagged key on entry and holds the hash once computed.
//
// t1 - used to hold the capacity mask of the dictionary
//
// t2 - used for the index into the dictionary.
Label done;
// Compute the hash code from the untagged key. This must be kept in sync
// with ComputeIntegerHash in utils.h.
//
// hash = ~hash + (hash << 15);
__ mvn(t1, Operand(t0));
__ add(t0, t1, Operand(t0, LSL, 15));
// hash = hash ^ (hash >> 12);
__ eor(t0, t0, Operand(t0, LSR, 12));
// hash = hash + (hash << 2);
__ add(t0, t0, Operand(t0, LSL, 2));
// hash = hash ^ (hash >> 4);
__ eor(t0, t0, Operand(t0, LSR, 4));
// hash = hash * 2057;
__ mov(t1, Operand(2057));
__ mul(t0, t0, t1);
// hash = hash ^ (hash >> 16);
__ eor(t0, t0, Operand(t0, LSR, 16));
// Compute the capacity mask.
__ ldr(t1, FieldMemOperand(elements, NumberDictionary::kCapacityOffset));
__ mov(t1, Operand(t1, ASR, kSmiTagSize)); // convert smi to int
__ sub(t1, t1, Operand(1));
// Generate an unrolled loop that performs a few probes before giving up.
static const int kProbes = 4;
for (int i = 0; i < kProbes; i++) {
// Use t2 for index calculations and keep the hash intact in t0.
__ mov(t2, t0);
// Compute the masked index: (hash + i + i * i) & mask.
if (i > 0) {
__ add(t2, t2, Operand(NumberDictionary::GetProbeOffset(i)));
}
__ and_(t2, t2, Operand(t1));
// Scale the index by multiplying by the element size.
ASSERT(NumberDictionary::kEntrySize == 3);
__ add(t2, t2, Operand(t2, LSL, 1)); // t2 = t2 * 3
// Check if the key is identical to the name.
__ add(t2, elements, Operand(t2, LSL, kPointerSizeLog2));
__ ldr(ip, FieldMemOperand(t2, NumberDictionary::kElementsStartOffset));
__ cmp(key, Operand(ip));
if (i != kProbes - 1) {
__ b(eq, &done);
} else {
__ b(ne, miss);
}
}
__ bind(&done);
// Check that the value is a normal property.
// t2: elements + (index * kPointerSize)
const int kDetailsOffset =
NumberDictionary::kElementsStartOffset + 2 * kPointerSize;
__ ldr(t1, FieldMemOperand(t2, kDetailsOffset));
__ tst(t1, Operand(Smi::FromInt(PropertyDetails::TypeField::mask())));
__ b(ne, miss);
// Get the value at the masked, scaled index and return.
const int kValueOffset =
NumberDictionary::kElementsStartOffset + kPointerSize;
__ ldr(result, FieldMemOperand(t2, kValueOffset));
}
void LoadIC::GenerateArrayLength(MacroAssembler* masm) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -- r0 : receiver
// -- sp[0] : receiver
// -----------------------------------
Label miss;
StubCompiler::GenerateLoadArrayLength(masm, r0, r3, &miss);
__ bind(&miss);
StubCompiler::GenerateLoadMiss(masm, Code::LOAD_IC);
}
void LoadIC::GenerateStringLength(MacroAssembler* masm) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -- r0 : receiver
// -- sp[0] : receiver
// -----------------------------------
Label miss;
StubCompiler::GenerateLoadStringLength(masm, r0, r1, r3, &miss);
// Cache miss: Jump to runtime.
__ bind(&miss);
StubCompiler::GenerateLoadMiss(masm, Code::LOAD_IC);
}
void LoadIC::GenerateFunctionPrototype(MacroAssembler* masm) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -- r0 : receiver
// -- sp[0] : receiver
// -----------------------------------
Label miss;
StubCompiler::GenerateLoadFunctionPrototype(masm, r0, r1, r3, &miss);
__ bind(&miss);
StubCompiler::GenerateLoadMiss(masm, Code::LOAD_IC);
}
// Checks the receiver for special cases (value type, slow case bits).
// Falls through for regular JS object.
static void GenerateKeyedLoadReceiverCheck(MacroAssembler* masm,
Register receiver,
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
Register map,
Register scratch,
int interceptor_bit,
Label* slow) {
// Check that the object isn't a smi.
__ BranchOnSmi(receiver, slow);
// Get the map of the receiver.
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
__ ldr(map, FieldMemOperand(receiver, HeapObject::kMapOffset));
// Check bit field.
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
__ ldrb(scratch, FieldMemOperand(map, Map::kBitFieldOffset));
__ tst(scratch,
Operand((1 << Map::kIsAccessCheckNeeded) | (1 << interceptor_bit)));
__ b(nz, slow);
// Check that the object is some kind of JS object EXCEPT JS Value type.
// In the case that the object is a value-wrapper object,
// we enter the runtime system to make sure that indexing into string
// objects work as intended.
ASSERT(JS_OBJECT_TYPE > JS_VALUE_TYPE);
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
__ ldrb(scratch, FieldMemOperand(map, Map::kInstanceTypeOffset));
__ cmp(scratch, Operand(JS_OBJECT_TYPE));
__ b(lt, slow);
}
// Loads an indexed element from a fast case array.
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
// If not_fast_array is NULL, doesn't perform the elements map check.
static void GenerateFastArrayLoad(MacroAssembler* masm,
Register receiver,
Register key,
Register elements,
Register scratch1,
Register scratch2,
Register result,
Label* not_fast_array,
Label* out_of_range) {
// Register use:
//
// receiver - holds the receiver on entry.
// Unchanged unless 'result' is the same register.
//
// key - holds the smi key on entry.
// Unchanged unless 'result' is the same register.
//
// elements - holds the elements of the receiver on exit.
//
// result - holds the result on exit if the load succeeded.
// Allowed to be the the same as 'receiver' or 'key'.
// Unchanged on bailout so 'receiver' and 'key' can be safely
// used by further computation.
//
// Scratch registers:
//
// scratch1 - used to hold elements map and elements length.
// Holds the elements map if not_fast_array branch is taken.
//
// scratch2 - used to hold the loaded value.
__ ldr(elements, FieldMemOperand(receiver, JSObject::kElementsOffset));
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
if (not_fast_array != NULL) {
// Check that the object is in fast mode and writable.
__ ldr(scratch1, FieldMemOperand(elements, HeapObject::kMapOffset));
__ LoadRoot(ip, Heap::kFixedArrayMapRootIndex);
__ cmp(scratch1, ip);
__ b(ne, not_fast_array);
} else {
__ AssertFastElements(elements);
}
// Check that the key (index) is within bounds.
__ ldr(scratch1, FieldMemOperand(elements, FixedArray::kLengthOffset));
__ cmp(key, Operand(scratch1));
__ b(hs, out_of_range);
// Fast case: Do the load.
__ add(scratch1, elements, Operand(FixedArray::kHeaderSize - kHeapObjectTag));
// The key is a smi.
ASSERT(kSmiTag == 0 && kSmiTagSize < kPointerSizeLog2);
__ ldr(scratch2,
MemOperand(scratch1, key, LSL, kPointerSizeLog2 - kSmiTagSize));
__ LoadRoot(ip, Heap::kTheHoleValueRootIndex);
__ cmp(scratch2, ip);
// In case the loaded value is the_hole we have to consult GetProperty
// to ensure the prototype chain is searched.
__ b(eq, out_of_range);
__ mov(result, scratch2);
}
// Checks whether a key is an array index string or a symbol string.
// Falls through if a key is a symbol.
static void GenerateKeyStringCheck(MacroAssembler* masm,
Register key,
Register map,
Register hash,
Label* index_string,
Label* not_symbol) {
// The key is not a smi.
// Is it a string?
__ CompareObjectType(key, map, hash, FIRST_NONSTRING_TYPE);
__ b(ge, not_symbol);
// Is the string an array index, with cached numeric value?
__ ldr(hash, FieldMemOperand(key, String::kHashFieldOffset));
__ tst(hash, Operand(String::kContainsCachedArrayIndexMask));
__ b(eq, index_string);
// Is the string a symbol?
// map: key map
__ ldrb(hash, FieldMemOperand(map, Map::kInstanceTypeOffset));
ASSERT(kSymbolTag != 0);
__ tst(hash, Operand(kIsSymbolMask));
__ b(eq, not_symbol);
}
// Defined in ic.cc.
Object* CallIC_Miss(Arguments args);
// The generated code does not accept smi keys.
// The generated code falls through if both probes miss.
static void GenerateMonomorphicCacheProbe(MacroAssembler* masm,
int argc,
Code::Kind kind) {
// ----------- S t a t e -------------
// -- r1 : receiver
// -- r2 : name
// -----------------------------------
Label number, non_number, non_string, boolean, probe, miss;
// Probe the stub cache.
Code::Flags flags =
Code::ComputeFlags(kind, NOT_IN_LOOP, MONOMORPHIC, NORMAL, argc);
StubCache::GenerateProbe(masm, flags, r1, r2, r3, r4, r5);
// If the stub cache probing failed, the receiver might be a value.
// For value objects, we use the map of the prototype objects for
// the corresponding JSValue for the cache and that is what we need
// to probe.
//
// Check for number.
__ tst(r1, Operand(kSmiTagMask));
__ b(eq, &number);
__ CompareObjectType(r1, r3, r3, HEAP_NUMBER_TYPE);
__ b(ne, &non_number);
__ bind(&number);
StubCompiler::GenerateLoadGlobalFunctionPrototype(
masm, Context::NUMBER_FUNCTION_INDEX, r1);
__ b(&probe);
// Check for string.
__ bind(&non_number);
__ cmp(r3, Operand(FIRST_NONSTRING_TYPE));
__ b(hs, &non_string);
StubCompiler::GenerateLoadGlobalFunctionPrototype(
masm, Context::STRING_FUNCTION_INDEX, r1);
__ b(&probe);
// Check for boolean.
__ bind(&non_string);
__ LoadRoot(ip, Heap::kTrueValueRootIndex);
__ cmp(r1, ip);
__ b(eq, &boolean);
__ LoadRoot(ip, Heap::kFalseValueRootIndex);
__ cmp(r1, ip);
__ b(ne, &miss);
__ bind(&boolean);
StubCompiler::GenerateLoadGlobalFunctionPrototype(
masm, Context::BOOLEAN_FUNCTION_INDEX, r1);
// Probe the stub cache for the value object.
__ bind(&probe);
StubCache::GenerateProbe(masm, flags, r1, r2, r3, r4, r5);
__ bind(&miss);
}
static void GenerateFunctionTailCall(MacroAssembler* masm,
int argc,
Label* miss,
Register scratch) {
// r1: function
// Check that the value isn't a smi.
__ tst(r1, Operand(kSmiTagMask));
__ b(eq, miss);
// Check that the value is a JSFunction.
__ CompareObjectType(r1, scratch, scratch, JS_FUNCTION_TYPE);
__ b(ne, miss);
// Invoke the function.
ParameterCount actual(argc);
__ InvokeFunction(r1, actual, JUMP_FUNCTION);
}
static void GenerateCallNormal(MacroAssembler* masm, int argc) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -----------------------------------
Label miss;
// Get the receiver of the function from the stack into r1.
__ ldr(r1, MemOperand(sp, argc * kPointerSize));
GenerateStringDictionaryReceiverCheck(masm, r1, r0, r3, r4, &miss);
// r0: elements
// Search the dictionary - put result in register r1.
GenerateDictionaryLoad(masm, &miss, r0, r2, r1, r3, r4);
GenerateFunctionTailCall(masm, argc, &miss, r4);
__ bind(&miss);
}
static void GenerateCallMiss(MacroAssembler* masm, int argc, IC::UtilityId id) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -----------------------------------
if (id == IC::kCallIC_Miss) {
__ IncrementCounter(&Counters::call_miss, 1, r3, r4);
} else {
__ IncrementCounter(&Counters::keyed_call_miss, 1, r3, r4);
}
// Get the receiver of the function from the stack.
__ ldr(r3, MemOperand(sp, argc * kPointerSize));
__ EnterInternalFrame();
// Push the receiver and the name of the function.
__ Push(r3, r2);
// Call the entry.
__ mov(r0, Operand(2));
__ mov(r1, Operand(ExternalReference(IC_Utility(id))));
CEntryStub stub(1);
__ CallStub(&stub);
// Move result to r1 and leave the internal frame.
__ mov(r1, Operand(r0));
__ LeaveInternalFrame();
// Check if the receiver is a global object of some sort.
// This can happen only for regular CallIC but not KeyedCallIC.
if (id == IC::kCallIC_Miss) {
Label invoke, global;
__ ldr(r2, MemOperand(sp, argc * kPointerSize)); // receiver
__ tst(r2, Operand(kSmiTagMask));
__ b(eq, &invoke);
__ CompareObjectType(r2, r3, r3, JS_GLOBAL_OBJECT_TYPE);
__ b(eq, &global);
__ cmp(r3, Operand(JS_BUILTINS_OBJECT_TYPE));
__ b(ne, &invoke);
// Patch the receiver on the stack.
__ bind(&global);
__ ldr(r2, FieldMemOperand(r2, GlobalObject::kGlobalReceiverOffset));
__ str(r2, MemOperand(sp, argc * kPointerSize));
__ bind(&invoke);
}
// Invoke the function.
ParameterCount actual(argc);
__ InvokeFunction(r1, actual, JUMP_FUNCTION);
}
void CallIC::GenerateMiss(MacroAssembler* masm, int argc) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -----------------------------------
GenerateCallMiss(masm, argc, IC::kCallIC_Miss);
}
void CallIC::GenerateMegamorphic(MacroAssembler* masm, int argc) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -----------------------------------
// Get the receiver of the function from the stack into r1.
__ ldr(r1, MemOperand(sp, argc * kPointerSize));
GenerateMonomorphicCacheProbe(masm, argc, Code::CALL_IC);
GenerateMiss(masm, argc);
}
void CallIC::GenerateNormal(MacroAssembler* masm, int argc) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -----------------------------------
GenerateCallNormal(masm, argc);
GenerateMiss(masm, argc);
}
void KeyedCallIC::GenerateMiss(MacroAssembler* masm, int argc) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -----------------------------------
GenerateCallMiss(masm, argc, IC::kKeyedCallIC_Miss);
}
void KeyedCallIC::GenerateMegamorphic(MacroAssembler* masm, int argc) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -----------------------------------
// Get the receiver of the function from the stack into r1.
__ ldr(r1, MemOperand(sp, argc * kPointerSize));
Label do_call, slow_call, slow_load, slow_reload_receiver;
Label check_number_dictionary, check_string, lookup_monomorphic_cache;
Label index_smi, index_string;
// Check that the key is a smi.
__ BranchOnNotSmi(r2, &check_string);
__ bind(&index_smi);
// Now the key is known to be a smi. This place is also jumped to from below
// where a numeric string is converted to a smi.
GenerateKeyedLoadReceiverCheck(
masm, r1, r0, r3, Map::kHasIndexedInterceptor, &slow_call);
GenerateFastArrayLoad(
masm, r1, r2, r4, r3, r0, r1, &check_number_dictionary, &slow_load);
__ IncrementCounter(&Counters::keyed_call_generic_smi_fast, 1, r0, r3);
__ bind(&do_call);
// receiver in r1 is not used after this point.
// r2: key
// r1: function
GenerateFunctionTailCall(masm, argc, &slow_call, r0);
__ bind(&check_number_dictionary);
// r2: key
// r3: elements map
// r4: elements
// Check whether the elements is a number dictionary.
__ LoadRoot(ip, Heap::kHashTableMapRootIndex);
__ cmp(r3, ip);
__ b(ne, &slow_load);
__ mov(r0, Operand(r2, ASR, kSmiTagSize));
// r0: untagged index
GenerateNumberDictionaryLoad(masm, &slow_load, r4, r2, r1, r0, r3, r5);
__ IncrementCounter(&Counters::keyed_call_generic_smi_dict, 1, r0, r3);
__ jmp(&do_call);
__ bind(&slow_load);
// This branch is taken when calling KeyedCallIC_Miss is neither required
// nor beneficial.
__ IncrementCounter(&Counters::keyed_call_generic_slow_load, 1, r0, r3);
__ EnterInternalFrame();
__ push(r2); // save the key
__ Push(r1, r2); // pass the receiver and the key
__ CallRuntime(Runtime::kKeyedGetProperty, 2);
__ pop(r2); // restore the key
__ LeaveInternalFrame();
__ mov(r1, r0);
__ jmp(&do_call);
__ bind(&check_string);
GenerateKeyStringCheck(masm, r2, r0, r3, &index_string, &slow_call);
// The key is known to be a symbol.
// If the receiver is a regular JS object with slow properties then do
// a quick inline probe of the receiver's dictionary.
// Otherwise do the monomorphic cache probe.
GenerateKeyedLoadReceiverCheck(
masm, r1, r0, r3, Map::kHasNamedInterceptor, &lookup_monomorphic_cache);
__ ldr(r0, FieldMemOperand(r1, JSObject::kPropertiesOffset));
__ ldr(r3, FieldMemOperand(r0, HeapObject::kMapOffset));
__ LoadRoot(ip, Heap::kHashTableMapRootIndex);
__ cmp(r3, ip);
__ b(ne, &lookup_monomorphic_cache);
GenerateDictionaryLoad(masm, &slow_load, r0, r2, r1, r3, r4);
__ IncrementCounter(&Counters::keyed_call_generic_lookup_dict, 1, r0, r3);
__ jmp(&do_call);
__ bind(&lookup_monomorphic_cache);
__ IncrementCounter(&Counters::keyed_call_generic_lookup_cache, 1, r0, r3);
GenerateMonomorphicCacheProbe(masm, argc, Code::KEYED_CALL_IC);
// Fall through on miss.
__ bind(&slow_call);
// This branch is taken if:
// - the receiver requires boxing or access check,
// - the key is neither smi nor symbol,
// - the value loaded is not a function,
// - there is hope that the runtime will create a monomorphic call stub
// that will get fetched next time.
__ IncrementCounter(&Counters::keyed_call_generic_slow, 1, r0, r3);
GenerateMiss(masm, argc);
__ bind(&index_string);
__ IndexFromHash(r3, r2);
// Now jump to the place where smi keys are handled.
__ jmp(&index_smi);
}
void KeyedCallIC::GenerateNormal(MacroAssembler* masm, int argc) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -----------------------------------
GenerateCallNormal(masm, argc);
GenerateMiss(masm, argc);
}
// Defined in ic.cc.
Object* LoadIC_Miss(Arguments args);
void LoadIC::GenerateMegamorphic(MacroAssembler* masm) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -- r0 : receiver
// -- sp[0] : receiver
// -----------------------------------
// Probe the stub cache.
Code::Flags flags = Code::ComputeFlags(Code::LOAD_IC,
NOT_IN_LOOP,
MONOMORPHIC);
StubCache::GenerateProbe(masm, flags, r0, r2, r3, r4, r5);
// Cache miss: Jump to runtime.
GenerateMiss(masm);
}
void LoadIC::GenerateNormal(MacroAssembler* masm) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -- r0 : receiver
// -- sp[0] : receiver
// -----------------------------------
Label miss;
GenerateStringDictionaryReceiverCheck(masm, r0, r1, r3, r4, &miss);
// r1: elements
GenerateDictionaryLoad(masm, &miss, r1, r2, r0, r3, r4);
__ Ret();
// Cache miss: Jump to runtime.
__ bind(&miss);
GenerateMiss(masm);
}
void LoadIC::GenerateMiss(MacroAssembler* masm) {
// ----------- S t a t e -------------
// -- r2 : name
// -- lr : return address
// -- r0 : receiver
// -- sp[0] : receiver
// -----------------------------------
__ IncrementCounter(&Counters::load_miss, 1, r3, r4);
__ mov(r3, r0);
__ Push(r3, r2);
// Perform tail call to the entry.
ExternalReference ref = ExternalReference(IC_Utility(kLoadIC_Miss));
__ TailCallExternalReference(ref, 2, 1);
}
static inline bool IsInlinedICSite(Address address,
Address* inline_end_address) {
// If the instruction after the call site is not the pseudo instruction nop1
// then this is not related to an inlined in-object property load. The nop1
// instruction is located just after the call to the IC in the deferred code
// handling the miss in the inlined code. After the nop1 instruction there is
// a branch instruction for jumping back from the deferred code.
Address address_after_call = address + Assembler::kCallTargetAddressOffset;
Instr instr_after_call = Assembler::instr_at(address_after_call);
if (!Assembler::IsNop(instr_after_call, PROPERTY_ACCESS_INLINED)) {
return false;
}
Address address_after_nop = address_after_call + Assembler::kInstrSize;
Instr instr_after_nop = Assembler::instr_at(address_after_nop);
// There may be some reg-reg move and frame merging code to skip over before
// the branch back from the DeferredReferenceGetKeyedValue code to the inlined
// code.
while (!Assembler::IsBranch(instr_after_nop)) {
address_after_nop += Assembler::kInstrSize;
instr_after_nop = Assembler::instr_at(address_after_nop);
}
// Find the end of the inlined code for handling the load.
int b_offset =
Assembler::GetBranchOffset(instr_after_nop) + Assembler::kPcLoadDelta;
ASSERT(b_offset < 0); // Jumping back from deferred code.
*inline_end_address = address_after_nop + b_offset;
return true;
}
bool LoadIC::PatchInlinedLoad(Address address, Object* map, int offset) {
// Find the end of the inlined code for handling the load if this is an
// inlined IC call site.
Address inline_end_address;
if (!IsInlinedICSite(address, &inline_end_address)) return false;
// Patch the offset of the property load instruction (ldr r0, [r1, #+XXX]).
// The immediate must be representable in 12 bits.
ASSERT((JSObject::kMaxInstanceSize - JSObject::kHeaderSize) < (1 << 12));
Address ldr_property_instr_address =
inline_end_address - Assembler::kInstrSize;
ASSERT(Assembler::IsLdrRegisterImmediate(
Assembler::instr_at(ldr_property_instr_address)));
Instr ldr_property_instr = Assembler::instr_at(ldr_property_instr_address);
ldr_property_instr = Assembler::SetLdrRegisterImmediateOffset(
ldr_property_instr, offset - kHeapObjectTag);
Assembler::instr_at_put(ldr_property_instr_address, ldr_property_instr);
// Indicate that code has changed.
CPU::FlushICache(ldr_property_instr_address, 1 * Assembler::kInstrSize);
// Patch the map check.
Address ldr_map_instr_address =
inline_end_address - 4 * Assembler::kInstrSize;
Assembler::set_target_address_at(ldr_map_instr_address,
reinterpret_cast<Address>(map));
return true;
}
bool LoadIC::PatchInlinedContextualLoad(Address address,
Object* map,
Object* cell,
bool is_dont_delete) {
// TODO(<bug#>): implement this.
return false;
}
bool StoreIC::PatchInlinedStore(Address address, Object* map, int offset) {
// Find the end of the inlined code for the store if there is an
// inlined version of the store.
Address inline_end_address;
if (!IsInlinedICSite(address, &inline_end_address)) return false;
// Compute the address of the map load instruction.
Address ldr_map_instr_address =
inline_end_address -
(CodeGenerator::GetInlinedNamedStoreInstructionsAfterPatch() *
Assembler::kInstrSize);
// Update the offsets if initializing the inlined store. No reason
// to update the offsets when clearing the inlined version because
// it will bail out in the map check.
if (map != Heap::null_value()) {
// Patch the offset in the actual store instruction.
Address str_property_instr_address =
ldr_map_instr_address + 3 * Assembler::kInstrSize;
Instr str_property_instr = Assembler::instr_at(str_property_instr_address);
ASSERT(Assembler::IsStrRegisterImmediate(str_property_instr));
str_property_instr = Assembler::SetStrRegisterImmediateOffset(
str_property_instr, offset - kHeapObjectTag);
Assembler::instr_at_put(str_property_instr_address, str_property_instr);
// Patch the offset in the add instruction that is part of the
// write barrier.
Address add_offset_instr_address =
str_property_instr_address + Assembler::kInstrSize;
Instr add_offset_instr = Assembler::instr_at(add_offset_instr_address);
ASSERT(Assembler::IsAddRegisterImmediate(add_offset_instr));
add_offset_instr = Assembler::SetAddRegisterImmediateOffset(
add_offset_instr, offset - kHeapObjectTag);
Assembler::instr_at_put(add_offset_instr_address, add_offset_instr);
// Indicate that code has changed.
CPU::FlushICache(str_property_instr_address, 2 * Assembler::kInstrSize);
}
// Patch the map check.
Assembler::set_target_address_at(ldr_map_instr_address,
reinterpret_cast<Address>(map));
return true;
}
bool KeyedLoadIC::PatchInlinedLoad(Address address, Object* map) {
Address inline_end_address;
if (!IsInlinedICSite(address, &inline_end_address)) return false;
// Patch the map check.
Address ldr_map_instr_address =
inline_end_address -
(CodeGenerator::GetInlinedKeyedLoadInstructionsAfterPatch() *
Assembler::kInstrSize);
Assembler::set_target_address_at(ldr_map_instr_address,
reinterpret_cast<Address>(map));
return true;
}
bool KeyedStoreIC::PatchInlinedStore(Address address, Object* map) {
// Find the end of the inlined code for handling the store if this is an
// inlined IC call site.
Address inline_end_address;
if (!IsInlinedICSite(address, &inline_end_address)) return false;
// Patch the map check.
Address ldr_map_instr_address =
inline_end_address -
(CodeGenerator::kInlinedKeyedStoreInstructionsAfterPatch *
Assembler::kInstrSize);
Assembler::set_target_address_at(ldr_map_instr_address,
reinterpret_cast<Address>(map));
return true;
}
Object* KeyedLoadIC_Miss(Arguments args);
void KeyedLoadIC::GenerateMiss(MacroAssembler* masm) {
// ---------- S t a t e --------------
// -- lr : return address
// -- r0 : key
// -- r1 : receiver
// -----------------------------------
__ IncrementCounter(&Counters::keyed_load_miss, 1, r3, r4);
__ Push(r1, r0);
ExternalReference ref = ExternalReference(IC_Utility(kKeyedLoadIC_Miss));
__ TailCallExternalReference(ref, 2, 1);
}
void KeyedLoadIC::GenerateRuntimeGetProperty(MacroAssembler* masm) {
// ---------- S t a t e --------------
// -- lr : return address
// -- r0 : key
// -- r1 : receiver
// -----------------------------------
__ Push(r1, r0);
__ TailCallRuntime(Runtime::kKeyedGetProperty, 2, 1);
}
void KeyedLoadIC::GenerateGeneric(MacroAssembler* masm) {
// ---------- S t a t e --------------
// -- lr : return address
// -- r0 : key
// -- r1 : receiver
// -----------------------------------
Label slow, check_string, index_smi, index_string, property_array_property;
Label check_pixel_array, probe_dictionary, check_number_dictionary;
Register key = r0;
Register receiver = r1;
// Check that the key is a smi.
__ BranchOnNotSmi(key, &check_string);
__ bind(&index_smi);
// Now the key is known to be a smi. This place is also jumped to from below
// where a numeric string is converted to a smi.
GenerateKeyedLoadReceiverCheck(
masm, receiver, r2, r3, Map::kHasIndexedInterceptor, &slow);
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
// Check the "has fast elements" bit in the receiver's map which is
// now in r2.
__ ldrb(r3, FieldMemOperand(r2, Map::kBitField2Offset));
__ tst(r3, Operand(1 << Map::kHasFastElements));
__ b(eq, &check_pixel_array);
GenerateFastArrayLoad(
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
masm, receiver, key, r4, r3, r2, r0, NULL, &slow);
__ IncrementCounter(&Counters::keyed_load_generic_smi, 1, r2, r3);
__ Ret();
// Check whether the elements is a pixel array.
// r0: key
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
// r1: receiver
__ bind(&check_pixel_array);
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
__ ldr(r4, FieldMemOperand(r1, JSObject::kElementsOffset));
__ ldr(r3, FieldMemOperand(r4, HeapObject::kMapOffset));
__ LoadRoot(ip, Heap::kPixelArrayMapRootIndex);
__ cmp(r3, ip);
__ b(ne, &check_number_dictionary);
__ ldr(ip, FieldMemOperand(r4, PixelArray::kLengthOffset));
__ mov(r2, Operand(key, ASR, kSmiTagSize));
__ cmp(r2, ip);
__ b(hs, &slow);
__ ldr(ip, FieldMemOperand(r4, PixelArray::kExternalPointerOffset));
__ ldrb(r2, MemOperand(ip, r2));
__ mov(r0, Operand(r2, LSL, kSmiTagSize)); // Tag result as smi.
__ Ret();
__ bind(&check_number_dictionary);
// Check whether the elements is a number dictionary.
// r0: key
// r3: elements map
// r4: elements
__ LoadRoot(ip, Heap::kHashTableMapRootIndex);
__ cmp(r3, ip);
__ b(ne, &slow);
__ mov(r2, Operand(r0, ASR, kSmiTagSize));
GenerateNumberDictionaryLoad(masm, &slow, r4, r0, r0, r2, r3, r5);
__ Ret();
// Slow case, key and receiver still in r0 and r1.
__ bind(&slow);
__ IncrementCounter(&Counters::keyed_load_generic_slow, 1, r2, r3);
GenerateRuntimeGetProperty(masm);
__ bind(&check_string);
GenerateKeyStringCheck(masm, key, r2, r3, &index_string, &slow);
GenerateKeyedLoadReceiverCheck(
masm, receiver, r2, r3, Map::kHasNamedInterceptor, &slow);
// If the receiver is a fast-case object, check the keyed lookup
// cache. Otherwise probe the dictionary.
__ ldr(r3, FieldMemOperand(r1, JSObject::kPropertiesOffset));
__ ldr(r4, FieldMemOperand(r3, HeapObject::kMapOffset));
__ LoadRoot(ip, Heap::kHashTableMapRootIndex);
__ cmp(r4, ip);
__ b(eq, &probe_dictionary);
// Load the map of the receiver, compute the keyed lookup cache hash
// based on 32 bits of the map pointer and the string hash.
__ ldr(r2, FieldMemOperand(r1, HeapObject::kMapOffset));
__ mov(r3, Operand(r2, ASR, KeyedLookupCache::kMapHashShift));
__ ldr(r4, FieldMemOperand(r0, String::kHashFieldOffset));
__ eor(r3, r3, Operand(r4, ASR, String::kHashShift));
__ And(r3, r3, Operand(KeyedLookupCache::kCapacityMask));
// Load the key (consisting of map and symbol) from the cache and
// check for match.
ExternalReference cache_keys = ExternalReference::keyed_lookup_cache_keys();
__ mov(r4, Operand(cache_keys));
__ add(r4, r4, Operand(r3, LSL, kPointerSizeLog2 + 1));
__ ldr(r5, MemOperand(r4, kPointerSize, PostIndex)); // Move r4 to symbol.
__ cmp(r2, r5);
__ b(ne, &slow);
__ ldr(r5, MemOperand(r4));
__ cmp(r0, r5);
__ b(ne, &slow);
// Get field offset.
// r0 : key
// r1 : receiver
// r2 : receiver's map
// r3 : lookup cache index
ExternalReference cache_field_offsets
= ExternalReference::keyed_lookup_cache_field_offsets();
__ mov(r4, Operand(cache_field_offsets));
__ ldr(r5, MemOperand(r4, r3, LSL, kPointerSizeLog2));
__ ldrb(r6, FieldMemOperand(r2, Map::kInObjectPropertiesOffset));
__ sub(r5, r5, r6, SetCC);
__ b(ge, &property_array_property);
// Load in-object property.
__ ldrb(r6, FieldMemOperand(r2, Map::kInstanceSizeOffset));
__ add(r6, r6, r5); // Index from start of object.
__ sub(r1, r1, Operand(kHeapObjectTag)); // Remove the heap tag.
__ ldr(r0, MemOperand(r1, r6, LSL, kPointerSizeLog2));
__ IncrementCounter(&Counters::keyed_load_generic_lookup_cache, 1, r2, r3);
__ Ret();
// Load property array property.
__ bind(&property_array_property);
__ ldr(r1, FieldMemOperand(r1, JSObject::kPropertiesOffset));
__ add(r1, r1, Operand(FixedArray::kHeaderSize - kHeapObjectTag));
__ ldr(r0, MemOperand(r1, r5, LSL, kPointerSizeLog2));
__ IncrementCounter(&Counters::keyed_load_generic_lookup_cache, 1, r2, r3);
__ Ret();
// Do a quick inline probe of the receiver's dictionary, if it
// exists.
__ bind(&probe_dictionary);
// r1: receiver
// r0: key
// r3: elements
__ ldr(r2, FieldMemOperand(r1, HeapObject::kMapOffset));
__ ldrb(r2, FieldMemOperand(r2, Map::kInstanceTypeOffset));
GenerateGlobalInstanceTypeCheck(masm, r2, &slow);
// Load the property to r0.
GenerateDictionaryLoad(masm, &slow, r3, r0, r0, r2, r4);
__ IncrementCounter(&Counters::keyed_load_generic_symbol, 1, r2, r3);
__ Ret();
__ bind(&index_string);
__ IndexFromHash(r3, key);
// Now jump to the place where smi keys are handled.
__ jmp(&index_smi);
}
void KeyedLoadIC::GenerateString(MacroAssembler* masm) {
// ---------- S t a t e --------------
// -- lr : return address
// -- r0 : key (index)
// -- r1 : receiver
// -----------------------------------
Label miss;
Register receiver = r1;
Register index = r0;
Register scratch1 = r2;
Register scratch2 = r3;
Register result = r0;
StringCharAtGenerator char_at_generator(receiver,
index,
scratch1,
scratch2,
result,
&miss, // When not a string.
&miss, // When not a number.
&miss, // When index out of range.
STRING_INDEX_IS_ARRAY_INDEX);
char_at_generator.GenerateFast(masm);
__ Ret();
ICRuntimeCallHelper call_helper;
char_at_generator.GenerateSlow(masm, call_helper);
__ bind(&miss);
GenerateMiss(masm);
}
// Convert unsigned integer with specified number of leading zeroes in binary
// representation to IEEE 754 double.
// Integer to convert is passed in register hiword.
// Resulting double is returned in registers hiword:loword.
// This functions does not work correctly for 0.
static void GenerateUInt2Double(MacroAssembler* masm,
Register hiword,
Register loword,
Register scratch,
int leading_zeroes) {
const int meaningful_bits = kBitsPerInt - leading_zeroes - 1;
const int biased_exponent = HeapNumber::kExponentBias + meaningful_bits;
const int mantissa_shift_for_hi_word =
meaningful_bits - HeapNumber::kMantissaBitsInTopWord;
const int mantissa_shift_for_lo_word =
kBitsPerInt - mantissa_shift_for_hi_word;
__ mov(scratch, Operand(biased_exponent << HeapNumber::kExponentShift));
if (mantissa_shift_for_hi_word > 0) {
__ mov(loword, Operand(hiword, LSL, mantissa_shift_for_lo_word));
__ orr(hiword, scratch, Operand(hiword, LSR, mantissa_shift_for_hi_word));
} else {
__ mov(loword, Operand(0, RelocInfo::NONE));
__ orr(hiword, scratch, Operand(hiword, LSL, mantissa_shift_for_hi_word));
}
// If least significant bit of biased exponent was not 1 it was corrupted
// by most significant bit of mantissa so we should fix that.
if (!(biased_exponent & 1)) {
__ bic(hiword, hiword, Operand(1 << HeapNumber::kExponentShift));
}
}
void KeyedLoadIC::GenerateExternalArray(MacroAssembler* masm,
ExternalArrayType array_type) {
// ---------- S t a t e --------------
// -- lr : return address
// -- r0 : key
// -- r1 : receiver
// -----------------------------------
Label slow, failed_allocation;
Register key = r0;
Register receiver = r1;
// Check that the object isn't a smi
__ BranchOnSmi(receiver, &slow);
// Check that the key is a smi.
__ BranchOnNotSmi(key, &slow);
// Check that the object is a JS object. Load map into r2.
__ CompareObjectType(receiver, r2, r3, FIRST_JS_OBJECT_TYPE);
__ b(lt, &slow);
// Check that the receiver does not require access checks. We need
// to check this explicitly since this generic stub does not perform
// map checks.
__ ldrb(r3, FieldMemOperand(r2, Map::kBitFieldOffset));
__ tst(r3, Operand(1 << Map::kIsAccessCheckNeeded));
__ b(ne, &slow);
// Check that the elements array is the appropriate type of
// ExternalArray.
__ ldr(r3, FieldMemOperand(receiver, JSObject::kElementsOffset));
__ ldr(r2, FieldMemOperand(r3, HeapObject::kMapOffset));
__ LoadRoot(ip, Heap::RootIndexForExternalArrayType(array_type));
__ cmp(r2, ip);
__ b(ne, &slow);
// Check that the index is in range.
__ ldr(ip, FieldMemOperand(r3, ExternalArray::kLengthOffset));
__ cmp(ip, Operand(key, ASR, kSmiTagSize));
// Unsigned comparison catches both negative and too-large values.
__ b(lo, &slow);
// r3: elements array
__ ldr(r3, FieldMemOperand(r3, ExternalArray::kExternalPointerOffset));
// r3: base pointer of external storage
// We are not untagging smi key and instead work with it
// as if it was premultiplied by 2.
ASSERT((kSmiTag == 0) && (kSmiTagSize == 1));
Register value = r2;
switch (array_type) {
case kExternalByteArray:
__ ldrsb(value, MemOperand(r3, key, LSR, 1));
break;
case kExternalUnsignedByteArray:
__ ldrb(value, MemOperand(r3, key, LSR, 1));
break;
case kExternalShortArray:
__ ldrsh(value, MemOperand(r3, key, LSL, 0));
break;
case kExternalUnsignedShortArray:
__ ldrh(value, MemOperand(r3, key, LSL, 0));
break;
case kExternalIntArray:
case kExternalUnsignedIntArray:
__ ldr(value, MemOperand(r3, key, LSL, 1));
break;
case kExternalFloatArray:
if (CpuFeatures::IsSupported(VFP3)) {
CpuFeatures::Scope scope(VFP3);
__ add(r2, r3, Operand(key, LSL, 1));
__ vldr(s0, r2, 0);
} else {
__ ldr(value, MemOperand(r3, key, LSL, 1));
}
break;
default:
UNREACHABLE();
break;
}
// For integer array types:
// r2: value
// For floating-point array type
// s0: value (if VFP3 is supported)
// r2: value (if VFP3 is not supported)
if (array_type == kExternalIntArray) {
// For the Int and UnsignedInt array types, we need to see whether
// the value can be represented in a Smi. If not, we need to convert
// it to a HeapNumber.
Label box_int;
__ cmp(value, Operand(0xC0000000));
__ b(mi, &box_int);
// Tag integer as smi and return it.
__ mov(r0, Operand(value, LSL, kSmiTagSize));
__ Ret();
__ bind(&box_int);
// Allocate a HeapNumber for the result and perform int-to-double
// conversion. Don't touch r0 or r1 as they are needed if allocation
// fails.
__ LoadRoot(r6, Heap::kHeapNumberMapRootIndex);
__ AllocateHeapNumber(r5, r3, r4, r6, &slow);
// Now we can use r0 for the result as key is not needed any more.
__ mov(r0, r5);
if (CpuFeatures::IsSupported(VFP3)) {
CpuFeatures::Scope scope(VFP3);
__ vmov(s0, value);
__ vcvt_f64_s32(d0, s0);
__ sub(r3, r0, Operand(kHeapObjectTag));
__ vstr(d0, r3, HeapNumber::kValueOffset);
__ Ret();
} else {
WriteInt32ToHeapNumberStub stub(value, r0, r3);
__ TailCallStub(&stub);
}
} else if (array_type == kExternalUnsignedIntArray) {
// The test is different for unsigned int values. Since we need
// the value to be in the range of a positive smi, we can't
// handle either of the top two bits being set in the value.
if (CpuFeatures::IsSupported(VFP3)) {
CpuFeatures::Scope scope(VFP3);
Label box_int, done;
__ tst(value, Operand(0xC0000000));
__ b(ne, &box_int);
// Tag integer as smi and return it.
__ mov(r0, Operand(value, LSL, kSmiTagSize));
__ Ret();
__ bind(&box_int);
__ vmov(s0, value);
// Allocate a HeapNumber for the result and perform int-to-double
// conversion. Don't use r0 and r1 as AllocateHeapNumber clobbers all
// registers - also when jumping due to exhausted young space.
__ LoadRoot(r6, Heap::kHeapNumberMapRootIndex);
__ AllocateHeapNumber(r2, r3, r4, r6, &slow);
__ vcvt_f64_u32(d0, s0);
__ sub(r1, r2, Operand(kHeapObjectTag));
__ vstr(d0, r1, HeapNumber::kValueOffset);
__ mov(r0, r2);
__ Ret();
} else {
// Check whether unsigned integer fits into smi.
Label box_int_0, box_int_1, done;
__ tst(value, Operand(0x80000000));
__ b(ne, &box_int_0);
__ tst(value, Operand(0x40000000));
__ b(ne, &box_int_1);
// Tag integer as smi and return it.
__ mov(r0, Operand(value, LSL, kSmiTagSize));
__ Ret();
Register hiword = value; // r2.
Register loword = r3;
__ bind(&box_int_0);
// Integer does not have leading zeros.
GenerateUInt2Double(masm, hiword, loword, r4, 0);
__ b(&done);
__ bind(&box_int_1);
// Integer has one leading zero.
GenerateUInt2Double(masm, hiword, loword, r4, 1);
__ bind(&done);
// Integer was converted to double in registers hiword:loword.
// Wrap it into a HeapNumber. Don't use r0 and r1 as AllocateHeapNumber
// clobbers all registers - also when jumping due to exhausted young
// space.
__ LoadRoot(r6, Heap::kHeapNumberMapRootIndex);
__ AllocateHeapNumber(r4, r5, r7, r6, &slow);
__ str(hiword, FieldMemOperand(r4, HeapNumber::kExponentOffset));
__ str(loword, FieldMemOperand(r4, HeapNumber::kMantissaOffset));
__ mov(r0, r4);
__ Ret();
}
} else if (array_type == kExternalFloatArray) {
// For the floating-point array type, we need to always allocate a
// HeapNumber.
if (CpuFeatures::IsSupported(VFP3)) {
CpuFeatures::Scope scope(VFP3);
// Allocate a HeapNumber for the result. Don't use r0 and r1 as
// AllocateHeapNumber clobbers all registers - also when jumping due to
// exhausted young space.
__ LoadRoot(r6, Heap::kHeapNumberMapRootIndex);
__ AllocateHeapNumber(r2, r3, r4, r6, &slow);
__ vcvt_f64_f32(d0, s0);
__ sub(r1, r2, Operand(kHeapObjectTag));
__ vstr(d0, r1, HeapNumber::kValueOffset);
__ mov(r0, r2);
__ Ret();
} else {
// Allocate a HeapNumber for the result. Don't use r0 and r1 as
// AllocateHeapNumber clobbers all registers - also when jumping due to
// exhausted young space.
__ LoadRoot(r6, Heap::kHeapNumberMapRootIndex);
__ AllocateHeapNumber(r3, r4, r5, r6, &slow);
// VFP is not available, do manual single to double conversion.
// r2: floating point value (binary32)
// r3: heap number for result
// Extract mantissa to r0. OK to clobber r0 now as there are no jumps to
// the slow case from here.
__ and_(r0, value, Operand(kBinary32MantissaMask));
// Extract exponent to r1. OK to clobber r1 now as there are no jumps to
// the slow case from here.
__ mov(r1, Operand(value, LSR, kBinary32MantissaBits));
__ and_(r1, r1, Operand(kBinary32ExponentMask >> kBinary32MantissaBits));
Label exponent_rebiased;
__ teq(r1, Operand(0x00));
__ b(eq, &exponent_rebiased);
__ teq(r1, Operand(0xff));
__ mov(r1, Operand(0x7ff), LeaveCC, eq);
__ b(eq, &exponent_rebiased);
// Rebias exponent.
__ add(r1,
r1,
Operand(-kBinary32ExponentBias + HeapNumber::kExponentBias));
__ bind(&exponent_rebiased);
__ and_(r2, value, Operand(kBinary32SignMask));
value = no_reg;
__ orr(r2, r2, Operand(r1, LSL, HeapNumber::kMantissaBitsInTopWord));
// Shift mantissa.
static const int kMantissaShiftForHiWord =
kBinary32MantissaBits - HeapNumber::kMantissaBitsInTopWord;
static const int kMantissaShiftForLoWord =
kBitsPerInt - kMantissaShiftForHiWord;
__ orr(r2, r2, Operand(r0, LSR, kMantissaShiftForHiWord));
__ mov(r0, Operand(r0, LSL, kMantissaShiftForLoWord));
__ str(r2, FieldMemOperand(r3, HeapNumber::kExponentOffset));
__ str(r0, FieldMemOperand(r3, HeapNumber::kMantissaOffset));
__ mov(r0, r3);
__ Ret();
}
} else {
// Tag integer as smi and return it.
__ mov(r0, Operand(value, LSL, kSmiTagSize));
__ Ret();
}
// Slow case, key and receiver still in r0 and r1.
__ bind(&slow);
__ IncrementCounter(&Counters::keyed_load_external_array_slow, 1, r2, r3);
GenerateRuntimeGetProperty(masm);
}
void KeyedLoadIC::GenerateIndexedInterceptor(MacroAssembler* masm) {
// ---------- S t a t e --------------
// -- lr : return address
// -- r0 : key
// -- r1 : receiver
// -----------------------------------
Label slow;
// Check that the receiver isn't a smi.
__ BranchOnSmi(r1, &slow);
// Check that the key is an array index, that is Uint32.
__ tst(r0, Operand(kSmiTagMask | kSmiSignMask));
__ b(ne, &slow);
// Get the map of the receiver.
__ ldr(r2, FieldMemOperand(r1, HeapObject::kMapOffset));
// Check that it has indexed interceptor and access checks
// are not enabled for this object.
__ ldrb(r3, FieldMemOperand(r2, Map::kBitFieldOffset));
__ and_(r3, r3, Operand(kSlowCaseBitFieldMask));
__ cmp(r3, Operand(1 << Map::kHasIndexedInterceptor));
__ b(ne, &slow);
// Everything is fine, call runtime.
__ Push(r1, r0); // Receiver, key.
// Perform tail call to the entry.
__ TailCallExternalReference(ExternalReference(
IC_Utility(kKeyedLoadPropertyWithInterceptor)), 2, 1);
__ bind(&slow);
GenerateMiss(masm);
}
void KeyedStoreIC::GenerateMiss(MacroAssembler* masm) {
// ---------- S t a t e --------------
// -- r0 : value
// -- r1 : key
// -- r2 : receiver
// -- lr : return address
// -----------------------------------
// Push receiver, key and value for runtime call.
__ Push(r2, r1, r0);
ExternalReference ref = ExternalReference(IC_Utility(kKeyedStoreIC_Miss));
__ TailCallExternalReference(ref, 3, 1);
}
void KeyedStoreIC::GenerateRuntimeSetProperty(MacroAssembler* masm) {
// ---------- S t a t e --------------
// -- r0 : value
// -- r1 : key
// -- r2 : receiver
// -- lr : return address
// -----------------------------------
// Push receiver, key and value for runtime call.
__ Push(r2, r1, r0);
__ TailCallRuntime(Runtime::kSetProperty, 3, 1);
}
void KeyedStoreIC::GenerateGeneric(MacroAssembler* masm) {
// ---------- S t a t e --------------
// -- r0 : value
// -- r1 : key
// -- r2 : receiver
// -- lr : return address
// -----------------------------------
Label slow, fast, array, extra, check_pixel_array;
// Register usage.
Register value = r0;
Register key = r1;
Register receiver = r2;
Register elements = r3; // Elements array of the receiver.
// r4 and r5 are used as general scratch registers.
// Check that the key is a smi.
__ tst(key, Operand(kSmiTagMask));
__ b(ne, &slow);
// Check that the object isn't a smi.
__ tst(receiver, Operand(kSmiTagMask));
__ b(eq, &slow);
// Get the map of the object.
__ ldr(r4, FieldMemOperand(receiver, HeapObject::kMapOffset));
// Check that the receiver does not require access checks. We need
// to do this because this generic stub does not perform map checks.
__ ldrb(ip, FieldMemOperand(r4, Map::kBitFieldOffset));
__ tst(ip, Operand(1 << Map::kIsAccessCheckNeeded));
__ b(ne, &slow);
// Check if the object is a JS array or not.
__ ldrb(r4, FieldMemOperand(r4, Map::kInstanceTypeOffset));
__ cmp(r4, Operand(JS_ARRAY_TYPE));
__ b(eq, &array);
// Check that the object is some kind of JS object.
__ cmp(r4, Operand(FIRST_JS_OBJECT_TYPE));
__ b(lt, &slow);
// Object case: Check key against length in the elements array.
__ ldr(elements, FieldMemOperand(receiver, JSObject::kElementsOffset));
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
// Check that the object is in fast mode and writable.
__ ldr(r4, FieldMemOperand(elements, HeapObject::kMapOffset));
__ LoadRoot(ip, Heap::kFixedArrayMapRootIndex);
__ cmp(r4, ip);
__ b(ne, &check_pixel_array);
// Check array bounds. Both the key and the length of FixedArray are smis.
__ ldr(ip, FieldMemOperand(elements, FixedArray::kLengthOffset));
__ cmp(key, Operand(ip));
__ b(lo, &fast);
// Slow case, handle jump to runtime.
__ bind(&slow);
// Entry registers are intact.
// r0: value.
// r1: key.
// r2: receiver.
GenerateRuntimeSetProperty(masm);
// Check whether the elements is a pixel array.
// r4: elements map.
__ bind(&check_pixel_array);
__ LoadRoot(ip, Heap::kPixelArrayMapRootIndex);
__ cmp(r4, ip);
__ b(ne, &slow);
// Check that the value is a smi. If a conversion is needed call into the
// runtime to convert and clamp.
__ BranchOnNotSmi(value, &slow);
__ mov(r4, Operand(key, ASR, kSmiTagSize)); // Untag the key.
__ ldr(ip, FieldMemOperand(elements, PixelArray::kLengthOffset));
__ cmp(r4, Operand(ip));
__ b(hs, &slow);
__ mov(r5, Operand(value, ASR, kSmiTagSize)); // Untag the value.
__ Usat(r5, 8, Operand(r5)); // Clamp the value to [0..255].
// Get the pointer to the external array. This clobbers elements.
__ ldr(elements,
FieldMemOperand(elements, PixelArray::kExternalPointerOffset));
__ strb(r5, MemOperand(elements, r4)); // Elements is now external array.
__ Ret();
// Extra capacity case: Check if there is extra capacity to
// perform the store and update the length. Used for adding one
// element to the array by writing to array[array.length].
__ bind(&extra);
// Condition code from comparing key and array length is still available.
__ b(ne, &slow); // Only support writing to writing to array[array.length].
// Check for room in the elements backing store.
// Both the key and the length of FixedArray are smis.
__ ldr(ip, FieldMemOperand(elements, FixedArray::kLengthOffset));
__ cmp(key, Operand(ip));
__ b(hs, &slow);
// Calculate key + 1 as smi.
ASSERT_EQ(0, kSmiTag);
__ add(r4, key, Operand(Smi::FromInt(1)));
__ str(r4, FieldMemOperand(receiver, JSArray::kLengthOffset));
__ b(&fast);
// Array case: Get the length and the elements array from the JS
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
// array. Check that the array is in fast mode (and writable); if it
// is the length is always a smi.
__ bind(&array);
__ ldr(elements, FieldMemOperand(receiver, JSObject::kElementsOffset));
__ ldr(r4, FieldMemOperand(elements, HeapObject::kMapOffset));
__ LoadRoot(ip, Heap::kFixedArrayMapRootIndex);
__ cmp(r4, ip);
__ b(ne, &slow);
// Check the key against the length in the array.
__ ldr(ip, FieldMemOperand(receiver, JSArray::kLengthOffset));
__ cmp(key, Operand(ip));
__ b(hs, &extra);
// Fall through to fast case.
__ bind(&fast);
// Fast case, store the value to the elements backing store.
__ add(r5, elements, Operand(FixedArray::kHeaderSize - kHeapObjectTag));
__ add(r5, r5, Operand(key, LSL, kPointerSizeLog2 - kSmiTagSize));
__ str(value, MemOperand(r5));
// Skip write barrier if the written value is a smi.
__ tst(value, Operand(kSmiTagMask));
__ Ret(eq);
// Update write barrier for the elements array address.
__ sub(r4, r5, Operand(elements));
__ RecordWrite(elements, Operand(r4), r5, r6);
__ Ret();
}
// Convert and store int passed in register ival to IEEE 754 single precision
// floating point value at memory location (dst + 4 * wordoffset)
// If VFP3 is available use it for conversion.
static void StoreIntAsFloat(MacroAssembler* masm,
Register dst,
Register wordoffset,
Register ival,
Register fval,
Register scratch1,
Register scratch2) {
if (CpuFeatures::IsSupported(VFP3)) {
CpuFeatures::Scope scope(VFP3);
__ vmov(s0, ival);
__ add(scratch1, dst, Operand(wordoffset, LSL, 2));
__ vcvt_f32_s32(s0, s0);
__ vstr(s0, scratch1, 0);
} else {
Label not_special, done;
// Move sign bit from source to destination. This works because the sign
// bit in the exponent word of the double has the same position and polarity
// as the 2's complement sign bit in a Smi.
ASSERT(kBinary32SignMask == 0x80000000u);
__ and_(fval, ival, Operand(kBinary32SignMask), SetCC);
// Negate value if it is negative.
__ rsb(ival, ival, Operand(0, RelocInfo::NONE), LeaveCC, ne);
// We have -1, 0 or 1, which we treat specially. Register ival contains
// absolute value: it is either equal to 1 (special case of -1 and 1),
// greater than 1 (not a special case) or less than 1 (special case of 0).
__ cmp(ival, Operand(1));
__ b(gt, &not_special);
// For 1 or -1 we need to or in the 0 exponent (biased).
static const uint32_t exponent_word_for_1 =
kBinary32ExponentBias << kBinary32ExponentShift;
__ orr(fval, fval, Operand(exponent_word_for_1), LeaveCC, eq);
__ b(&done);
__ bind(&not_special);
// Count leading zeros.
// Gets the wrong answer for 0, but we already checked for that case above.
Register zeros = scratch2;
__ CountLeadingZeros(zeros, ival, scratch1);
// Compute exponent and or it into the exponent register.
__ rsb(scratch1,
zeros,
Operand((kBitsPerInt - 1) + kBinary32ExponentBias));
__ orr(fval,
fval,
Operand(scratch1, LSL, kBinary32ExponentShift));
// Shift up the source chopping the top bit off.
__ add(zeros, zeros, Operand(1));
// This wouldn't work for 1 and -1 as the shift would be 32 which means 0.
__ mov(ival, Operand(ival, LSL, zeros));
// And the top (top 20 bits).
__ orr(fval,
fval,
Operand(ival, LSR, kBitsPerInt - kBinary32MantissaBits));
__ bind(&done);
__ str(fval, MemOperand(dst, wordoffset, LSL, 2));
}
}
static bool IsElementTypeSigned(ExternalArrayType array_type) {
switch (array_type) {
case kExternalByteArray:
case kExternalShortArray:
case kExternalIntArray:
return true;
case kExternalUnsignedByteArray:
case kExternalUnsignedShortArray:
case kExternalUnsignedIntArray:
return false;
default:
UNREACHABLE();
return false;
}
}
void KeyedStoreIC::GenerateExternalArray(MacroAssembler* masm,
ExternalArrayType array_type) {
// ---------- S t a t e --------------
// -- r0 : value
// -- r1 : key
// -- r2 : receiver
// -- lr : return address
// -----------------------------------
Label slow, check_heap_number;
// Register usage.
Register value = r0;
Register key = r1;
Register receiver = r2;
// r3 mostly holds the elements array or the destination external array.
// Check that the object isn't a smi.
__ BranchOnSmi(receiver, &slow);
// Check that the object is a JS object. Load map into r3.
__ CompareObjectType(receiver, r3, r4, FIRST_JS_OBJECT_TYPE);
__ b(le, &slow);
// Check that the receiver does not require access checks. We need
// to do this because this generic stub does not perform map checks.
__ ldrb(ip, FieldMemOperand(r3, Map::kBitFieldOffset));
__ tst(ip, Operand(1 << Map::kIsAccessCheckNeeded));
__ b(ne, &slow);
// Check that the key is a smi.
__ BranchOnNotSmi(key, &slow);
// Check that the elements array is the appropriate type of ExternalArray.
__ ldr(r3, FieldMemOperand(receiver, JSObject::kElementsOffset));
__ ldr(r4, FieldMemOperand(r3, HeapObject::kMapOffset));
__ LoadRoot(ip, Heap::RootIndexForExternalArrayType(array_type));
__ cmp(r4, ip);
__ b(ne, &slow);
// Check that the index is in range.
__ mov(r4, Operand(key, ASR, kSmiTagSize)); // Untag the index.
__ ldr(ip, FieldMemOperand(r3, ExternalArray::kLengthOffset));
__ cmp(r4, ip);
// Unsigned comparison catches both negative and too-large values.
__ b(hs, &slow);
// Handle both smis and HeapNumbers in the fast path. Go to the
// runtime for all other kinds of values.
// r3: external array.
// r4: key (integer).
__ BranchOnNotSmi(value, &check_heap_number);
__ mov(r5, Operand(value, ASR, kSmiTagSize)); // Untag the value.
__ ldr(r3, FieldMemOperand(r3, ExternalArray::kExternalPointerOffset));
// r3: base pointer of external storage.
// r4: key (integer).
// r5: value (integer).
switch (array_type) {
case kExternalByteArray:
case kExternalUnsignedByteArray:
__ strb(r5, MemOperand(r3, r4, LSL, 0));
break;
case kExternalShortArray:
case kExternalUnsignedShortArray:
__ strh(r5, MemOperand(r3, r4, LSL, 1));
break;
case kExternalIntArray:
case kExternalUnsignedIntArray:
__ str(r5, MemOperand(r3, r4, LSL, 2));
break;
case kExternalFloatArray:
// Perform int-to-float conversion and store to memory.
StoreIntAsFloat(masm, r3, r4, r5, r6, r7, r9);
break;
default:
UNREACHABLE();
break;
}
// Entry registers are intact, r0 holds the value which is the return value.
__ Ret();
// r3: external array.
// r4: index (integer).
__ bind(&check_heap_number);
__ CompareObjectType(value, r5, r6, HEAP_NUMBER_TYPE);
__ b(ne, &slow);
__ ldr(r3, FieldMemOperand(r3, ExternalArray::kExternalPointerOffset));
// r3: base pointer of external storage.
// r4: key (integer).
// The WebGL specification leaves the behavior of storing NaN and
// +/-Infinity into integer arrays basically undefined. For more
// reproducible behavior, convert these to zero.
if (CpuFeatures::IsSupported(VFP3)) {
CpuFeatures::Scope scope(VFP3);
if (array_type == kExternalFloatArray) {
// vldr requires offset to be a multiple of 4 so we can not
// include -kHeapObjectTag into it.
__ sub(r5, r0, Operand(kHeapObjectTag));
__ vldr(d0, r5, HeapNumber::kValueOffset);
__ add(r5, r3, Operand(r4, LSL, 2));
__ vcvt_f32_f64(s0, d0);
__ vstr(s0, r5, 0);
} else {
// Need to perform float-to-int conversion.
// Test for NaN or infinity (both give zero).
__ ldr(r6, FieldMemOperand(r5, HeapNumber::kExponentOffset));
// Hoisted load. vldr requires offset to be a multiple of 4 so we can not
// include -kHeapObjectTag into it.
__ sub(r5, r0, Operand(kHeapObjectTag));
__ vldr(d0, r5, HeapNumber::kValueOffset);
__ Sbfx(r6, r6, HeapNumber::kExponentShift, HeapNumber::kExponentBits);
// NaNs and Infinities have all-one exponents so they sign extend to -1.
__ cmp(r6, Operand(-1));
__ mov(r5, Operand(Smi::FromInt(0)), LeaveCC, eq);
// Not infinity or NaN simply convert to int.
if (IsElementTypeSigned(array_type)) {
__ vcvt_s32_f64(s0, d0, Assembler::RoundToZero, ne);
} else {
__ vcvt_u32_f64(s0, d0, Assembler::RoundToZero, ne);
}
__ vmov(r5, s0, ne);
switch (array_type) {
case kExternalByteArray:
case kExternalUnsignedByteArray:
__ strb(r5, MemOperand(r3, r4, LSL, 0));
break;
case kExternalShortArray:
case kExternalUnsignedShortArray:
__ strh(r5, MemOperand(r3, r4, LSL, 1));
break;
case kExternalIntArray:
case kExternalUnsignedIntArray:
__ str(r5, MemOperand(r3, r4, LSL, 2));
break;
default:
UNREACHABLE();
break;
}
}
// Entry registers are intact, r0 holds the value which is the return value.
__ Ret();
} else {
// VFP3 is not available do manual conversions.
__ ldr(r5, FieldMemOperand(value, HeapNumber::kExponentOffset));
__ ldr(r6, FieldMemOperand(value, HeapNumber::kMantissaOffset));
if (array_type == kExternalFloatArray) {
Label done, nan_or_infinity_or_zero;
static const int kMantissaInHiWordShift =
kBinary32MantissaBits - HeapNumber::kMantissaBitsInTopWord;
static const int kMantissaInLoWordShift =
kBitsPerInt - kMantissaInHiWordShift;
// Test for all special exponent values: zeros, subnormal numbers, NaNs
// and infinities. All these should be converted to 0.
__ mov(r7, Operand(HeapNumber::kExponentMask));
__ and_(r9, r5, Operand(r7), SetCC);
__ b(eq, &nan_or_infinity_or_zero);
__ teq(r9, Operand(r7));
__ mov(r9, Operand(kBinary32ExponentMask), LeaveCC, eq);
__ b(eq, &nan_or_infinity_or_zero);
// Rebias exponent.
__ mov(r9, Operand(r9, LSR, HeapNumber::kExponentShift));
__ add(r9,
r9,
Operand(kBinary32ExponentBias - HeapNumber::kExponentBias));
__ cmp(r9, Operand(kBinary32MaxExponent));
__ and_(r5, r5, Operand(HeapNumber::kSignMask), LeaveCC, gt);
__ orr(r5, r5, Operand(kBinary32ExponentMask), LeaveCC, gt);
__ b(gt, &done);
__ cmp(r9, Operand(kBinary32MinExponent));
__ and_(r5, r5, Operand(HeapNumber::kSignMask), LeaveCC, lt);
__ b(lt, &done);
__ and_(r7, r5, Operand(HeapNumber::kSignMask));
__ and_(r5, r5, Operand(HeapNumber::kMantissaMask));
__ orr(r7, r7, Operand(r5, LSL, kMantissaInHiWordShift));
__ orr(r7, r7, Operand(r6, LSR, kMantissaInLoWordShift));
__ orr(r5, r7, Operand(r9, LSL, kBinary32ExponentShift));
__ bind(&done);
__ str(r5, MemOperand(r3, r4, LSL, 2));
// Entry registers are intact, r0 holds the value which is the return
// value.
__ Ret();
__ bind(&nan_or_infinity_or_zero);
__ and_(r7, r5, Operand(HeapNumber::kSignMask));
__ and_(r5, r5, Operand(HeapNumber::kMantissaMask));
__ orr(r9, r9, r7);
__ orr(r9, r9, Operand(r5, LSL, kMantissaInHiWordShift));
__ orr(r5, r9, Operand(r6, LSR, kMantissaInLoWordShift));
__ b(&done);
} else {
bool is_signed_type = IsElementTypeSigned(array_type);
int meaningfull_bits = is_signed_type ? (kBitsPerInt - 1) : kBitsPerInt;
int32_t min_value = is_signed_type ? 0x80000000 : 0x00000000;
Label done, sign;
// Test for all special exponent values: zeros, subnormal numbers, NaNs
// and infinities. All these should be converted to 0.
__ mov(r7, Operand(HeapNumber::kExponentMask));
__ and_(r9, r5, Operand(r7), SetCC);
__ mov(r5, Operand(0, RelocInfo::NONE), LeaveCC, eq);
__ b(eq, &done);
__ teq(r9, Operand(r7));
__ mov(r5, Operand(0, RelocInfo::NONE), LeaveCC, eq);
__ b(eq, &done);
// Unbias exponent.
__ mov(r9, Operand(r9, LSR, HeapNumber::kExponentShift));
__ sub(r9, r9, Operand(HeapNumber::kExponentBias), SetCC);
// If exponent is negative than result is 0.
__ mov(r5, Operand(0, RelocInfo::NONE), LeaveCC, mi);
__ b(mi, &done);
// If exponent is too big than result is minimal value.
__ cmp(r9, Operand(meaningfull_bits - 1));
__ mov(r5, Operand(min_value), LeaveCC, ge);
__ b(ge, &done);
__ and_(r7, r5, Operand(HeapNumber::kSignMask), SetCC);
__ and_(r5, r5, Operand(HeapNumber::kMantissaMask));
__ orr(r5, r5, Operand(1u << HeapNumber::kMantissaBitsInTopWord));
__ rsb(r9, r9, Operand(HeapNumber::kMantissaBitsInTopWord), SetCC);
__ mov(r5, Operand(r5, LSR, r9), LeaveCC, pl);
__ b(pl, &sign);
__ rsb(r9, r9, Operand(0, RelocInfo::NONE));
__ mov(r5, Operand(r5, LSL, r9));
__ rsb(r9, r9, Operand(meaningfull_bits));
__ orr(r5, r5, Operand(r6, LSR, r9));
__ bind(&sign);
__ teq(r7, Operand(0, RelocInfo::NONE));
__ rsb(r5, r5, Operand(0, RelocInfo::NONE), LeaveCC, ne);
__ bind(&done);
switch (array_type) {
case kExternalByteArray:
case kExternalUnsignedByteArray:
__ strb(r5, MemOperand(r3, r4, LSL, 0));
break;
case kExternalShortArray:
case kExternalUnsignedShortArray:
__ strh(r5, MemOperand(r3, r4, LSL, 1));
break;
case kExternalIntArray:
case kExternalUnsignedIntArray:
__ str(r5, MemOperand(r3, r4, LSL, 2));
break;
default:
UNREACHABLE();
break;
}
}
}
// Slow case: call runtime.
__ bind(&slow);
// Entry registers are intact.
// r0: value
// r1: key
// r2: receiver
GenerateRuntimeSetProperty(masm);
}
void StoreIC::GenerateMegamorphic(MacroAssembler* masm) {
// ----------- S t a t e -------------
// -- r0 : value
// -- r1 : receiver
// -- r2 : name
// -- lr : return address
// -----------------------------------
// Get the receiver from the stack and probe the stub cache.
Code::Flags flags = Code::ComputeFlags(Code::STORE_IC,
NOT_IN_LOOP,
MONOMORPHIC);
StubCache::GenerateProbe(masm, flags, r1, r2, r3, r4, r5);
// Cache miss: Jump to runtime.
GenerateMiss(masm);
}
void StoreIC::GenerateMiss(MacroAssembler* masm) {
// ----------- S t a t e -------------
// -- r0 : value
// -- r1 : receiver
// -- r2 : name
// -- lr : return address
// -----------------------------------
__ Push(r1, r2, r0);
// Perform tail call to the entry.
ExternalReference ref = ExternalReference(IC_Utility(kStoreIC_Miss));
__ TailCallExternalReference(ref, 3, 1);
}
void StoreIC::GenerateArrayLength(MacroAssembler* masm) {
// ----------- S t a t e -------------
// -- r0 : value
// -- r1 : receiver
// -- r2 : name
// -- lr : return address
// -----------------------------------
//
// This accepts as a receiver anything JSObject::SetElementsLength accepts
// (currently anything except for external and pixel arrays which means
// anything with elements of FixedArray type.), but currently is restricted
// to JSArray.
// Value must be a number, but only smis are accepted as the most common case.
Label miss;
Register receiver = r1;
Register value = r0;
Register scratch = r3;
// Check that the receiver isn't a smi.
__ BranchOnSmi(receiver, &miss);
// Check that the object is a JS array.
__ CompareObjectType(receiver, scratch, scratch, JS_ARRAY_TYPE);
__ b(ne, &miss);
// Check that elements are FixedArray.
Copy-on-write arrays. Object model changes ---------------------------------------- New fixed_cow_array_map is used for the elements array of a JSObject to mark it as COW. The JSObject's map and other fields are not affected. The JSObject's map still has the "fast elements" bit set. It means we can do only the receiver map check in keyed loads and the receiver and the elements map checks in keyed stores. So introducing COW arrays doesn't hurt performance of these operations. But note that the elements map check is necessary in all mutating operations because the "has fast elements" bit now means "has fast elements for reading". EnsureWritableFastElements can be used in runtime functions to perform the necessary lazy copying. Generated code changes ---------------------------------------- Generic keyed load is updated to only do the receiver map check (this could have been done earlier). FastCloneShallowArrayStub now has two modes: clone elements and use COW elements. AssertFastElements macro is added to check the elements when necessary. The custom call IC generators for Array.prototype.{push,pop} are updated to avoid going to the slow case (and patching the IC) when calling the builtin should work. COW enablement ---------------------------------------- Currently we only put shallow and simple literal arrays in the COW mode. This is done by the parser. Review URL: http://codereview.chromium.org/3144002 git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@5275 ce2b1a6d-e550-0410-aec6-3dcde31c8c00
2010-08-16 16:06:46 +00:00
// We rely on StoreIC_ArrayLength below to deal with all types of
// fast elements (including COW).
__ ldr(scratch, FieldMemOperand(receiver, JSArray::kElementsOffset));
__ CompareObjectType(scratch, scratch, scratch, FIXED_ARRAY_TYPE);
__ b(ne, &miss);
// Check that value is a smi.
__ BranchOnNotSmi(value, &miss);
// Prepare tail call to StoreIC_ArrayLength.
__ Push(receiver, value);
ExternalReference ref = ExternalReference(IC_Utility(kStoreIC_ArrayLength));
__ TailCallExternalReference(ref, 2, 1);
__ bind(&miss);
GenerateMiss(masm);
}
void StoreIC::GenerateNormal(MacroAssembler* masm) {
// ----------- S t a t e -------------
// -- r0 : value
// -- r1 : receiver
// -- r2 : name
// -- lr : return address
// -----------------------------------
Label miss;
GenerateStringDictionaryReceiverCheck(masm, r1, r3, r4, r5, &miss);
GenerateDictionaryStore(masm, &miss, r3, r2, r0, r4, r5);
__ IncrementCounter(&Counters::store_normal_hit, 1, r4, r5);
__ Ret();
__ bind(&miss);
__ IncrementCounter(&Counters::store_normal_miss, 1, r4, r5);
GenerateMiss(masm);
}
#undef __
} } // namespace v8::internal
#endif // V8_TARGET_ARCH_ARM