Porting r10023 and r10054 to arm (pointer cache for external strings).

Review URL: http://codereview.chromium.org/8682010

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@10059 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

src/arm/code-stubs-arm.cc

@@ -5107,74 +5107,14 @@ void StringCharCodeAtGenerator::GenerateFast(MacroAssembler* masm) {
   __ cmp(ip, Operand(index_));
   __ b(ls, index_out_of_range_);
 
-  // We need special handling for non-flat strings.
-  STATIC_ASSERT(kSeqStringTag == 0);
-  __ tst(result_, Operand(kStringRepresentationMask));
-  __ b(eq, &flat_string);
+  __ mov(index_, Operand(index_, ASR, kSmiTagSize));
+
+  StringCharLoadGenerator::Generate(masm,
+                                    object_,
+                                    index_,
+                                    result_,
+                                    &call_runtime_);
 
-  // Handle non-flat strings.
-  __ and_(result_, result_, Operand(kStringRepresentationMask));
-  STATIC_ASSERT(kConsStringTag < kExternalStringTag);
-  STATIC_ASSERT(kSlicedStringTag > kExternalStringTag);
-  __ cmp(result_, Operand(kExternalStringTag));
-  __ b(gt, &sliced_string);
-  __ b(eq, &call_runtime_);
-
-  // ConsString.
-  // Check whether the right hand side is the empty string (i.e. if
-  // this is really a flat string in a cons string). If that is not
-  // the case we would rather go to the runtime system now to flatten
-  // the string.
-  Label assure_seq_string;
-  __ ldr(result_, FieldMemOperand(object_, ConsString::kSecondOffset));
-  __ LoadRoot(ip, Heap::kEmptyStringRootIndex);
-  __ cmp(result_, Operand(ip));
-  __ b(ne, &call_runtime_);
-  // Get the first of the two parts.
-  __ ldr(object_, FieldMemOperand(object_, ConsString::kFirstOffset));
-  __ jmp(&assure_seq_string);
-
-  // SlicedString, unpack and add offset.
-  __ bind(&sliced_string);
-  __ ldr(result_, FieldMemOperand(object_, SlicedString::kOffsetOffset));
-  __ add(index_, index_, result_);
-  __ ldr(object_, FieldMemOperand(object_, SlicedString::kParentOffset));
-
-  // Assure that we are dealing with a sequential string. Go to runtime if not.
-  __ bind(&assure_seq_string);
-  __ ldr(result_, FieldMemOperand(object_, HeapObject::kMapOffset));
-  __ ldrb(result_, FieldMemOperand(result_, Map::kInstanceTypeOffset));
-  // Check that parent is not an external string. Go to runtime otherwise.
-  // Note that if the original string is a cons or slice with an external
-  // string as underlying string, we pass that unpacked underlying string with
-  // the adjusted index to the runtime function.
-  STATIC_ASSERT(kSeqStringTag == 0);
-  __ tst(result_, Operand(kStringRepresentationMask));
-  __ b(ne, &call_runtime_);
-
-  // Check for 1-byte or 2-byte string.
-  __ bind(&flat_string);
-  STATIC_ASSERT((kStringEncodingMask & kAsciiStringTag) != 0);
-  STATIC_ASSERT((kStringEncodingMask & kTwoByteStringTag) == 0);
-  __ tst(result_, Operand(kStringEncodingMask));
-  __ b(ne, &ascii_string);
-
-  // 2-byte string.
-  // Load the 2-byte character code into the result register. We can
-  // add without shifting since the smi tag size is the log2 of the
-  // number of bytes in a two-byte character.
-  STATIC_ASSERT(kSmiTag == 0 && kSmiTagSize == 1 && kSmiShiftSize == 0);
-  __ add(index_, object_, Operand(index_));
-  __ ldrh(result_, FieldMemOperand(index_, SeqTwoByteString::kHeaderSize));
-  __ jmp(&got_char_code);
-
-  // ASCII string.
-  // Load the byte into the result register.
-  __ bind(&ascii_string);
-  __ add(index_, object_, Operand(index_, LSR, kSmiTagSize));
-  __ ldrb(result_, FieldMemOperand(index_, SeqAsciiString::kHeaderSize));
-
-  __ bind(&got_char_code);
   __ mov(result_, Operand(result_, LSL, kSmiTagSize));
   __ bind(&exit_);
 }
@@ -5221,6 +5161,7 @@ void StringCharCodeAtGenerator::GenerateSlow(
   // is too complex (e.g., when the string needs to be flattened).
   __ bind(&call_runtime_);
   call_helper.BeforeCall(masm);
+  __ mov(index_, Operand(index_, LSL, kSmiTagSize));
   __ Push(object_, index_);
   __ CallRuntime(Runtime::kStringCharCodeAt, 2);
   __ Move(result_, r0);

src/arm/codegen-arm.cc

@@ -302,6 +302,94 @@ void ElementsTransitionGenerator::GenerateDoubleToObject(
   __ pop(lr);
 }
 
+
+void StringCharLoadGenerator::Generate(MacroAssembler* masm,
+                                       Register string,
+                                       Register index,
+                                       Register result,
+                                       Label* call_runtime) {
+  // Fetch the instance type of the receiver into result register.
+  __ ldr(result, FieldMemOperand(string, HeapObject::kMapOffset));
+  __ ldrb(result, FieldMemOperand(result, Map::kInstanceTypeOffset));
+
+  // We need special handling for indirect strings.
+  Label check_sequential;
+  __ tst(result, Operand(kIsIndirectStringMask));
+  __ b(eq, &check_sequential);
+
+  // Dispatch on the indirect string shape: slice or cons.
+  Label cons_string;
+  __ tst(result, Operand(kSlicedNotConsMask));
+  __ b(eq, &cons_string);
+
+  // Handle slices.
+  Label indirect_string_loaded;
+  __ ldr(result, FieldMemOperand(string, SlicedString::kOffsetOffset));
+  __ add(index, index, Operand(result, ASR, kSmiTagSize));
+  __ ldr(string, FieldMemOperand(string, SlicedString::kParentOffset));
+  __ jmp(&indirect_string_loaded);
+
+  // Handle cons strings.
+  // Check whether the right hand side is the empty string (i.e. if
+  // this is really a flat string in a cons string). If that is not
+  // the case we would rather go to the runtime system now to flatten
+  // the string.
+  __ bind(&cons_string);
+  __ ldr(result, FieldMemOperand(string, ConsString::kSecondOffset));
+  __ LoadRoot(ip, Heap::kEmptyStringRootIndex);
+  __ cmp(result, ip);
+  __ b(ne, call_runtime);
+  // Get the first of the two strings and load its instance type.
+  __ ldr(string, FieldMemOperand(string, ConsString::kFirstOffset));
+
+  __ bind(&indirect_string_loaded);
+  __ ldr(result, FieldMemOperand(string, HeapObject::kMapOffset));
+  __ ldrb(result, FieldMemOperand(result, Map::kInstanceTypeOffset));
+
+  // Distinguish sequential and external strings. Only these two string
+  // representations can reach here (slices and flat cons strings have been
+  // reduced to the underlying sequential or external string).
+  Label external_string, check_encoding;
+  __ bind(&check_sequential);
+  STATIC_ASSERT(kSeqStringTag == 0);
+  __ tst(result, Operand(kStringRepresentationMask));
+  __ b(ne, &external_string);
+
+  // Prepare sequential strings
+  STATIC_ASSERT(SeqTwoByteString::kHeaderSize == SeqAsciiString::kHeaderSize);
+  __ add(string,
+         string,
+         Operand(SeqTwoByteString::kHeaderSize - kHeapObjectTag));
+  __ jmp(&check_encoding);
+
+  // Handle external strings.
+  __ bind(&external_string);
+  if (FLAG_debug_code) {
+    // Assert that we do not have a cons or slice (indirect strings) here.
+    // Sequential strings have already been ruled out.
+    __ tst(result, Operand(kIsIndirectStringMask));
+    __ Assert(eq, "external string expected, but not found");
+  }
+  // Rule out short external strings.
+  STATIC_CHECK(kShortExternalStringTag != 0);
+  __ tst(result, Operand(kShortExternalStringMask));
+  __ b(ne, call_runtime);
+  __ ldr(string, FieldMemOperand(string, ExternalString::kResourceDataOffset));
+
+  Label ascii, done;
+  __ bind(&check_encoding);
+  STATIC_ASSERT(kTwoByteStringTag == 0);
+  __ tst(result, Operand(kStringEncodingMask));
+  __ b(ne, &ascii);
+  // Two-byte string.
+  __ ldrh(result, MemOperand(string, index, LSL, 1));
+  __ jmp(&done);
+  __ bind(&ascii);
+  // Ascii string.
+  __ ldrb(result, MemOperand(string, index));
+  __ bind(&done);
+}
+
 #undef __
 
 } }  // namespace v8::internal

src/arm/codegen-arm.h

@@ -73,6 +73,21 @@ class CodeGenerator: public AstVisitor {
 };
 
 
+class StringCharLoadGenerator : public AllStatic {
+ public:
+  // Generates the code for handling different string types and loading the
+  // indexed character into |result|.  We expect |index| as untagged input and
+  // |result| as untagged output.
+  static void Generate(MacroAssembler* masm,
+                       Register string,
+                       Register index,
+                       Register result,
+                       Label* call_runtime);
+
+ private:
+  DISALLOW_COPY_AND_ASSIGN(StringCharLoadGenerator);
+};
+
 } }  // namespace v8::internal
 
 #endif  // V8_ARM_CODEGEN_ARM_H_

src/arm/lithium-codegen-arm.cc

@@ -3592,86 +3592,14 @@ void LCodeGen::DoStringCharCodeAt(LStringCharCodeAt* instr) {
     LStringCharCodeAt* instr_;
   };
 
-  Register string = ToRegister(instr->string());
-  Register index = ToRegister(instr->index());
-  Register result = ToRegister(instr->result());
-
   DeferredStringCharCodeAt* deferred =
       new DeferredStringCharCodeAt(this, instr);
 
-  // Fetch the instance type of the receiver into result register.
-  __ ldr(result, FieldMemOperand(string, HeapObject::kMapOffset));
-  __ ldrb(result, FieldMemOperand(result, Map::kInstanceTypeOffset));
-
-  // We need special handling for indirect strings.
-  Label check_sequential;
-  __ tst(result, Operand(kIsIndirectStringMask));
-  __ b(eq, &check_sequential);
-
-  // Dispatch on the indirect string shape: slice or cons.
-  Label cons_string;
-  __ tst(result, Operand(kSlicedNotConsMask));
-  __ b(eq, &cons_string);
-
-  // Handle slices.
-  Label indirect_string_loaded;
-  __ ldr(result, FieldMemOperand(string, SlicedString::kOffsetOffset));
-  __ add(index, index, Operand(result, ASR, kSmiTagSize));
-  __ ldr(string, FieldMemOperand(string, SlicedString::kParentOffset));
-  __ jmp(&indirect_string_loaded);
-
-  // Handle conses.
-  // Check whether the right hand side is the empty string (i.e. if
-  // this is really a flat string in a cons string). If that is not
-  // the case we would rather go to the runtime system now to flatten
-  // the string.
-  __ bind(&cons_string);
-  __ ldr(result, FieldMemOperand(string, ConsString::kSecondOffset));
-  __ LoadRoot(ip, Heap::kEmptyStringRootIndex);
-  __ cmp(result, ip);
-  __ b(ne, deferred->entry());
-  // Get the first of the two strings and load its instance type.
-  __ ldr(string, FieldMemOperand(string, ConsString::kFirstOffset));
-
-  __ bind(&indirect_string_loaded);
-  __ ldr(result, FieldMemOperand(string, HeapObject::kMapOffset));
-  __ ldrb(result, FieldMemOperand(result, Map::kInstanceTypeOffset));
-
-  // Check whether the string is sequential. The only non-sequential
-  // shapes we support have just been unwrapped above.
-  // Note that if the original string is a cons or slice with an external
-  // string as underlying string, we pass that unpacked underlying string with
-  // the adjusted index to the runtime function.
-  __ bind(&check_sequential);
-  STATIC_ASSERT(kSeqStringTag == 0);
-  __ tst(result, Operand(kStringRepresentationMask));
-  __ b(ne, deferred->entry());
-
-  // Dispatch on the encoding: ASCII or two-byte.
-  Label ascii_string;
-  STATIC_ASSERT((kStringEncodingMask & kAsciiStringTag) != 0);
-  STATIC_ASSERT((kStringEncodingMask & kTwoByteStringTag) == 0);
-  __ tst(result, Operand(kStringEncodingMask));
-  __ b(ne, &ascii_string);
-
-  // Two-byte string.
-  // Load the two-byte character code into the result register.
-  Label done;
-  __ add(result,
-         string,
-         Operand(SeqTwoByteString::kHeaderSize - kHeapObjectTag));
-  __ ldrh(result, MemOperand(result, index, LSL, 1));
-  __ jmp(&done);
-
-  // ASCII string.
-  // Load the byte into the result register.
-  __ bind(&ascii_string);
-  __ add(result,
-         string,
-         Operand(SeqAsciiString::kHeaderSize - kHeapObjectTag));
-  __ ldrb(result, MemOperand(result, index));
-
-  __ bind(&done);
+  StringCharLoadGenerator::Generate(masm(),
+                                    ToRegister(instr->string()),
+                                    ToRegister(instr->index()),
+                                    ToRegister(instr->result()),
+                                    deferred->entry());
   __ bind(deferred->exit());
 }
 

src/ia32/codegen-ia32.cc

@@ -538,12 +538,12 @@ void StringCharLoadGenerator::Generate(MacroAssembler* masm,
   // We need special handling for indirect strings.
   Label check_sequential;
   __ test(result, Immediate(kIsIndirectStringMask));
-  __ j(zero, &check_sequential);
+  __ j(zero, &check_sequential, Label::kNear);
 
   // Dispatch on the indirect string shape: slice or cons.
   Label cons_string;
   __ test(result, Immediate(kSlicedNotConsMask));
-  __ j(zero, &cons_string);
+  __ j(zero, &cons_string, Label::kNear);
 
   // Handle slices.
   Label indirect_string_loaded;
@@ -551,33 +551,7 @@ void StringCharLoadGenerator::Generate(MacroAssembler* masm,
   __ SmiUntag(result);
   __ add(index, result);
   __ mov(string, FieldOperand(string, SlicedString::kParentOffset));
-  __ jmp(&indirect_string_loaded);
-
-  // Handle external strings.
-  Label external_string, ascii_external, done;
-  __ bind(&external_string);
-  if (FLAG_debug_code) {
-    // Assert that we do not have a cons or slice (indirect strings) here.
-    // Sequential strings have already been ruled out.
-    __ test(result, Immediate(kIsIndirectStringMask));
-    __ Assert(zero, "external string expected, but not found");
-  }
-  // Rule out short external strings.
-  STATIC_CHECK(kShortExternalStringTag != 0);
-  __ test_b(result, kShortExternalStringMask);
-  __ j(not_zero, call_runtime);
-  // Check encoding.
-  STATIC_ASSERT(kTwoByteStringTag == 0);
-  __ test_b(result, kStringEncodingMask);
-  __ mov(result, FieldOperand(string, ExternalString::kResourceDataOffset));
-  __ j(not_equal, &ascii_external, Label::kNear);
-  // Two-byte string.
-  __ movzx_w(result, Operand(result, index, times_2, 0));
-  __ jmp(&done);
-  __ bind(&ascii_external);
-  // Ascii string.
-  __ movzx_b(result, Operand(result, index, times_1, 0));
-  __ jmp(&done);
+  __ jmp(&indirect_string_loaded, Label::kNear);
 
   // Handle cons strings.
   // Check whether the right hand side is the empty string (i.e. if
@@ -597,21 +571,47 @@ void StringCharLoadGenerator::Generate(MacroAssembler* masm,
   // Distinguish sequential and external strings. Only these two string
   // representations can reach here (slices and flat cons strings have been
   // reduced to the underlying sequential or external string).
+  Label seq_string;
   __ bind(&check_sequential);
   STATIC_ASSERT(kSeqStringTag == 0);
   __ test(result, Immediate(kStringRepresentationMask));
-  __ j(not_zero, &external_string);
+  __ j(zero, &seq_string, Label::kNear);
+
+  // Handle external strings.
+  Label ascii_external, done;
+  if (FLAG_debug_code) {
+    // Assert that we do not have a cons or slice (indirect strings) here.
+    // Sequential strings have already been ruled out.
+    __ test(result, Immediate(kIsIndirectStringMask));
+    __ Assert(zero, "external string expected, but not found");
+  }
+  // Rule out short external strings.
+  STATIC_CHECK(kShortExternalStringTag != 0);
+  __ test_b(result, kShortExternalStringMask);
+  __ j(not_zero, call_runtime);
+  // Check encoding.
+  STATIC_ASSERT(kTwoByteStringTag == 0);
+  __ test_b(result, kStringEncodingMask);
+  __ mov(result, FieldOperand(string, ExternalString::kResourceDataOffset));
+  __ j(not_equal, &ascii_external, Label::kNear);
+  // Two-byte string.
+  __ movzx_w(result, Operand(result, index, times_2, 0));
+  __ jmp(&done, Label::kNear);
+  __ bind(&ascii_external);
+  // Ascii string.
+  __ movzx_b(result, Operand(result, index, times_1, 0));
+  __ jmp(&done, Label::kNear);
 
   // Dispatch on the encoding: ASCII or two-byte.
-  Label ascii_string;
+  Label ascii;
+  __ bind(&seq_string);
   STATIC_ASSERT((kStringEncodingMask & kAsciiStringTag) != 0);
   STATIC_ASSERT((kStringEncodingMask & kTwoByteStringTag) == 0);
   __ test(result, Immediate(kStringEncodingMask));
-  __ j(not_zero, &ascii_string, Label::kNear);
+  __ j(not_zero, &ascii, Label::kNear);
 
   // Two-byte string.
   // Load the two-byte character code into the result register.
-  STATIC_ASSERT(kSmiTag == 0 && kSmiTagSize == 1);
   __ movzx_w(result, FieldOperand(string,
                                   index,
                                   times_2,
@@ -620,7 +620,7 @@ void StringCharLoadGenerator::Generate(MacroAssembler* masm,
 
   // Ascii string.
   // Load the byte into the result register.
-  __ bind(&ascii_string);
+  __ bind(&ascii);
   __ movzx_b(result, FieldOperand(string,
                                   index,
                                   times_1,

src/x64/codegen-x64.cc

@@ -380,7 +380,7 @@ void StringCharLoadGenerator::Generate(MacroAssembler* masm,
   // We need special handling for indirect strings.
   Label check_sequential;
   __ testb(result, Immediate(kIsIndirectStringMask));
-  __ j(zero, &check_sequential);
+  __ j(zero, &check_sequential, Label::kNear);
 
   // Dispatch on the indirect string shape: slice or cons.
   Label cons_string;
@@ -394,32 +394,6 @@ void StringCharLoadGenerator::Generate(MacroAssembler* masm,
   __ movq(string, FieldOperand(string, SlicedString::kParentOffset));
   __ jmp(&indirect_string_loaded, Label::kNear);
 
-  // Handle external strings.
-  Label external_string, ascii_external, done;
-  __ bind(&external_string);
-  if (FLAG_debug_code) {
-    // Assert that we do not have a cons or slice (indirect strings) here.
-    // Sequential strings have already been ruled out.
-    __ testb(result, Immediate(kIsIndirectStringMask));
-    __ Assert(zero, "external string expected, but not found");
-  }
-  // Rule out short external strings.
-  STATIC_CHECK(kShortExternalStringTag != 0);
-  __ testb(result, Immediate(kShortExternalStringTag));
-  __ j(not_zero, call_runtime);
-  // Check encoding.
-  STATIC_ASSERT(kTwoByteStringTag == 0);
-  __ testb(result, Immediate(kStringEncodingMask));
-  __ movq(result, FieldOperand(string, ExternalString::kResourceDataOffset));
-  __ j(not_equal, &ascii_external, Label::kNear);
-  // Two-byte string.
-  __ movzxwl(result, Operand(result, index, times_2, 0));
-  __ jmp(&done);
-  __ bind(&ascii_external);
-  // Ascii string.
-  __ movzxbl(result, Operand(result, index, times_1, 0));
-  __ jmp(&done);
-
   // Handle cons strings.
   // Check whether the right hand side is the empty string (i.e. if
   // this is really a flat string in a cons string). If that is not
@@ -438,17 +412,44 @@ void StringCharLoadGenerator::Generate(MacroAssembler* masm,
   // Distinguish sequential and external strings. Only these two string
   // representations can reach here (slices and flat cons strings have been
   // reduced to the underlying sequential or external string).
+  Label seq_string;
   __ bind(&check_sequential);
   STATIC_ASSERT(kSeqStringTag == 0);
   __ testb(result, Immediate(kStringRepresentationMask));
-  __ j(not_zero, &external_string);
+  __ j(zero, &seq_string, Label::kNear);
+
+  // Handle external strings.
+  Label ascii_external, done;
+  if (FLAG_debug_code) {
+    // Assert that we do not have a cons or slice (indirect strings) here.
+    // Sequential strings have already been ruled out.
+    __ testb(result, Immediate(kIsIndirectStringMask));
+    __ Assert(zero, "external string expected, but not found");
+  }
+  // Rule out short external strings.
+  STATIC_CHECK(kShortExternalStringTag != 0);
+  __ testb(result, Immediate(kShortExternalStringTag));
+  __ j(not_zero, call_runtime);
+  // Check encoding.
+  STATIC_ASSERT(kTwoByteStringTag == 0);
+  __ testb(result, Immediate(kStringEncodingMask));
+  __ movq(result, FieldOperand(string, ExternalString::kResourceDataOffset));
+  __ j(not_equal, &ascii_external, Label::kNear);
+  // Two-byte string.
+  __ movzxwl(result, Operand(result, index, times_2, 0));
+  __ jmp(&done, Label::kNear);
+  __ bind(&ascii_external);
+  // Ascii string.
+  __ movzxbl(result, Operand(result, index, times_1, 0));
+  __ jmp(&done, Label::kNear);
 
   // Dispatch on the encoding: ASCII or two-byte.
-  Label ascii_string;
+  Label ascii;
+  __ bind(&seq_string);
   STATIC_ASSERT((kStringEncodingMask & kAsciiStringTag) != 0);
   STATIC_ASSERT((kStringEncodingMask & kTwoByteStringTag) == 0);
   __ testb(result, Immediate(kStringEncodingMask));
-  __ j(not_zero, &ascii_string, Label::kNear);
+  __ j(not_zero, &ascii, Label::kNear);
 
   // Two-byte string.
   // Load the two-byte character code into the result register.
@@ -461,7 +462,7 @@ void StringCharLoadGenerator::Generate(MacroAssembler* masm,
 
   // ASCII string.
   // Load the byte into the result register.
-  __ bind(&ascii_string);
+  __ bind(&ascii);
   __ movzxbl(result, FieldOperand(string,
                                   index,
                                   times_1,