More compact string representation on 64 bit.

runtime/vm/raw_object.h

 // Copyright (c) 2012, the Dart project authors.  Please see the AUTHORS file
 // for details. All rights reserved. Use of this source code is governed by a
 // BSD-style license that can be found in the LICENSE file.
 
 #ifndef RUNTIME_VM_RAW_OBJECT_H_
 #define RUNTIME_VM_RAW_OBJECT_H_
 
 #include "platform/assert.h"
 #include "vm/atomic.h"
 #include "vm/exceptions.h"
 #include "vm/globals.h"
 #include "vm/snapshot.h"
 #include "vm/token.h"
 #include "vm/token_position.h"
 
+#if defined(ARCH_IS_64_BIT)
+#define HASH_IN_OBJECT_HEADER 1

[Vyacheslav Egorov (Google), 2017/05/18 06:14:43]

maybe this needs to go to globals.h?

[erikcorry, 2017/05/18 14:28:38]

Done.

+#endif
+
 namespace dart {
 
 // Macrobatics to define the Object hierarchy of VM implementation classes.
 #define CLASS_LIST_NO_OBJECT_NOR_STRING_NOR_ARRAY(V)                           \
   V(Class)                                                                     \
   V(UnresolvedClass)                                                           \
   V(TypeArguments)                                                             \
   V(PatchClass)                                                                \
   V(Function)                                                                  \
   V(ClosureData)                                                               \
@@ skipping 233 matching lines @@
 class RawObject {
  public:
   // The tags field which is a part of the object header uses the following
   // bit fields for storing tags.
   enum TagBits {
     kMarkBit = 0,
     kCanonicalBit = 1,
     kVMHeapObjectBit = 2,
     kRememberedBit = 3,
     kReservedTagPos = 4,  // kReservedBit{100K,1M,10M}
-#if defined(ARCH_IS_32_BIT)
     kReservedTagSize = 4,
     kSizeTagPos = kReservedTagPos + kReservedTagSize,  // = 8
     kSizeTagSize = 8,
     kClassIdTagPos = kSizeTagPos + kSizeTagSize,  // = 16
     kClassIdTagSize = 16,
-#elif defined(ARCH_IS_64_BIT)
-    kReservedTagSize = 12,
-    kSizeTagPos = kReservedTagPos + kReservedTagSize,  // = 16
-    kSizeTagSize = 16,
-    kClassIdTagPos = kSizeTagPos + kSizeTagSize,  // = 32
-    kClassIdTagSize = 32,
-#else
-#error Unexpected architecture word size
-#endif
   };
 
   COMPILE_ASSERT(kClassIdTagSize == (sizeof(classid_t) * kBitsPerByte));
 
   // Encodes the object size in the tag in units of object alignment.
   class SizeTag {
    public:
     static const intptr_t kMaxSizeTag = ((1 << RawObject::kSizeTagSize) - 1)
                                         << kObjectAlignmentLog2;
 
@@ skipping 234 matching lines @@
   intptr_t GetClassId() const {
     uword tags = ptr()->tags_;
     return ClassIdTag::decode(tags);
   }
 
   void SetClassId(intptr_t new_cid) {
     uword tags = ptr()->tags_;
     ptr()->tags_ = ClassIdTag::update(new_cid, tags);
   }
 
+#if defined(HASH_IN_OBJECT_HEADER)
+  int32_t GetHeaderHash() const {
+    ASSERT(IsHeapObject());
+    uword tags = ptr()->tags_;
+    return tags >> 32;
+  }
+
+  void SetHeaderHash(uintptr_t hash) {
+    ASSERT(GetHeaderHash() == 0);
+    ASSERT(hash >> 32 == 0);
+    ptr()->tags_ |= hash << 32ul;
+  }
+#endif
+
   template <class TagBitField>
   void UpdateTagBit(bool value) {
     uword tags = ptr()->tags_;
     uword old_tags;
     do {
       old_tags = tags;
       uword new_tags = TagBitField::update(value, old_tags);
       tags = AtomicOperations::CompareAndSwapWord(&ptr()->tags_, old_tags,
                                                   new_tags);
     } while (tags != old_tags);
@@ skipping 59 matching lines @@
   friend class ClassStatsVisitor;
   template <bool>
   friend class MarkingVisitorBase;
   friend class Mint;
   friend class Object;
   friend class OneByteString;  // StoreSmi
   friend class RawCode;
   friend class RawExternalTypedData;
   friend class RawInstructions;
   friend class RawInstance;
+  friend class RawString;
   friend class RawTypedData;
   friend class Scavenger;
   friend class ScavengerVisitor;
   friend class SizeExcludingClassVisitor;  // GetClassId
   friend class InstanceAccumulator;        // GetClassId
   friend class RetainingPathVisitor;       // GetClassId
   friend class SkippedCodeFunctions;       // StorePointer
   friend class ImageReader;                // tags_ check
   friend class ImageWriter;
   friend class AssemblyImageWriter;
@@ skipping 1186 matching lines @@
   RAW_OBJECT_IMPLEMENTATION(Number);
 };
 
 
 class RawInteger : public RawNumber {
   RAW_OBJECT_IMPLEMENTATION(Integer);
 };
 
 
 class RawSmi : public RawInteger {
+ public:
+  static RawSmi* New(intptr_t value) {

[Vyacheslav Egorov (Google), 2017/05/18 06:14:42]

It's very uncommon to have these here. Especially because it essentially
duplicates Smi::New(...).

[erikcorry, 2017/05/18 14:28:38]

I managed to move them

+    intptr_t raw_smi = (value << kSmiTagShift) | kSmiTag;
+    return reinterpret_cast<RawSmi*>(raw_smi);
+  }
+
+  intptr_t Value() {
+    ASSERT(kSmiTag == 0);
+    return reinterpret_cast<intptr_t>(this) >> kSmiTagShift;
+  }
+
   RAW_OBJECT_IMPLEMENTATION(Smi);
 };
 
 
 class RawMint : public RawInteger {
   RAW_HEAP_OBJECT_IMPLEMENTATION(Mint);
 
   ALIGN8 int64_t value_;
 
   friend class Api;
@@ skipping 23 matching lines @@
 };
 COMPILE_ASSERT(sizeof(RawDouble) == 16);
 
 
 class RawString : public RawInstance {
   RAW_HEAP_OBJECT_IMPLEMENTATION(String);
 
  protected:
   RawObject** from() { return reinterpret_cast<RawObject**>(&ptr()->length_); }
   RawSmi* length_;
+#if !defined(HASH_IN_OBJECT_HEADER)
   RawSmi* hash_;
   RawObject** to() { return reinterpret_cast<RawObject**>(&ptr()->hash_); }
+#else
+  RawObject** to() { return reinterpret_cast<RawObject**>(&ptr()->length_); }
+#endif
+
+ private:
+#if defined(HASH_IN_OBJECT_HEADER)

[Vyacheslav Egorov (Google), 2017/05/18 06:14:43]

I think this needs to go to object.h you can have

static ... GetCachedHash(RawString* str)
static  ... SetCachedHash() there

then you also don't need to introduce RawSmi::New

[erikcorry, 2017/05/18 14:28:38]

Done.

+  uint32_t GetRawHash() const { return GetHeaderHash(); }
+  void SetRawHash(uintptr_t hash) {

[Vyacheslav Egorov (Google), 2017/05/18 06:14:43]

can this be uint32_t hash for consistency with SetRawHash?

[erikcorry, 2017/05/18 14:28:38]

No, then the assert doesn't work.

I want to ensure that the assert catches the case where we generate a hash code
that is out of range, store the truncated version, then return the untruncated
version.

+    ASSERT(hash >> 32 == 0);
+    SetHeaderHash(hash);

[Vyacheslav Egorov (Google), 2017/05/18 06:14:43]

I would better call is SetHashInHeader() or SetHashInTags(...)

[erikcorry, 2017/05/18 14:28:38]

This no longer exists.

+  }
+#else
+  uint32_t GetRawHash() const { return ptr()->hash_->Value(); }
+  void SetRawHash(uint32_t hash) { ptr()->hash_ = RawSmi::New(hash); }
+#endif
 
   friend class Library;
+  friend class OneByteStringSerializationCluster;
+  friend class TwoByteStringSerializationCluster;
+  friend class OneByteStringDeserializationCluster;
+  friend class TwoByteStringDeserializationCluster;
+  friend class RODataSerializationCluster;
 };
 
 
 class RawOneByteString : public RawString {
   RAW_HEAP_OBJECT_IMPLEMENTATION(OneByteString);
 
   // Variable length data follows here.
   uint8_t* data() { OPEN_ARRAY_START(uint8_t, uint8_t); }
   const uint8_t* data() const { OPEN_ARRAY_START(uint8_t, uint8_t); }
 
@@ skipping 566 matching lines @@
 inline intptr_t RawObject::NumberOfTypedDataClasses() {
   // Make sure this is updated when new TypedData types are added.
   COMPILE_ASSERT(kTypedDataInt8ArrayViewCid == kTypedDataInt8ArrayCid + 14);
   COMPILE_ASSERT(kExternalTypedDataInt8ArrayCid ==
                  kTypedDataInt8ArrayViewCid + 15);
   COMPILE_ASSERT(kByteBufferCid == kExternalTypedDataInt8ArrayCid + 14);
   COMPILE_ASSERT(kNullCid == kByteBufferCid + 1);
   return (kNullCid - kTypedDataInt8ArrayCid);
 }
 
+
 }  // namespace dart
 
 #endif  // RUNTIME_VM_RAW_OBJECT_H_