More compact string representation on 64 bit.

runtime/vm/raw_object.h

Index: runtime/vm/raw_object.h
diff --git a/runtime/vm/raw_object.h b/runtime/vm/raw_object.h
index 95288b1e1407513aceb2514fe2abea4016cd6c76..dc0aa7e7bc3a50767c89e435015ddb7879163623 100644
--- a/runtime/vm/raw_object.h
+++ b/runtime/vm/raw_object.h
@@ -13,6 +13,10 @@
 #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.
@@ -266,21 +270,11 @@ class RawObject {
     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));
@@ -535,6 +529,20 @@ class RawObject {
     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_;
@@ -614,6 +622,7 @@ class RawObject {
   friend class RawExternalTypedData;
   friend class RawInstructions;
   friend class RawInstance;
+  friend class RawString;
   friend class RawTypedData;
   friend class Scavenger;
   friend class ScavengerVisitor;
@@ -1820,6 +1829,17 @@ class RawInteger : public RawNumber {
 
 
 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);
 };
 
@@ -1863,10 +1883,31 @@ class RawString : public RawInstance {
  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;
 };
 
 
@@ -2453,6 +2494,7 @@ inline intptr_t RawObject::NumberOfTypedDataClasses() {
   return (kNullCid - kTypedDataInt8ArrayCid);
 }
 
+
 }  // namespace dart
 
 #endif  // RUNTIME_VM_RAW_OBJECT_H_