Merge pull request #4 from JetBrains/mem_mgmt

Add memory mgmt.

backend.native/compiler/ir/backend.native/src/org/jetbrains/kotlin/backend/native/llvm/LlvmUtils.kt

@@ -39,6 +39,10 @@ internal open class Struct(val type: LLVMOpaqueType?, val elements: List<Compile
     }
 }
 
+internal class Int8(val value: Byte) : CompileTimeValue() {
+    override fun getLlvmValue() = LLVMConstInt(LLVMInt8Type(), value.toLong(), 1)
+}
+
 internal class Int32(val value: Int) : CompileTimeValue() {
     override fun getLlvmValue() = LLVMConstInt(LLVMInt32Type(), value.toLong(), 1)
 }
@@ -63,11 +67,11 @@ internal fun pointerType(pointeeType: LLVMOpaqueType?) = LLVMPointerType(pointee
 internal fun getLlvmFunctionType(function: FunctionDescriptor): LLVMOpaqueType? {
     val returnType = getLLVMType(function.returnType!!)
     val params = function.dispatchReceiverParameter.singletonOrEmptyList() +
-                 function.extensionReceiverParameter.singletonOrEmptyList() +
-                 function.valueParameters
+            function.extensionReceiverParameter.singletonOrEmptyList() +
+            function.valueParameters
 
     val paramTypes = params.map { getLLVMType(it.type) }.toTypedArray()
 
     val paramTypesPtr = mallocNativeArrayOf(LLVMOpaqueType, *paramTypes)[0] // TODO: dispose
     return LLVMFunctionType(returnType, paramTypesPtr, paramTypes.size, 0)
-}
+}

backend.native/compiler/ir/backend.native/src/org/jetbrains/kotlin/backend/native/llvm/RTTIGenerator.kt

@@ -39,7 +39,7 @@ internal class RTTIGenerator(override val context: Context) : ContextUtils {
             Struct(
                     runtime.typeInfoType,
 
-                    Int64(name),
+                    Struct(runtime.globalhHashType, ConstArray(LLVMInt8Type(), Array(20, {i -> Int8(1)}).toList())),
                     Int32(size),
 
                     superType,
@@ -102,7 +102,7 @@ internal class RTTIGenerator(override val context: Context) : ContextUtils {
 
     private fun getMethodTableEntries(classDesc: ClassDescriptor): List<FunctionDescriptor> {
         val contributedDescriptors = classDesc.unsubstitutedMemberScope.getContributedDescriptors()
-         // (includes declarations from supers)
+        // (includes declarations from supers)
 
         val functions = contributedDescriptors.filterIsInstance<FunctionDescriptor>()
 
@@ -166,16 +166,16 @@ internal class RTTIGenerator(override val context: Context) : ContextUtils {
         val methodsPtr = addGlobalConstArray("kmethods:$className", runtime.methodTableRecordType, methods)
 
         val typeInfo = TypeInfo(name, size,
-                                superType,
-                                objOffsetsPtr, objOffsets.size,
-                                interfacesPtr, interfaces.size,
-                                vtablePtr,
-                                methodsPtr, methods.size,
-                                fieldsPtr, fields.size)
+                superType,
+                objOffsetsPtr, objOffsets.size,
+                interfacesPtr, interfaces.size,
+                vtablePtr,
+                methodsPtr, methods.size,
+                fieldsPtr, fields.size)
 
         val typeInfoGlobal = classDesc.llvmTypeInfoPtr.getLlvmValue() // TODO: it is a hack
         LLVMSetInitializer(typeInfoGlobal, typeInfo.getLlvmValue())
         LLVMSetGlobalConstant(typeInfoGlobal, 1)
     }
 
-}
+}

backend.native/compiler/ir/backend.native/src/org/jetbrains/kotlin/backend/native/llvm/Runtime.kt

@@ -33,6 +33,7 @@ class Runtime(private val bitcodeFile: String) {
     val typeInfoType = LLVMGetTypeByName(llvmModule, "struct.TypeInfo")
     val fieldTableRecordType = LLVMGetTypeByName(llvmModule, "struct.FieldTableRecord")
     val methodTableRecordType = LLVMGetTypeByName(llvmModule, "struct.MethodTableRecord")
+    val globalhHashType = LLVMGetTypeByName(llvmModule, "struct.GlobalHash")
 
     val target = LLVMGetTarget(llvmModule)!!.asCString().toString()
 
@@ -40,4 +41,4 @@ class Runtime(private val bitcodeFile: String) {
 
     val targetData = LLVMCreateTargetData(dataLayout)
 
-}
+}

backend.native/tests/codegen/klass/basic.kt

@@ -0,0 +1,3 @@
+class Foo {
+  var x : Int = 3
+}

runtime/src/main/cpp/City.cpp

@@ -0,0 +1,218 @@
+#include "City.h"
+
+#include <string.h>
+
+#include <algorithm>
+
+namespace {
+
+// Some primes between 2^63 and 2^64 for various uses.
+static const uint64_t k0 = 0xc3a5c85c97cb3127ULL;
+static const uint64_t k1 = 0xb492b66fbe98f273ULL;
+static const uint64_t k2 = 0x9ae16a3b2f90404fULL;
+
+// Magic numbers for 32-bit hashing.  Copied from Murmur3.
+static const uint32_t c1 = 0xcc9e2d51;
+static const uint32_t c2 = 0x1b873593;
+
+uint64_t UNALIGNED_LOAD64(const char *p) {
+  uint64_t result;
+  memcpy(&result, p, sizeof(result));
+  return result;
+}
+
+uint32_t UNALIGNED_LOAD32(const char *p) {
+  uint32_t result;
+  memcpy(&result, p, sizeof(result));
+  return result;
+}
+
+#define bswap32(x) __builtin_bswap32(x)
+#define bswap64(x) __builtin_bswap64(x)
+
+#ifdef WORDS_BIGENDIAN
+#define uint32_in_expected_order(x) (bswap32(x))
+#define uint64_in_expected_order(x) (bswap64(x))
+#else
+#define uint32_in_expected_order(x) (x)
+#define uint64_in_expected_order(x) (x)
+#endif
+
+uint64_t Fetch64(const char *p) {
+  return uint64_in_expected_order(UNALIGNED_LOAD64(p));
+}
+
+uint32_t Fetch32(const char *p) {
+  return uint32_in_expected_order(UNALIGNED_LOAD32(p));
+}
+
+uint32_t Rotate32(uint32_t val, int shift) {
+  // Avoid shifting by 32: doing so yields an undefined result.
+  return shift == 0 ? val : ((val >> shift) | (val << (32 - shift)));
+}
+
+// Bitwise right rotate.  Normally this will compile to a single
+// instruction, especially if the shift is a manifest constant.
+uint64_t Rotate(uint64_t val, int shift) {
+  // Avoid shifting by 64: doing so yields an undefined result.
+  return shift == 0 ? val : ((val >> shift) | (val << (64 - shift)));
+}
+
+uint64_t ShiftMix(uint64_t val) {
+  return val ^ (val >> 47);
+}
+
+uint64_t HashLen16(uint64_t u, uint64_t v, uint64_t mul) {
+  // Murmur-inspired hashing.
+  uint64_t a = (u ^ v) * mul;
+  a ^= (a >> 47);
+  uint64_t b = (v ^ a) * mul;
+  b ^= (b >> 47);
+  b *= mul;
+  return b;
+}
+
+typedef std::pair<uint64_t, uint64_t> uint128_t;
+
+uint64_t Hash128to64(const uint128_t& x) {
+  // Murmur-inspired hashing.
+  const uint64_t kMul = 0x9ddfea08eb382d69ULL;
+  uint64_t a = (x.first ^ x.second) * kMul;
+  a ^= (a >> 47);
+  uint64_t b = (x.second ^ a) * kMul;
+  b ^= (b >> 47);
+  b *= kMul;
+  return b;
+}
+
+uint64_t HashLen16(uint64_t u, uint64_t v) {
+  return Hash128to64(uint128_t(u, v));
+}
+
+uint64_t HashLen0to16(const char *s, size_t len) {
+  if (len >= 8) {
+    uint64_t mul = k2 + len * 2;
+    uint64_t a = Fetch64(s) + k2;
+    uint64_t b = Fetch64(s + len - 8);
+    uint64_t c = Rotate(b, 37) * mul + a;
+    uint64_t d = (Rotate(a, 25) + b) * mul;
+    return HashLen16(c, d, mul);
+  }
+  if (len >= 4) {
+    uint64_t mul = k2 + len * 2;
+    uint64_t a = Fetch32(s);
+    return HashLen16(len + (a << 3), Fetch32(s + len - 4), mul);
+  }
+  if (len > 0) {
+    uint8_t a = s[0];
+    uint8_t b = s[len >> 1];
+    uint8_t c = s[len - 1];
+    uint32_t y = static_cast<uint32_t>(a) + (static_cast<uint32_t>(b) << 8);
+    uint32_t z = len + (static_cast<uint32_t>(c) << 2);
+    return ShiftMix(y * k2 ^ z * k0) * k2;
+  }
+  return k2;
+}
+
+// This probably works well for 16-byte strings as well, but it may be overkill
+// in that case.
+static uint64_t HashLen17to32(const char *s, size_t len) {
+  uint64_t mul = k2 + len * 2;
+  uint64_t a = Fetch64(s) * k1;
+  uint64_t b = Fetch64(s + 8);
+  uint64_t c = Fetch64(s + len - 8) * mul;
+  uint64_t d = Fetch64(s + len - 16) * k2;
+  return HashLen16(Rotate(a + b, 43) + Rotate(c, 30) + d,
+                   a + Rotate(b + k2, 18) + c, mul);
+}
+
+// Return a 16-byte hash for 48 bytes.  Quick and dirty.
+// Callers do best to use "random-looking" values for a and b.
+std::pair<uint64_t, uint64_t> WeakHashLen32WithSeeds(
+    uint64_t w, uint64_t x, uint64_t y, uint64_t z, uint64_t a, uint64_t b) {
+  a += w;
+  b = Rotate(b + a + z, 21);
+  uint64_t c = a;
+  a += x;
+  a += y;
+  b += Rotate(a, 44);
+  return std::make_pair(a + z, b + c);
+}
+
+// Return a 16-byte hash for s[0] ... s[31], a, and b.  Quick and dirty.
+std::pair<uint64_t, uint64_t> WeakHashLen32WithSeeds(
+    const char* s, uint64_t a, uint64_t b) {
+  return WeakHashLen32WithSeeds(Fetch64(s),
+                                Fetch64(s + 8),
+                                Fetch64(s + 16),
+                                Fetch64(s + 24),
+                                a,
+                                b);
+}
+
+// Return an 8-byte hash for 33 to 64 bytes.
+uint64_t HashLen33to64(const char *s, size_t len) {
+  uint64_t mul = k2 + len * 2;
+  uint64_t a = Fetch64(s) * k2;
+  uint64_t b = Fetch64(s + 8);
+  uint64_t c = Fetch64(s + len - 24);
+  uint64_t d = Fetch64(s + len - 32);
+  uint64_t e = Fetch64(s + 16) * k2;
+  uint64_t f = Fetch64(s + 24) * 9;
+  uint64_t g = Fetch64(s + len - 8);
+  uint64_t h = Fetch64(s + len - 16) * mul;
+  uint64_t u = Rotate(a + g, 43) + (Rotate(b, 30) + c) * 9;
+  uint64_t v = ((a + g) ^ d) + f + 1;
+  uint64_t w = bswap64((u + v) * mul) + h;
+  uint64_t x = Rotate(e + f, 42) + c;
+  uint64_t y = (bswap64((v + w) * mul) + g) * mul;
+  uint64_t z = e + f + c;
+  a = bswap64((x + z) * mul + y) + b;
+  b = ShiftMix((z + a) * mul + d + h) * mul;
+  return b + x;
+}
+
+} // namespace
+
+extern "C" {
+
+uint64_t CityHash64(const void* data, size_t len) {
+  const char* s = reinterpret_cast<const char*>(data);
+  if (len <= 32) {
+    if (len <= 16) {
+      return HashLen0to16(s, len);
+    } else {
+      return HashLen17to32(s, len);
+    }
+  } else if (len <= 64) {
+    return HashLen33to64(s, len);
+  }
+
+  // For strings over 64 bytes we hash the end first, and then as we
+  // loop we keep 56 bytes of state: v, w, x, y, and z.
+  uint64_t x = Fetch64(s + len - 40);
+  uint64_t y = Fetch64(s + len - 16) + Fetch64(s + len - 56);
+  uint64_t z = HashLen16(Fetch64(s + len - 48) + len, Fetch64(s + len - 24));
+  std::pair<uint64_t, uint64_t> v = WeakHashLen32WithSeeds(s + len - 64, len, z);
+  std::pair<uint64_t, uint64_t> w = WeakHashLen32WithSeeds(s + len - 32, y + k1, x);
+  x = x * k1 + Fetch64(s);
+
+  // Decrease len to the nearest multiple of 64, and operate on 64-byte chunks.
+  len = (len - 1) & ~static_cast<size_t>(63);
+  do {
+    x = Rotate(x + y + v.first + Fetch64(s + 8), 37) * k1;
+    y = Rotate(y + v.second + Fetch64(s + 48), 42) * k1;
+    x ^= w.second;
+    y += v.first + Fetch64(s + 40);
+    z = Rotate(z + w.first, 33) * k1;
+    v = WeakHashLen32WithSeeds(s, v.second * k1, x + w.first);
+    w = WeakHashLen32WithSeeds(s + 32, z + w.second, y + Fetch64(s + 16));
+    std::swap(z, x);
+    s += 64;
+    len -= 64;
+  } while (len != 0);
+  return HashLen16(HashLen16(v.first, w.first) + ShiftMix(y) * k1 + z,
+                   HashLen16(v.second, w.second) + x);
+}
+
+} // extern "C"

runtime/src/main/cpp/City.h

@@ -0,0 +1,20 @@
+#ifndef RUNTIME_CITY_H
+#define RUNTIME_CITY_H
+
+// CityHash, by Geoff Pike and Jyrki Alakuijala.
+
+#include <stddef.h>
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+// Hash function for a byte array.
+uint64_t CityHash64(const void* buf, size_t len);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // RUNTIME_CITY_H

runtime/src/main/cpp/Memory.cpp

@@ -0,0 +1,68 @@
+#include <stdlib.h>
+
+#include <cassert>
+
+#include "Memory.h"
+
+void FreeObject(ContainerHeader* header) {
+  free(header);
+}
+
+ArenaContainer::ArenaContainer(uint32_t size) {
+    ArenaContainerHeader* header = reinterpret_cast<ArenaContainerHeader*>(
+        calloc(size + sizeof(ArenaContainerHeader), 1));
+    header_ = header;
+    header->ref_count_ = 1;
+    header->current_ = reinterpret_cast<uint8_t*>(header_) + sizeof(ArenaContainerHeader);
+    header->end_ = header->current_ + size;
+}
+
+void ObjectContainer::Init(const TypeInfo* type_info, uint32_t elements) {
+  header_ = reinterpret_cast<ContainerHeader*>(
+      calloc(sizeof(ContainerHeader) + sizeof(ObjHeader) +
+             type_info->size_ * elements, 1));
+  header_->ref_count_ = 1;
+  SetMeta(GetPlace(), type_info);
+}
+
+ObjHeader* ArenaContainer::PlaceObject(const TypeInfo* type_info) {
+  int size = type_info->size_ + sizeof(ObjHeader);
+  ObjHeader* result = reinterpret_cast<ObjHeader*>(Place(size));
+  if (!result) {
+      return nullptr;
+  }
+  SetMeta(result, type_info);
+  return result;
+}
+
+ArrayHeader* ArenaContainer::PlaceArray(const TypeInfo* type_info, int count) {
+  int size = sizeof(ArrayHeader) + type_info->size_ * count;
+  ArrayHeader* result = reinterpret_cast<ArrayHeader*>(Place(size));
+  if (!result) {
+    return nullptr;
+  }
+  SetMeta(result, type_info);
+  result->count_ = count;
+  return result;
+}
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void InitMemory() {
+  // TODO: initialize heap here.
+}
+
+// Now we ignore all placement hints and always allocate heap space for new object.
+void* AllocInstance(const TypeInfo* type_info, PlacementHint hint) {
+  return ObjectContainer(type_info).GetPlace();
+}
+
+void* AllocArrayInstance(const TypeInfo* type_info, PlacementHint hint, uint32_t elements) {
+  return ObjectContainer(type_info, elements).GetPlace();
+}
+
+#ifdef __cplusplus
+}
+#endif

runtime/src/main/cpp/Memory.h

@@ -0,0 +1,313 @@
+#ifndef RUNTIME_MEMORY_H
+#define RUNTIME_MEMORY_H
+
+#include <cassert>
+
+#include "TypeInfo.h"
+
+typedef enum {
+  FRAME_SCOPE = 0,
+  GLOBAL_SCOPE = 1,
+  ARENA_SCOPE = 2
+} PlacementHint;
+
+// Could be made 64-bit for large memory configs.
+typedef uint32_t container_offset_t;
+
+// Header of every object.
+struct ObjHeader {
+  const TypeInfo* type_info_;
+  container_offset_t container_offset_negative_;
+};
+
+// Header of value type array objects.
+struct ArrayHeader : public ObjHeader {
+  uint32_t count_;
+};
+
+// Header of all container objects. Contains reference counter.
+struct ContainerHeader {
+  // Reference counter of container. Maybe use some upper bit of counter for
+  // container type (for polymorphism in ::Release()).
+  uint32_t ref_count_;
+};
+
+struct ArenaContainerHeader : public ContainerHeader {
+  // Current allocation limit.
+  uint8_t* current_;
+  // Allocation end. Maybe consider having chunked backing storage
+  // at cost of smarter ::Release() polymorphic on container type.
+  uint8_t* end_;
+};
+
+// Thos two operations are implemented by translator when storing references
+// to objects.
+inline void AddRef(ContainerHeader* header) {
+  // Looking at container type we may want to skip AddRef() totally
+  // (non-escaping stack objects).
+  header->ref_count_++;
+}
+
+void FreeObject(ContainerHeader* header);
+
+inline void Release(ContainerHeader* header) {
+  // Looking at container type we may want to skip Release() totally
+  // (non-escaping stack objects, permanent objects).
+  if (--header->ref_count_ == 0) {
+    FreeObject(header);
+  }
+}
+
+// Class representing arbitrary placement container.
+class Container {
+ protected:
+  // Data where everything is being stored.
+  ContainerHeader* header_;
+
+  void SetMeta(ObjHeader* obj, const TypeInfo* type_info) {
+    obj->container_offset_negative_ =
+        reinterpret_cast<uintptr_t>(obj) - reinterpret_cast<uintptr_t>(header_);
+    obj->type_info_ = type_info;
+  }
+
+ public:
+  // Increment reference counter associated with container.
+  void AddRef() {
+    if (header_) ::AddRef(header_);
+  }
+
+  // Decrement reference counter associated with container.
+  // For objects whith tricky lifetime (such as ones shared between threads objects)
+  // individual container per object (ObjectContainer) shall be created.
+  // As an alternative, such objects could be evacuated from short-lived containers.
+  void Release() {
+    if (header_) ::Release(header_);
+  }
+};
+
+// Container for a single object.
+class ObjectContainer : public Container {
+ public:
+  explicit ObjectContainer(const TypeInfo* type_info) {
+    Init(type_info, 1);
+  }
+
+  ObjectContainer(const TypeInfo* type_info, uint32_t elements) {
+    Init(type_info, elements);
+  }
+
+  // Object container shalln't have any dtor, as it's being freed by ::Release().
+  ObjHeader* GetPlace() const {
+    return reinterpret_cast<ObjHeader*>(
+        reinterpret_cast<uint8_t*>(header_) + sizeof(ContainerHeader));
+  }
+
+ private:
+  void Init(const TypeInfo* type_info, uint32_t elements);
+};
+
+// Class representing arena-style placement container.
+// Container is used for reference counting,
+// and it is assumed that objects with related placement will share container. Only
+// whole container can be freed, individual objects are not taken into account.
+class ArenaContainer : public Container {
+ public:
+  explicit ArenaContainer(uint32_t size);
+
+  ~ArenaContainer() {
+    if (header_) {
+      assert(header_->ref_count_ == 0);
+      Dispose();
+    }
+  }
+
+  // Allocation function.
+  void* Place(int size) {
+    ArenaContainerHeader* header = reinterpret_cast<ArenaContainerHeader*>(header_);
+    if (header->current_ + size > header->end_) {
+      return nullptr;
+    }
+    void* result = header->current_;
+    header->current_ += size;
+    return result;
+  }
+
+  // Place individual object in this container.
+  ObjHeader* PlaceObject(const TypeInfo* type_info);
+
+  // Places an array of certain type in this container. Note that array_type_info
+  // is type info for an array, not for an individual element. Also note that exactly
+  // same operation could be used to place strings.
+  ArrayHeader* PlaceArray(const TypeInfo* array_type_info, int count);
+
+  // Dispose whole container ignoring non-zero refcount. Use with care.
+  void Dispose() {
+    if (header_) {
+      FreeObject(header_);
+      header_ = nullptr;
+    }
+  }
+};
+
+// Raw reference to data, meaning T*, invented only for cleaness of intentions.
+template <class T>
+class RawRef {
+ private:
+  T* ptr_;
+ public:
+  RawRef(T* ptr) : ptr_(ptr) {}
+  const T& get() const { return *ptr_; }
+  void set(const T& value) { *ptr_ = value; }
+};
+
+// Object reference, adds reference counting in container and type information.
+class AnyObjRef {
+ protected:
+  ObjHeader* ptr_;
+
+  explicit AnyObjRef(ObjHeader* ptr) : ptr_(ptr) {
+    if (ptr_) {
+      AddRef(container_header());
+    }
+  }
+
+ public:
+  ~AnyObjRef() {
+    if (ptr_) {
+      Release(container_header());
+    }
+  }
+
+  ContainerHeader* container_header() const {
+    return reinterpret_cast<ContainerHeader*>(
+        reinterpret_cast<uint8_t*>(ptr_) - ptr_->container_offset_negative_);
+  }
+
+  const TypeInfo* type_info() const {
+    return ptr_->type_info_;
+  }
+
+  // Accesses raw data inside object specified by offset. Typing by M is optional and
+  // will be replaced by translator typing.
+  template<typename M, int offset>
+  RawRef<M> at() const {
+    return RawRef<M>(
+      reinterpret_cast<M*>(reinterpret_cast<uint8_t*>(any_ref()) + offset));
+  }
+
+  // Assign reference to certain object. Releases currently held object in its container and
+  // adds reference to container storing given object.
+  void Assign(const AnyObjRef& other) {
+    // TODO: optimize for an important case where containers match?
+    if (ptr_) {
+      Release(container_header());
+    }
+    ptr_ = other.ptr_;
+    if (ptr_) {
+      AddRef(container_header());
+    }
+  }
+
+  // Returns pointer to the raw data referred by this reference.
+  uint8_t* any_ref() const {
+    if (!ptr_) return nullptr;
+    return reinterpret_cast<uint8_t*>(ptr_) + sizeof(ObjHeader);
+  }
+
+  // Uses pointer stored in object's field to create a reference to that object.
+  AnyObjRef any_obj_at(int offset) const {
+    assert(ptr_);
+    return AnyObjRef(*reinterpret_cast<ObjHeader**>(any_ref() + offset));
+  }
+
+  // Checks if given reference has null value.
+  bool null() const { return ptr_ == nullptr; }
+};
+
+// Returns typeinfo for array of type T. Specialize for types which are allowed as array elements.
+template <typename T>
+const TypeInfo* GetArrayTypeInfo() {
+  return nullptr;
+}
+
+// Reference to an object with particular memory layout specified by T.
+// In real runtime will be compile time only type, on runtime all references are
+// AnyObjRef.
+template <class T>
+class ObjRef : public AnyObjRef {
+ private:
+  explicit ObjRef(ObjHeader* ptr) : AnyObjRef(ptr) {}
+
+  // Reference to raw data in owned class.
+  T* ref() const {
+    if (!ptr_) return nullptr;
+    return reinterpret_cast<T*>(any_ref());
+  }
+
+  template <typename T1> friend class ArrayRef;
+
+ public:
+  // Assigns reference, compile time type-safe.
+  ObjRef(const ObjRef& other) : AnyObjRef(nullptr) {
+    Assign(other);
+  }
+  void Assign(const ObjRef<T>& other) {
+    AnyObjRef::Assign(other);
+  }
+
+  // Copies data bits to another place, reference counting is properly accounted for
+  // by consulting type information.
+  void CopyTo(ObjRef<T> other) const;
+
+  // Clones object to given container.
+  ObjRef<T> Clone(ArenaContainer* container) {
+    ObjRef<T> result = Alloc(container);
+    CopyTo(result);
+    return result;
+  }
+
+  // Takes typed object reference at offset.
+  template<typename M, int offset>
+  ObjRef<M> obj_at() const {
+    return ObjRef<M>(reinterpret_cast<M*>(any_ref() + offset));
+  }
+
+  // Allocates properly typed object in container.
+  static ObjRef<T> Alloc(ArenaContainer* container) {
+    return ObjRef<T>(container->PlaceObject(T::GetTypeInfo()));
+  }
+};
+
+// This is an array of value types only, no object references here.
+template <class T>
+class ArrayRef : public AnyObjRef {
+ protected:
+  explicit ArrayRef(ArrayHeader* ptr) : AnyObjRef(ptr) {}
+  ArrayHeader* header() { return reinterpret_cast<ArrayHeader*>(ptr_); }
+
+ public:
+   static ArrayRef<T> Alloc(ArenaContainer* container, int count) {
+     auto result = ArrayRef<T>(container->PlaceArray(GetArrayTypeInfo<T>(), count));
+     result.header()->count_ = count;
+     return result;
+  }
+
+  RawRef<T> element_at(int index) const {
+    assert(header() && index >= 0 && index < header()->count_);
+    return reinterpret_cast<T*>(any_ref() + index * sizeof(T));
+  }
+};
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+void InitMemory();
+void* AllocInstance(const TypeInfo* type_info, PlacementHint hint);
+void* AllocArrayInstance(const TypeInfo* type_info, PlacementHint hint, uint32_t elements);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // RUNTIME_MEMORY_H

runtime/src/main/cpp/Names.cpp

@@ -0,0 +1,54 @@
+#include <cassert>
+
+#include "Names.h"
+
+#include "City.h"
+#include "Sha1.h"
+
+namespace {
+
+void Printable(const uint8_t* data, uint32_t data_length, char* hex) {
+  static const char* hex_digits = "0123456789ABCDEF";
+  int i = 0;
+  for(int i = 0; i < data_length; ++i) {
+    *hex++ = hex_digits[(*data >> 4) & 0xf];
+    *hex++ = hex_digits[(*data++) & 0xf];
+  }
+}
+
+} // namespace
+
+extern "C" {
+
+// Make local hash out of arbitrary data.
+void MakeLocalHash(const void* data, uint32_t size, LocalHash* hash) {
+  *hash = CityHash64(data, size);
+}
+
+// Make global hash out of arbitrary data.
+void MakeGlobalHash(const void* data, uint32_t size, GlobalHash* hash) {
+  SHA1_CTX ctx;
+  SHA1Init(&ctx);
+  SHA1Update(&ctx, reinterpret_cast<const unsigned char *>(data), size);
+  SHA1Final(&hash->bits[0], &ctx);
+}
+
+// Make printable C string out of local hash.
+void PrintableLocalHash(const LocalHash* hash, char* buffer, uint32_t size) {
+  if (size < sizeof(*hash) * 2) {
+    assert(false);
+    return;
+  }
+  Printable(reinterpret_cast<const uint8_t*>(&hash), sizeof(*hash), buffer);
+}
+
+// Make printable C string out of global hash.
+void PrintableGlobalHash(const GlobalHash* hash, char* buffer, uint32_t size) {
+  if (size < sizeof(*hash) * 2) {
+    assert(false);
+    return;
+  }
+  Printable(hash->bits, sizeof(*hash), buffer);
+}
+
+} // extern "C"

runtime/src/main/cpp/Names.h

@@ -0,0 +1,43 @@
+#ifndef RUNTIME_NAMES_H
+#define RUNTIME_NAMES_H
+
+#include <cstdint>
+
+// All names in system are stored as hashes (or maybe, for debug builds,
+// as pointers to uniqued C strings containing names?).
+// There are two types of hashes:
+//  - local hash, must be unique per class (CityHash64 is being used)
+//  - global hash, must be unique globally (SHA1 is being used)
+// Generic guideline is that global hash is being used in global persistent context, while local
+// hashes are more local in scope.
+// Local hash.
+typedef uint64_t LocalHash;
+// Hash of field name.
+typedef LocalHash FieldNameHash;
+// Hash of open method name.
+typedef LocalHash MethodNameHash;
+// Global hash.
+typedef struct {
+  uint8_t bits[20];
+} GlobalHash;
+// Hash of function name.
+typedef GlobalHash FunctionNameHash;
+// Hash of class name.
+typedef GlobalHash ClassNameHash;
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+// Make local hash out of arbitrary data.
+void MakeLocalHash(const void* data, uint32_t size, LocalHash* hash);
+// Make global hash out of arbitrary data.
+void MakeGlobalHash(const void* data, uint32_t size, GlobalHash* hash);
+// Make printable C string out of local hash.
+void PrintableLocalHash(const LocalHash* hash, char* buffer, uint32_t size);
+// Make printable C string out of global hash.
+void PrintableGlobalHash(const GlobalHash* hash, char* buffer, uint32_t size);
+#ifdef __cplusplus
+} // extern "C"
+#endif
+
+#endif // RUNTIME_NAMES_H

runtime/src/main/cpp/Sha1.cpp

@@ -0,0 +1,175 @@
+/*
+SHA-1 in C
+By Steve Reid <steve@edmweb.com>
+100% Public Domain
+
+Test Vectors (from FIPS PUB 180-1)
+"abc"
+  A9993E36 4706816A BA3E2571 7850C26C 9CD0D89D
+"abcdbcdecdefdefgefghfghighijhijkijkljklmklmnlmnomnopnopq"
+  84983E44 1C3BD26E BAAE4AA1 F95129E5 E54670F1
+A million repetitions of "a"
+  34AA973C D4C4DAA4 F61EEB2B DBAD2731 6534016F
+*/
+
+/* #define LITTLE_ENDIAN * This should be #define'd already, if true. */
+/* #define SHA1HANDSOFF * Copies data before messing with it. */
+
+#define SHA1HANDSOFF
+
+#include <stdio.h>
+#include <string.h>
+#include <stdint.h>
+
+#include "Sha1.h"
+
+#define rol(value, bits) (((value) << (bits)) | ((value) >> (32 - (bits))))
+
+/* blk0() and blk() perform the initial expand. */
+/* I got the idea of expanding during the round function from SSLeay */
+#if BYTE_ORDER == LITTLE_ENDIAN
+#define blk0(i) (block->l[i] = (rol(block->l[i],24)&0xFF00FF00) \
+    |(rol(block->l[i],8)&0x00FF00FF))
+#elif BYTE_ORDER == BIG_ENDIAN
+#define blk0(i) block->l[i]
+#else
+#error "Endianness not defined!"
+#endif
+#define blk(i) (block->l[i&15] = rol(block->l[(i+13)&15]^block->l[(i+8)&15] \
+    ^block->l[(i+2)&15]^block->l[i&15],1))
+
+/* (R0+R1), R2, R3, R4 are the different operations used in SHA1 */
+#define R0(v,w,x,y,z,i) z+=((w&(x^y))^y)+blk0(i)+0x5A827999+rol(v,5);w=rol(w,30);
+#define R1(v,w,x,y,z,i) z+=((w&(x^y))^y)+blk(i)+0x5A827999+rol(v,5);w=rol(w,30);
+#define R2(v,w,x,y,z,i) z+=(w^x^y)+blk(i)+0x6ED9EBA1+rol(v,5);w=rol(w,30);
+#define R3(v,w,x,y,z,i) z+=(((w|x)&y)|(w&x))+blk(i)+0x8F1BBCDC+rol(v,5);w=rol(w,30);
+#define R4(v,w,x,y,z,i) z+=(w^x^y)+blk(i)+0xCA62C1D6+rol(v,5);w=rol(w,30);
+
+/* Hash a single 512-bit block. This is the core of the algorithm. */
+static void SHA1Transform(uint32_t state[5], const unsigned char buffer[64])
+{
+    uint32_t a, b, c, d, e;
+    typedef union {
+        unsigned char c[64];
+        uint32_t l[16];
+    } CHAR64LONG16;
+#ifdef SHA1HANDSOFF
+    CHAR64LONG16 block[1];  /* use array to appear as a pointer */
+    memcpy(block, buffer, 64);
+#else
+    /* The following had better never be used because it causes the
+     * pointer-to-const buffer to be cast into a pointer to non-const.
+     * And the result is written through.  I threw a "const" in, hoping
+     * this will cause a diagnostic.
+     */
+    CHAR64LONG16* block = (const CHAR64LONG16*)buffer;
+#endif
+    /* Copy context->state[] to working vars */
+    a = state[0];
+    b = state[1];
+    c = state[2];
+    d = state[3];
+    e = state[4];
+    /* 4 rounds of 20 operations each. Loop unrolled. */
+    R0(a,b,c,d,e, 0); R0(e,a,b,c,d, 1); R0(d,e,a,b,c, 2); R0(c,d,e,a,b, 3);
+    R0(b,c,d,e,a, 4); R0(a,b,c,d,e, 5); R0(e,a,b,c,d, 6); R0(d,e,a,b,c, 7);
+    R0(c,d,e,a,b, 8); R0(b,c,d,e,a, 9); R0(a,b,c,d,e,10); R0(e,a,b,c,d,11);
+    R0(d,e,a,b,c,12); R0(c,d,e,a,b,13); R0(b,c,d,e,a,14); R0(a,b,c,d,e,15);
+    R1(e,a,b,c,d,16); R1(d,e,a,b,c,17); R1(c,d,e,a,b,18); R1(b,c,d,e,a,19);
+    R2(a,b,c,d,e,20); R2(e,a,b,c,d,21); R2(d,e,a,b,c,22); R2(c,d,e,a,b,23);
+    R2(b,c,d,e,a,24); R2(a,b,c,d,e,25); R2(e,a,b,c,d,26); R2(d,e,a,b,c,27);
+    R2(c,d,e,a,b,28); R2(b,c,d,e,a,29); R2(a,b,c,d,e,30); R2(e,a,b,c,d,31);
+    R2(d,e,a,b,c,32); R2(c,d,e,a,b,33); R2(b,c,d,e,a,34); R2(a,b,c,d,e,35);
+    R2(e,a,b,c,d,36); R2(d,e,a,b,c,37); R2(c,d,e,a,b,38); R2(b,c,d,e,a,39);
+    R3(a,b,c,d,e,40); R3(e,a,b,c,d,41); R3(d,e,a,b,c,42); R3(c,d,e,a,b,43);
+    R3(b,c,d,e,a,44); R3(a,b,c,d,e,45); R3(e,a,b,c,d,46); R3(d,e,a,b,c,47);
+    R3(c,d,e,a,b,48); R3(b,c,d,e,a,49); R3(a,b,c,d,e,50); R3(e,a,b,c,d,51);
+    R3(d,e,a,b,c,52); R3(c,d,e,a,b,53); R3(b,c,d,e,a,54); R3(a,b,c,d,e,55);
+    R3(e,a,b,c,d,56); R3(d,e,a,b,c,57); R3(c,d,e,a,b,58); R3(b,c,d,e,a,59);
+    R4(a,b,c,d,e,60); R4(e,a,b,c,d,61); R4(d,e,a,b,c,62); R4(c,d,e,a,b,63);
+    R4(b,c,d,e,a,64); R4(a,b,c,d,e,65); R4(e,a,b,c,d,66); R4(d,e,a,b,c,67);
+    R4(c,d,e,a,b,68); R4(b,c,d,e,a,69); R4(a,b,c,d,e,70); R4(e,a,b,c,d,71);
+    R4(d,e,a,b,c,72); R4(c,d,e,a,b,73); R4(b,c,d,e,a,74); R4(a,b,c,d,e,75);
+    R4(e,a,b,c,d,76); R4(d,e,a,b,c,77); R4(c,d,e,a,b,78); R4(b,c,d,e,a,79);
+    /* Add the working vars back into context.state[] */
+    state[0] += a;
+    state[1] += b;
+    state[2] += c;
+    state[3] += d;
+    state[4] += e;
+    /* Wipe variables */
+    a = b = c = d = e = 0;
+#ifdef SHA1HANDSOFF
+    memset(block, '\0', sizeof(block));
+#endif
+}
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* SHA1Init - Initialize new context */
+void SHA1Init(SHA1_CTX* context)
+{
+    /* SHA1 initialization constants */
+    context->state[0] = 0x67452301;
+    context->state[1] = 0xEFCDAB89;
+    context->state[2] = 0x98BADCFE;
+    context->state[3] = 0x10325476;
+    context->state[4] = 0xC3D2E1F0;
+    context->count[0] = context->count[1] = 0;
+}
+
+
+/* Run your data through this. */
+void SHA1Update(SHA1_CTX* context, const unsigned char* data, uint32_t len)
+{
+    uint32_t i, j;
+
+    j = context->count[0];
+    if ((context->count[0] += len << 3) < j)
+        context->count[1]++;
+    context->count[1] += (len>>29);
+    j = (j >> 3) & 63;
+    if ((j + len) > 63) {
+        memcpy(&context->buffer[j], data, (i = 64-j));
+        SHA1Transform(context->state, context->buffer);
+        for ( ; i + 63 < len; i += 64) {
+            SHA1Transform(context->state, &data[i]);
+        }
+        j = 0;
+    }
+    else i = 0;
+    memcpy(&context->buffer[j], &data[i], len - i);
+}
+
+/* Add padding and return the message digest. */
+void SHA1Final(unsigned char digest[20], SHA1_CTX* context)
+{
+    unsigned i;
+    unsigned char finalcount[8];
+    unsigned char c;
+
+    for (i = 0; i < 8; i++) {
+        finalcount[i] = (unsigned char)((context->count[(i >= 4 ? 0 : 1)]
+         >> ((3-(i & 3)) * 8) ) & 255);  /* Endian independent */
+    }
+    c = 0200;
+    SHA1Update(context, &c, 1);
+    while ((context->count[0] & 504) != 448) {
+        c = 0000;
+        SHA1Update(context, &c, 1);
+    }
+    SHA1Update(context, finalcount, 8);  /* Should cause a SHA1Transform() */
+    for (i = 0; i < 20; i++) {
+        digest[i] = (unsigned char)
+         ((context->state[i>>2] >> ((3-(i & 3)) * 8) ) & 255);
+    }
+    /* Wipe variables */
+    memset(context, '\0', sizeof(*context));
+    memset(&finalcount, '\0', sizeof(finalcount));
+}
+
+#ifdef __cplusplus
+}
+#endif

runtime/src/main/cpp/Sha1.h

@@ -0,0 +1,27 @@
+#ifndef RUNTIME_SHA1_H
+#define RUNTIME_SHA1_H
+/*
+SHA-1 in C
+By Steve Reid <steve@edmweb.com>
+100% Public Domain
+*/
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+typedef struct SHA1_CTX {
+    uint32_t state[5];
+    uint32_t count[2];
+    unsigned char buffer[64];
+} SHA1_CTX;
+
+void SHA1Init(SHA1_CTX* context);
+void SHA1Update(SHA1_CTX* context, const unsigned char* data, uint32_t len);
+void SHA1Final(unsigned char digest[20], SHA1_CTX* context);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // RUNTIME_UTIL_SHA1_H

runtime/src/main/cpp/TypeInfo.cpp

@@ -2,8 +2,15 @@
 #include "TypeInfo.h"
 
 extern "C" {
-    int lookupField(TypeInfo* info, NameHash nameSignature) {
-        assert(false); // not implemented yet
-        return -1;
-    }
-}
+
+int LookupFieldOffset(const TypeInfo* info, FieldNameHash nameSignature) {
+  assert(false); // not implemented yet
+  return -1;
+}
+
+void* LookupMethod(const TypeInfo* info, MethodNameHash nameSignature) {
+  assert(false); // not implemented yet
+  return nullptr;
+}
+
+}

runtime/src/main/cpp/TypeInfo.h

@@ -3,38 +3,48 @@
 
 #include <cstdint>
 
-// All names in system are stored as hashes (or maybe, for debug builds,
-// as pointers to uniqued C strings containing names?).
-typedef int64_t NameHash;
+#include "Names.h"
 
 // An element of sorted by hash in-place array representing methods.
 struct MethodTableRecord {
-    NameHash nameSignature;
-    void* methodEntryPoint;
+    MethodNameHash nameSignature_;
+    void* methodEntryPoint_;
 };
 
 // An element of sorted by hash in-place array representing field offsets.
 struct FieldTableRecord {
-    NameHash nameSignature;
-    int fieldOffset;
+    FieldNameHash nameSignature_;
+    int fieldOffset_;
 };
 
 // This struct represents runtime type information and by itself is compile time
 // constant.
 struct TypeInfo {
-    NameHash name;
-    int size;
-    const TypeInfo* superType;
-    const int* objOffsets;
-    int objOffsetsCount;
-    TypeInfo* const* implementedInterfaces;
-    int implementedInterfacesCount;
-    void* const* vtable; // TODO: place vtable at the end of TypeInfo to eliminate the indirection
-    const MethodTableRecord* methods;
-    int methodsCount;
-    const FieldTableRecord* fields;
-    int fieldsCount;
+    ClassNameHash name_;
+    int size_;
+    const TypeInfo* superType_;
+    const int* objOffsets_;
+    int objOffsetsCount_;
+    TypeInfo* const* implementedInterfaces_;
+    int implementedInterfacesCount_;
+    void* const* vtable_; // TODO: place vtable at the end of TypeInfo to eliminate the indirection
+    const MethodTableRecord* methods_;
+    int methodsCount_;
+    const FieldTableRecord* fields_;
+    int fieldsCount_;
 };
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+// Find offset of given hash in table.
+int LookupFieldOffset(const TypeInfo* type_info, LocalHash hash);
 
-#endif //RUNTIME_TYPEINFO_H
+// Find method by its hash.
+void* LookupMethod(const TypeInfo* info, MethodNameHash nameSignature);
+
+#ifdef __cplusplus
+} // extern "C"
+#endif
+
+#endif // RUNTIME_TYPEINFO_H

runtime/src/main/cpp/launcher.cpp

@@ -1,6 +1,9 @@
+#include "Memory.h"
+
 extern "C" void kotlinNativeMain();
 
 int main() {
-    kotlinNativeMain();
-    return 0;
-}
+  InitMemory();
+  kotlinNativeMain();
+  return 0;
+}