Merge branch 'ios' into armvfp

src/arm.cpp

@@ -1847,6 +1847,10 @@ class MyArchitecture: public Assembler::Architecture {
     return 0;
   }
 
+  virtual int scratch() {
+    return 5;
+  }
+
   virtual int stack() {
     return StackRegister;
   }

src/assembler.h

@@ -329,6 +329,7 @@ class Assembler {
     virtual uint32_t generalRegisterMask() = 0;
     virtual uint32_t floatRegisterMask() = 0;
 
+    virtual int scratch() = 0;
     virtual int stack() = 0;
     virtual int thread() = 0;
     virtual int returnLow() = 0;

src/bootimage.cpp

@@ -496,8 +496,10 @@ makeCodeImage(Thread* t, Zone* zone, BootImage* image, uint8_t* code,
             }
 
             if (fieldFlags(t, field) & ACC_STATIC) {
-              while (targetStaticOffset % targetSize) {
+              unsigned excess = (targetStaticOffset % targetSize)
-                ++ targetStaticOffset;
+                % TargetBytesPerWord;
+              if (excess) {
+                targetStaticOffset += TargetBytesPerWord - excess;
               }
 
               buildStaticOffset = fieldOffset(t, field);
@@ -672,7 +674,7 @@ makeCodeImage(Thread* t, Zone* zone, BootImage* image, uint8_t* code,
     expect(t, value >= code);
 
     addresses->listener->resolve
-      (targetVW(static_cast<target_intptr_t>(value - code)), 0);
+      (static_cast<target_intptr_t>(value - code), 0);
   }
 
   for (; methods; methods = pairSecond(t, methods)) {
@@ -987,7 +989,7 @@ copy(Thread* t, object typeMaps, object p, uint8_t* dst)
         if (field->type == Type_object) {
           unsigned offset = field->targetOffset / TargetBytesPerWord;
           reinterpret_cast<uint32_t*>(mask)[offset / 32]
-            |= static_cast<uint32_t>(1) << (offset % 32);
+            |= targetV4(static_cast<uint32_t>(1) << (offset % 32));
         }
       }
 
@@ -1027,14 +1029,15 @@ copy(Thread* t, object typeMaps, object p, uint8_t* dst)
         switch (field->type) {
         case Type_object:
           reinterpret_cast<uint32_t*>(objectMask)[i / 32]
-            |= static_cast<uint32_t>(1) << (i % 32);
+            |= targetV4(static_cast<uint32_t>(1) << (i % 32));
           break;
 
         case Type_float:
         case Type_double:
           reinterpret_cast<target_uintptr_t*>(poolMask)
             [i / TargetBitsPerWord]
-            |= static_cast<target_uintptr_t>(1) << (i % TargetBitsPerWord);
+            |= targetVW
+            (static_cast<target_uintptr_t>(1) << (i % TargetBitsPerWord));
           break;
 
         default:
@@ -1083,7 +1086,7 @@ copy(Thread* t, object typeMaps, object referer, unsigned refererOffset,
       if (field->type == Type_object) {
         unsigned offset = field->targetOffset / TargetBytesPerWord;
         reinterpret_cast<uint32_t*>(dst + (TargetBytesPerWord * 2))
-          [offset / 32] |= static_cast<uint32_t>(1) << (offset % 32);
+          [offset / 32] |= targetV4(static_cast<uint32_t>(1) << (offset % 32));
       }
     }
 
@@ -1092,7 +1095,7 @@ copy(Thread* t, object typeMaps, object referer, unsigned refererOffset,
     {
       unsigned offset = map->targetFixedSizeInWords;
       reinterpret_cast<uint32_t*>(dst + (TargetBytesPerWord * 2))
-        [offset / 32] |= static_cast<uint32_t>(1) << (offset % 32);
+        [offset / 32] |= targetV4(static_cast<uint32_t>(1) << (offset % 32));
     }
   } else {
     copy(t, typeMaps, p, dst);
@@ -1241,7 +1244,7 @@ makeHeapImage(Thread* t, BootImage* image, target_uintptr_t* heap,
   HeapWalker* w = makeHeapWalker(t, &visitor);
   visitRoots(t, image, w, constants);
   
-  image->heapSize = visitor.position * BytesPerWord;
+  image->heapSize = visitor.position * TargetBytesPerWord;
 
   return w;
 }
@@ -1642,7 +1645,7 @@ main(int ac, const char** av)
 {
   if (ac < 4 or ac > 7) {
     fprintf(stderr, "usage: %s <classpath> <bootimage file> <code file>"
-            "[<class name> [<method name> [<method spec>]]]\n", av[0]);
+            " [<class name> [<method name> [<method spec>]]]\n", av[0]);
     return -1;
   }
 

src/compile.cpp

@@ -380,7 +380,7 @@ compareIpToMethodBounds(Thread* t, intptr_t ip, object method)
   if (ip < start) {
     return -1;
   } else if (ip < start + static_cast<intptr_t>
-             (compiledSize(start) + BytesPerWord))
+             (compiledSize(start) + TargetBytesPerWord))
   {
     return 0;
   } else {
@@ -6554,10 +6554,11 @@ simpleFrameMapTableSize(MyThread* t, object method, object map)
 }
 
 uint8_t*
-finish(MyThread* t, Allocator* allocator, Assembler* a, const char* name,
+finish(MyThread* t, FixedAllocator* allocator, Assembler* a, const char* name,
        unsigned length)
 {
-  uint8_t* start = static_cast<uint8_t*>(allocator->allocate(pad(length)));
+  uint8_t* start = static_cast<uint8_t*>
+    (allocator->allocate(length, TargetBytesPerWord));
 
   a->setDestination(start);
   a->write();
@@ -6872,10 +6873,11 @@ finish(MyThread* t, FixedAllocator* allocator, Context* context)
   unsigned codeSize = c->resolve
     (allocator->base + allocator->offset + TargetBytesPerWord);
 
-  unsigned total = pad(codeSize) + pad(c->poolSize()) + TargetBytesPerWord;
+  unsigned total = pad(codeSize, TargetBytesPerWord)
+    + pad(c->poolSize(), TargetBytesPerWord) + TargetBytesPerWord;
 
   target_uintptr_t* code = static_cast<target_uintptr_t*>
-    (allocator->allocate(total));
+    (allocator->allocate(total, TargetBytesPerWord));
   code[0] = codeSize;
   uint8_t* start = reinterpret_cast<uint8_t*>(code + 1);
 
@@ -8291,7 +8293,7 @@ MyProcessor*
 processor(MyThread* t);
 
 void
-compileThunks(MyThread* t, Allocator* allocator);
+compileThunks(MyThread* t, FixedAllocator* allocator);
 
 class MyProcessor: public Processor {
  public:
@@ -8347,8 +8349,14 @@ class MyProcessor: public Processor {
 #define THUNK(s) thunkTable[s##Index] = voidPointer(s);
 #include "thunks.cpp"
 #undef THUNK
+    // Set the dummyIndex entry to a constant which should require the
+    // maximum number of bytes to represent in assembly code
+    // (i.e. can't be represented by a smaller number of bytes and
+    // implicitly sign- or zero-extended).  We'll use this property
+    // later to determine the maximum size of a thunk in the thunk
+    // table.
     thunkTable[dummyIndex] = reinterpret_cast<void*>
-      (~static_cast<uintptr_t>(0));
+      (static_cast<uintptr_t>(UINT64_C(0x5555555555555555)));
   }
 
   virtual Thread*
@@ -9396,10 +9404,7 @@ compileCall(MyThread* t, Context* c, ThunkIndex index, bool call = true)
 
   if (processor(t)->bootImage) {
     Assembler::Memory table(t->arch->thread(), TargetThreadThunkTable);
-    // use Architecture::virtualCallTarget register here as a scratch
+    Assembler::Register scratch(t->arch->scratch());
-    // register; any register that isn't used to pass arguments would
-    // be acceptable:
-    Assembler::Register scratch(t->arch->virtualCallTarget());
     a->apply(Move, TargetBytesPerWord, MemoryOperand, &table,
              TargetBytesPerWord, RegisterOperand, &scratch);
     Assembler::Memory proc(scratch.low, index * TargetBytesPerWord);
@@ -9418,7 +9423,7 @@ compileCall(MyThread* t, Context* c, ThunkIndex index, bool call = true)
 }
 
 void
-compileThunks(MyThread* t, Allocator* allocator)
+compileThunks(MyThread* t, FixedAllocator* allocator)
 {
   MyProcessor* p = processor(t);
 
@@ -9559,7 +9564,8 @@ compileThunks(MyThread* t, Allocator* allocator)
       p->thunks.table.length = a->endBlock(false)->resolve(0, 0);
 
       p->thunks.table.start = static_cast<uint8_t*>
-        (allocator->allocate(p->thunks.table.length * ThunkCount));
+        (allocator->allocate
+         (p->thunks.table.length * ThunkCount, TargetBytesPerWord));
     }
 
     uint8_t* start = p->thunks.table.start;
@@ -9678,7 +9684,8 @@ compileVirtualThunk(MyThread* t, unsigned index, unsigned* size)
 
   *size = a->endBlock(false)->resolve(0, 0);
 
-  uint8_t* start = static_cast<uint8_t*>(codeAllocator(t)->allocate(*size));
+  uint8_t* start = static_cast<uint8_t*>
+    (codeAllocator(t)->allocate(*size, TargetBytesPerWord));
 
   a->setDestination(start);
   a->write();

src/compiler.cpp

@@ -6953,14 +6953,14 @@ class MyCompiler: public Compiler {
         (c.machineCode + pad(c.machineCodeSize, TargetBytesPerWord) + i);
 
       if (n->promise->resolved()) {
-        *target = n->promise->value();
+        *target = targetVW(n->promise->value());
       } else {
         class Listener: public Promise::Listener {
          public:
           Listener(target_intptr_t* target): target(target){ }
 
           virtual bool resolve(int64_t value, void** location) {
-            *target = value;
+            *target = targetVW(value);
             if (location) *location = target;
             return true;
           }

src/machine.cpp

@@ -1145,8 +1145,9 @@ parseFieldTable(Thread* t, Stream& s, object class_, object pool)
 
       unsigned size = fieldSize(t, code);
       if (flags & ACC_STATIC) {
-        while (staticOffset % size) {
+        unsigned excess = (staticOffset % size) % BytesPerWord;
-          ++ staticOffset;
+        if (excess) {
+          staticOffset += BytesPerWord - excess;
         }
 
         fieldOffset(t, field) = staticOffset;

src/machine.h

@@ -1796,8 +1796,8 @@ class FixedAllocator: public Allocator {
     abort(s);
   }
 
-  virtual void* allocate(unsigned size) {
+  void* allocate(unsigned size, unsigned padAlignment) {
-    unsigned paddedSize = pad(size);
+    unsigned paddedSize = pad(size, padAlignment);
     expect(s, offset + paddedSize < capacity);
 
     void* p = base + offset;
@@ -1805,6 +1805,10 @@ class FixedAllocator: public Allocator {
     return p;
   }
 
+  virtual void* allocate(unsigned size) {
+    return allocate(size, BytesPerWord);
+  }
+
   virtual void free(const void* p, unsigned size) {
     if (p >= base and static_cast<const uint8_t*>(p) + size == base + offset) {
       offset -= size;

src/powerpc.cpp

@@ -789,14 +789,14 @@ updateImmediate(System* s, void* dst, int32_t src, unsigned size, bool address)
   switch (size) {
   case 4: {
     int32_t* p = static_cast<int32_t*>(dst);
-    int r = (p[1] >> 21) & 31;
+    int r = (targetV4(p[1]) >> 21) & 31;
 
     if (address) {
-      p[0] = lis(r, ha16(src));
+      p[0] = targetV4(lis(r, ha16(src)));
-      p[1] |= (src & 0xFFFF);
+      p[1] |= targetV4(src & 0xFFFF);
     } else {
-      p[0] = lis(r, src >> 16);
+      p[0] = targetV4(lis(r, src >> 16));
-      p[1] = ori(r, r, src);
+      p[1] = targetV4(ori(r, r, src));
     }
   } break;
 
@@ -2076,6 +2076,10 @@ class MyArchitecture: public Assembler::Architecture {
     return 0;
   }
 
+  virtual int scratch() {
+    return 31;
+  }
+
   virtual int stack() {
     return StackRegister;
   }

src/x86.cpp

@@ -2763,6 +2763,10 @@ class MyArchitecture: public Assembler::Architecture {
     return useSSE(&c) ? FloatRegisterMask : 0;
   }
 
+  virtual int scratch() {
+    return rax;
+  }
+
   virtual int stack() {
     return rsp;
   }