[IRMemoryMap] Avoid redundant zero-init in the Allocation constructor (NFC)

In the lldb-bench/arithmetic benchmark, 1.7% of the total running time
is spent zero-initializing a std::vector that has already been zeroed.

llvm-svn: 339051

lldb/include/lldb/Utility/DataBufferHeap.h

@@ -90,7 +90,8 @@ public:
   /// Set the number of bytes in the data buffer.
   ///
   /// Sets the number of bytes that this object should be able to contain.
-  /// This can be used prior to copying data into the buffer.
+  /// This can be used prior to copying data into the buffer. Note that this
+  /// zero-initializes up to \p byte_size bytes.
   ///
   /// @param[in] byte_size
   ///     The new size in bytes that this data buffer should attempt

lldb/source/Expression/IRMemoryMap.cpp

@@ -278,14 +278,10 @@ IRMemoryMap::Allocation::Allocation(lldb::addr_t process_alloc,
   default:
     assert(0 && "We cannot reach this!");
   case eAllocationPolicyHostOnly:
-    m_data.SetByteSize(size);
-    memset(m_data.GetBytes(), 0, size);
-    break;
-  case eAllocationPolicyProcessOnly:
-    break;
   case eAllocationPolicyMirror:
     m_data.SetByteSize(size);
-    memset(m_data.GetBytes(), 0, size);
+    break;
+  case eAllocationPolicyProcessOnly:
     break;
   }
 }