[tracing] Optimize AllocationRegister and increase max backtrace depth.

base/trace_event/heap_profiler_allocation_context.cc

 // Copyright 2015 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "base/trace_event/heap_profiler_allocation_context.h"
 
 #include <cstring>
 
 #include "base/hash.h"
 #include "base/macros.h"
@@ skipping 13 matching lines @@
   return !(lhs.value == rhs.value);
 }
 
 Backtrace::Backtrace(): frame_count(0) {}
 
 bool operator==(const Backtrace& lhs, const Backtrace& rhs) {
   if (lhs.frame_count != rhs.frame_count) return false;
   return std::equal(lhs.frames, lhs.frames + lhs.frame_count, rhs.frames);
 }
 
+bool operator!=(const Backtrace& lhs, const Backtrace& rhs) {
+  return !(lhs == rhs);
+}
+
 AllocationContext::AllocationContext(): type_name(nullptr) {}
 
+AllocationContext::AllocationContext(const Backtrace& backtrace,
+                                     const char* type_name)
+  : backtrace(backtrace), type_name(type_name) {}
+
 bool operator==(const AllocationContext& lhs, const AllocationContext& rhs) {
   return (lhs.backtrace == rhs.backtrace) && (lhs.type_name == rhs.type_name);
 }
 
+bool operator!=(const AllocationContext& lhs, const AllocationContext& rhs) {
+  return !(lhs == rhs);
+}
 }  // namespace trace_event
 }  // namespace base
 
 namespace BASE_HASH_NAMESPACE {
 using base::trace_event::AllocationContext;
 using base::trace_event::Backtrace;
 using base::trace_event::StackFrame;
 
 size_t hash<StackFrame>::operator()(const StackFrame& frame) const {
   return hash<const void*>()(frame.value);
 }
 
 size_t hash<Backtrace>::operator()(const Backtrace& backtrace) const {
-  const void* values[Backtrace::kMaxFrameCount];
-  for (size_t i = 0; i != backtrace.frame_count; ++i) {
-    values[i] = backtrace.frames[i].value;
+  // This function needs to be fast, because AllocationRegister checks
+  // its Backtrace hash map on every Insert() call. Surprisingly, Knuth's

[Primiano Tucci (use gerrit), 2016/06/23 20:46:24]

Not sure I'm that "surprised" :)

+  // fast multiplicative hash produces great results.
+  const uintptr_t kKnuthConstant = 2654435761;

[Primiano Tucci (use gerrit), 2016/06/23 20:46:24]

s/2654435761/2654435761u/ (trailing u)
I never remember what to expect otherwise on 32-bit for this that wraps over
2^31

[Dmitry Skiba, 2016/06/28 10:54:58]

Done.

+  const size_t kHashableCount = 10;
+
+  uintptr_t hash = 0;
+
+  size_t head_end = std::min(backtrace.frame_count, kHashableCount);
+  for (size_t i = 0; i != head_end; ++i) {
+    hash += reinterpret_cast<uintptr_t>(
+        backtrace.frames[i].value) * kKnuthConstant;
   }
-  return base::SuperFastHash(
-      reinterpret_cast<const char*>(values),
-      static_cast<int>(backtrace.frame_count * sizeof(*values)));
+
+  size_t tail_start = backtrace.frame_count -
+      std::min(backtrace.frame_count - head_end, kHashableCount);
+  for (size_t i = tail_start; i != backtrace.frame_count; ++i) {
+    hash += reinterpret_cast<uintptr_t>(
+        backtrace.frames[i].value) * kKnuthConstant;
+  }
+
+  // Also include number of frames.
+  hash += backtrace.frame_count * kKnuthConstant;
+
+  return hash;
 }
 
 size_t hash<AllocationContext>::operator()(const AllocationContext& ctx) const {
   size_t backtrace_hash = hash<Backtrace>()(ctx.backtrace);
 
   // Multiplicative hash from [Knuth 1998]. Works best if |size_t| is 32 bits,
   // because the magic number is a prime very close to 2^32 / golden ratio, but
   // will still redistribute keys bijectively on 64-bit architectures because
   // the magic number is coprime to 2^64.
   size_t type_hash = reinterpret_cast<size_t>(ctx.type_name) * 2654435761;
 
   // Multiply one side to break the commutativity of +. Multiplication with a
   // number coprime to |numeric_limits<size_t>::max() + 1| is bijective so
   // randomness is preserved.
   return (backtrace_hash * 3) + type_hash;
 }
 
 }  // BASE_HASH_NAMESPACE