[heap] Rework Worklist base type

Worklist uses a singly-linked list of segments to hold entries.
Segment size was based on a compile-time constant but already stored
in the segment itself.

Rework the segments to query `malloc_usable_size()` on allocation and
adjust the capacity properly. For PartitionAlloc, it turns out that
there's ~20% more capacity available for the 64-element segments.

This slows down actual allocation of the segments with the upside of
improving utilization and requiring 20% less segments.

Change-Id: Ib8595c3fb9fb75b02e4022f6c525bb59a2df7ab7
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3826047
Commit-Queue: Anton Bikineev <bikineev@chromium.org>
Commit-Queue: Michael Lippautz <mlippautz@chromium.org>
Reviewed-by: Anton Bikineev <bikineev@chromium.org>
Cr-Commit-Position: refs/heads/main@{#82432}

src/base/platform/platform-posix.cc

@@ -52,6 +52,9 @@
 
 #if V8_OS_DARWIN
 #include <mach/mach.h>
+#include <malloc/malloc.h>
+#else
+#include <malloc.h>
 #endif
 
 #if V8_OS_LINUX
@@ -1262,5 +1265,14 @@ Stack::StackSlot Stack::GetCurrentStackPosition() {
 #undef MAP_ANONYMOUS
 #undef MADV_FREE
 
+// static
+size_t Malloc::GetUsableSize(void* ptr) {
+#if defined(V8_OS_DARWIN)
+  return malloc_size(ptr);
+#else   // defined(V8_OS_DARWIN)
+  return malloc_usable_size(ptr);
+#endif  // !defined(V8_OS_DARWIN)
+}
+
 }  // namespace base
 }  // namespace v8

src/base/platform/platform-win32.cc

@@ -19,7 +19,8 @@
 
 // This has to come after windows.h.
 #include <VersionHelpers.h>
-#include <dbghelp.h>   // For SymLoadModule64 and al.
+#include <dbghelp.h>  // For SymLoadModule64 and al.
+#include <malloc.h>   // For _msize()
 #include <mmsystem.h>  // For timeGetTime().
 #include <tlhelp32.h>  // For Module32First and al.
 
@@ -1765,5 +1766,8 @@ Stack::StackSlot Stack::GetCurrentStackPosition() {
 #endif
 }
 
+// static
+size_t Malloc::GetUsableSize(void* ptr) { return _msize(ptr); }
+
 }  // namespace base
 }  // namespace v8

src/base/platform/platform.h

@@ -661,6 +661,12 @@ class V8_BASE_EXPORT Stack {
   }
 };
 
+class V8_BASE_EXPORT Malloc final {
+ public:
+  // Returns the usable size in bytes for a `ptr` allocated using `malloc()`.
+  static size_t GetUsableSize(void* ptr);
+};
+
 }  // namespace base
 }  // namespace v8
 

src/heap/base/worklist.cc

@@ -4,16 +4,12 @@
 
 #include "src/heap/base/worklist.h"
 
-namespace heap {
-namespace base {
-namespace internal {
+namespace heap::base::internal {
 
 // static
 SegmentBase* SegmentBase::GetSentinelSegmentAddress() {
-static SegmentBase kSentinelSegment(0);
-return &kSentinelSegment;
+  static SegmentBase sentinel_segment(0);
+  return &sentinel_segment;
 }
 
-}  // namespace internal
-}  // namespace base
-}  // namespace heap
+}  // namespace heap::base::internal

src/heap/scavenger.cc

@@ -704,7 +704,7 @@ void Scavenger::Process(JobDelegate* delegate) {
       scavenge_visitor.Visit(object_and_size.first);
       done = false;
       if (delegate && ((++objects % kInterruptThreshold) == 0)) {
-        if (!copied_list_local_.IsEmpty()) {
+        if (!copied_list_local_.IsLocalEmpty()) {
           delegate->NotifyConcurrencyIncrease();
         }
       }

test/cctest/heap/test-concurrent-marking.cc

@@ -20,10 +20,10 @@ namespace heap {
 
 void PublishSegment(MarkingWorklist* worklist, HeapObject object) {
   MarkingWorklist::Local local(worklist);
-  for (size_t i = 0; i <= MarkingWorklist::kSegmentSize; i++) {
+  for (size_t i = 0; i < MarkingWorklist::kMinSegmentSizeForTesting; i++) {
     local.Push(object);
   }
-  CHECK(local.Pop(&object));
+  local.Publish();
 }
 
 TEST(ConcurrentMarking) {