diff --git a/compiler-rt/lib/tsan/rtl/tsan_debugging.cpp b/compiler-rt/lib/tsan/rtl/tsan_debugging.cpp
--- a/compiler-rt/lib/tsan/rtl/tsan_debugging.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_debugging.cpp
@@ -215,9 +215,9 @@
     } else {
       // TODO(kuba.brecka): We should not lock. This is supposed to be called
       // from within the debugger when other threads are stopped.
-      ctx->thread_registry->Lock();
+      ctx->thread_registry.Lock();
       ThreadContext *tctx = IsThreadStackOrTls(addr, &is_stack);
-      ctx->thread_registry->Unlock();
+      ctx->thread_registry.Unlock();
       if (tctx) {
         region_kind = is_stack ? "stack" : "tls";
       } else {
@@ -252,7 +252,7 @@
   *thread_id = b->tid;
   // No locking.  This is supposed to be called from within the debugger when
   // other threads are stopped.
-  ThreadContextBase *tctx = ctx->thread_registry->GetThreadLocked(b->tid);
+  ThreadContextBase *tctx = ctx->thread_registry.GetThreadLocked(b->tid);
   *os_id = tctx->os_id;
 
   StackTrace stack = StackDepotGet(b->stk);
diff --git a/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp b/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp
--- a/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_interceptors_posix.cpp
@@ -1986,7 +1986,7 @@
     // StackTrace::GetNestInstructionPc(pc) is used because return address is
     // expected, OutputReport() will undo this.
     ObtainCurrentStack(thr, StackTrace::GetNextInstructionPc(pc), &stack);
-    ThreadRegistryLock l(ctx->thread_registry);
+    ThreadRegistryLock l(&ctx->thread_registry);
     ScopedReport rep(ReportTypeErrnoInSignal);
     if (!IsFiredSuppression(ctx, ReportTypeErrnoInSignal, stack)) {
       rep.AddStack(stack, true);
@@ -2348,7 +2348,7 @@
   ThreadSetName(((TsanInterceptorContext *) ctx)->thr, name)
 
 #define COMMON_INTERCEPTOR_SET_PTHREAD_NAME(ctx, thread, name) \
-  __tsan::ctx->thread_registry->SetThreadNameByUserId(thread, name)
+  __tsan::ctx->thread_registry.SetThreadNameByUserId(thread, name)
 
 #define COMMON_INTERCEPTOR_BLOCK_REAL(name) BLOCK_REAL(name)
 
diff --git a/compiler-rt/lib/tsan/rtl/tsan_mman.cpp b/compiler-rt/lib/tsan/rtl/tsan_mman.cpp
--- a/compiler-rt/lib/tsan/rtl/tsan_mman.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_mman.cpp
@@ -148,7 +148,7 @@
   ObtainCurrentStack(thr, pc, &stack);
   if (IsFiredSuppression(ctx, ReportTypeSignalUnsafe, stack))
     return;
-  ThreadRegistryLock l(ctx->thread_registry);
+  ThreadRegistryLock l(&ctx->thread_registry);
   ScopedReport rep(ReportTypeSignalUnsafe);
   rep.AddStack(stack, true);
   OutputReport(thr, rep);
diff --git a/compiler-rt/lib/tsan/rtl/tsan_rtl.h b/compiler-rt/lib/tsan/rtl/tsan_rtl.h
--- a/compiler-rt/lib/tsan/rtl/tsan_rtl.h
+++ b/compiler-rt/lib/tsan/rtl/tsan_rtl.h
@@ -530,7 +530,7 @@
   void *background_thread;
   atomic_uint32_t stop_background_thread;
 
-  ThreadRegistry *thread_registry;
+  ThreadRegistry thread_registry;
 
   Mutex racy_mtx;
   Vector<RacyStacks> racy_stacks;
diff --git a/compiler-rt/lib/tsan/rtl/tsan_rtl.cpp b/compiler-rt/lib/tsan/rtl/tsan_rtl.cpp
--- a/compiler-rt/lib/tsan/rtl/tsan_rtl.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_rtl.cpp
@@ -77,8 +77,6 @@
 }
 #endif
 
-static ALIGNED(64) char thread_registry_placeholder[sizeof(ThreadRegistry)];
-
 static ThreadContextBase *CreateThreadContext(u32 tid) {
   // Map thread trace when context is created.
   char name[50];
@@ -118,8 +116,8 @@
       report_mtx(MutexTypeReport),
       nreported(),
       nmissed_expected(),
-      thread_registry(new (thread_registry_placeholder) ThreadRegistry(
-          CreateThreadContext, kMaxTid, kThreadQuarantineSize, kMaxTidReuse)),
+      thread_registry(CreateThreadContext, kMaxTid, kThreadQuarantineSize,
+                      kMaxTidReuse),
       racy_mtx(MutexTypeRacy),
       racy_stacks(),
       racy_addresses(),
@@ -161,7 +159,7 @@
 static void MemoryProfiler(Context *ctx, fd_t fd, int i) {
   uptr n_threads;
   uptr n_running_threads;
-  ctx->thread_registry->GetNumberOfThreads(&n_threads, &n_running_threads);
+  ctx->thread_registry.GetNumberOfThreads(&n_threads, &n_running_threads);
   InternalMmapVector<char> buf(4096);
   WriteMemoryProfile(buf.data(), buf.size(), n_threads, n_running_threads);
   WriteToFile(fd, buf.data(), internal_strlen(buf.data()));
@@ -526,7 +524,7 @@
 
 #if !SANITIZER_GO
 void ForkBefore(ThreadState *thr, uptr pc) NO_THREAD_SAFETY_ANALYSIS {
-  ctx->thread_registry->Lock();
+  ctx->thread_registry.Lock();
   ctx->report_mtx.Lock();
   ScopedErrorReportLock::Lock();
   // Suppress all reports in the pthread_atfork callbacks.
@@ -545,7 +543,7 @@
   thr->ignore_interceptors--;
   ScopedErrorReportLock::Unlock();
   ctx->report_mtx.Unlock();
-  ctx->thread_registry->Unlock();
+  ctx->thread_registry.Unlock();
 }
 
 void ForkChildAfter(ThreadState *thr, uptr pc) NO_THREAD_SAFETY_ANALYSIS {
@@ -553,10 +551,10 @@
   thr->ignore_interceptors--;
   ScopedErrorReportLock::Unlock();
   ctx->report_mtx.Unlock();
-  ctx->thread_registry->Unlock();
+  ctx->thread_registry.Unlock();
 
   uptr nthread = 0;
-  ctx->thread_registry->GetNumberOfThreads(0, 0, &nthread /* alive threads */);
+  ctx->thread_registry.GetNumberOfThreads(0, 0, &nthread /* alive threads */);
   VPrintf(1, "ThreadSanitizer: forked new process with pid %d,"
       " parent had %d threads\n", (int)internal_getpid(), (int)nthread);
   if (nthread == 1) {
diff --git a/compiler-rt/lib/tsan/rtl/tsan_rtl_mutex.cpp b/compiler-rt/lib/tsan/rtl/tsan_rtl_mutex.cpp
--- a/compiler-rt/lib/tsan/rtl/tsan_rtl_mutex.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_rtl_mutex.cpp
@@ -53,7 +53,7 @@
     return;
   if (!ShouldReport(thr, typ))
     return;
-  ThreadRegistryLock l(ctx->thread_registry);
+  ThreadRegistryLock l(&ctx->thread_registry);
   ScopedReport rep(typ);
   rep.AddMutex(mid);
   VarSizeStackTrace trace;
@@ -107,7 +107,7 @@
     s->Reset(thr->proc());  // must not reset it before the report is printed
   s->mtx.Unlock();
   if (unlock_locked && ShouldReport(thr, ReportTypeMutexDestroyLocked)) {
-    ThreadRegistryLock l(ctx->thread_registry);
+    ThreadRegistryLock l(&ctx->thread_registry);
     ScopedReport rep(ReportTypeMutexDestroyLocked);
     rep.AddMutex(mid);
     VarSizeStackTrace trace;
@@ -417,9 +417,8 @@
   DPrintf("#%d: AcquireGlobal\n", thr->tid);
   if (thr->ignore_sync)
     return;
-  ThreadRegistryLock l(ctx->thread_registry);
-  ctx->thread_registry->RunCallbackForEachThreadLocked(
-      UpdateClockCallback, thr);
+  ThreadRegistryLock l(&ctx->thread_registry);
+  ctx->thread_registry.RunCallbackForEachThreadLocked(UpdateClockCallback, thr);
 }
 
 void ReleaseStoreAcquire(ThreadState *thr, uptr pc, uptr addr) NO_THREAD_SAFETY_ANALYSIS {
@@ -473,9 +472,9 @@
   if (thr->ignore_sync)
     return;
   thr->last_sleep_stack_id = CurrentStackId(thr, pc);
-  ThreadRegistryLock l(ctx->thread_registry);
-  ctx->thread_registry->RunCallbackForEachThreadLocked(
-      UpdateSleepClockCallback, thr);
+  ThreadRegistryLock l(&ctx->thread_registry);
+  ctx->thread_registry.RunCallbackForEachThreadLocked(UpdateSleepClockCallback,
+                                                      thr);
 }
 #endif
 
@@ -521,7 +520,7 @@
 void ReportDeadlock(ThreadState *thr, uptr pc, DDReport *r) {
   if (r == 0 || !ShouldReport(thr, ReportTypeDeadlock))
     return;
-  ThreadRegistryLock l(ctx->thread_registry);
+  ThreadRegistryLock l(&ctx->thread_registry);
   ScopedReport rep(ReportTypeDeadlock);
   for (int i = 0; i < r->n; i++) {
     rep.AddMutex(r->loop[i].mtx_ctx0);
diff --git a/compiler-rt/lib/tsan/rtl/tsan_rtl_report.cpp b/compiler-rt/lib/tsan/rtl/tsan_rtl_report.cpp
--- a/compiler-rt/lib/tsan/rtl/tsan_rtl_report.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_rtl_report.cpp
@@ -134,7 +134,7 @@
   CheckedMutex::CheckNoLocks();
   // For the same reason check we didn't lock thread_registry yet.
   if (SANITIZER_DEBUG)
-    ThreadRegistryLock l(ctx->thread_registry);
+    ThreadRegistryLock l(&ctx->thread_registry);
   if (!flags()->report_bugs || thr->suppress_reports)
     return false;
   switch (typ) {
@@ -156,7 +156,7 @@
 }
 
 ScopedReportBase::ScopedReportBase(ReportType typ, uptr tag) {
-  ctx->thread_registry->CheckLocked();
+  ctx->thread_registry.CheckLocked();
   void *mem = internal_alloc(sizeof(ReportDesc));
   rep_ = new(mem) ReportDesc;
   rep_->typ = typ;
@@ -229,16 +229,16 @@
 }
 
 static ThreadContext *FindThreadByUidLocked(int unique_id) {
-  ctx->thread_registry->CheckLocked();
+  ctx->thread_registry.CheckLocked();
   return static_cast<ThreadContext *>(
-      ctx->thread_registry->FindThreadContextLocked(
+      ctx->thread_registry.FindThreadContextLocked(
           FindThreadByUidLockedCallback, &unique_id));
 }
 
 static ThreadContext *FindThreadByTidLocked(int tid) {
-  ctx->thread_registry->CheckLocked();
-  return static_cast<ThreadContext*>(
-      ctx->thread_registry->GetThreadLocked(tid));
+  ctx->thread_registry.CheckLocked();
+  return static_cast<ThreadContext *>(
+      ctx->thread_registry.GetThreadLocked(tid));
 }
 
 static bool IsInStackOrTls(ThreadContextBase *tctx_base, void *arg) {
@@ -253,10 +253,10 @@
 }
 
 ThreadContext *IsThreadStackOrTls(uptr addr, bool *is_stack) {
-  ctx->thread_registry->CheckLocked();
-  ThreadContext *tctx = static_cast<ThreadContext*>(
-      ctx->thread_registry->FindThreadContextLocked(IsInStackOrTls,
-                                                    (void*)addr));
+  ctx->thread_registry.CheckLocked();
+  ThreadContext *tctx =
+      static_cast<ThreadContext *>(ctx->thread_registry.FindThreadContextLocked(
+          IsInStackOrTls, (void *)addr));
   if (!tctx)
     return 0;
   ThreadState *thr = tctx->thr;
@@ -697,7 +697,7 @@
     }
   }
 
-  ThreadRegistryLock l0(ctx->thread_registry);
+  ThreadRegistryLock l0(&ctx->thread_registry);
   ScopedReport rep(typ, tag);
   for (uptr i = 0; i < kMop; i++) {
     Shadow s(thr->racy_state[i]);
@@ -707,8 +707,8 @@
 
   for (uptr i = 0; i < kMop; i++) {
     FastState s(thr->racy_state[i]);
-    ThreadContext *tctx = static_cast<ThreadContext*>(
-        ctx->thread_registry->GetThreadLocked(s.tid()));
+    ThreadContext *tctx = static_cast<ThreadContext *>(
+        ctx->thread_registry.GetThreadLocked(s.tid()));
     if (s.epoch() < tctx->epoch0 || s.epoch() > tctx->epoch1)
       continue;
     rep.AddThread(tctx);
diff --git a/compiler-rt/lib/tsan/rtl/tsan_rtl_thread.cpp b/compiler-rt/lib/tsan/rtl/tsan_rtl_thread.cpp
--- a/compiler-rt/lib/tsan/rtl/tsan_rtl_thread.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_rtl_thread.cpp
@@ -205,10 +205,10 @@
 #if !SANITIZER_GO
   if (!ShouldReport(thr, ReportTypeThreadLeak))
     return;
-  ThreadRegistryLock l(ctx->thread_registry);
+  ThreadRegistryLock l(&ctx->thread_registry);
   Vector<ThreadLeak> leaks;
-  ctx->thread_registry->RunCallbackForEachThreadLocked(
-      MaybeReportThreadLeak, &leaks);
+  ctx->thread_registry.RunCallbackForEachThreadLocked(MaybeReportThreadLeak,
+                                                      &leaks);
   for (uptr i = 0; i < leaks.Size(); i++) {
     ScopedReport rep(ReportTypeThreadLeak);
     rep.AddThread(leaks[i].tctx, true);
@@ -220,15 +220,14 @@
 
 int ThreadCount(ThreadState *thr) {
   uptr result;
-  ctx->thread_registry->GetNumberOfThreads(0, 0, &result);
+  ctx->thread_registry.GetNumberOfThreads(0, 0, &result);
   return (int)result;
 }
 
 int ThreadCreate(ThreadState *thr, uptr pc, uptr uid, bool detached) {
   OnCreatedArgs args = { thr, pc };
   u32 parent_tid = thr ? thr->tid : kInvalidTid;  // No parent for GCD workers.
-  int tid =
-      ctx->thread_registry->CreateThread(uid, detached, parent_tid, &args);
+  int tid = ctx->thread_registry.CreateThread(uid, detached, parent_tid, &args);
   DPrintf("#%d: ThreadCreate tid=%d uid=%zu\n", parent_tid, tid, uid);
   return tid;
 }
@@ -252,7 +251,7 @@
   }
 #endif
 
-  ThreadRegistry *tr = ctx->thread_registry;
+  ThreadRegistry *tr = &ctx->thread_registry;
   OnStartedArgs args = { thr, stk_addr, stk_size, tls_addr, tls_size };
   tr->StartThread(tid, os_id, thread_type, &args);
 
@@ -276,7 +275,7 @@
   if (thr->tls_addr && thr->tls_size)
     DontNeedShadowFor(thr->tls_addr, thr->tls_size);
   thr->is_dead = true;
-  ctx->thread_registry->FinishThread(thr->tid);
+  ctx->thread_registry.FinishThread(thr->tid);
 }
 
 struct ConsumeThreadContext {
@@ -303,7 +302,7 @@
 
 int ThreadConsumeTid(ThreadState *thr, uptr pc, uptr uid) {
   ConsumeThreadContext findCtx = {uid, nullptr};
-  ctx->thread_registry->FindThread(ConsumeThreadByUid, &findCtx);
+  ctx->thread_registry.FindThread(ConsumeThreadByUid, &findCtx);
   int tid = findCtx.tctx ? findCtx.tctx->tid : kInvalidTid;
   DPrintf("#%d: ThreadTid uid=%zu tid=%d\n", thr->tid, uid, tid);
   return tid;
@@ -313,23 +312,23 @@
   CHECK_GT(tid, 0);
   CHECK_LT(tid, kMaxTid);
   DPrintf("#%d: ThreadJoin tid=%d\n", thr->tid, tid);
-  ctx->thread_registry->JoinThread(tid, thr);
+  ctx->thread_registry.JoinThread(tid, thr);
 }
 
 void ThreadDetach(ThreadState *thr, uptr pc, int tid) {
   CHECK_GT(tid, 0);
   CHECK_LT(tid, kMaxTid);
-  ctx->thread_registry->DetachThread(tid, thr);
+  ctx->thread_registry.DetachThread(tid, thr);
 }
 
 void ThreadNotJoined(ThreadState *thr, uptr pc, int tid, uptr uid) {
   CHECK_GT(tid, 0);
   CHECK_LT(tid, kMaxTid);
-  ctx->thread_registry->SetThreadUserId(tid, uid);
+  ctx->thread_registry.SetThreadUserId(tid, uid);
 }
 
 void ThreadSetName(ThreadState *thr, const char *name) {
-  ctx->thread_registry->SetThreadName(thr->tid, name);
+  ctx->thread_registry.SetThreadName(thr->tid, name);
 }
 
 void MemoryAccessRange(ThreadState *thr, uptr pc, uptr addr,