src/hotspot/share/gc/shared/allocTracer.cpp

/*
 * Copyright (c) 2013, 2020, Oracle and/or its affiliates. All rights reserved.
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This code is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 only, as
 * published by the Free Software Foundation.
 *
 * This code is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 * version 2 for more details (a copy is included in the LICENSE file that
 * accompanied this code).
 *
 * You should have received a copy of the GNU General Public License version
 * 2 along with this work; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 *
 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 * or visit www.oracle.com if you need additional information or have any
 * questions.
 *
 */

#include "precompiled.hpp"
#include "gc/shared/allocTracer.hpp"
#include "gc/shared/threadLocalAllocBuffer.inline.hpp"
#include "jfr/jfrEvents.hpp"
#include "utilities/globalDefinitions.hpp"
#include "utilities/macros.hpp"
#if INCLUDE_JFR
#include "jfr/support/jfrAllocationTracer.hpp"
#endif

static THREAD_LOCAL int64_t _last_allocated_bytes = 0;

inline void send_allocation_sample(const Klass* klass, int64_t allocated_bytes) {
  assert(allocated_bytes > 0, "invariant");
  EventObjectAllocationSample event;
  if (event.should_commit()) {
    const size_t weight = allocated_bytes - _last_allocated_bytes;
    assert(weight > 0, "invariant");
    event.set_objectClass(klass);
    event.set_weight(weight);
    event.commit();
    _last_allocated_bytes = allocated_bytes;
  }
}

inline bool send_allocation_sample_with_result(const Klass* klass, int64_t allocated_bytes) {
  assert(allocated_bytes > 0, "invariant");
  EventObjectAllocationSample event;
  if (event.should_commit()) {
    const size_t weight = allocated_bytes - _last_allocated_bytes;
    assert(weight > 0, "invariant");
    event.set_objectClass(klass);
    event.set_weight(weight);
    event.commit();
    _last_allocated_bytes = allocated_bytes;
    return true;
  }
  return false;
}

inline intptr_t estimate_tlab_size_bytes(Thread* thread) {
  assert(thread != NULL, "invariant");
  const size_t desired_tlab_size_bytes = thread->tlab().desired_size() * HeapWordSize;
  const size_t alignment_reserve_bytes = thread->tlab().alignment_reserve_in_bytes();
  assert(desired_tlab_size_bytes > alignment_reserve_bytes, "invariant");
  return static_cast<intptr_t>(desired_tlab_size_bytes - alignment_reserve_bytes);
}

inline int64_t load_allocated_bytes(Thread* thread) {
  const int64_t allocated_bytes = thread->allocated_bytes();
  if (allocated_bytes < _last_allocated_bytes) {
    // A hw thread can detach and reattach to the VM, and when it does,
    // it gets a new JavaThread representation. The thread local variable
    // tracking _last_allocated_bytes is mapped to the existing hw thread,
    // so it needs to be reset.
    _last_allocated_bytes = 0;
  }
  return allocated_bytes == _last_allocated_bytes ? 0 : allocated_bytes;
}

// To avoid large objects from being undersampled compared to the regular TLAB samples,
// the data amount is normalized as if it was a TLAB, giving a number of TLAB sampling attempts to the large object.
static void normalize_as_tlab_and_send_allocation_samples(Klass* klass, intptr_t obj_alloc_size_bytes, Thread* thread) {
  const int64_t allocated_bytes = load_allocated_bytes(thread);
  assert(allocated_bytes > 0, "invariant"); // obj_alloc_size_bytes is already attributed to allocated_bytes at this point.
  if (!UseTLAB) {
    send_allocation_sample(klass, allocated_bytes);
    return;
  }
  const intptr_t tlab_size_bytes = estimate_tlab_size_bytes(thread);
  if (allocated_bytes - _last_allocated_bytes < tlab_size_bytes) {
    return;
  }
  assert(obj_alloc_size_bytes > 0, "invariant");
  do {
    if (send_allocation_sample_with_result(klass, allocated_bytes)) {
      return;
    }
    obj_alloc_size_bytes -= tlab_size_bytes;
  } while (obj_alloc_size_bytes > 0);
}

void AllocTracer::send_allocation_outside_tlab(Klass* klass, HeapWord* obj, size_t alloc_size, Thread* thread) {
  JFR_ONLY(JfrAllocationTracer tracer(obj, alloc_size, thread);)
  EventObjectAllocationOutsideTLAB event;
  if (event.should_commit()) {
    event.set_objectClass(klass);
    event.set_allocationSize(alloc_size);
    event.commit();
  }
  normalize_as_tlab_and_send_allocation_samples(klass, static_cast<intptr_t>(alloc_size), thread);
}

void AllocTracer::send_allocation_in_new_tlab(Klass* klass, HeapWord* obj, size_t tlab_size, size_t alloc_size, Thread* thread) {
  JFR_ONLY(JfrAllocationTracer tracer(obj, alloc_size, thread);)
  EventObjectAllocationInNewTLAB event;
  if (event.should_commit()) {
    event.set_objectClass(klass);
    event.set_allocationSize(alloc_size);
    event.set_tlabSize(tlab_size);
    event.commit();
  }
  const int64_t allocated_bytes = load_allocated_bytes(thread);
  if (allocated_bytes == 0) {
    return;
  }
  send_allocation_sample(klass, allocated_bytes);
}

void AllocTracer::send_allocation_requiring_gc_event(size_t size, uint gcId) {
  EventAllocationRequiringGC event;
  if (event.should_commit()) {
    event.set_gcId(gcId);
    event.set_size(size);
    event.commit();
  }
}