view src/share/vm/code/vtableStubs.cpp @ 2149:7e37af9d69ef

7011379: G1: overly long concurrent marking cycles Summary: This changeset introduces filtering of SATB buffers at the point when they are about to be enqueued. If this filtering clears enough entries on each buffer, the buffer can then be re-used and not enqueued. This cuts down the number of SATB buffers that need to be processed by the concurrent marking threads. Reviewed-by: johnc, ysr
author tonyp
date Wed, 19 Jan 2011 09:35:17 -0500
parents f95d63e2154a
children da91efe96a93
line wrap: on
line source

/*
 * Copyright (c) 1997, 2010, 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 "code/vtableStubs.hpp"
#include "compiler/disassembler.hpp"
#include "memory/allocation.inline.hpp"
#include "memory/resourceArea.hpp"
#include "oops/instanceKlass.hpp"
#include "oops/klassVtable.hpp"
#include "oops/oop.inline.hpp"
#include "prims/forte.hpp"
#include "prims/jvmtiExport.hpp"
#include "runtime/handles.inline.hpp"
#include "runtime/mutexLocker.hpp"
#include "runtime/sharedRuntime.hpp"
#ifdef COMPILER2
#include "opto/matcher.hpp"
#endif

// -----------------------------------------------------------------------------------------
// Implementation of VtableStub

address VtableStub::_chunk             = NULL;
address VtableStub::_chunk_end         = NULL;
VMReg   VtableStub::_receiver_location = VMRegImpl::Bad();

static int num_vtable_chunks = 0;


void* VtableStub::operator new(size_t size, int code_size) {
  assert(size == sizeof(VtableStub), "mismatched size");
  num_vtable_chunks++;
  // compute real VtableStub size (rounded to nearest word)
  const int real_size = round_to(code_size + sizeof(VtableStub), wordSize);
  // malloc them in chunks to minimize header overhead
  const int chunk_factor = 32;
  if (_chunk == NULL || _chunk + real_size > _chunk_end) {
    const int bytes = chunk_factor * real_size + pd_code_alignment();
    BufferBlob* blob = BufferBlob::create("vtable chunks", bytes);
    if (blob == NULL) {
      vm_exit_out_of_memory(bytes, "CodeCache: no room for vtable chunks");
    }
    _chunk = blob->content_begin();
    _chunk_end = _chunk + bytes;
    Forte::register_stub("vtable stub", _chunk, _chunk_end);
    // Notify JVMTI about this stub. The event will be recorded by the enclosing
    // JvmtiDynamicCodeEventCollector and posted when this thread has released
    // all locks.
    if (JvmtiExport::should_post_dynamic_code_generated()) {
      JvmtiExport::post_dynamic_code_generated_while_holding_locks("vtable stub", _chunk, _chunk_end);
    }
    align_chunk();
  }
  assert(_chunk + real_size <= _chunk_end, "bad allocation");
  void* res = _chunk;
  _chunk += real_size;
  align_chunk();
 return res;
}


void VtableStub::print_on(outputStream* st) const {
  st->print("vtable stub (index = %d, receiver_location = %d, code = [" INTPTR_FORMAT ", " INTPTR_FORMAT "[)",
             index(), receiver_location(), code_begin(), code_end());
}


// -----------------------------------------------------------------------------------------
// Implementation of VtableStubs
//
// For each hash value there's a linked list of vtable stubs (with that
// hash value). Each list is anchored in a little hash _table, indexed
// by that hash value.

VtableStub* VtableStubs::_table[VtableStubs::N];
int VtableStubs::_number_of_vtable_stubs = 0;


void VtableStubs::initialize() {
  VtableStub::_receiver_location = SharedRuntime::name_for_receiver();
  {
    MutexLocker ml(VtableStubs_lock);
    assert(_number_of_vtable_stubs == 0, "potential performance bug: VtableStubs initialized more than once");
    assert(is_power_of_2(N), "N must be a power of 2");
    for (int i = 0; i < N; i++) {
      _table[i] = NULL;
    }
  }
}


address VtableStubs::create_stub(bool is_vtable_stub, int vtable_index, methodOop method) {
  assert(vtable_index >= 0, "must be positive");

  VtableStub* s = ShareVtableStubs ? lookup(is_vtable_stub, vtable_index) : NULL;
  if (s == NULL) {
    if (is_vtable_stub) {
      s = create_vtable_stub(vtable_index);
    } else {
      s = create_itable_stub(vtable_index);
    }
    enter(is_vtable_stub, vtable_index, s);
    if (PrintAdapterHandlers) {
      tty->print_cr("Decoding VtableStub %s[%d]@%d",
                    is_vtable_stub? "vtbl": "itbl", vtable_index, VtableStub::receiver_location());
      Disassembler::decode(s->code_begin(), s->code_end());
    }
  }
  return s->entry_point();
}


inline uint VtableStubs::hash(bool is_vtable_stub, int vtable_index){
  // Assumption: receiver_location < 4 in most cases.
  int hash = ((vtable_index << 2) ^ VtableStub::receiver_location()->value()) + vtable_index;
  return (is_vtable_stub ? ~hash : hash)  & mask;
}


VtableStub* VtableStubs::lookup(bool is_vtable_stub, int vtable_index) {
  MutexLocker ml(VtableStubs_lock);
  unsigned hash = VtableStubs::hash(is_vtable_stub, vtable_index);
  VtableStub* s = _table[hash];
  while( s && !s->matches(is_vtable_stub, vtable_index)) s = s->next();
  return s;
}


void VtableStubs::enter(bool is_vtable_stub, int vtable_index, VtableStub* s) {
  MutexLocker ml(VtableStubs_lock);
  assert(s->matches(is_vtable_stub, vtable_index), "bad vtable stub");
  unsigned int h = VtableStubs::hash(is_vtable_stub, vtable_index);
  // enter s at the beginning of the corresponding list
  s->set_next(_table[h]);
  _table[h] = s;
  _number_of_vtable_stubs++;
}


bool VtableStubs::is_entry_point(address pc) {
  MutexLocker ml(VtableStubs_lock);
  VtableStub* stub = (VtableStub*)(pc - VtableStub::entry_offset());
  uint hash = VtableStubs::hash(stub->is_vtable_stub(), stub->index());
  VtableStub* s;
  for (s = _table[hash]; s != NULL && s != stub; s = s->next()) {}
  return s == stub;
}


bool VtableStubs::contains(address pc) {
  // simple solution for now - we may want to use
  // a faster way if this function is called often
  return stub_containing(pc) != NULL;
}


VtableStub* VtableStubs::stub_containing(address pc) {
  // Note: No locking needed since any change to the data structure
  //       happens with an atomic store into it (we don't care about
  //       consistency with the _number_of_vtable_stubs counter).
  for (int i = 0; i < N; i++) {
    for (VtableStub* s = _table[i]; s != NULL; s = s->next()) {
      if (s->contains(pc)) return s;
    }
  }
  return NULL;
}

void vtableStubs_init() {
  VtableStubs::initialize();
}


//-----------------------------------------------------------------------------------------------------
// Non-product code
#ifndef PRODUCT

extern "C" void bad_compiled_vtable_index(JavaThread* thread, oop receiver, int index) {
  ResourceMark rm;
  HandleMark hm;
  klassOop klass = receiver->klass();
  instanceKlass* ik = instanceKlass::cast(klass);
  klassVtable* vt = ik->vtable();
  klass->print();
  fatal(err_msg("bad compiled vtable dispatch: receiver " INTPTR_FORMAT ", "
                "index %d (vtable length %d)",
                (address)receiver, index, vt->length()));
}

#endif // Product