/*

 * Copyright (c) 1999, 2018, 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 "asm/macroAssembler.inline.hpp"

#include "c1/c1_MacroAssembler.hpp"

#include "c1/c1_Runtime1.hpp"

#include "classfile/systemDictionary.hpp"

#include "gc/shared/collectedHeap.hpp"

#include "interpreter/interpreter.hpp"

#include "oops/arrayOop.hpp"

#include "oops/markOop.hpp"

#include "runtime/basicLock.hpp"

#include "runtime/biasedLocking.hpp"

#include "runtime/os.hpp"

#include "runtime/sharedRuntime.hpp"

#include "runtime/stubRoutines.hpp"

void C1_MacroAssembler::inline_cache_check(Register receiver, Register iCache) {

  Label L;

  const Register temp_reg = G3_scratch;

  // Note: needs more testing of out-of-line vs. inline slow case

  verify_oop(receiver);

  load_klass(receiver, temp_reg);

  cmp_and_brx_short(temp_reg, iCache, Assembler::equal, Assembler::pt, L);

  AddressLiteral ic_miss(SharedRuntime::get_ic_miss_stub());

  jump_to(ic_miss, temp_reg);

  delayed()->nop();

  align(CodeEntryAlignment);

  bind(L);

}

void C1_MacroAssembler::explicit_null_check(Register base) {

  Unimplemented();

}

void C1_MacroAssembler::build_frame(int frame_size_in_bytes, int bang_size_in_bytes) {

  assert(bang_size_in_bytes >= frame_size_in_bytes, "stack bang size incorrect");

  generate_stack_overflow_check(bang_size_in_bytes);

  // Create the frame.

  save_frame_c1(frame_size_in_bytes);

}

void C1_MacroAssembler::verified_entry() {

  if (C1Breakpoint) breakpoint_trap();

  // build frame

  verify_FPU(0, "method_entry");

}

void C1_MacroAssembler::lock_object(Register Rmark, Register Roop, Register Rbox, Register Rscratch, Label& slow_case) {

  assert_different_registers(Rmark, Roop, Rbox, Rscratch);

  Label done;

  Address mark_addr(Roop, oopDesc::mark_offset_in_bytes());

  // The following move must be the first instruction of emitted since debug

  // information may be generated for it.

  // Load object header

  ld_ptr(mark_addr, Rmark);

  verify_oop(Roop);

  // save object being locked into the BasicObjectLock

  st_ptr(Roop, Rbox, BasicObjectLock::obj_offset_in_bytes());

  if (UseBiasedLocking) {

    biased_locking_enter(Roop, Rmark, Rscratch, done, &slow_case);

  }

  // Save Rbox in Rscratch to be used for the cas operation

  mov(Rbox, Rscratch);

  // and mark it unlocked

  // save unlocked object header into the displaced header location on the stack

  st_ptr(Rmark, Rbox, BasicLock::displaced_header_offset_in_bytes());

  assert(mark_addr.disp() == 0, "cas must take a zero displacement");

  cas_ptr(mark_addr.base(), Rmark, Rscratch);

  // if compare/exchange succeeded we found an unlocked object and we now have locked it

  // hence we are done

  cmp(Rmark, Rscratch);

  brx(Assembler::equal, false, Assembler::pt, done);

  delayed()->sub(Rscratch, SP, Rscratch);  //pull next instruction into delay slot

  // we did not find an unlocked object so see if this is a recursive case

  // sub(Rscratch, SP, Rscratch);

  assert(os::vm_page_size() > 0xfff, "page size too small - change the constant");

  andcc(Rscratch, 0xfffff003, Rscratch);

  brx(Assembler::notZero, false, Assembler::pn, slow_case);

  delayed()->st_ptr(Rscratch, Rbox, BasicLock::displaced_header_offset_in_bytes());

  bind(done);

}

void C1_MacroAssembler::unlock_object(Register Rmark, Register Roop, Register Rbox, Label& slow_case) {

  assert_different_registers(Rmark, Roop, Rbox);

  Label done;

  Address mark_addr(Roop, oopDesc::mark_offset_in_bytes());

  assert(mark_addr.disp() == 0, "cas must take a zero displacement");

  if (UseBiasedLocking) {

    // load the object out of the BasicObjectLock

    ld_ptr(Rbox, BasicObjectLock::obj_offset_in_bytes(), Roop);

    verify_oop(Roop);

    biased_locking_exit(mark_addr, Rmark, done);

  }

  // Test first it it is a fast recursive unlock

  ld_ptr(Rbox, BasicLock::displaced_header_offset_in_bytes(), Rmark);

  br_null_short(Rmark, Assembler::pt, done);

  if (!UseBiasedLocking) {

    // load object

    ld_ptr(Rbox, BasicObjectLock::obj_offset_in_bytes(), Roop);

    verify_oop(Roop);

  }

  // Check if it is still a light weight lock, this is is true if we see

  cas_ptr(mark_addr.base(), Rbox, Rmark);

  cmp(Rbox, Rmark);

  brx(Assembler::notEqual, false, Assembler::pn, slow_case);

  delayed()->nop();

  // Done

  bind(done);

}

void C1_MacroAssembler::try_allocate(

  Register obj,                        // result: pointer to object after successful allocation

  Register var_size_in_bytes,          // object size in bytes if unknown at compile time; invalid otherwise

  int      con_size_in_bytes,          // object size in bytes if   known at compile time

  Register t1,                         // temp register, must be global register for incr_allocated_bytes

  Register t2,                         // temp register

  Label&   slow_case                   // continuation point if fast allocation fails

) {

  RegisterOrConstant size_in_bytes = var_size_in_bytes->is_valid()

    ? RegisterOrConstant(var_size_in_bytes) : RegisterOrConstant(con_size_in_bytes);

  if (UseTLAB) {

    tlab_allocate(obj, var_size_in_bytes, con_size_in_bytes, t1, slow_case);

  } else {

    eden_allocate(obj, var_size_in_bytes, con_size_in_bytes, t1, t2, slow_case);

    incr_allocated_bytes(size_in_bytes, t1, t2);

  }

}

void C1_MacroAssembler::initialize_header(Register obj, Register klass, Register len, Register t1, Register t2) {

  assert_different_registers(obj, klass, len, t1, t2);

  if (UseBiasedLocking && !len->is_valid()) {

    ld_ptr(klass, in_bytes(Klass::prototype_header_offset()), t1);

  } else {

  }

  st_ptr(t1, obj, oopDesc::mark_offset_in_bytes());

  if (UseCompressedClassPointers) {

    // Save klass

    mov(klass, t1);

    encode_klass_not_null(t1);

    stw(t1, obj, oopDesc::klass_offset_in_bytes());

  } else {

    st_ptr(klass, obj, oopDesc::klass_offset_in_bytes());

  }

  if (len->is_valid()) {

    st(len, obj, arrayOopDesc::length_offset_in_bytes());

  } else if (UseCompressedClassPointers) {

    // otherwise length is in the class gap

    store_klass_gap(G0, obj);

  }

}

void C1_MacroAssembler::initialize_body(Register base, Register index) {

  zero_memory(base, index);

}

void C1_MacroAssembler::allocate_object(

  Register obj,                        // result: pointer to object after successful allocation

  Register t1,                         // temp register

  Register t2,                         // temp register, must be a global register for try_allocate

  Register t3,                         // temp register

  int      hdr_size,                   // object header size in words

  int      obj_size,                   // object size in words

  Register klass,                      // object klass

  Label&   slow_case                   // continuation point if fast allocation fails

) {

  assert_different_registers(obj, t1, t2, t3, klass);

  assert(klass == G5, "must be G5");

  // allocate space & initialize header

  if (!is_simm13(obj_size * wordSize)) {

    // would need to use extra register to load

    // object size => go the slow case for now

    ba(slow_case);

    delayed()->nop();

    return;

  }

  try_allocate(obj, noreg, obj_size * wordSize, t2, t3, slow_case);

  initialize_object(obj, klass, noreg, obj_size * HeapWordSize, t1, t2, /* is_tlab_allocated */ UseTLAB);

}

void C1_MacroAssembler::initialize_object(

  Register obj,                        // result: pointer to object after successful allocation

  Register klass,                      // object klass

  Register var_size_in_bytes,          // object size in bytes if unknown at compile time; invalid otherwise

  int      con_size_in_bytes,          // object size in bytes if   known at compile time

  Register t1,                         // temp register

  Register t2,                         // temp register

  bool     is_tlab_allocated           // the object was allocated in a TLAB; relevant for the implementation of ZeroTLAB

  ) {

  const int hdr_size_in_bytes = instanceOopDesc::header_size() * HeapWordSize;

  initialize_header(obj, klass, noreg, t1, t2);

#ifdef ASSERT

  {

    Label ok;

    ld(klass, in_bytes(Klass::layout_helper_offset()), t1);

    if (var_size_in_bytes != noreg) {

      cmp_and_brx_short(t1, var_size_in_bytes, Assembler::equal, Assembler::pt, ok);

    } else {

      cmp_and_brx_short(t1, con_size_in_bytes, Assembler::equal, Assembler::pt, ok);

    }

    stop("bad size in initialize_object");

    should_not_reach_here();

    bind(ok);

  }

#endif

  if (!(UseTLAB && ZeroTLAB && is_tlab_allocated)) {

    // initialize body

    const int threshold = 5 * HeapWordSize;              // approximate break even point for code size

    if (var_size_in_bytes != noreg) {

      // use a loop

      add(obj, hdr_size_in_bytes, t1);               // compute address of first element

      sub(var_size_in_bytes, hdr_size_in_bytes, t2); // compute size of body

      initialize_body(t1, t2);

    } else if (con_size_in_bytes <= threshold) {

      // use explicit NULL stores

      for (int i = hdr_size_in_bytes; i < con_size_in_bytes; i += HeapWordSize)     st_ptr(G0, obj, i);

    } else if (con_size_in_bytes > hdr_size_in_bytes) {

      // use a loop

      const Register base  = t1;

      const Register index = t2;

      add(obj, hdr_size_in_bytes, base);               // compute address of first element

      // compute index = number of words to clear

      set(con_size_in_bytes - hdr_size_in_bytes, index);

      initialize_body(base, index);

    }

  }

  if (CURRENT_ENV->dtrace_alloc_probes()) {

    assert(obj == O0, "must be");

    call(CAST_FROM_FN_PTR(address, Runtime1::entry_for(Runtime1::dtrace_object_alloc_id)),

         relocInfo::runtime_call_type);

    delayed()->nop();

  }

  verify_oop(obj);

}

void C1_MacroAssembler::allocate_array(

  Register obj,                        // result: pointer to array after successful allocation

  Register len,                        // array length

  Register t1,                         // temp register

  Register t2,                         // temp register

  Register t3,                         // temp register

  int      hdr_size,                   // object header size in words

  int      elt_size,                   // element size in bytes

  Register klass,                      // object klass

  Label&   slow_case                   // continuation point if fast allocation fails

) {

  assert_different_registers(obj, len, t1, t2, t3, klass);

  assert(klass == G5, "must be G5");

  assert(t1 == G1, "must be G1");

  // determine alignment mask

  assert(!(BytesPerWord & 1), "must be a multiple of 2 for masking code to work");

  // check for negative or excessive length

  // note: the maximum length allowed is chosen so that arrays of any

  //       element size with this length are always smaller or equal

  //       to the largest integer (i.e., array size computation will

  //       not overflow)

  set(max_array_allocation_length, t1);

  cmp(len, t1);

  br(Assembler::greaterUnsigned, false, Assembler::pn, slow_case);

  // compute array size

  // note: if 0 <= len <= max_length, len*elt_size + header + alignment is

  //       smaller or equal to the largest integer; also, since top is always

  //       aligned, we can do the alignment here instead of at the end address

  //       computation

  const Register arr_size = t1;

  switch (elt_size) {

    case  1: delayed()->mov(len,    arr_size); break;

    case  2: delayed()->sll(len, 1, arr_size); break;

    case  4: delayed()->sll(len, 2, arr_size); break;

    case  8: delayed()->sll(len, 3, arr_size); break;

    default: ShouldNotReachHere();

  }

  add(arr_size, hdr_size * wordSize + MinObjAlignmentInBytesMask, arr_size); // add space for header & alignment

  and3(arr_size, ~MinObjAlignmentInBytesMask, arr_size);                     // align array size

  // allocate space & initialize header

  if (UseTLAB) {

    tlab_allocate(obj, arr_size, 0, t2, slow_case);

  } else {

    eden_allocate(obj, arr_size, 0, t2, t3, slow_case);

  }

  initialize_header(obj, klass, len, t2, t3);

  // initialize body

  const Register base  = t2;

  const Register index = t3;

  add(obj, hdr_size * wordSize, base);               // compute address of first element

  sub(arr_size, hdr_size * wordSize, index);         // compute index = number of words to clear

  initialize_body(base, index);

  if (CURRENT_ENV->dtrace_alloc_probes()) {

    assert(obj == O0, "must be");

    call(CAST_FROM_FN_PTR(address, Runtime1::entry_for(Runtime1::dtrace_object_alloc_id)),

         relocInfo::runtime_call_type);

    delayed()->nop();

  }

  verify_oop(obj);

}

#ifndef PRODUCT

void C1_MacroAssembler::verify_stack_oop(int stack_offset) {

  if (!VerifyOops) return;

  verify_oop_addr(Address(SP, stack_offset + STACK_BIAS));

}

void C1_MacroAssembler::verify_not_null_oop(Register r) {

  Label not_null;

  br_notnull_short(r, Assembler::pt, not_null);

  stop("non-null oop required");

  bind(not_null);

  if (!VerifyOops) return;

  verify_oop(r);

}

void C1_MacroAssembler::invalidate_registers(bool iregisters, bool lregisters, bool oregisters,

                                             Register preserve1, Register preserve2) {

  if (iregisters) {

    for (int i = 0; i < 6; i++) {

      Register r = as_iRegister(i);

      if (r != preserve1 && r != preserve2)  set(0xdead, r);

    }

  }

  if (oregisters) {

    for (int i = 0; i < 6; i++) {

      Register r = as_oRegister(i);

      if (r != preserve1 && r != preserve2)  set(0xdead, r);

    }

  }

  if (lregisters) {

    for (int i = 0; i < 8; i++) {

      Register r = as_lRegister(i);

      if (r != preserve1 && r != preserve2)  set(0xdead, r);

    }

  }

}

#endif