/*

 * Copyright (c) 2008, 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_Compilation.hpp"

#include "c1/c1_LIRAssembler.hpp"

#include "c1/c1_MacroAssembler.hpp"

#include "c1/c1_Runtime1.hpp"

#include "c1/c1_ValueStack.hpp"

#include "ci/ciArrayKlass.hpp"

#include "ci/ciInstance.hpp"

#include "gc/shared/barrierSet.hpp"

#include "gc/shared/cardTableBarrierSet.hpp"

#include "gc/shared/collectedHeap.hpp"

#include "nativeInst_arm.hpp"

#include "oops/objArrayKlass.hpp"

#include "runtime/frame.inline.hpp"

#include "runtime/sharedRuntime.hpp"

#include "vmreg_arm.inline.hpp"

#define __ _masm->

// Note: Rtemp usage is this file should not impact C2 and should be

// correct as long as it is not implicitly used in lower layers (the

// arm [macro]assembler) and used with care in the other C1 specific

// files.

bool LIR_Assembler::is_small_constant(LIR_Opr opr) {

  ShouldNotCallThis(); // Not used on ARM

  return false;

}

LIR_Opr LIR_Assembler::receiverOpr() {

  // The first register in Java calling conventions

  return FrameMap::R0_oop_opr;

}

LIR_Opr LIR_Assembler::osrBufferPointer() {

  return FrameMap::as_pointer_opr(R0);

}

#ifndef PRODUCT

void LIR_Assembler::verify_reserved_argument_area_size(int args_count) {

  assert(args_count * wordSize <= frame_map()->reserved_argument_area_size(), "not enough space for arguments");

}

#endif // !PRODUCT

void LIR_Assembler::store_parameter(jint c, int offset_from_sp_in_words) {

  assert(offset_from_sp_in_words >= 0, "invalid offset from sp");

  int offset_from_sp_in_bytes = offset_from_sp_in_words * BytesPerWord;

  assert(offset_from_sp_in_bytes < frame_map()->reserved_argument_area_size(), "not enough space");

  __ mov_slow(Rtemp, c);

  __ str(Rtemp, Address(SP, offset_from_sp_in_bytes));

}

void LIR_Assembler::store_parameter(Metadata* m, int offset_from_sp_in_words) {

  assert(offset_from_sp_in_words >= 0, "invalid offset from sp");

  int offset_from_sp_in_bytes = offset_from_sp_in_words * BytesPerWord;

  assert(offset_from_sp_in_bytes < frame_map()->reserved_argument_area_size(), "not enough space");

  __ mov_metadata(Rtemp, m);

  __ str(Rtemp, Address(SP, offset_from_sp_in_bytes));

}

//--------------fpu register translations-----------------------

void LIR_Assembler::set_24bit_FPU() {

  ShouldNotReachHere();

}

void LIR_Assembler::reset_FPU() {

  ShouldNotReachHere();

}

void LIR_Assembler::fpop() {

  Unimplemented();

}

void LIR_Assembler::fxch(int i) {

  Unimplemented();

}

void LIR_Assembler::fld(int i) {

  Unimplemented();

}

void LIR_Assembler::ffree(int i) {

  Unimplemented();

}

void LIR_Assembler::breakpoint() {

  __ breakpoint();

}

void LIR_Assembler::push(LIR_Opr opr) {

  Unimplemented();

}

void LIR_Assembler::pop(LIR_Opr opr) {

  Unimplemented();

}

//-------------------------------------------

Address LIR_Assembler::as_Address(LIR_Address* addr) {

  Register base = addr->base()->as_pointer_register();

#ifdef AARCH64

  int align = exact_log2(type2aelembytes(addr->type(), true));

#endif

  if (addr->index()->is_illegal() || addr->index()->is_constant()) {

    int offset = addr->disp();

    if (addr->index()->is_constant()) {

      offset += addr->index()->as_constant_ptr()->as_jint() << addr->scale();

    }

#ifdef AARCH64

    if (!Assembler::is_unsigned_imm_in_range(offset, 12, align) && !Assembler::is_imm_in_range(offset, 9, 0)) {

      BAILOUT_("offset not in range", Address(base));

    }

    assert(UseUnalignedAccesses || (offset & right_n_bits(align)) == 0, "offset should be aligned");

#else

    if ((offset <= -4096) || (offset >= 4096)) {

      BAILOUT_("offset not in range", Address(base));

    }

#endif // AARCH64

    return Address(base, offset);

  } else {

    assert(addr->disp() == 0, "can't have both");

    int scale = addr->scale();

#ifdef AARCH64

    assert((scale == 0) || (scale == align), "scale should be zero or equal to embedded shift");

    bool is_index_extended = (addr->index()->type() == T_INT);

    if (is_index_extended) {

      assert(addr->index()->is_single_cpu(), "should be");

      return Address(base, addr->index()->as_register(), ex_sxtw, scale);

    } else {

      assert(addr->index()->is_double_cpu(), "should be");

      return Address(base, addr->index()->as_register_lo(), ex_lsl, scale);

    }

#else

    assert(addr->index()->is_single_cpu(), "should be");

    return scale >= 0 ? Address(base, addr->index()->as_register(), lsl, scale) :

                        Address(base, addr->index()->as_register(), lsr, -scale);

#endif // AARCH64

  }

}

Address LIR_Assembler::as_Address_hi(LIR_Address* addr) {

#ifdef AARCH64

  ShouldNotCallThis(); // Not used on AArch64

  return Address();

#else

  Address base = as_Address(addr);

  assert(base.index() == noreg, "must be");

  if (base.disp() + BytesPerWord >= 4096) { BAILOUT_("offset not in range", Address(base.base(),0)); }

  return Address(base.base(), base.disp() + BytesPerWord);

#endif // AARCH64

}

Address LIR_Assembler::as_Address_lo(LIR_Address* addr) {

#ifdef AARCH64

  ShouldNotCallThis(); // Not used on AArch64

  return Address();

#else

  return as_Address(addr);

#endif // AARCH64

}

void LIR_Assembler::osr_entry() {

  offsets()->set_value(CodeOffsets::OSR_Entry, code_offset());

  BlockBegin* osr_entry = compilation()->hir()->osr_entry();

  ValueStack* entry_state = osr_entry->end()->state();

  int number_of_locks = entry_state->locks_size();

  __ build_frame(initial_frame_size_in_bytes(), bang_size_in_bytes());

  Register OSR_buf = osrBufferPointer()->as_pointer_register();

  assert(frame::interpreter_frame_monitor_size() == BasicObjectLock::size(), "adjust code below");

  int monitor_offset = (method()->max_locals() + 2 * (number_of_locks - 1)) * BytesPerWord;

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

    int slot_offset = monitor_offset - (i * 2 * BytesPerWord);

    __ ldr(R1, Address(OSR_buf, slot_offset + 0*BytesPerWord));

    __ ldr(R2, Address(OSR_buf, slot_offset + 1*BytesPerWord));

    __ str(R1, frame_map()->address_for_monitor_lock(i));

    __ str(R2, frame_map()->address_for_monitor_object(i));

  }

}

int LIR_Assembler::check_icache() {

  Register receiver = LIR_Assembler::receiverOpr()->as_register();

  int offset = __ offset();

  __ inline_cache_check(receiver, Ricklass);

  return offset;

}

void LIR_Assembler::jobject2reg_with_patching(Register reg, CodeEmitInfo* info) {

  jobject o = (jobject)Universe::non_oop_word();

  int index = __ oop_recorder()->allocate_oop_index(o);

  PatchingStub* patch = new PatchingStub(_masm, patching_id(info), index);

  __ patchable_mov_oop(reg, o, index);

  patching_epilog(patch, lir_patch_normal, reg, info);

}

void LIR_Assembler::klass2reg_with_patching(Register reg, CodeEmitInfo* info) {

  Metadata* o = (Metadata*)Universe::non_oop_word();

  int index = __ oop_recorder()->allocate_metadata_index(o);

  PatchingStub* patch = new PatchingStub(_masm, PatchingStub::load_klass_id, index);

  __ patchable_mov_metadata(reg, o, index);

  patching_epilog(patch, lir_patch_normal, reg, info);

}

int LIR_Assembler::initial_frame_size_in_bytes() const {

  // Subtracts two words to account for return address and link

  return frame_map()->framesize()*VMRegImpl::stack_slot_size - 2*wordSize;

}

int LIR_Assembler::emit_exception_handler() {

  // TODO: ARM

  __ nop(); // See comments in other ports

  address handler_base = __ start_a_stub(exception_handler_size());

  if (handler_base == NULL) {

    bailout("exception handler overflow");

    return -1;

  }

  int offset = code_offset();

  // check that there is really an exception

  __ verify_not_null_oop(Rexception_obj);

  __ call(Runtime1::entry_for(Runtime1::handle_exception_from_callee_id), relocInfo::runtime_call_type);

  __ should_not_reach_here();

  assert(code_offset() - offset <= exception_handler_size(), "overflow");

  __ end_a_stub();

  return offset;

}

// Emit the code to remove the frame from the stack in the exception

// unwind path.

int LIR_Assembler::emit_unwind_handler() {

#ifndef PRODUCT

  if (CommentedAssembly) {

    _masm->block_comment("Unwind handler");

  }

#endif

  int offset = code_offset();

  // Fetch the exception from TLS and clear out exception related thread state

  Register zero = __ zero_register(Rtemp);

  __ ldr(Rexception_obj, Address(Rthread, JavaThread::exception_oop_offset()));

  __ str(zero, Address(Rthread, JavaThread::exception_oop_offset()));

  __ str(zero, Address(Rthread, JavaThread::exception_pc_offset()));

  __ bind(_unwind_handler_entry);

  __ verify_not_null_oop(Rexception_obj);

  // Preform needed unlocking

  MonitorExitStub* stub = NULL;

  if (method()->is_synchronized()) {

    monitor_address(0, FrameMap::R0_opr);

    stub = new MonitorExitStub(FrameMap::R0_opr, true, 0);

    __ unlock_object(R2, R1, R0, Rtemp, *stub->entry());

    __ bind(*stub->continuation());

  }

  // remove the activation and dispatch to the unwind handler

  __ remove_frame(initial_frame_size_in_bytes()); // restores FP and LR

  __ jump(Runtime1::entry_for(Runtime1::unwind_exception_id), relocInfo::runtime_call_type, Rtemp);

  // Emit the slow path assembly

  if (stub != NULL) {

    stub->emit_code(this);

  }

  return offset;

}

int LIR_Assembler::emit_deopt_handler() {

  address handler_base = __ start_a_stub(deopt_handler_size());

  if (handler_base == NULL) {

    bailout("deopt handler overflow");

    return -1;

  }

  int offset = code_offset();

  __ mov_relative_address(LR, __ pc());

#ifdef AARCH64

  __ raw_push(LR, LR);

  __ jump(SharedRuntime::deopt_blob()->unpack(), relocInfo::runtime_call_type, Rtemp);

#else

  __ push(LR); // stub expects LR to be saved

  __ jump(SharedRuntime::deopt_blob()->unpack(), relocInfo::runtime_call_type, noreg);

#endif // AARCH64

  assert(code_offset() - offset <= deopt_handler_size(), "overflow");

  __ end_a_stub();

  return offset;

}

void LIR_Assembler::return_op(LIR_Opr result) {

  // Pop the frame before safepoint polling

  __ remove_frame(initial_frame_size_in_bytes());

  // mov_slow here is usually one or two instruction

  // TODO-AARCH64 3 instructions on AArch64, so try to load polling page by ldr_literal

  __ mov_address(Rtemp, os::get_polling_page(), symbolic_Relocation::polling_page_reference);

  __ relocate(relocInfo::poll_return_type);

  __ ldr(Rtemp, Address(Rtemp));

  __ ret();

}

int LIR_Assembler::safepoint_poll(LIR_Opr tmp, CodeEmitInfo* info) {

  __ mov_address(Rtemp, os::get_polling_page(), symbolic_Relocation::polling_page_reference);

  if (info != NULL) {

    add_debug_info_for_branch(info);

  }

  int offset = __ offset();

  __ relocate(relocInfo::poll_type);

  __ ldr(Rtemp, Address(Rtemp));

  return offset;

}

void LIR_Assembler::move_regs(Register from_reg, Register to_reg) {

  if (from_reg != to_reg) {

    __ mov(to_reg, from_reg);

  }

}

void LIR_Assembler::const2reg(LIR_Opr src, LIR_Opr dest, LIR_PatchCode patch_code, CodeEmitInfo* info) {

  assert(src->is_constant() && dest->is_register(), "must be");

  LIR_Const* c = src->as_constant_ptr();

  switch (c->type()) {

    case T_ADDRESS:

    case T_INT:

      assert(patch_code == lir_patch_none, "no patching handled here");

      __ mov_slow(dest->as_register(), c->as_jint());

      break;

    case T_LONG:

      assert(patch_code == lir_patch_none, "no patching handled here");

#ifdef AARCH64

      __ mov_slow(dest->as_pointer_register(), (intptr_t)c->as_jlong());

#else

      __ mov_slow(dest->as_register_lo(), c->as_jint_lo());

      __ mov_slow(dest->as_register_hi(), c->as_jint_hi());

#endif // AARCH64

      break;

    case T_OBJECT:

      if (patch_code == lir_patch_none) {

        __ mov_oop(dest->as_register(), c->as_jobject());

      } else {

        jobject2reg_with_patching(dest->as_register(), info);

      }

      break;

    case T_METADATA:

      if (patch_code == lir_patch_none) {

        __ mov_metadata(dest->as_register(), c->as_metadata());

      } else {

        klass2reg_with_patching(dest->as_register(), info);

      }

      break;

    case T_FLOAT:

      if (dest->is_single_fpu()) {

        __ mov_float(dest->as_float_reg(), c->as_jfloat());

      } else {

#ifdef AARCH64

        ShouldNotReachHere();

#else

        // Simple getters can return float constant directly into r0

        __ mov_slow(dest->as_register(), c->as_jint_bits());

#endif // AARCH64

      }

      break;

    case T_DOUBLE:

      if (dest->is_double_fpu()) {

        __ mov_double(dest->as_double_reg(), c->as_jdouble());

      } else {

#ifdef AARCH64

        ShouldNotReachHere();

#else

        // Simple getters can return double constant directly into r1r0

        __ mov_slow(dest->as_register_lo(), c->as_jint_lo_bits());

        __ mov_slow(dest->as_register_hi(), c->as_jint_hi_bits());

#endif // AARCH64

      }

      break;

    default:

      ShouldNotReachHere();

  }

}

void LIR_Assembler::const2stack(LIR_Opr src, LIR_Opr dest) {

  assert(src->is_constant(), "must be");

  assert(dest->is_stack(), "must be");

  LIR_Const* c = src->as_constant_ptr();

  switch (c->type()) {

    case T_INT:  // fall through

    case T_FLOAT:

      __ mov_slow(Rtemp, c->as_jint_bits());

      __ str_32(Rtemp, frame_map()->address_for_slot(dest->single_stack_ix()));

      break;

    case T_ADDRESS:

      __ mov_slow(Rtemp, c->as_jint());

      __ str(Rtemp, frame_map()->address_for_slot(dest->single_stack_ix()));

      break;

    case T_OBJECT:

      __ mov_oop(Rtemp, c->as_jobject());

      __ str(Rtemp, frame_map()->address_for_slot(dest->single_stack_ix()));

      break;

    case T_LONG:  // fall through

    case T_DOUBLE:

#ifdef AARCH64

      __ mov_slow(Rtemp, c->as_jlong_bits());

      __ str(Rtemp, frame_map()->address_for_slot(dest->double_stack_ix()));

#else

      __ mov_slow(Rtemp, c->as_jint_lo_bits());

      __ str(Rtemp, frame_map()->address_for_slot(dest->double_stack_ix(), lo_word_offset_in_bytes));

      if (c->as_jint_hi_bits() != c->as_jint_lo_bits()) {

        __ mov_slow(Rtemp, c->as_jint_hi_bits());

      }

      __ str(Rtemp, frame_map()->address_for_slot(dest->double_stack_ix(), hi_word_offset_in_bytes));

#endif // AARCH64

      break;

    default:

      ShouldNotReachHere();

  }

}

void LIR_Assembler::const2mem(LIR_Opr src, LIR_Opr dest, BasicType type,

                              CodeEmitInfo* info, bool wide) {

#ifdef AARCH64

  assert((src->as_constant_ptr()->type() == T_OBJECT && src->as_constant_ptr()->as_jobject() == NULL) ||

         (src->as_constant_ptr()->type() == T_INT && src->as_constant_ptr()->as_jint() == 0) ||

         (src->as_constant_ptr()->type() == T_LONG && src->as_constant_ptr()->as_jlong() == 0) ||

         (src->as_constant_ptr()->type() == T_FLOAT && src->as_constant_ptr()->as_jint_bits() == 0) ||

         (src->as_constant_ptr()->type() == T_DOUBLE && src->as_constant_ptr()->as_jlong_bits() == 0),

        "cannot handle otherwise");

  assert(dest->as_address_ptr()->type() == type, "should be");

  Address addr = as_Address(dest->as_address_ptr());

  int null_check_offset = code_offset();

  switch (type) {

    case T_OBJECT:  // fall through

    case T_ARRAY:

        if (UseCompressedOops && !wide) {

          __ str_w(ZR, addr);

        } else {

          __ str(ZR, addr);

        }

        break;

    case T_ADDRESS: // fall through

    case T_DOUBLE:  // fall through

    case T_LONG:    __ str(ZR, addr);   break;

    case T_FLOAT:   // fall through

    case T_INT:     __ str_w(ZR, addr); break;

    case T_BOOLEAN: // fall through

    case T_BYTE:    __ strb(ZR, addr);  break;

    case T_CHAR:    // fall through

    case T_SHORT:   __ strh(ZR, addr);  break;

    default: ShouldNotReachHere();

  }

#else

  assert((src->as_constant_ptr()->type() == T_OBJECT && src->as_constant_ptr()->as_jobject() == NULL),"cannot handle otherwise");

  __ mov(Rtemp, 0);

  int null_check_offset = code_offset();

  __ str(Rtemp, as_Address(dest->as_address_ptr()));

#endif // AARCH64

  if (info != NULL) {