/*

 * Copyright (c) 2000, 2018, Oracle and/or its affiliates. All rights reserved.

 * Copyright (c) 2012, 2017, SAP SE. 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/collectedHeap.hpp"

#include "gc/shared/barrierSet.hpp"

#include "gc/shared/cardTableBarrierSet.hpp"

#include "nativeInst_ppc.hpp"

#include "oops/objArrayKlass.hpp"

#include "runtime/frame.inline.hpp"

#include "runtime/safepointMechanism.inline.hpp"

#include "runtime/sharedRuntime.hpp"

#define __ _masm->

const ConditionRegister LIR_Assembler::BOOL_RESULT = CCR5;

bool LIR_Assembler::is_small_constant(LIR_Opr opr) {

  Unimplemented(); return false; // Currently not used on this platform.

}

LIR_Opr LIR_Assembler::receiverOpr() {

  return FrameMap::R3_oop_opr;

}

LIR_Opr LIR_Assembler::osrBufferPointer() {

  return FrameMap::R3_opr;

}

// This specifies the stack pointer decrement needed to build the frame.

int LIR_Assembler::initial_frame_size_in_bytes() const {

  return in_bytes(frame_map()->framesize_in_bytes());

}

// Inline cache check: the inline cached class is in inline_cache_reg;

// we fetch the class of the receiver and compare it with the cached class.

// If they do not match we jump to slow case.

int LIR_Assembler::check_icache() {

  int offset = __ offset();

  __ inline_cache_check(R3_ARG1, R19_inline_cache_reg);

  return offset;

}

void LIR_Assembler::osr_entry() {

  // On-stack-replacement entry sequence:

  //

  //   1. Create a new compiled activation.

  //   2. Initialize local variables in the compiled activation. The expression

  //      stack must be empty at the osr_bci; it is not initialized.

  //   3. Jump to the continuation address in compiled code to resume execution.

  // OSR entry point

  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();

  // Create a frame for the compiled activation.

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

  // OSR buffer is

  //

  // locals[nlocals-1..0]

  // monitors[number_of_locks-1..0]

  //

  // Locals is a direct copy of the interpreter frame so in the osr buffer

  // the first slot in the local array is the last local from the interpreter

  // and the last slot is local[0] (receiver) from the interpreter.

  //

  // Similarly with locks. The first lock slot in the osr buffer is the nth lock

  // from the interpreter frame, the nth lock slot in the osr buffer is 0th lock

  // in the interpreter frame (the method lock if a sync method).

  // Initialize monitors in the compiled activation.

  //   R3: pointer to osr buffer

  //

  // All other registers are dead at this point and the locals will be

  // copied into place by code emitted in the IR.

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

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

    int monitor_offset = BytesPerWord * method()->max_locals() +

      (2 * BytesPerWord) * (number_of_locks - 1);

    // SharedRuntime::OSR_migration_begin() packs BasicObjectLocks in

    // the OSR buffer using 2 word entries: first the lock and then

    // the oop.

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

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

#ifdef ASSERT

      // Verify the interpreter's monitor has a non-null object.

      {

        Label L;

        __ ld(R0, slot_offset + 1*BytesPerWord, OSR_buf);

        __ cmpdi(CCR0, R0, 0);

        __ bne(CCR0, L);

        __ stop("locked object is NULL");

        __ bind(L);

      }

#endif // ASSERT

      // Copy the lock field into the compiled activation.

      Address ml = frame_map()->address_for_monitor_lock(i),

              mo = frame_map()->address_for_monitor_object(i);

      assert(ml.index() == noreg && mo.index() == noreg, "sanity");

      __ ld(R0, slot_offset + 0, OSR_buf);

      __ std(R0, ml.disp(), ml.base());

      __ ld(R0, slot_offset + 1*BytesPerWord, OSR_buf);

      __ std(R0, mo.disp(), mo.base());

    }

  }

}

int LIR_Assembler::emit_exception_handler() {

  // If the last instruction is a call (typically to do a throw which

  // is coming at the end after block reordering) the return address

  // must still point into the code area in order to avoid assertion

  // failures when searching for the corresponding bci => add a nop

  // (was bug 5/14/1999 - gri).

  __ nop();

  // Generate code for the exception handler.

  address handler_base = __ start_a_stub(exception_handler_size());

  if (handler_base == NULL) {

    // Not enough space left for the handler.

    bailout("exception handler overflow");

    return -1;

  }

  int offset = code_offset();

  address entry_point = CAST_FROM_FN_PTR(address, Runtime1::entry_for(Runtime1::handle_exception_from_callee_id));

  //__ load_const_optimized(R0, entry_point);

  __ add_const_optimized(R0, R29_TOC, MacroAssembler::offset_to_global_toc(entry_point));

  __ mtctr(R0);

  __ bctr();

  guarantee(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() {

  _masm->block_comment("Unwind handler");

  int offset = code_offset();

  bool preserve_exception = method()->is_synchronized() || compilation()->env()->dtrace_method_probes();

  const Register Rexception = R3 /*LIRGenerator::exceptionOopOpr()*/, Rexception_save = R31;

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

  __ ld(Rexception, in_bytes(JavaThread::exception_oop_offset()), R16_thread);

  __ li(R0, 0);

  __ std(R0, in_bytes(JavaThread::exception_oop_offset()), R16_thread);

  __ std(R0, in_bytes(JavaThread::exception_pc_offset()), R16_thread);

  __ bind(_unwind_handler_entry);

  __ verify_not_null_oop(Rexception);

  if (preserve_exception) { __ mr(Rexception_save, Rexception); }

  // Perform needed unlocking

  MonitorExitStub* stub = NULL;

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

    monitor_address(0, FrameMap::R4_opr);

    stub = new MonitorExitStub(FrameMap::R4_opr, true, 0);

    __ unlock_object(R5, R6, R4, *stub->entry());

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

  }

  if (compilation()->env()->dtrace_method_probes()) {

    Unimplemented();

  }

  // Dispatch to the unwind logic.

  address unwind_stub = Runtime1::entry_for(Runtime1::unwind_exception_id);

  //__ load_const_optimized(R0, unwind_stub);

  __ add_const_optimized(R0, R29_TOC, MacroAssembler::offset_to_global_toc(unwind_stub));

  if (preserve_exception) { __ mr(Rexception, Rexception_save); }

  __ mtctr(R0);

  __ bctr();

  // Emit the slow path assembly.

  if (stub != NULL) {

    stub->emit_code(this);

  }

  return offset;

}

int LIR_Assembler::emit_deopt_handler() {

  // If the last instruction is a call (typically to do a throw which

  // is coming at the end after block reordering) the return address

  // must still point into the code area in order to avoid assertion

  // failures when searching for the corresponding bci => add a nop

  // (was bug 5/14/1999 - gri).

  __ nop();

  // Generate code for deopt handler.

  address handler_base = __ start_a_stub(deopt_handler_size());

  if (handler_base == NULL) {

    // Not enough space left for the handler.

    bailout("deopt handler overflow");

    return -1;

  }

  int offset = code_offset();

  __ bl64_patchable(SharedRuntime::deopt_blob()->unpack(), relocInfo::runtime_call_type);

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

  __ end_a_stub();

  return offset;

}

void LIR_Assembler::jobject2reg(jobject o, Register reg) {

  if (o == NULL) {

    __ li(reg, 0);

  } else {

    AddressLiteral addrlit = __ constant_oop_address(o);

    __ load_const(reg, addrlit, (reg != R0) ? R0 : noreg);

  }

}

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

  // Allocate a new index in table to hold the object once it's been patched.

  int oop_index = __ oop_recorder()->allocate_oop_index(NULL);

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

  AddressLiteral addrlit((address)NULL, oop_Relocation::spec(oop_index));

  __ load_const(reg, addrlit, R0);

  patching_epilog(patch, lir_patch_normal, reg, info);

}

void LIR_Assembler::metadata2reg(Metadata* o, Register reg) {

  AddressLiteral md = __ constant_metadata_address(o); // Notify OOP recorder (don't need the relocation)

  __ load_const_optimized(reg, md.value(), (reg != R0) ? R0 : noreg);

}

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

  // Allocate a new index in table to hold the klass once it's been patched.

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

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

  AddressLiteral addrlit((address)NULL, metadata_Relocation::spec(index));

  assert(addrlit.rspec().type() == relocInfo::metadata_type, "must be an metadata reloc");

  __ load_const(reg, addrlit, R0);

  patching_epilog(patch, lir_patch_normal, reg, info);

}

void LIR_Assembler::arithmetic_idiv(LIR_Code code, LIR_Opr left, LIR_Opr right, LIR_Opr temp, LIR_Opr result, CodeEmitInfo* info) {

  const bool is_int = result->is_single_cpu();

  Register Rdividend = is_int ? left->as_register() : left->as_register_lo();

  Register Rdivisor  = noreg;

  Register Rscratch  = temp->as_register();

  Register Rresult   = is_int ? result->as_register() : result->as_register_lo();

  long divisor = -1;

  if (right->is_register()) {

    Rdivisor = is_int ? right->as_register() : right->as_register_lo();

  } else {

    divisor = is_int ? right->as_constant_ptr()->as_jint()

                     : right->as_constant_ptr()->as_jlong();

  }

  assert(Rdividend != Rscratch, "");

  assert(Rdivisor  != Rscratch, "");

  assert(code == lir_idiv || code == lir_irem, "Must be irem or idiv");

  if (Rdivisor == noreg) {

    if (divisor == 1) { // stupid, but can happen

      if (code == lir_idiv) {

        __ mr_if_needed(Rresult, Rdividend);

      } else {

        __ li(Rresult, 0);

      }

    } else if (is_power_of_2(divisor)) {

      // Convert division by a power of two into some shifts and logical operations.

      int log2 = log2_intptr(divisor);

      // Round towards 0.

      if (divisor == 2) {

        if (is_int) {

          __ srwi(Rscratch, Rdividend, 31);

        } else {

          __ srdi(Rscratch, Rdividend, 63);

        }

      } else {

        if (is_int) {

          __ srawi(Rscratch, Rdividend, 31);

        } else {

          __ sradi(Rscratch, Rdividend, 63);

        }

        __ clrldi(Rscratch, Rscratch, 64-log2);

      }

      __ add(Rscratch, Rdividend, Rscratch);

      if (code == lir_idiv) {

        if (is_int) {

          __ srawi(Rresult, Rscratch, log2);

        } else {

          __ sradi(Rresult, Rscratch, log2);

        }

      } else { // lir_irem

        __ clrrdi(Rscratch, Rscratch, log2);

        __ sub(Rresult, Rdividend, Rscratch);

      }

    } else if (divisor == -1) {

      if (code == lir_idiv) {

        __ neg(Rresult, Rdividend);

      } else {

        __ li(Rresult, 0);

      }

    } else {

      __ load_const_optimized(Rscratch, divisor);

      if (code == lir_idiv) {

        if (is_int) {

          __ divw(Rresult, Rdividend, Rscratch); // Can't divide minint/-1.

        } else {

          __ divd(Rresult, Rdividend, Rscratch); // Can't divide minint/-1.

        }

      } else {

        assert(Rscratch != R0, "need both");

        if (is_int) {

          __ divw(R0, Rdividend, Rscratch); // Can't divide minint/-1.

          __ mullw(Rscratch, R0, Rscratch);

        } else {

          __ divd(R0, Rdividend, Rscratch); // Can't divide minint/-1.

          __ mulld(Rscratch, R0, Rscratch);

        }

        __ sub(Rresult, Rdividend, Rscratch);

      }

    }

    return;

  }

  Label regular, done;

  if (is_int) {

    __ cmpwi(CCR0, Rdivisor, -1);

  } else {

    __ cmpdi(CCR0, Rdivisor, -1);

  }

  __ bne(CCR0, regular);

  if (code == lir_idiv) {

    __ neg(Rresult, Rdividend);

    __ b(done);

    __ bind(regular);

    if (is_int) {

      __ divw(Rresult, Rdividend, Rdivisor); // Can't divide minint/-1.

    } else {

      __ divd(Rresult, Rdividend, Rdivisor); // Can't divide minint/-1.

    }

  } else { // lir_irem

    __ li(Rresult, 0);

    __ b(done);

    __ bind(regular);

    if (is_int) {

      __ divw(Rscratch, Rdividend, Rdivisor); // Can't divide minint/-1.

      __ mullw(Rscratch, Rscratch, Rdivisor);

    } else {

      __ divd(Rscratch, Rdividend, Rdivisor); // Can't divide minint/-1.

      __ mulld(Rscratch, Rscratch, Rdivisor);

    }

    __ sub(Rresult, Rdividend, Rscratch);

  }

  __ bind(done);

}

void LIR_Assembler::emit_op3(LIR_Op3* op) {

  switch (op->code()) {

  case lir_idiv:

  case lir_irem:

    arithmetic_idiv(op->code(), op->in_opr1(), op->in_opr2(), op->in_opr3(),

                    op->result_opr(), op->info());

    break;

  case lir_fmad:

    __ fmadd(op->result_opr()->as_double_reg(), op->in_opr1()->as_double_reg(),

             op->in_opr2()->as_double_reg(), op->in_opr3()->as_double_reg());

    break;

  case lir_fmaf:

    __ fmadds(op->result_opr()->as_float_reg(), op->in_opr1()->as_float_reg(),

              op->in_opr2()->as_float_reg(), op->in_opr3()->as_float_reg());

    break;

  default: ShouldNotReachHere(); break;

  }

}

void LIR_Assembler::emit_opBranch(LIR_OpBranch* op) {

#ifdef ASSERT

  assert(op->block() == NULL || op->block()->label() == op->label(), "wrong label");

  if (op->block() != NULL)  _branch_target_blocks.append(op->block());

  if (op->ublock() != NULL) _branch_target_blocks.append(op->ublock());

  assert(op->info() == NULL, "shouldn't have CodeEmitInfo");

#endif

  Label *L = op->label();

  if (op->cond() == lir_cond_always) {

    __ b(*L);

  } else {

    Label done;

    bool is_unordered = false;

    if (op->code() == lir_cond_float_branch) {

      assert(op->ublock() != NULL, "must have unordered successor");

      is_unordered = true;

    } else {

      assert(op->code() == lir_branch, "just checking");

    }

    bool positive = false;

    Assembler::Condition cond = Assembler::equal;

    switch (op->cond()) {

      case lir_cond_equal:        positive = true ; cond = Assembler::equal  ; is_unordered = false; break;

      case lir_cond_notEqual:     positive = false; cond = Assembler::equal  ; is_unordered = false; break;

      case lir_cond_less:         positive = true ; cond = Assembler::less   ; break;

      case lir_cond_belowEqual:   assert(op->code() != lir_cond_float_branch, ""); // fallthru

      case lir_cond_lessEqual:    positive = false; cond = Assembler::greater; break;

      case lir_cond_greater:      positive = true ; cond = Assembler::greater; break;

      case lir_cond_aboveEqual:   assert(op->code() != lir_cond_float_branch, ""); // fallthru

      case lir_cond_greaterEqual: positive = false; cond = Assembler::less   ; break;

      default:                    ShouldNotReachHere();

    }

    int bo = positive ? Assembler::bcondCRbiIs1 : Assembler::bcondCRbiIs0;

    int bi = Assembler::bi0(BOOL_RESULT, cond);

    if (is_unordered) {

      if (positive) {

        if (op->ublock() == op->block()) {

          __ bc_far_optimized(Assembler::bcondCRbiIs1, __ bi0(BOOL_RESULT, Assembler::summary_overflow), *L);

        }

      } else {

        if (op->ublock() != op->block()) { __ bso(BOOL_RESULT, done); }

      }

    }

    __ bc_far_optimized(bo, bi, *L);

    __ bind(done);

  }

}

void LIR_Assembler::emit_opConvert(LIR_OpConvert* op) {

  Bytecodes::Code code = op->bytecode();

  LIR_Opr src = op->in_opr(),

          dst = op->result_opr();

  switch(code) {

    case Bytecodes::_i2l: {

      __ extsw(dst->as_register_lo(), src->as_register());

      break;

    }

    case Bytecodes::_l2i: {

      __ mr_if_needed(dst->as_register(), src->as_register_lo()); // high bits are garbage

      break;

    }

    case Bytecodes::_i2b: {

      __ extsb(dst->as_register(), src->as_register());

      break;

    }

    case Bytecodes::_i2c: {

      __ clrldi(dst->as_register(), src->as_register(), 64-16);

      break;

    }

    case Bytecodes::_i2s: {

      __ extsh(dst->as_register(), src->as_register());

      break;

    }

    case Bytecodes::_i2d:

    case Bytecodes::_l2d: {

      bool src_in_memory = !VM_Version::has_mtfprd();

      FloatRegister rdst = dst->as_double_reg();

      FloatRegister rsrc;