/*

 * Copyright (c) 1997, 2014, Oracle and/or its affiliates. All rights reserved.

 * Copyright 2012, 2014 SAP AG. 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/assembler.hpp"

#include "asm/macroAssembler.inline.hpp"

#include "interpreter/bytecodeHistogram.hpp"

#include "interpreter/interpreter.hpp"

#include "interpreter/interpreterGenerator.hpp"

#include "interpreter/interpreterRuntime.hpp"

#include "interpreter/templateTable.hpp"

#include "oops/arrayOop.hpp"

#include "oops/methodData.hpp"

#include "oops/method.hpp"

#include "oops/oop.inline.hpp"

#include "prims/jvmtiExport.hpp"

#include "prims/jvmtiThreadState.hpp"

#include "prims/methodHandles.hpp"

#include "runtime/arguments.hpp"

#include "runtime/deoptimization.hpp"

#include "runtime/frame.inline.hpp"

#include "runtime/sharedRuntime.hpp"

#include "runtime/stubRoutines.hpp"

#include "runtime/synchronizer.hpp"

#include "runtime/timer.hpp"

#include "runtime/vframeArray.hpp"

#include "utilities/debug.hpp"

#ifdef COMPILER1

#include "c1/c1_Runtime1.hpp"

#endif

#define __ _masm->

#ifdef PRODUCT

#define BLOCK_COMMENT(str) // nothing

#else

#define BLOCK_COMMENT(str) __ block_comment(str)

#endif

#define BIND(label) bind(label); BLOCK_COMMENT(#label ":")

int AbstractInterpreter::BasicType_as_index(BasicType type) {

  int i = 0;

  switch (type) {

    case T_BOOLEAN: i = 0; break;

    case T_CHAR   : i = 1; break;

    case T_BYTE   : i = 2; break;

    case T_SHORT  : i = 3; break;

    case T_INT    : i = 4; break;

    case T_LONG   : i = 5; break;

    case T_VOID   : i = 6; break;

    case T_FLOAT  : i = 7; break;

    case T_DOUBLE : i = 8; break;

    case T_OBJECT : i = 9; break;

    case T_ARRAY  : i = 9; break;

    default       : ShouldNotReachHere();

  }

  assert(0 <= i && i < AbstractInterpreter::number_of_result_handlers, "index out of bounds");

  return i;

}

address AbstractInterpreterGenerator::generate_slow_signature_handler() {

  // Slow_signature handler that respects the PPC C calling conventions.

  //

  // We get called by the native entry code with our output register

  // area == 8. First we call InterpreterRuntime::get_result_handler

  // to copy the pointer to the signature string temporarily to the

  // first C-argument and to return the result_handler in

  // R3_RET. Since native_entry will copy the jni-pointer to the

  // first C-argument slot later on, it is OK to occupy this slot

  // temporarilly. Then we copy the argument list on the java

  // expression stack into native varargs format on the native stack

  // and load arguments into argument registers. Integer arguments in

  // the varargs vector will be sign-extended to 8 bytes.

  //

  // On entry:

  //   R3_ARG1        - intptr_t*     Address of java argument list in memory.

  //   R15_prev_state - BytecodeInterpreter* Address of interpreter state for

  //     this method

  //   R19_method

  //

  // On exit (just before return instruction):

  //   R3_RET            - contains the address of the result_handler.

  //   R4_ARG2           - is not updated for static methods and contains "this" otherwise.

  //   R5_ARG3-R10_ARG8: - When the (i-2)th Java argument is not of type float or double,

  //                       ARGi contains this argument. Otherwise, ARGi is not updated.

  //   F1_ARG1-F13_ARG13 - contain the first 13 arguments of type float or double.

  const int LogSizeOfTwoInstructions = 3;

  // FIXME: use Argument:: GL: Argument names different numbers!

  const int max_fp_register_arguments  = 13;

  const int max_int_register_arguments = 6;  // first 2 are reserved

  const Register arg_java       = R21_tmp1;

  const Register arg_c          = R22_tmp2;

  const Register signature      = R23_tmp3;  // is string

  const Register sig_byte       = R24_tmp4;

  const Register fpcnt          = R25_tmp5;

  const Register argcnt         = R26_tmp6;

  const Register intSlot        = R27_tmp7;

  const Register target_sp      = R28_tmp8;

  const FloatRegister floatSlot = F0;

  address entry = __ function_entry();

  __ save_LR_CR(R0);

  __ save_nonvolatile_gprs(R1_SP, _spill_nonvolatiles_neg(r14));

  // We use target_sp for storing arguments in the C frame.

  __ mr(target_sp, R1_SP);

  __ push_frame_reg_args_nonvolatiles(0, R11_scratch1);

  __ mr(arg_java, R3_ARG1);

  __ call_VM_leaf(CAST_FROM_FN_PTR(address, InterpreterRuntime::get_signature), R16_thread, R19_method);

  // Signature is in R3_RET. Signature is callee saved.

  __ mr(signature, R3_RET);

  // Get the result handler.

  __ call_VM_leaf(CAST_FROM_FN_PTR(address, InterpreterRuntime::get_result_handler), R16_thread, R19_method);

  {

    Label L;

    // test if static

    // _access_flags._flags must be at offset 0.

    // TODO PPC port: requires change in shared code.

    //assert(in_bytes(AccessFlags::flags_offset()) == 0,

    //       "MethodDesc._access_flags == MethodDesc._access_flags._flags");

    // _access_flags must be a 32 bit value.

    assert(sizeof(AccessFlags) == 4, "wrong size");

    __ lwa(R11_scratch1/*access_flags*/, method_(access_flags));

    // testbit with condition register.

    __ testbitdi(CCR0, R0, R11_scratch1/*access_flags*/, JVM_ACC_STATIC_BIT);

    __ btrue(CCR0, L);

    // For non-static functions, pass "this" in R4_ARG2 and copy it

    // to 2nd C-arg slot.

    // We need to box the Java object here, so we use arg_java

    // (address of current Java stack slot) as argument and don't

    // dereference it as in case of ints, floats, etc.

    __ mr(R4_ARG2, arg_java);

    __ addi(arg_java, arg_java, -BytesPerWord);

    __ std(R4_ARG2, _abi(carg_2), target_sp);

    __ bind(L);

  }

  // Will be incremented directly after loop_start. argcnt=0

  // corresponds to 3rd C argument.

  __ li(argcnt, -1);

  // arg_c points to 3rd C argument

  __ addi(arg_c, target_sp, _abi(carg_3));

  // no floating-point args parsed so far

  __ li(fpcnt, 0);

  Label move_intSlot_to_ARG, move_floatSlot_to_FARG;

  Label loop_start, loop_end;

  Label do_int, do_long, do_float, do_double, do_dontreachhere, do_object, do_array, do_boxed;

  // signature points to '(' at entry

#ifdef ASSERT

  __ lbz(sig_byte, 0, signature);

  __ cmplwi(CCR0, sig_byte, '(');

  __ bne(CCR0, do_dontreachhere);

#endif

  __ bind(loop_start);

  __ addi(argcnt, argcnt, 1);

  __ lbzu(sig_byte, 1, signature);

  __ cmplwi(CCR0, sig_byte, ')'); // end of signature

  __ beq(CCR0, loop_end);

  __ cmplwi(CCR0, sig_byte, 'B'); // byte

  __ beq(CCR0, do_int);

  __ cmplwi(CCR0, sig_byte, 'C'); // char

  __ beq(CCR0, do_int);

  __ cmplwi(CCR0, sig_byte, 'D'); // double

  __ beq(CCR0, do_double);

  __ cmplwi(CCR0, sig_byte, 'F'); // float

  __ beq(CCR0, do_float);

  __ cmplwi(CCR0, sig_byte, 'I'); // int

  __ beq(CCR0, do_int);

  __ cmplwi(CCR0, sig_byte, 'J'); // long

  __ beq(CCR0, do_long);

  __ cmplwi(CCR0, sig_byte, 'S'); // short

  __ beq(CCR0, do_int);

  __ cmplwi(CCR0, sig_byte, 'Z'); // boolean

  __ beq(CCR0, do_int);

  __ cmplwi(CCR0, sig_byte, 'L'); // object

  __ beq(CCR0, do_object);

  __ cmplwi(CCR0, sig_byte, '['); // array

  __ beq(CCR0, do_array);

  //  __ cmplwi(CCR0, sig_byte, 'V'); // void cannot appear since we do not parse the return type

  //  __ beq(CCR0, do_void);

  __ bind(do_dontreachhere);

  __ unimplemented("ShouldNotReachHere in slow_signature_handler", 120);

  __ bind(do_array);

  {

    Label start_skip, end_skip;

    __ bind(start_skip);

    __ lbzu(sig_byte, 1, signature);

    __ cmplwi(CCR0, sig_byte, '[');

    __ beq(CCR0, start_skip); // skip further brackets

    __ cmplwi(CCR0, sig_byte, '9');

    __ bgt(CCR0, end_skip);   // no optional size

    __ cmplwi(CCR0, sig_byte, '0');

    __ bge(CCR0, start_skip); // skip optional size

    __ bind(end_skip);

    __ cmplwi(CCR0, sig_byte, 'L');

    __ beq(CCR0, do_object);  // for arrays of objects, the name of the object must be skipped

    __ b(do_boxed);          // otherwise, go directly to do_boxed

  }

  __ bind(do_object);

  {

    Label L;

    __ bind(L);

    __ lbzu(sig_byte, 1, signature);

    __ cmplwi(CCR0, sig_byte, ';');

    __ bne(CCR0, L);

   }

  // Need to box the Java object here, so we use arg_java (address of

  // current Java stack slot) as argument and don't dereference it as

  // in case of ints, floats, etc.

  Label do_null;

  __ bind(do_boxed);

  __ ld(R0,0, arg_java);

  __ cmpdi(CCR0, R0, 0);

  __ li(intSlot,0);

  __ beq(CCR0, do_null);

  __ mr(intSlot, arg_java);

  __ bind(do_null);

  __ std(intSlot, 0, arg_c);

  __ addi(arg_java, arg_java, -BytesPerWord);

  __ addi(arg_c, arg_c, BytesPerWord);

  __ cmplwi(CCR0, argcnt, max_int_register_arguments);

  __ blt(CCR0, move_intSlot_to_ARG);

  __ b(loop_start);

  __ bind(do_int);

  __ lwa(intSlot, 0, arg_java);

  __ std(intSlot, 0, arg_c);

  __ addi(arg_java, arg_java, -BytesPerWord);

  __ addi(arg_c, arg_c, BytesPerWord);

  __ cmplwi(CCR0, argcnt, max_int_register_arguments);

  __ blt(CCR0, move_intSlot_to_ARG);

  __ b(loop_start);

  __ bind(do_long);

  __ ld(intSlot, -BytesPerWord, arg_java);

  __ std(intSlot, 0, arg_c);

  __ addi(arg_java, arg_java, - 2 * BytesPerWord);

  __ addi(arg_c, arg_c, BytesPerWord);

  __ cmplwi(CCR0, argcnt, max_int_register_arguments);

  __ blt(CCR0, move_intSlot_to_ARG);

  __ b(loop_start);

  __ bind(do_float);

  __ lfs(floatSlot, 0, arg_java);

#if defined(LINUX)

  __ stfs(floatSlot, 4, arg_c);

#elif defined(AIX)

  __ stfs(floatSlot, 0, arg_c);

#else

#error "unknown OS"

#endif

  __ addi(arg_java, arg_java, -BytesPerWord);

  __ addi(arg_c, arg_c, BytesPerWord);

  __ cmplwi(CCR0, fpcnt, max_fp_register_arguments);

  __ blt(CCR0, move_floatSlot_to_FARG);

  __ b(loop_start);

  __ bind(do_double);

  __ lfd(floatSlot, - BytesPerWord, arg_java);

  __ stfd(floatSlot, 0, arg_c);

  __ addi(arg_java, arg_java, - 2 * BytesPerWord);

  __ addi(arg_c, arg_c, BytesPerWord);

  __ cmplwi(CCR0, fpcnt, max_fp_register_arguments);

  __ blt(CCR0, move_floatSlot_to_FARG);

  __ b(loop_start);

  __ bind(loop_end);

  __ pop_frame();

  __ restore_nonvolatile_gprs(R1_SP, _spill_nonvolatiles_neg(r14));

  __ restore_LR_CR(R0);

  __ blr();

  Label move_int_arg, move_float_arg;

  __ bind(move_int_arg); // each case must consist of 2 instructions (otherwise adapt LogSizeOfTwoInstructions)

  __ mr(R5_ARG3, intSlot);  __ b(loop_start);

  __ mr(R6_ARG4, intSlot);  __ b(loop_start);

  __ mr(R7_ARG5, intSlot);  __ b(loop_start);

  __ mr(R8_ARG6, intSlot);  __ b(loop_start);

  __ mr(R9_ARG7, intSlot);  __ b(loop_start);

  __ mr(R10_ARG8, intSlot); __ b(loop_start);

  __ bind(move_float_arg); // each case must consist of 2 instructions (otherwise adapt LogSizeOfTwoInstructions)

  __ fmr(F1_ARG1, floatSlot);   __ b(loop_start);

  __ fmr(F2_ARG2, floatSlot);   __ b(loop_start);

  __ fmr(F3_ARG3, floatSlot);   __ b(loop_start);

  __ fmr(F4_ARG4, floatSlot);   __ b(loop_start);

  __ fmr(F5_ARG5, floatSlot);   __ b(loop_start);

  __ fmr(F6_ARG6, floatSlot);   __ b(loop_start);

  __ fmr(F7_ARG7, floatSlot);   __ b(loop_start);

  __ fmr(F8_ARG8, floatSlot);   __ b(loop_start);

  __ fmr(F9_ARG9, floatSlot);   __ b(loop_start);

  __ fmr(F10_ARG10, floatSlot); __ b(loop_start);

  __ fmr(F11_ARG11, floatSlot); __ b(loop_start);

  __ fmr(F12_ARG12, floatSlot); __ b(loop_start);

  __ fmr(F13_ARG13, floatSlot); __ b(loop_start);

  __ bind(move_intSlot_to_ARG);

  __ sldi(R0, argcnt, LogSizeOfTwoInstructions);

  __ load_const(R11_scratch1, move_int_arg); // Label must be bound here.

  __ add(R11_scratch1, R0, R11_scratch1);

  __ mtctr(R11_scratch1/*branch_target*/);

  __ bctr();

  __ bind(move_floatSlot_to_FARG);

  __ sldi(R0, fpcnt, LogSizeOfTwoInstructions);

  __ addi(fpcnt, fpcnt, 1);

  __ load_const(R11_scratch1, move_float_arg); // Label must be bound here.

  __ add(R11_scratch1, R0, R11_scratch1);

  __ mtctr(R11_scratch1/*branch_target*/);

  __ bctr();

  return entry;

}

address AbstractInterpreterGenerator::generate_result_handler_for(BasicType type) {

  //

  // Registers alive

  //   R3_RET

  //   LR

  //

  // Registers updated

  //   R3_RET

  //

  Label done;

  address entry = __ pc();

  switch (type) {

  case T_BOOLEAN:

    // convert !=0 to 1

    __ neg(R0, R3_RET);

    __ orr(R0, R3_RET, R0);

    __ srwi(R3_RET, R0, 31);

    break;

  case T_BYTE:

     // sign extend 8 bits

     __ extsb(R3_RET, R3_RET);

     break;

  case T_CHAR:

     // zero extend 16 bits

     __ clrldi(R3_RET, R3_RET, 48);

     break;

  case T_SHORT:

     // sign extend 16 bits

     __ extsh(R3_RET, R3_RET);

     break;

  case T_INT:

     // sign extend 32 bits

     __ extsw(R3_RET, R3_RET);

     break;

  case T_LONG:

     break;

  case T_OBJECT:

    // unbox result if not null

    __ cmpdi(CCR0, R3_RET, 0);

    __ beq(CCR0, done);

    __ ld(R3_RET, 0, R3_RET);

    __ verify_oop(R3_RET);

    break;

  case T_FLOAT:

     break;

  case T_DOUBLE:

     break;

  case T_VOID:

     break;

  default: ShouldNotReachHere();

  }

  __ BIND(done);

  __ blr();

  return entry;

}

// Abstract method entry.

//

address InterpreterGenerator::generate_abstract_entry(void) {

  address entry = __ pc();

  //

  // Registers alive

  //   R16_thread     - JavaThread*

  //   R19_method     - callee's method (method to be invoked)

  //   R1_SP          - SP prepared such that caller's outgoing args are near top

  //   LR             - return address to caller

  //

  // Stack layout at this point:

  //

  //   0       [TOP_IJAVA_FRAME_ABI]         <-- R1_SP

  //           alignment (optional)

  //           [outgoing Java arguments]

  //           ...

  //   PARENT  [PARENT_IJAVA_FRAME_ABI]

  //            ...

  //

  // Can't use call_VM here because we have not set up a new

  // interpreter state. Make the call to the vm and make it look like

  // our caller set up the JavaFrameAnchor.

  __ set_top_ijava_frame_at_SP_as_last_Java_frame(R1_SP, R12_scratch2/*tmp*/);

  // Push a new C frame and save LR.

  __ save_LR_CR(R0);

  __ push_frame_reg_args(0, R11_scratch1);

  // This is not a leaf but we have a JavaFrameAnchor now and we will

  // check (create) exceptions afterward so this is ok.

  __ call_VM_leaf(CAST_FROM_FN_PTR(address, InterpreterRuntime::throw_AbstractMethodError));

  // Pop the C frame and restore LR.

  __ pop_frame();

  __ restore_LR_CR(R0);

  // Reset JavaFrameAnchor from call_VM_leaf above.

  __ reset_last_Java_frame();

#ifdef CC_INTERP

  // Return to frame manager, it will handle the pending exception.

  __ blr();

#else

  // We don't know our caller, so jump to the general forward exception stub,

  // which will also pop our full frame off. Satisfy the interface of

  // SharedRuntime::generate_forward_exception()

  __ load_const_optimized(R11_scratch1, StubRoutines::forward_exception_entry(), R0);

  __ mtctr(R11_scratch1);

  __ bctr();

#endif

  return entry;

}

// Call an accessor method (assuming it is resolved, otherwise drop into

// vanilla (slow path) entry.

address InterpreterGenerator::generate_accessor_entry(void) {

  if (!UseFastAccessorMethods && (!FLAG_IS_ERGO(UseFastAccessorMethods))) {

    return NULL;

  }

  Label Lslow_path, Lacquire;

  const Register

         Rclass_or_obj = R3_ARG1,

         Rconst_method = R4_ARG2,

         Rcodes        = Rconst_method,

         Rcpool_cache  = R5_ARG3,

         Rscratch      = R11_scratch1,

         Rjvmti_mode   = Rscratch,

         Roffset       = R12_scratch2,

         Rflags        = R6_ARG4,

         Rbtable       = R7_ARG5;

  static address branch_table[number_of_states];

  address entry = __ pc();

  // Check for safepoint:

  // Ditch this, real man don't need safepoint checks.

  // Also check for JVMTI mode

  // Check for null obj, take slow path if so.

  __ ld(Rclass_or_obj, Interpreter::stackElementSize, CC_INTERP_ONLY(R17_tos) NOT_CC_INTERP(R15_esp));

  __ lwz(Rjvmti_mode, thread_(interp_only_mode));

  __ cmpdi(CCR1, Rclass_or_obj, 0);

  __ cmpwi(CCR0, Rjvmti_mode, 0);

  __ crorc(/*CCR0 eq*/2, /*CCR1 eq*/4+2, /*CCR0 eq*/2);

  __ beq(CCR0, Lslow_path); // this==null or jvmti_mode!=0

  // Do 2 things in parallel: