// -*- C -*-
//
// NEC specific instructions
//

:%s::::MFHI:int hi
{
  return hi ? "hi" : "";
}

:%s::::SAT:int s
{
  return s ? "s" : "";
}

:%s::::UNS:int u
{
  return u ? "u" : "";
}

// Simulate the various kinds of multiply and multiply-accumulate instructions.
// Perform an operation of the form:
//
//	LHS (+/-) GPR[RS] * GPR[RT]
//
// and store it in the 64-bit accumulator.  Optionally copy either LO or
// HI into a general purpose register.
//
// - RD is the destination register of the LO or HI move
// - RS are RT are the multiplication source registers
// - ACCUMULATE_P is true if LHS should be the value of the 64-bit accumulator,
//     false if it should be 0.
// - STORE_HI_P is true if HI should be stored in RD, false if LO should be.
// - UNSIGNED_P is true if the operation should be unsigned.
// - SATURATE_P is true if the result should be saturated to a 32-bit value.
// - SUBTRACT_P is true if the right hand side should be subtraced from LHS,
//     false if it should be added.
// - SHORT_P is true if RS and RT must be 16-bit numbers.
// - DOUBLE_P is true if the 64-bit accumulator is in LO, false it is a
//     concatenation of the low 32 bits of HI and LO.
:function:::void:do_vr_mul_op:int rd, int rs, int rt, int accumulate_p, int store_hi_p, int unsigned_p, int saturate_p, int subtract_p, int short_p, int double_p
{
  uint64_t lhs, x, y, xcut, ycut, product, result;

  check_mult_hilo (SD_, HIHISTORY, LOHISTORY);

  lhs = (!accumulate_p ? 0 : double_p ? LO : U8_4 (HI, LO));
  x = GPR[rs];
  y = GPR[rt];

  /* Work out the canonical form of X and Y from their significant bits.  */
  if (!short_p)
    {
      /* Normal sign-extension rule for 32-bit operands.  */
      xcut = EXTEND32 (x);
      ycut = EXTEND32 (y);
    }
  else if (unsigned_p)
    {
      /* Operands must be zero-extended 16-bit numbers.  */
      xcut = x & 0xffff;
      ycut = y & 0xffff;
    }
  else
    {
      /* Likewise but sign-extended.  */
      xcut = EXTEND16 (x);
      ycut = EXTEND16 (y);
    }
  if (x != xcut || y != ycut)
    sim_engine_abort (SD, CPU, CIA,
		      "invalid multiplication operand at 0x%08lx\n",
		      (long) CIA);

  TRACE_ALU_INPUT2 (x, y);
  product = (unsigned_p
	     ? V8_4 (x, 1) * V8_4 (y, 1)
	     : EXTEND32 (x) * EXTEND32 (y));
  result = (subtract_p ? lhs - product : lhs + product);
  if (saturate_p)
    {
      /* Saturate the result to 32 bits.  An unsigned, unsaturated
	 result is zero-extended to 64 bits, but unsigned overflow
	 causes all 64 bits to be set.  */
      if (!unsigned_p && (uint64_t) EXTEND32 (result) != result)
	result = ((int64_t) result < 0 ? -0x7fffffff - 1 : 0x7fffffff);
      else if (unsigned_p && (result >> 32) != 0)
	result = (uint64_t) 0 - 1;
    }
  TRACE_ALU_RESULT (result);

  if (double_p)
    LO = result;
  else
    {
      LO = EXTEND32 (result);
      HI = EXTEND32 (VH4_8 (result));
    }
  if (rd != 0)
    GPR[rd] = store_hi_p ? HI : LO;
}

// VR4100 instructions.

000000,5.RS,5.RT,00000,00000,101000::32::MADD16
"madd16 r<RS>, r<RT>"
*vr4100:
{
  do_vr_mul_op (SD_, 0, RS, RT,
		1 /* accumulate */,
		0 /* store in LO */,
		0 /* signed arithmetic */,
		0 /* don't saturate */,
		0 /* don't subtract */,
		1 /* short */,
		0 /* single */);
}

000000,5.RS,5.RT,00000,00000,101001::64::DMADD16
"dmadd16 r<RS>, r<RT>"
*vr4100:
{
  do_vr_mul_op (SD_, 0, RS, RT,
		1 /* accumulate */,
		0 /* store in LO */,
		0 /* signed arithmetic */,
		0 /* don't saturate */,
		0 /* don't subtract */,
		1 /* short */,
		1 /* double */);
}


// VR4120 and VR4130 instructions.

000000,5.RS,5.RT,5.RD,1.SAT,1.MFHI,00,1.UNS,101001::64::DMACC
"dmacc%s<MFHI>%s<UNS>%s<SAT> r<RD>, r<RS>, r<RT>"
*vr4120:
{
  do_vr_mul_op (SD_, RD, RS, RT,
		1 /* accumulate */,
		MFHI, UNS, SAT,
		0 /* don't subtract */,
		SAT /* short */,
		1 /* double */);
}

000000,5.RS,5.RT,5.RD,1.SAT,1.MFHI,00,1.UNS,101000::32::MACC_4120
"macc%s<MFHI>%s<UNS>%s<SAT> r<RD>, r<RS>, r<RT>"
*vr4120:
{
  do_vr_mul_op (SD_, RD, RS, RT,
		1 /* accumulate */,
		MFHI, UNS, SAT,
		0 /* don't subtract */,
		SAT /* short */,
		0 /* single */);
}


// VR5400 and VR5500 instructions.

000000,5.RS,5.RT,5.RD,0,1.MFHI,001,01100,1.UNS::32::MUL
"mul%s<MFHI>%s<UNS> r<RD>, r<RS>, r<RT>"
*vr5400:
*vr5500:
{
  do_vr_mul_op (SD_, RD, RS, RT,
		0 /* don't accumulate */,
		MFHI, UNS,
		0 /* don't saturate */,
		0 /* don't subtract */,
		0 /* not short */,
		0 /* single */);
}

000000,5.RS,5.RT,5.RD,0,1.MFHI,011,01100,1.UNS::32::MULS
"muls%s<MFHI>%s<UNS> r<RD>, r<RS>, r<RT>"
*vr5400:
*vr5500:
{
  do_vr_mul_op (SD_, RD, RS, RT,
		0 /* don't accumulate */,
		MFHI, UNS,
		0 /* don't saturate */,
		1 /* subtract */,
		0 /* not short */,
		0 /* single */);
}

000000,5.RS,5.RT,5.RD,0,1.MFHI,101,01100,1.UNS::32::MACC_5xxx
"macc%s<MFHI>%s<UNS> r<RD>, r<RS>, r<RT>"
*vr5400:
*vr5500:
{
  do_vr_mul_op (SD_, RD, RS, RT,
		1 /* accumulate */,
		MFHI, UNS,
		0 /* don't saturate */,
		0 /* don't subtract */,
		0 /* not short */,
		0 /* single */);
}

000000,5.RS,5.RT,5.RD,0,1.MFHI,111,01100,1.UNS::32::MSAC
"msac%s<MFHI>%s<UNS> r<RD>, r<RS>, r<RT>"
*vr5400:
*vr5500:
{
  do_vr_mul_op (SD_, RD, RS, RT,
		1 /* accumulate */,
		MFHI, UNS,
		0 /* don't saturate */,
		1 /* subtract */,
		0 /* not short */,
		0 /* single */);
}


010011,5.BASE,5.INDEX,5.0,5.FD,000101:COP1X:64::LUXC1
"luxc1 f<FD>, r<INDEX>(r<BASE>)"
*vr5500:
{
  check_fpu (SD_);
  COP_LD (1, FD, do_load (SD_, AccessLength_DOUBLEWORD,
			  (GPR[BASE] + GPR[INDEX]) & ~MASK64 (2, 0), 0));
}

010011,5.BASE,5.INDEX,5.FS,00000,001101:COP1X:64::SUXC1
"suxc1 f<FS>, r<INDEX>(r<BASE>)"
*vr5500:
{
  check_fpu (SD_);
  do_store (SD_, AccessLength_DOUBLEWORD,
	    (GPR[BASE] + GPR[INDEX]) & ~MASK64 (2, 0), 0,
	    COP_SD (1, FS));
}

010000,1,19.*,100000:COP0:32::WAIT
"wait"
*vr5500:

011100,00000,5.RT,5.DR,00000,111101:SPECIAL:64::MFDR
"mfdr r<RT>, r<DR>"
*vr5400:
*vr5500:

011100,00100,5.RT,5.DR,00000,111101:SPECIAL:64::MTDR
"mtdr r<RT>, r<DR>"
*vr5400:
*vr5500:

011100,00000,00000,00000,00000,111110:SPECIAL:64::DRET
"dret"
*vr5400:
*vr5500: