}

func nextTerm(k int64) {
	y2 := k*2 + 1
	accum.Add(accum, tmp1.Lsh(numer, 1))
	accum.Mul(accum, tmp1.SetInt64(y2))
	numer.Mul(numer, tmp1.SetInt64(k))
	denom.Mul(denom, tmp1.SetInt64(y2))
}

func eliminateDigit(d int64) {
	accum.Sub(accum, tmp1.Mul(denom, tmp1.SetInt64(d)))
	accum.Mul(accum, ten)
	numer.Mul(numer, ten)
}

func main() {
	i := 0
	k := int64(0)
	for {
		d := int64(-1)
		for d < 0 {

    /* Given upstream grad U and a mul op A*B, the gradients are:
     *
     *    dA = U * B
     *    dB = A * U
     *
     */

    AbstractTensorHandle* upstream_grad = grad_outputs[0];

    // Gradient for A
    std::string name = "Mul_Grad_A";
    TF_RETURN_IF_ERROR(Mul(ctx, upstream_grad, forward_inputs_[1],

  AbstractTensorHandle* expand_dims_outputs;
  TF_RETURN_IF_ERROR(
      ops::ExpandDims(ctx, vec, dim.get(), &expand_dims_outputs, "ExpandDims"));
  TF_RETURN_IF_ERROR(
      ops::Mul(ctx, expand_dims_outputs, mat, &outputs[0], "Mul"));
  expand_dims_outputs->Unref();
  return absl::OkStatus();
}

class SparseSoftmaxCrossEntropyWithLogitsGradientFunction
    : public GradientFunction {
 public:

          command: [
            "/bin/sh",
            "-ce",

	//	}
	ADD	R1, R2

	//	LADDW imm ',' rreg
	//	{
	//		outcode(int($1), &$2, 0, &$4);
	//	}
	ADD	$4, R1

	//	LMUL rreg ',' rreg
	//	{
	//		outcode(int($1), &$2, 0, &$4);
	//	}
	MUL	R1, R2

	//	LSHW rreg ',' sreg ',' rreg
	//	{
	//		outcode(int($1), &$2, int($4), &$6);
	//	}
	SLL	R1, R2, R3

	//	LSHW rreg ',' rreg
	//	{
	//		outcode(int($1), &$2, 0, &$4);
	//	}

	MOVV	F4, R5			// 85b81401
	WORD	$74565			// 45230100
	BREAK	R4, result+16(FP)	// 64600006
	BREAK	R4, 1(R5)		// a4040006
	BREAK				// 00002a00
	UNDEF				// 00002a00

	// mul
	MUL	R4, R5	  		// a5101c00
	MUL	R4, R5, R6	  	// a6101c00
	MULV	R4, R5	   		// a5901d00
	MULV	R4, R5, R6	   	// a6901d00
	MULVU	R4, R5			// a5901d00
	MULVU	R4, R5, R6		// a6901d00
	MULHV	R4, R5			// a5101e00

}

Status MulModel(AbstractContext* ctx,
                absl::Span<AbstractTensorHandle* const> inputs,
                absl::Span<AbstractTensorHandle*> outputs) {
  return ops::Mul(ctx, inputs[0], inputs[1], &outputs[0], "Mul");
}

Status Log1pModel(AbstractContext* ctx,
                  absl::Span<AbstractTensorHandle* const> inputs,
                  absl::Span<AbstractTensorHandle*> outputs) {

		FieldD string `gorm:"uniqueIndex;index"` // uniqueIndex and index

		FieldE1 string `gorm:"uniqueIndex:uniq_field_e1_e2"` // mul uniqueIndex
		FieldE2 string `gorm:"uniqueIndex:uniq_field_e1_e2"`

		FieldF1 string `gorm:"uniqueIndex:uniq_field_f1_f2;index"` // mul uniqueIndex and index
		FieldF2 string `gorm:"uniqueIndex:uniq_field_f1_f2;"`

		FieldG string `gorm:"unique;uniqueIndex"` // unique and uniqueIndex

}

Status MulModel(AbstractContext* ctx,
                absl::Span<AbstractTensorHandle* const> inputs,
                absl::Span<AbstractTensorHandle*> outputs) {
  return ops::Mul(ctx, inputs[0], inputs[1], &outputs[0], "Mul");
}

// TODO(vnvo2409): Add more tests from `python/ops/gradient_checker_v2_test.py`.
// These tests should not be confused with `[*]_grad_test` which compare the

		loong64.ACMPGTF, loong64.ACMPGTD:
		return true
	}
	return false
}

// IsLoong64MUL reports whether the op (as defined by an loong64.A* constant) is
// one of the MUL/DIV/REM instructions that require special handling.
func IsLoong64MUL(op obj.As) bool {
	switch op {
	case loong64.AMUL, loong64.AMULU, loong64.AMULV, loong64.AMULVU,
		loong64.ADIV, loong64.ADIVU, loong64.ADIVV, loong64.ADIVVU,