VMOVDQU XDWORD3, XDWORD1

	CMPQ SRND, $4*4*32
	JB   avx2_loop2

	MOVQ dig+0(FP), CTX // d.h[8]
	MOVQ _INP(SP), INP

	addm(  0(CTX), a)
	addm(  4(CTX), b)
	addm(  8(CTX), c)
	addm( 12(CTX), d)
	addm( 16(CTX), e)
	addm( 20(CTX), f)
	addm( 24(CTX), g)
	addm( 28(CTX), h)

	CMPQ _INP_END(SP), INP
	JB   done_hash

	XORQ SRND, SRND

	ADDQ R15, AX

	ADDQ DI, AX

	VMOVDQU Y6, Y4
	VMOVDQU Y7, Y5

	SUBQ $1, frame_SRND(SP)
	JNE  loop2

	addm(8*0(SI),AX)
	addm(8*1(SI),BX)
	addm(8*2(SI),CX)
	addm(8*3(SI),R8)
	addm(8*4(SI),DX)
	addm(8*5(SI),R9)
	addm(8*6(SI),R10)
	addm(8*7(SI),R11)

	MOVQ frame_INP(SP), DI
	ADDQ $128, DI
	CMPQ DI, frame_INPEND(SP)
	JNE  loop0

done_hash:

  return tensorflow::TensorCApi::Buffer(tensor_)->data();
}

void TensorInterface::SetShape(const int64_t* dims, int num_dims) {
  tensorflow::TensorShape s;
  for (int i = 0; i < num_dims; ++i) {
    s.AddDim(dims[i]);
  }
  tensor_.set_shape(s);
}

Status TensorInterface::BitcastFrom(const TensorInterface& from, DataType type,
                                    const int64_t* new_dims, int num_new_dims) {

	// loong64: "ADDV", -"SGTU"
	// ppc64x: "ADDC", "ADDE", -"ADDZE"
	// s390x:"ADDE","ADDC\t[$]-1,"
	// mips64:"ADDV",-"SGTU"
	// riscv64: "ADD",-"SLTU"
	r, _ := bits.Add64(x, y, ci)
	return r
}

func Add64M(p, q, r *[3]uint64) {
	var c uint64
	r[0], c = bits.Add64(p[0], q[0], c)
	// arm64:"ADCS",-"ADD\t",-"CMP"
	// amd64:"ADCQ",-"NEGL",-"SBBQ",-"NEGQ"
	// ppc64x: -"ADDC", "ADDE", -"ADDZE"

this.max_[0]=Math.max(this.max_[0],x);this.max_[1]=Math.max(this.max_[1],y);this.min_[0]=Math...