VORPS X9, X9, K1, X25                              // 6241340956c9
	VORPS 7(SI)(DI*8), X9, K1, X25                     // 62613409568cfe07000000
	VORPS -15(R14), X9, K1, X25                        // 62413409568ef1ffffff
	VORPS X2, X13, K1, X25                             // 6261140956ca
	VORPS X8, X13, K1, X25                             // 6241140956c8

	x21, x20 = bits.Mul64(x11, 0xffffffff00000001)
	var x22 uint64
	var x23 uint64
	x23, x22 = bits.Mul64(x11, 0xffffffff)
	var x24 uint64
	var x25 uint64
	x25, x24 = bits.Mul64(x11, 0xffffffffffffffff)
	var x26 uint64
	var x27 uint64
	x26, x27 = bits.Add64(x25, x22, uint64(0x0))
	x28 := (uint64(p256Uint1(x27)) + x23)
	var x30 uint64
	_, x30 = bits.Add64(x11, x24, uint64(0x0))
	var x31 uint64
	var x32 uint64

	var x23 uint64
	x23, x22 = bits.Mul64(x20, 0x1000000000000000)
	var x24 uint64
	var x25 uint64
	x25, x24 = bits.Mul64(x20, 0x14def9dea2f79cd6)
	var x26 uint64
	var x27 uint64
	x27, x26 = bits.Mul64(x20, 0x5812631a5cf5d3ed)
	var x28 uint64
	var x29 uint64
	x28, x29 = bits.Add64(x27, x24, uint64(0x0))
	x30 := (uint64(fiatScalarUint1(x29)) + x25)
	var x32 uint64
	_, x32 = bits.Add64(x11, x26, uint64(0x0))
	var x33 uint64

	MOV	X8, (8*4)(X2)
	MOV	X9, (8*5)(X2)
	MOV	X18, (8*6)(X2)
	MOV	X19, (8*7)(X2)
	MOV	X20, (8*8)(X2)
	MOV	X21, (8*9)(X2)
	MOV	X22, (8*10)(X2)
	MOV	X23, (8*11)(X2)
	MOV	X24, (8*12)(X2)
	MOV	X25, (8*13)(X2)
	MOV	X26, (8*14)(X2)
	MOV	g, (8*15)(X2)
	MOV	X1, (8*16)(X2)
	MOVD	F8, (8*17)(X2)
	MOVD	F9, (8*18)(X2)
	MOVD	F18, (8*19)(X2)
	MOVD	F19, (8*20)(X2)
	MOVD	F20, (8*21)(X2)
	MOVD	F21, (8*22)(X2)

	MOV X14, 80(X2)
	MOV X15, 88(X2)
	MOV X16, 96(X2)
	MOV X17, 104(X2)
	MOV X18, 112(X2)
	MOV X19, 120(X2)
	MOV X20, 128(X2)
	MOV X21, 136(X2)
	MOV X22, 144(X2)
	MOV X23, 152(X2)
	MOV X24, 160(X2)
	MOV X25, 168(X2)
	MOV X26, 176(X2)
	MOV X28, 184(X2)
	MOV X29, 192(X2)
	MOV X30, 200(X2)
	MOVD F0, 208(X2)
	MOVD F1, 216(X2)
	MOVD F2, 224(X2)
	MOVD F3, 232(X2)
	MOVD F4, 240(X2)
	MOVD F5, 248(X2)

	x23, x22 = bits.Mul64(x20, 0xffffffff)
	var x24 uint64
	var x25 uint64
	x25, x24 = bits.Mul64(x20, 0xffffffffffffffff)
	var x26 uint64
	var x27 uint64
	x27, x26 = bits.Mul64(x20, 0xffffffff00000000)
	var x28 uint64
	var x29 uint64
	x28, x29 = bits.Add64(x27, x24, uint64(0x0))
	var x30 uint64
	var x31 uint64
	x30, x31 = bits.Add64(x25, x22, uint64(p224Uint1(x29)))
	x32 := (uint64(p224Uint1(x31)) + x23)

	MOV	X1, (8*3)(X2)
	MOV	X8, (8*4)(X2)
	MOV	X9, (8*5)(X2)
	MOV	X18, (8*6)(X2)
	MOV	X19, (8*7)(X2)
	MOV	X20, (8*8)(X2)
	MOV	X21, (8*9)(X2)
	MOV	X22, (8*10)(X2)
	MOV	X23, (8*11)(X2)
	MOV	X24, (8*12)(X2)
	MOV	X25, (8*13)(X2)
	MOV	X26, (8*14)(X2)
	MOV	g, (8*15)(X2)
	MOVD	F8, (8*16)(X2)
	MOVD	F9, (8*17)(X2)
	MOVD	F18, (8*18)(X2)
	MOVD	F19, (8*19)(X2)
	MOVD	F20, (8*20)(X2)
	MOVD	F21, (8*21)(X2)
	MOVD	F22, (8*22)(X2)

	VPOPCNTW X25, K3, X11                              // 6212fd0b54d9
	VPOPCNTW (CX), K3, X11                             // 6272fd0b5419
	VPOPCNTW 99(R15), K3, X11                          // 6252fd0b549f63000000
	VPOPCNTW X20, K3, X23                              // 62a2fd0b54fc
	VPOPCNTW X5, K3, X23                               // 62e2fd0b54fd
	VPOPCNTW X25, K3, X23                              // 6282fd0b54f9

	VRCP28PS -7(CX)(DX*2), K3, Z6                      // 62f27d4bcab451f9ffffff
	VRCP28SD X25, X14, K4, X19                         // 62828d0ccbd9
	VRCP28SD X11, X14, K4, X19                         // 62c28d0ccbdb
	VRCP28SD X17, X14, K4, X19                         // 62a28d0ccbd9
	VRCP28SD X25, X0, K4, X19                          // 6282fd0ccbd9
	VRCP28SD X11, X0, K4, X19                          // 62c2fd0ccbdb

	VPDPBUSDS X25, X14, K4, X19                        // 62820d0c51d9
	VPDPBUSDS X11, X14, K4, X19                        // 62c20d0c51db
	VPDPBUSDS X17, X14, K4, X19                        // 62a20d0c51d9
	VPDPBUSDS 7(AX)(CX*4), X14, K4, X19                // 62e20d0c519c8807000000
	VPDPBUSDS 7(AX)(CX*1), X14, K4, X19                // 62e20d0c519c0807000000
	VPDPBUSDS X25, X0, K4, X19                         // 62827d0c51d9