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)
	MOVD	F23, (8*23)(X2)

	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
	x31, x32 = bits.Add64(x13, x26, uint64(p256Uint1(x30)))
	var x33 uint64
	var x34 uint64

	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
	var x34 uint64

	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)
	MOVD F6, 256(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)
	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)
	MOVD	F22, (8*23)(X2)

	x22, x21 = bits.Mul64(x19, 0xffffffff)
	var x23 uint64
	var x24 uint64
	x24, x23 = bits.Mul64(x19, 0xffffffffffffffff)
	var x25 uint64
	var x26 uint64
	x26, x25 = bits.Mul64(x19, 0xffffffff00000000)
	var x27 uint64
	var x28 uint64
	x27, x28 = bits.Add64(x26, x23, uint64(0x0))
	var x29 uint64
	var x30 uint64
	x29, x30 = bits.Add64(x24, x21, uint64(p224Uint1(x28)))
	var x32 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)
	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)
	MOVD	F23, (8*23)(X2)

TEXT asmtest_avx512_vbmi(SB), NOSPLIT, $0
	VPERMB X26, X20, K1, X23                           // 62825d018dfa
	VPERMB X19, X20, K1, X23                           // 62a25d018dfb
	VPERMB X0, X20, K1, X23                            // 62e25d018df8
	VPERMB 7(SI)(DI*4), X20, K1, X23                   // 62e25d018dbcbe07000000
	VPERMB -7(DI)(R8*2), X20, K1, X23                  // 62a25d018dbc47f9ffffff
	VPERMB X26, X2, K1, X23                            // 62826d098dfa

	CLZ	X20, X21				// 931a0a60
	CLZW	X21, X22				// 1b9b0a60
	CPOP	X22, X23				// 931b2b60
	CPOPW	X23, X24				// 1b9c2b60
	CTZ	X24, X25				// 931c1c60
	CTZW	X25, X26				// 1b9d1c60
	MAX	X26, X28, X29				// b36eae0b
	MAX	X26, X28				// 336eae0b
	MAXU	X28, X29, X30				// 33ffce0b
	MAXU	X28, X29				// b3fece0b
	MIN	X29, X30, X5				// b342df0b
	MIN	X29, X30				// 334fdf0b
	MINU	X30, X5, X6				// 33d3e20b

	VCVTTPS2DQ 15(DX)(BX*8), K3, Z30                   // 62617e4b5bb4da0f000000
	VCVTTPS2UDQ X22, K2, X26                           // 62217c0a78d6
	VCVTTPS2UDQ 15(DX)(BX*1), K2, X26                  // 62617c0a78941a0f000000
	VCVTTPS2UDQ -7(CX)(DX*2), K2, X26                  // 62617c0a789451f9ffffff
	VCVTTPS2UDQ Y13, K1, Y24                           // 62417c2978c5