ADDL -2147483648(AX), AX // 038000000080
	// Make sure MOV CR/DR continues to work after changing its movtabs.
	MOVL CR0, AX // 0f20c0
	MOVL CR0, DX // 0f20c2
	MOVL CR4, DI // 0f20e7
	MOVL AX, CR0 // 0f22c0
	MOVL DX, CR0 // 0f22c2
	MOVL DI, CR4 // 0f22e7
	MOVL DR0, AX // 0f21c0
	MOVL DR6, DX // 0f21f2
	MOVL DR7, SI // 0f21fe
	// Test other movtab entries.
	PUSHL SS // 16
	PUSHL FS // 0fa0
	POPL FS  // 0fa1
	POPL SS  // 17

	VADDPD 2032(DX), X29, X0        // 62f1950058427f
	VADDPD 2032(DX), X1, X29        // 6261f508586a7f
	VADDPD 2032(DX), X29, X28       // 6261950058627f
	VADDPD 2032(DX)(AX*2), X29, X0  // 62f195005844427f
	VADDPD 2032(DX)(AX*2), X1, X29  // 6261f508586c427f
	VADDPD 2032(DX)(AX*2), X29, X28 // 626195005864427f
	VADDPD 4064(DX), Y0, Y29        // 6261fd28586a7f
	VADDPD 4064(DX), Y29, Y1        // 62f19520584a7f

		lines(
			"#define LOAD(off, reg) \\",
			"\tMOVBLZX	(off*4)(R12),	reg \\",
			"\tADDB	reg,		DX",
			"",
			"LOAD(8, AX)",
		),
		"\n.\n.MOVBLZX.(.8.*.4.).(.R12.).,.AX.\n.ADDB.AX.,.DX.\n",
	},
	{
		"nested multiline macro",
		lines(
			"#define KEYROUND(xmm, load, off, r1, r2, index) \\",
			"\tMOVBLZX	(BP)(DX*4),	R8 \\",
			"\tload((off+1), r2) \\",
			"\tMOVB	R8,		(off*4)(R12) \\",

		{"VADDPD.SAE.SAE X0, X1, X2", `duplicate suffix "SAE"`},
		{"VADDPD.RZ_SAE.SAE X0, X1, X2", `bad suffix combination`},

		// BSWAP on 16-bit registers is undefined. See #29167,
		{"BSWAPW DX", `unrecognized instruction`},
		{"BSWAPW R11", `unrecognized instruction`},
	})
}

func testBadInstParser(t *testing.T, goarch string, tests []badInstTest) {
	for i, test := range tests {

// i.e., d(y_1 + y_2 + ...)/dx_1, d(y_1 + y_2 + ...)/dx_2...
//
// `dx` are used as initial gradients (which represent the symbolic partial
// derivatives of some loss function `L` w.r.t. `y`).
// `dx` must be nullptr or have size `ny`.
// If `dx` is nullptr, the implementation will use dx of `OnesLike` for all
// shapes in `y`.

	{"(AX)(CX*8)", "(AX)(CX*8)"},
	{"(BP)(CX*4)", "(BP)(CX*4)"},
	{"(BP)(DX*4)", "(BP)(DX*4)"},
	{"(BP)(R8*4)", "(BP)(R8*4)"},
	{"(BX)", "(BX)"},
	{"(DI)", "(DI)"},
	{"(DI)(BX*1)", "(DI)(BX*1)"},
	{"(DX)", "(DX)"},
	{"(R9)", "(R9)"},
	{"(R9)(BX*8)", "(R9)(BX*8)"},
	{"(SI)", "(SI)"},
	{"(SI)(BX*1)", "(SI)(BX*1)"},
	{"(SI)(DX*1)", "(SI)(DX*1)"},
	{"(SP)", "(SP)"},
	{"(SP)(AX*4)", "(SP)(AX*4)"},

     *
     *    dX = U / Y
     *    dY = -U*X / Y^2 = (X/Y) * -U / Y = -U*Z / Y
     *
     */

    AbstractTensorHandle* upstream_grad = grad_outputs[0];
    AbstractTensorHandle* Y = forward_inputs_[1];
    AbstractTensorHandle* Z = forward_outputs_[0];

    // Calculate dX =  U / Y
    std::string name = "Div_Grad_X";
    TF_RETURN_IF_ERROR(

	VPGATHERDQ X2, 664(BX*1), X1    // ERROR "invalid instruction"
	VPGATHERDQ Y2, (BP)(AX*2), Y1   // ERROR "invalid instruction"
	VPGATHERDQ Y5, 664(DX*8), Y6    // ERROR "invalid instruction"
	VPGATHERDQ Y5, (DX), Y0         // ERROR "invalid instruction"
	// VM/X rejects Y index register.
	VPGATHERDQ Y5, 664(Y14*8), Y6   // ERROR "invalid instruction"

    ASSERT_EQ(errors::OK, s.code()) << s.message();
    y.reset(y_raw);
  }

  float expected_dx[1] = {7.0f};
  ASSERT_NO_FATAL_FAILURE(CompareNumericalAndManualGradients(
      MulModel, ctx_.get(), {x.get(), y.get()}, 0, expected_dx, 1,
      UseFunction()));
}

#ifdef PLATFORM_GOOGLE
INSTANTIATE_TEST_SUITE_P(
    UnifiedCAPI, GradientCheckerTest,

<mxfile host="65bd71144e">
    <diagram id="jyERGzDynktFHFRGN0ph" name="Page-1">
        <mxGraphModel dx="2738" dy="2173" grid="1" gridSize="10" guides="1" tooltips="1" connect="1" arrows="1" fold="1" page="1" pageScale="1" pageWidth="1920" pageHeight="1200" math="0" shadow="0">
            <root>
                <mxCell id="0"/>
                <mxCell id="1" parent="0"/>