MOVD F10, 288(X2)
	MOVD F11, 296(X2)
	MOVD F12, 304(X2)
	MOVD F13, 312(X2)
	MOVD F14, 320(X2)
	MOVD F15, 328(X2)
	MOVD F16, 336(X2)
	MOVD F17, 344(X2)
	MOVD F18, 352(X2)
	MOVD F19, 360(X2)
	MOVD F20, 368(X2)
	MOVD F21, 376(X2)
	MOVD F22, 384(X2)
	MOVD F23, 392(X2)
	MOVD F24, 400(X2)
	MOVD F25, 408(X2)
	MOVD F26, 416(X2)
	MOVD F27, 424(X2)
	MOVD F28, 432(X2)
	MOVD F29, 440(X2)
	MOVD F30, 448(X2)

	MOVD	R28, 160(R1)
	MOVD	R29, 168(R1)
	MOVD	g, 176(R1) // R30
	MOVD	R31, 184(R1)
	FMOVD	F14, 192(R1)
	FMOVD	F15, 200(R1)
	FMOVD	F16, 208(R1)
	FMOVD	F17, 216(R1)
	FMOVD	F18, 224(R1)
	FMOVD	F19, 232(R1)
	FMOVD	F20, 240(R1)
	FMOVD	F21, 248(R1)
	FMOVD	F22, 256(R1)
	FMOVD	F23, 264(R1)
	FMOVD	F24, 272(R1)
	FMOVD	F25, 280(R1)
	FMOVD	F26, 288(R1)
	FMOVD	F27, 296(R1)
	FMOVD	F28, 304(R1)
	FMOVD	F29, 312(R1)

	MOVD	k_base+48(FP), R2                            // k constants first address
	MOVD	p_len+32(FP), R3                             // message length
	VLD1.P	16(R0), [V0.S4]
	FMOVS	(R0), F20
	SUB	$16, R0, R0

blockloop:

	VLD1.P	16(R1), [V4.B16]                             // load message
	VLD1.P	16(R1), [V5.B16]
	VLD1.P	16(R1), [V6.B16]
	VLD1.P	16(R1), [V7.B16]

func f17[E2 ~byte, E1 ~byte, S1 ~[]E1, S2 ~[]E2](S1, S2) {}
func f18[E2 ~byte, E1 ~byte, S2 ~[]E2, S1 ~[]E1](S1, S2) {}
func f19[E1 ~byte, E2 ~byte, S2 ~[]E2, S1 ~[]E1](S1, S2) {}
func f20[S2 ~[]E2, E2 ~byte, E1 ~byte, S1 ~[]E1](S1, S2) {}
func f21[E2 ~byte, S2 ~[]E2, E1 ~byte, S1 ~[]E1](S1, S2) {}
func f22[E1 ~byte, S2 ~[]E2, E2 ~byte, S1 ~[]E1](S1, S2) {}
func f23[S2 ~[]E2, E1 ~byte, E2 ~byte, S1 ~[]E1](S1, S2) {}

	FMOVD	F14, (offset+8*0)(R1)  \
	FMOVD	F15, (offset+8*1)(R1)  \
	FMOVD	F16, (offset+8*2)(R1)  \
	FMOVD	F17, (offset+8*3)(R1)  \
	FMOVD	F18, (offset+8*4)(R1)  \
	FMOVD	F19, (offset+8*5)(R1)  \
	FMOVD	F20, (offset+8*6)(R1)  \
	FMOVD	F21, (offset+8*7)(R1)  \
	FMOVD	F22, (offset+8*8)(R1)  \
	FMOVD	F23, (offset+8*9)(R1)  \
	FMOVD	F24, (offset+8*10)(R1) \
	FMOVD	F25, (offset+8*11)(R1) \
	FMOVD	F26, (offset+8*12)(R1) \

	SMSUBL R10, R3, R17, R27    <=>    smsubl x27, w17, w10, x3

(3) FMADDD, FMADDS, FMSUBD, FMSUBS, FNMADDD, FNMADDS, FNMSUBD, FNMSUBS <Fm>, <Fa>, <Fn>, <Fd>

Examples:

	FMADDD F30, F20, F3, F29    <=>    fmadd d29, d3, d30, d20
	FNMSUBS F7, F25, F7, F22    <=>    fnmsub s22, s7, s7, s25

(4) BFI, BFXIL, SBFIZ, SBFX, UBFIZ, UBFX $<lsb>, <Rn>, $<width>, <Rd>

Examples:

	MOV	X23, (8*8)(X2)
	MOV	X24, (9*8)(X2)
	MOV	X25, (10*8)(X2)
	MOV	X26, (11*8)(X2)
	MOV	g, (12*8)(X2)
	MOVF	F8, (13*8)(X2)
	MOVF	F9, (14*8)(X2)
	MOVF	F18, (15*8)(X2)
	MOVF	F19, (16*8)(X2)
	MOVF	F20, (17*8)(X2)
	MOVF	F21, (18*8)(X2)
	MOVF	F22, (19*8)(X2)
	MOVF	F23, (20*8)(X2)
	MOVF	F24, (21*8)(X2)
	MOVF	F25, (22*8)(X2)
	MOVF	F26, (23*8)(X2)
	MOVF	F27, (24*8)(X2)

	MOV	m_g0(X10), g

	"F0",
	"F1",
	"F2",
	"F3",
	"F4",
	"F5",
	"F6",
	"F7",
	"F8",
	"F9",
	"F10",
	"F11",
	"F12",
	"F13",
	"F14",
	"F15",

	"F16",
	"F17",
	"F18",
	"F19",
	"F20",
	"F21",
	"F22",
	"F23",
	"F24",
	"F25",
	"F26",
	"F27",
	"F28",
	"F29",
	"F30",
	"F31",

	"SP",
	"g",

	// pseudo-registers
	"SB",
}

func init() {

// pre/post-indexed
	FMOVS.P	F20, 4(R0)                                      // 144400bc
	FMOVS.W	F20, 4(R0)                                      // 144c00bc
	FMOVD.P	F20, 8(R1)                                      // 348400fc
	FMOVQ.P	F13, 11(R10)                                    // 4db5803c
	FMOVQ.W	F15, 11(R20)                                    // 8fbe803c

	FMOVS.P	8(R0), F20                                      // 148440bc

		c: `#include <stdlib.h>
		    struct s20a { char *a[32769]; };
		    struct s20b { struct s20a f; };
		    struct s20b *f20() { return malloc(sizeof(struct s20b)); }
		    void f20b(struct s20b *p) {}
		    void f20c(void *p) {}`,
		imports:   []string{"unsafe"},
		body:      `p := C.f20(); n := &C.struct_s20a{[32769]*C.char{new(C.char)}}; p.f = *n; C.f20b(p); n.a[0] = nil; C.f20c(unsafe.Pointer(n))`,
		fail:      true,