func (c *sigctxt) r26() uint32       { return uint32(c.regs().sc_regs[26]) }
func (c *sigctxt) r27() uint32       { return uint32(c.regs().sc_regs[27]) }
func (c *sigctxt) r28() uint32       { return uint32(c.regs().sc_regs[28]) }
func (c *sigctxt) r29() uint32       { return uint32(c.regs().sc_regs[29]) }
func (c *sigctxt) r30() uint32       { return uint32(c.regs().sc_regs[30]) }

	MOVD	$off,idx; \
	MOVWBR	(idx)(ptr), dst
#endif

#define M00 R18
#define M01 R19
#define M02 R20
#define M03 R24
#define M04 R25
#define M05 R26
#define M06 R27
#define M07 R28
#define M08 R29
#define M09 R21
#define M10 R11
#define M11 R8
#define M12 R7
#define M13 R12
#define M14 R23
#define M15 R10

#define ROUND1(a, b, c, d, index, const, shift) \
	ADD	$const, index, R9; \

// A brief recap of the arm64 calling convention.
// Arguments are passed in R0...R7, the rest is on stack.
// Callee-saved registers are: R19...R28.
// Temporary registers are: R9...R15
// SP must be 16-byte aligned.

// When calling racecalladdr, R9 is the call target address.

// The race ctx, ThreadState *thr below, is passed in R0 and loaded in racecalladdr.

	MOVD	R19, 200(R1)
	MOVD	R20, 208(R1)
	MOVD	R21, 216(R1)
	MOVD	R22, 224(R1)
	MOVD	R23, 232(R1)
	MOVD	R24, 240(R1)
	MOVD	R25, 248(R1)
	MOVD	R26, 256(R1)
	MOVD	R27, 264(R1)
	MOVD	R28, 272(R1)
	MOVD	R29, 280(R1)
	MOVD	g, 288(R1)
	MOVD	LR, R31
	MOVD	R31, 32(R1)
	CALL	runtime·debugCallCheck(SB)
	MOVD	40(R1), R22
	XOR	R0, R0
	CMP	R22, $0
	BEQ	good
	MOVD	48(R1), R22

	"R11",
	"R12",
	"R13",
	"R14",
	"R15",
	"R16",
	"R17",
	"R18",
	"R19",
	"R20",
	"R21",
	"R22",
	//REGTMP
	"R24",
	"R25",
	// R26 reserved by kernel
	// R27 reserved by kernel
	"R28",
	"SP",  // aka R29
	"g",   // aka R30
	"R31", // REGLINK

	// odd FP registers contain high parts of 64-bit FP values
	"F0",
	"F2",
	"F4",
	"F6",
	"F8",
	"F10",
	"F12",
	"F14",
	"F16",

	// R22 is g.
	MOVV	R23, 160(R3)
	MOVV	R24, 168(R3)
	MOVV	R25, 176(R3)
	MOVV	R26, 184(R3)
	// R27 already saved
	// R28 already saved.
	MOVV	R29, 192(R3)
	// R30 is tmp register.
	MOVV	R31, 200(R3)

	CALL	runtime·wbBufFlush(SB)

	MOVV	8(R3), R27
	MOVV	16(R3), R28
	MOVV	24(R3), R2
	MOVV	32(R3), R4
	MOVV	40(R3), R5
	MOVV	48(R3), R6
	MOVV	56(R3), R7
	MOVV	64(R3), R8
	MOVV	72(R3), R9

	"R9",
	"R10",
	"R11",
	"R12",
	"R13",
	"R14",
	"R15",
	"R16",
	"R17",
	"R18",
	"R19",
	"R20",
	"R21",
	"g", // aka R22
	"R23",
	"R24",
	"R25",
	"R26",
	"R27",
	"R28",
	"R29",
	// R30 is REGTMP not used in regalloc
	"R31",

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

#define R_x050	R15
#define R_x060	R16
#define R_x070	R17
#define R_x080	R18
#define R_x090	R19
#define R_x0a0	R20
#define R_x0b0	R21
#define R_x0c0	R22
#define R_x0d0	R23
#define R_x0e0	R24
#define R_x0f0	R28
#define R_x100	R29
#define R_x110	R27


// V0-V7 are A-H
// V8-V23 are used for the message schedule
#define KI	V24
#define FUNC	V25
#define S0	V26
#define S1	V27
#define s0	V28
#define s1	V29

	LDAXRB (R19), R16                          // 70fe5f08
	LDAXRH (R5), R8                            // a8fc5f48
	//TODO LDNP 0xcc(RSP), ZR, R12             // ecff5928
	//TODO LDNP 0x40(R28), R9, R5              // 852744a8
	//TODO LDPSW -0xd0(R2), R0, R12            // 4c00e668
	//TODO LDPSW 0x5c(R4), R8, R5              // 85a0cb69
	//TODO LDPSW 0x6c(R12), R2, R27            // 9b894d69

	// R21 already saved
	// R22 already saved.
	MOVW	R22, 84(R29)
	// R23 is tmp register.
	MOVW	R24, 88(R29)
	MOVW	R25, 92(R29)
	// R26 is reserved by kernel.
	// R27 is reserved by kernel.
	MOVW	R28, 96(R29)
	// R29 is SP.
	// R30 is g.
	// R31 is LR, which was saved by the prologue.

	CALL	runtime·wbBufFlush(SB)

	MOVW	4(R29), R20
	MOVW	8(R29), R21
	MOVW	12(R29), R3
	MOVW	16(R29), R4