MOVD R30, (RSP)
	#endif
	STP (R0, R1), 8(RSP)
	STP (R2, R3), 24(RSP)
	STP (R4, R5), 40(RSP)
	STP (R6, R7), 56(RSP)
	STP (R8, R9), 72(RSP)
	STP (R10, R11), 88(RSP)
	STP (R12, R13), 104(RSP)
	STP (R14, R15), 120(RSP)
	STP (R16, R17), 136(RSP)
	STP (R19, R20), 152(RSP)
	STP (R21, R22), 168(RSP)
	STP (R23, R24), 184(RSP)
	STP (R25, R26), 200(RSP)
	MOVD NZCV, R0
	MOVD R0, 216(RSP)
	MOVD FPSR, R0
	MOVD R0, 224(RSP)

loop:
	CBZ	R0, done

	LDP.P	16(R2), (R5, R6)
	LDP.P	16(R2), (R7, R8)

	LDP	(R1), (R9, R10)
	ADDS	R4, R9
	MUL	R6, R3, R14
	ADCS	R14, R10
	MUL	R7, R3, R15
	LDP	16(R1), (R11, R12)
	ADCS	R15, R11
	MUL	R8, R3, R16
	ADCS	R16, R12
	UMULH	R8, R3, R20
	ADC	$0, R20

	MUL	R5, R3, R13
	ADDS	R13, R9
	UMULH	R5, R3, R17
	ADCS	R17, R10
	UMULH	R6, R3, R21
	STP.P	(R9, R10), 16(R1)

	print("r10    ", hex(c.r10()), "\n")
	print("r11    ", hex(c.r11()), "\n")
	print("r12    ", hex(c.r12()), "\n")
	print("r13    ", hex(c.r13()), "\n")
	print("r14    ", hex(c.r14()), "\n")
	print("r15    ", hex(c.r15()), "\n")
	print("rip    ", hex(c.rip()), "\n")
	print("rflags ", hex(c.rflags()), "\n")
	print("cs     ", hex(c.cs()), "\n")
	print("fs     ", hex(c.fs()), "\n")
	print("gs     ", hex(c.gs()), "\n")
}

	VLD1R.P	(R15)(R1), [V15.H4]                             // efc5c10d
	VLD2R	(R15), [V15.H4, V16.H4]                         // efc5600d
	VLD2R.P	16(R0), [V0.D2, V1.D2]                          // 00ccff4d
	VLD2R.P	(R0)(R5), [V31.D1, V0.D1]                       // 1fcce50d
	VLD3R	(RSP), [V31.S2, V0.S2, V1.S2]                   // ffeb400d
	VLD3R.P	6(R15), [V15.H4, V16.H4, V17.H4]                // efe5df0d

	MOVD.P	R12, 8(R0)
less_than8:
	TBZ	$2, R3, less_than4
	MOVWU.P	4(R1), R13
	MOVWU.P	4(R2), R14
	EORW	R13, R14, R14
	MOVWU.P	R14, 4(R0)
less_than4:
	TBZ	$1, R3, less_than2
	MOVHU.P	2(R1), R15
	MOVHU.P	2(R2), R16
	EORW	R15, R16, R16
	MOVHU.P	R16, 2(R0)
less_than2:
	TBZ	$0, R3, end
	MOVBU	(R1), R17
	MOVBU	(R2), R19
	EORW	R17, R19, R19
	MOVBU	R19, (R0)
end:

rest:
	// Transition from C ABI to Go ABI.
	PUSH_REGS_HOST_TO_ABI0()
	// Set g = g0.
	get_tls(R12)
	MOVQ	g(R12), R14
	MOVQ	g_m(R14), R13
	MOVQ	m_g0(R13), R15
	CMPQ	R13, R15
	JEQ	noswitch	// branch if already on g0
	MOVQ	R15, g(R12)	// g = m->g0
	MOVQ	R15, R14	// set g register
	PUSHQ	RARG1	// func arg
	PUSHQ	RARG0	// func arg
	CALL	runtime·racecallback(SB)
	POPQ	R12
	POPQ	R12

	print("r10  ", hex(c.r10()), "\t")
	print("r11  ", hex(c.r11()), "\n")
	print("r12  ", hex(c.r12()), "\t")
	print("r13  ", hex(c.r13()), "\n")
	print("r14  ", hex(c.r14()), "\t")
	print("r15  ", hex(c.r15()), "\n")
	print("r16  ", hex(c.r16()), "\t")
	print("r17  ", hex(c.r17()), "\n")
	print("r18  ", hex(c.r18()), "\t")
	print("r19  ", hex(c.r19()), "\n")
	print("r20  ", hex(c.r20()), "\t")

	print("r10  ", hex(c.r10()), "\t")
	print("r11  ", hex(c.r11()), "\n")
	print("r12  ", hex(c.r12()), "\t")
	print("r13  ", hex(c.r13()), "\n")
	print("r14  ", hex(c.r14()), "\t")
	print("r15  ", hex(c.r15()), "\n")
	print("r16  ", hex(c.r16()), "\t")
	print("r17  ", hex(c.r17()), "\n")
	print("r18  ", hex(c.r18()), "\t")
	print("r19  ", hex(c.r19()), "\n")
	print("r20  ", hex(c.r20()), "\t")

	MOVV R5, 40(R29)
	MOVV R6, 48(R29)
	MOVV R7, 56(R29)
	MOVV R8, 64(R29)
	MOVV R9, 72(R29)
	MOVV R10, 80(R29)
	MOVV R11, 88(R29)
	MOVV R12, 96(R29)
	MOVV R13, 104(R29)
	MOVV R14, 112(R29)
	MOVV R15, 120(R29)
	MOVV R16, 128(R29)
	MOVV R17, 136(R29)
	MOVV R18, 144(R29)
	MOVV R19, 152(R29)
	MOVV R20, 160(R29)
	MOVV R21, 168(R29)
	MOVV R22, 176(R29)
	MOVV R24, 184(R29)
	MOVV R25, 192(R29)

	STP	(R12, R13), -16(R5)
	RET

	// Copy 65..128 bytes.
copy128:
	LDP	32(R1), (R14, R15)
	LDP	48(R1), (R16, R17)
	CMP	$96, R2
	BLS	copy96
	LDP	-64(R4), (R2, R3)
	LDP	-48(R4), (R1, R4)
	STP	(R2, R3), -64(R5)
	STP	(R1, R4), -48(R5)

copy96:
	STP	(R6, R7), (R0)
	STP	(R8, R9), 16(R0)
	STP	(R14, R15), 32(R0)
	STP	(R16, R17), 48(R0)
	STP	(R10, R11), -32(R5)
	STP	(R12, R13), -16(R5)
	RET