// TODO: Maybe the constants are named wrong.
	SPB: "spl",
	BPB: "bpl",
	SIB: "sil",
	DIB: "dil",

	R8L:  "r8d",
	R9L:  "r9d",
	R10L: "r10d",
	R11L: "r11d",
	R12L: "r12d",
	R13L: "r13d",
	R14L: "r14d",
	R15L: "r15d",

	R8L:  "%r8d",
	R9L:  "%r9d",
	R10L: "%r10d",
	R11L: "%r11d",
	R12L: "%r12d",
	R13L: "%r13d",
	R14L: "%r14d",
	R15L: "%r15d",
	RAX:  "%rax",
	RCX:  "%rcx",
	RDX:  "%rdx",
	RBX:  "%rbx",
	RSP:  "%rsp",
	RBP:  "%rbp",
	RSI:  "%rsi",
	RDI:  "%rdi",
	R8:   "%r8",
	R9:   "%r9",
	R10:  "%r10",
	R11:  "%r11",
	R12:  "%r12",
	R13:  "%r13",
	R14:  "%r14",
	R15:  "%r15",
	IP:   "%ip",

	R10B: "R10B",
	R11B: "R11B",
	R12B: "R12B",
	R13B: "R13B",
	R14B: "R14B",
	R15B: "R15B",
	AX:   "AX",
	CX:   "CX",
	BX:   "BX",
	DX:   "DX",
	SP:   "SP",
	BP:   "BP",
	SI:   "SI",
	DI:   "DI",
	R8W:  "R8W",
	R9W:  "R9W",
	R10W: "R10W",
	R11W: "R11W",
	R12W: "R12W",
	R13W: "R13W",
	R14W: "R14W",
	R15W: "R15W",
	EAX:  "EAX",
	ECX:  "ECX",
	EDX:  "EDX",

	R9B:  "R9",
	R10B: "R10",
	R11B: "R11",
	R12B: "R12",
	R13B: "R13",
	R14B: "R14",
	R15B: "R15",
	AX:   "AX",
	CX:   "CX",
	BX:   "BX",
	DX:   "DX",
	SP:   "SP",
	BP:   "BP",
	SI:   "SI",
	DI:   "DI",
	R8W:  "R8",
	R9W:  "R9",
	R10W: "R10",
	R11W: "R11",
	R12W: "R12",
	R13W: "R13",
	R14W: "R14",
	R15W: "R15",
	EAX:  "AX",
	ECX:  "CX",
	EDX:  "DX",
	EBX:  "BX",

	"DL",
	"BL",
	"SPB",
	"BPB",
	"SIB",
	"DIB",
	"R8B",
	"R9B",
	"R10B",
	"R11B",
	"R12B",
	"R13B",
	"R14B",
	"R15B",
	"AX", // [D_AX]
	"CX",
	"DX",
	"BX",
	"SP",
	"BP",
	"SI",
	"DI",
	"R8",
	"R9",
	"R10",
	"R11",
	"R12",
	"R13",
	"R14",
	"R15",
	"AH",
	"CH",
	"DH",
	"BH",
	"F0", // [D_F0]
	"F1",
	"F2",
	"F3",
	"F4",
	"F5",
	"F6",

	MOVD    R3, R14			// pointer to args

	MOVD	(0*8)(R14), R12		// fn
	MOVD	(1*8)(R14), R3		// a1
	MOVD	(2*8)(R14), R4		// a2
	MOVD	(3*8)(R14), R5		// a3
	MOVD	(4*8)(R14), R6		// a4
	MOVD	(5*8)(R14), R7		// a5
	MOVD	(6*8)(R14), R8		// a6
	MOVD	$0, R9			// vararg

	MOVD	R12, CTR
	CALL	(CTR)

	MOVD	R3, (7*8)(R14)		// r1
	MOVD	R4, (8*8)(R14)		// r2

	ADDQ  1*8+frame_YFER(SP), R10
	ORQ   BX, DI

	MOVQ  DX, R15
	RORXQ $34, R11, R12
	XORQ  R14, R13
	XORQ  R9, R15

	RORXQ $14, R8, R14
	XORQ  R14, R13
	RORXQ $39, R11, R14
	ANDQ  R8, R15
	ADDQ  R10, CX

	ANDQ AX, DI
	XORQ R12, R14

	RORXQ $28, R11, R12
	XORQ  R9, R15

	XORQ R12, R14
	MOVQ R11, R12
	ANDQ BX, R12
	ADDQ R13, R15

	ORQ  R12, DI
	ADDQ R14, R10

	VCNTMBB	V0, $1, R14	// Sum the value of bit 0 of each byte of the compare into R14.
	SRD	$56, R14, R14	// The result of VCNTMBB is shifted. Unshift it.
	ADD	R14, R18, R3
	RET

#else
tail:	// Count the last 0 - 31 bytes.
	CMP	R4, $16
	BLT	tail_8
	MOVD	(R3), R12
	MOVD	8(R3), R14
	CMPB	R12, R5, R12
	CMPB	R14, R5, R14
	POPCNTD	R12, R12
	POPCNTD	R14, R14
	ADD	R12, R18, R18
	ADD	R14, R18, R18

	// First reduction chain
	MOVQ   $0x0007ffffffffffff, AX
	SHLQ   $0x0d, DI, SI
	SHLQ   $0x0d, R9, R8
	SHLQ   $0x0d, R11, R10
	SHLQ   $0x0d, R13, R12
	SHLQ   $0x0d, R15, R14
	ANDQ   AX, DI
	IMUL3Q $0x13, R14, R14
	ADDQ   R14, DI
	ANDQ   AX, R9
	ADDQ   SI, R9
	ANDQ   AX, R11
	ADDQ   R8, R11
	ANDQ   AX, R13
	ADDQ   R10, R13
	ANDQ   AX, R15
	ADDQ   R12, R15

	BEQ	R4, R6, samebytes // same start of a and b

	SGTU	R5, R7, R9
	BNE	R0, R9, r2_lt_r1
	MOVV	R5, R14
	JMP	entry
r2_lt_r1:
	MOVV	R7, R14	// R14 is min(R4, R5)
entry:
	ADDV	R4, R14, R12	// R6 start of a, R14 end of a
	BEQ	R4, R12, samebytes // length is 0

	SRLV	$4, R14		// R14 is number of chunks
	BEQ	R0, R14, byte_loop

	// make sure both a and b are aligned.
	OR	R4, R6, R15
	AND	$7, R15
	BNE	R0, R15, byte_loop