XORQ R15, R15
	RET
TEXT ·a5(SB), 0, $0-0
	CMPL runtime·writeBarrier(SB), $0
	XORL R15, R15
	RET
TEXT ·a6(SB), 0, $0-0
	CMPL runtime·writeBarrier(SB), $0
	POPQ R15
	PUSHQ R15
	RET
TEXT ·a7(SB), 0, $0-0
	CMPL runtime·writeBarrier(SB), $0
	MOVQ R15, AX // ERROR "when dynamic linking, R15 is clobbered by a global variable access and is used here"
	RET
TEXT ·a8(SB), 0, $0-0
	CMPL runtime·writeBarrier(SB), $0

	std     %f9, 8(%r15)
	std     %f10, 16(%r15)
	std     %f11, 24(%r15)
	std     %f12, 32(%r15)
	std     %f13, 40(%r15)
	std     %f14, 48(%r15)
	std     %f15, 56(%r15)

	/* restore g pointer */
	lgr     %r13, %r3

	/* call fn */
	basr    %r14, %r2

	/* restore floating point registers */
	ld      %f8, 0(%r15)
	ld      %f9, 8(%r15)
	ld      %f10, 16(%r15)
	ld      %f11, 24(%r15)
	ld      %f12, 32(%r15)

	V4FNMADDPS 99(R15)(R15*1), [Z1-Z4], K3, Z15        // 6212774baabc3f63000000
	V4FNMADDPS (DX), [Z1-Z4], K3, Z15                  // 6272774baa3a
	V4FNMADDPS 99(R15)(R15*1), [Z11-Z14], K3, Z15      // 6212274baabc3f63000000
	V4FNMADDPS (DX), [Z11-Z14], K3, Z15                // 6272274baa3a
	V4FNMADDPS 99(R15)(R15*1), [Z21-Z24], K3, Z15      // 62125743aabc3f63000000

// for more details.
// No arg size here, runtime pulls arg map out of the func value.
TEXT ·makeFuncStub(SB),(NOSPLIT|WRAPPER),$40
	NO_LOCAL_POINTERS
	MOVD	R12, 8(R15)
	MOVD	$argframe+0(FP), R3
	MOVD	R3, 16(R15)
	MOVB	$0, 40(R15)
	ADD	$40, R15, R3
	MOVD	R3, 24(R15)
	MOVD	$0, 32(R15)
	BL	·callReflect(SB)
	RET

// methodValueCall is the code half of the function returned by makeMethodValue.

				{1, 49151}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R15
			},
		},
	},
	{
		name:   "CMPW",
		argLen: 2,
		asm:    x86.ACMPW,
		reg: regInfo{
			inputs: []inputInfo{
				{0, 49151}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R15
				{1, 49151}, // AX CX DX BX SP BP SI DI R8 R9 R10 R11 R12 R13 R15
			},
		},
	},
	{

	VPCLMULQDQ $127, (R15), X28, X16                   // 62c31d0044077f or 62c39d0044077f
	VPCLMULQDQ $0, Y15, Y2, Y31                        // 62436d2844ff00 or 6243ed2844ff00
	VPCLMULQDQ $0, Y22, Y2, Y31                        // 62236d2844fe00 or 6223ed2844fe00
	VPCLMULQDQ $0, Y20, Y2, Y31                        // 62236d2844fc00 or 6223ed2844fc00

	// Case for 8 ≤ n < 16 bytes
	MOVD	(R4)(R8), R14   // R14 = a[i,...,i+7]
	MOVD	(R5)(R8), R15   // R15 = b[i,...,i+7]
	XOR	R14, R15, R16   // R16 = a[] ^ b[]
	SUB	$8, R6          // n = n - 8
	MOVD	R16, (R3)(R8)   // Store to dst
	ADD	$8, R8
xor4:
	CMP	R6, $4
	BLT	xor2
	MOVWZ	(R4)(R8), R14
	MOVWZ	(R5)(R8), R15
	XOR	R14, R15, R16
	MOVW	R16, (R3)(R8)
	ADD	$4,R8
	ADD	$-4,R6
xor2:
	CMP	R6, $2
	BLT	xor1

	// r4 = a0×b4
	MOVQ (CX), AX
	MULQ 32(BX)
	MOVQ AX, R15
	MOVQ DX, R14

	// r4 += a1×b3
	MOVQ 8(CX), AX
	MULQ 24(BX)
	ADDQ AX, R15
	ADCQ DX, R14

	// r4 += a2×b2
	MOVQ 16(CX), AX
	MULQ 16(BX)
	ADDQ AX, R15
	ADCQ DX, R14

	// r4 += a3×b1
	MOVQ 24(CX), AX
	MULQ 8(BX)
	ADDQ AX, R15
	ADCQ DX, R14

	// r4 += a4×b0
	MOVQ 32(CX), AX
	MULQ (BX)

	MOVQ msg_base+8(FP), SI
	MOVQ msg_len+16(FP), R15

	MOVQ 0(DI), R8   // h0
	MOVQ 8(DI), R9   // h1
	MOVQ 16(DI), R10 // h2
	MOVQ 24(DI), R11 // r0
	MOVQ 32(DI), R12 // r1

	CMPQ R15, $16
	JB   bytes_between_0_and_15

loop:
	POLY1305_ADD(SI, R8, R9, R10)

multiply:
	POLY1305_MUL(R8, R9, R10, R11, R12, BX, CX, R13, R14)
	SUBQ $16, R15
	CMPQ R15, $16
	JAE  loop

bytes_between_0_and_15:

	VPCOMPRESSW Z3, K7, -15(R14)(R15*2)                // 6292fd4f639c7ef1ffffff
	VPCOMPRESSW Z0, K7, -15(R14)(R15*2)                // 6292fd4f63847ef1ffffff
	VPEXPANDB X16, K1, X6                              // 62b27d0962f0
	VPEXPANDB X28, K1, X6                              // 62927d0962f4
	VPEXPANDB X8, K1, X6                               // 62d27d0962f0
	VPEXPANDB 99(R15)(R15*4), K1, X6                   // 62927d096274bf63