TEXT runtime·memclrNoHeapPointers(SB), NOSPLIT, $0-8
	MOVL	ptr+0(FP), DI
	MOVL	n+4(FP), BX
	XORL	AX, AX

tail:
	TESTL	BX, BX
	JEQ	_0
	CMPL	BX, $2
	JBE	_1or2
	CMPL	BX, $4
	JB	_3
	JE	_4
	CMPL	BX, $8
	JBE	_5through8
	CMPL	BX, $16
	JBE	_9through16
	MOVL	BX, CX
	SHRL	$2, CX
	REP
	STOSL
	ANDL	$3, BX
	JNE	tail
	RET

_1or2:
	MOVB	AX, (DI)
	MOVB	AX, -1(DI)(BX*1)

// output:
//   AX = output (-1/0/1)
TEXT cmpbody<>(SB),NOSPLIT,$0-0
	CMPQ	SI, DI
	JEQ	allsame
	CMPQ	BX, DX
	MOVQ	DX, R8
	CMOVQLT	BX, R8 // R8 = min(alen, blen) = # of bytes to compare
	CMPQ	R8, $8
	JB	small

	CMPQ	R8, $63
	JBE	loop
#ifndef hasAVX2
	CMPB	internal∕cpu·X86+const_offsetX86HasAVX2(SB), $1
	JEQ     big_loop_avx2
	JMP	big_loop
#else
	JMP	big_loop_avx2
#endif
loop:
	CMPQ	R8, $16

	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:

		want string
	}{
		{Name(""), byte(7), "7"},
		{Name(""), jsonbyte(7), `{"JB":7}`},
		{Name(""), textbyte(4), `"TB:4"`},
		{Name(""), jsonint(5), `{"JI":5}`},
		{Name(""), textint(1), `"TI:1"`},
		{Name(""), []byte{0, 1}, `"AAE="`},
		{Name(""), []jsonbyte{0, 1}, `[{"JB":0},{"JB":1}]`},
		{Name(""), [][]jsonbyte{{0, 1}, {3}}, `[[{"JB":0},{"JB":1}],[{"JB":3}]]`},
		{Name(""), []textbyte{2, 3}, `["TB:2","TB:3"]`},

	MOVOU  32(SI), X3
	MOVOU  48(SI), X4
	PXOR   X0, X1
	ADDQ   $64, SI                  // buf+=64
	SUBQ   $64, CX                  // len-=64
	CMPQ   CX, $64                  // Less than 64 bytes left
	JB     remain64

	MOVOA  r2r1<>+0(SB), X0
loopback64:
	MOVOA  X1, X5
	MOVOA  X2, X6
	MOVOA  X3, X7
	MOVOA  X4, X8

	PCLMULQDQ $0, X0, X1
	PCLMULQDQ $0, X0, X2
	PCLMULQDQ $0, X0, X3

	shiftB1Right; shiftC1Right; shiftD1Right
	shiftB2Right; shiftC2Right; shiftD2Right

	CMPQ itr2, itr1
	JB   openSSLTail192LoopA

	CMPQ itr2, $160
	JNE  openSSLTail192LoopB

	CMPQ inl, $176
	JB   openSSLTail192Store

	polyAdd(160(inp))
	polyMul

	CMPQ inl, $192
	JB   openSSLTail192Store

	polyAdd(176(inp))
	polyMul

openSSLTail192Store:

	// MOVOU seems always faster than REP STOSL.
tail:
	// BSR+branch table make almost all memmove/memclr benchmarks worse. Not worth doing.
	TESTL	BX, BX
	JEQ	_0
	CMPL	BX, $2
	JBE	_1or2
	CMPL	BX, $4
	JB	_3
	JE	_4
	CMPL	BX, $8
	JBE	_5through8
	CMPL	BX, $16
	JBE	_9through16
#ifdef GO386_softfloat
	JMP	nosse2
#endif
	PXOR	X0, X0
	CMPL	BX, $32
	JBE	_17through32
	CMPL	BX, $64
	JBE	_33through64

function compressDashboard() {
  < "${DASHBOARDS}/$1" jq -c  > "${TMP}/$1"
}

# Set up grafana
{
  # Generate all dynamic dashboards
  (
    pushd "${DASHBOARDS}" > /dev/null
    jb install
    for file in *.libsonnet; do
      dashboard="${file%.*}"
      jsonnet -J vendor -J lib "${file}" > "${dashboard}-dashboard.gen.json"
    done
  )
  helm3 template grafana grafana \

	// for large sizes. The cutover is approximately 1K.
tail:
	TESTQ	BX, BX
	JEQ	move_0
	CMPQ	BX, $2
	JBE	move_1or2
	CMPQ	BX, $4
	JBE	move_3or4
	CMPQ	BX, $8
	JB	move_5through7
	JE	move_8
	CMPQ	BX, $16
	JBE	move_9through16

/*
 * check and set for backwards
 */
	CMPQ	SI, DI
	JLS	back

/*
 * forward copy loop
 */
forward:
	MOVQ	BX, CX

	iz    interface{} = z
	ip    interface{} = p
	ipp   interface{} = pp

	ju16  J = u16
	ju32  J = u32
	ju64  J = u64
	ju128 J = u128
	jf32  J = f32
	jf64  J = f64
	jc128 J = c128
	js    J = s
	jb    J = b
	jm    J = m
	jc    J = c
	jz J = z
	jp J = p // The method set for *T contains the methods for T.
	// pp does not implement error.
)

func second(a ...interface{}) interface{} {
	return a[1]
}