{name: "SHRQ", argLength: 2, reg: gp21shift, asm: "SHRQ", resultInArg0: true, clobberFlags: true},
		{name: "SHRL", argLength: 2, reg: gp21shift, asm: "SHRL", resultInArg0: true, clobberFlags: true},
		{name: "SHRW", argLength: 2, reg: gp21shift, asm: "SHRW", resultInArg0: true, clobberFlags: true},
		{name: "SHRB", argLength: 2, reg: gp21shift, asm: "SHRB", resultInArg0: true, clobberFlags: true},

	ADCQ $0, mul0
	TESTQ $1, t0

	CMOVQEQ t0, acc4
	CMOVQEQ t1, acc5
	CMOVQEQ t2, acc6
	CMOVQEQ t3, acc7
	ANDQ t0, mul0

	SHRQ $1, acc5, acc4
	SHRQ $1, acc6, acc5
	SHRQ $1, acc7, acc6
	SHRQ $1, mul0, acc7
	ST (y)
	/////////////////////////
	LDacc (x)
	LDt (s)
	CALL p256MulInternal(SB)
	ST (s)
	p256MulBy2Inline
	STt (tmp)

	LDacc (m)

 * forward copy loop
 */
forward:
	MOVQ	BX, CX
	SHRQ	$3, CX
	ANDQ	$7, BX

	REP;	MOVSQ
	JMP	tail

back:
/*
 * check overlap
 */
	MOVQ	SI, CX
	ADDQ	BX, CX
	CMPQ	CX, DI
	JLS	forward

/*
 * whole thing backwards has
 * adjusted addresses
 */
	ADDQ	BX, DI
	ADDQ	BX, SI
	STD

/*
 * copy
 */
	MOVQ	BX, CX
	SHRQ	$3, CX
	ANDQ	$7, BX

	SUBQ	$8, DI
	SUBQ	$8, SI

		{[]string{}, "foo", 0, false},
		{[]string{}, "", 0, false},

		{str1, "foo", 0, true},
		{str1, "bar", 0, false},
		{str1, "zx", 1, false},

		{str2, "aa", 0, false},
		{str2, "ab", 0, true},
		{str2, "ad", 1, false},
		{str2, "ca", 1, true},
		{str2, "ra", 2, false},

		{str3, "bb", 0, false},
		{str3, "mo", 0, true},
		{str3, "nb", 1, false},
		{str3, "qo", 1, true},
		{str3, "tr", 2, false},

    val b1: Int
      get() =
        when {
          codepointDelta < 0 -> 0x40 or (absoluteDelta shr 14)
          codepointDelta > 0 -> 0x50 or (absoluteDelta shr 14)
          else -> error("Unexpected codepointDelta of 0")
        }

    val b2: Int
      get() = absoluteDelta shr 7 and 0x7f

    val b3: Int
      get() = absoluteDelta and 0x7f

    companion object {

	JMP	si_finish
si_high:
	// address ends in 11111xxx. Load up to bytes we want, move to correct position.
	MOVQ	-8(SI)(BX*1), SI
	SHRQ	CX, SI
si_finish:

	// same for DI.
	CMPB	DI, $0xf8
	JA	di_high
	MOVQ	(DI), DI
	JMP	di_finish
di_high:
	MOVQ	-8(DI)(BX*1), DI
	SHRQ	CX, DI
di_finish:

	SUBQ	SI, DI
	SHLQ	CX, DI
equal:
	SETEQ	AX

/*
 * forward copy loop
 */
forward:
	MOVL	BX, CX
	SHRL	$2, CX
	ANDL	$3, BX

	REP;	MOVSL
	JMP	tail
/*
 * check overlap
 */
back:
	MOVL	SI, CX
	ADDL	BX, CX
	CMPL	CX, DI
	JLS	forward
/*
 * whole thing backwards has
 * adjusted addresses
 */

	ADDL	BX, DI
	ADDL	BX, SI
	STD

/*
 * copy
 */
	MOVL	BX, CX
	SHRL	$2, CX
	ANDL	$3, BX

	SUBL	$4, DI
	SUBL	$4, SI

	ANDQ  $3, h2;                  \
	MOVQ  t2, t0;                  \
	ANDQ  $0xFFFFFFFFFFFFFFFC, t0; \
	ADDQ  t0, h0;                  \
	ADCQ  t3, h1;                  \
	ADCQ  $0, h2;                  \
	SHRQ  $2, t3, t2;              \
	SHRQ  $2, t3;                  \
	ADDQ  t2, h0;                  \
	ADCQ  t3, h1;                  \
	ADCQ  $0, h2

// func update(state *[7]uint64, msg []byte)
TEXT ·update(SB), $0-32

	x.mant = append(x.mant[:0], s[:n]...)

	// Do any (remaining) shift right in decimal representation.
	if shift < 0 {
		for shift < -maxShift {
			shr(x, maxShift)
			shift += maxShift
		}
		shr(x, uint(-shift))
	}
}

// shr implements x >> s, for s <= maxShift.
func shr(x *decimal, s uint) {
	// Division by 1<<s using shift-and-subtract algorithm.

	// pick up enough leading digits to cover first shift

	MOVL $0, DX
	SHRL CX, AX, DX		// w1<<ŝ
	MOVL DX, c+28(FP)

	MOVL $0, BX		// i = 0
	JMP E9

	// i < n-1
L9:	MOVL AX, DX		// w = w1
	MOVL 4(SI)(BX*4), AX	// w1 = x[i+1]
	SHRL CX, AX, DX		// w>>s | w1<<ŝ
	MOVL DX, (DI)(BX*4)	// z[i] = w>>s | w1<<ŝ
	ADDL $1, BX		// i++

E9:	CMPL BX, BP
	JL L9			// i < n-1

	// i >= n-1
X9a:	SHRL CX, AX		// w1>>s