diff --git a/decimal.go b/decimal.go
index 7ee6a15..0e0df15 100644
--- a/decimal.go
+++ b/decimal.go
@@ -1121,10 +1121,6 @@ func (x *Decimal) Signbit() bool {
 	return x.neg
 }
 
-func (z *Decimal) Sqrt(x *Decimal) *Decimal {
-	panic("not implemented")
-}
-
 // Sub sets z to the rounded difference x-y and returns z.
 // Precision, rounding, and accuracy reporting are as for Add.
 // Sub panics with ErrNaN if x and y are infinities with equal
diff --git a/decimal_test.go b/decimal_test.go
index 26ef8d8..86d5207 100644
--- a/decimal_test.go
+++ b/decimal_test.go
@@ -106,3 +106,10 @@ func BenchmarkDecimal_dnorm(b *testing.B) {
 		benchU = uint(dnorm(d))
 	}
 }
+func BenchmarkDecimal_Sqrt(b *testing.B) {
+	x := new(Decimal).SetUint64(2)
+	z := new(Decimal).SetPrec(34)
+	for i := 0; i < b.N; i++ {
+		z.Sqrt(x)
+	}
+}
diff --git a/decsqrt.go b/decsqrt.go
new file mode 100644
index 0000000..0afd7d3
--- /dev/null
+++ b/decsqrt.go
@@ -0,0 +1,110 @@
+package decimal
+
+// Copyright 2017 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+import "math"
+
+var (
+	three = new(Decimal).SetUint64(3)
+	two   = new(Decimal).SetUint64(2)
+)
+
+// Sqrt sets z to the rounded square root of x, and returns it.
+//
+// If z's precision is 0, it is changed to x's precision before the
+// operation. Rounding is performed according to z's precision and
+// rounding mode.
+//
+// The function panics if z < 0. The value of z is undefined in that
+// case.
+func (z *Decimal) Sqrt(x *Decimal) *Decimal {
+	if debugDecimal {
+		x.validate()
+	}
+
+	if z.prec == 0 {
+		z.prec = x.prec
+	}
+
+	if x.Sign() == -1 {
+		// following IEEE754-2008 (section 7.2)
+		panic(ErrNaN{"square root of negative operand"})
+	}
+
+	// handle ±0 and +∞
+	if x.form != finite {
+		z.acc = Exact
+		z.form = x.form
+		z.neg = x.neg // IEEE754-2008 requires √±0 = ±0
+		return z
+	}
+
+	// MantExp sets the argument's precision to the receiver's, and
+	// when z.prec > x.prec this will lower z.prec. Restore it after
+	// the MantExp call.
+	prec := z.prec
+	b := x.MantExp(z)
+	z.prec = prec
+
+	// Compute √(z·10**b) as
+	//   √( z)·10**(½b)     if b is even
+	//   √(10z)·10**(⌊½b⌋)   if b > 0 is odd
+	//   √(z/10)·10**(⌈½b⌉)   if b < 0 is odd
+	switch b % 2 {
+	case 0:
+		// nothing to do
+	case 1:
+		z.exp++
+	case -1:
+		z.exp--
+	}
+	// 0.01 <= z < 10.0
+
+	// Unlike with big.Float, solving x² - z = 0 directly is faster only for
+	// very small precisions (<_DW/2).
+
+	// Compute √x (to z.prec precision) by solving
+	//   1/t² - x = 0
+	// for t (using Newton's method), and then inverting.
+
+	// let
+	//   f(t) = 1/t² - x
+	// then
+	//   g(t) = f(t)/f'(t) = -½t(1 - xt²)
+	// and the next guess is given by
+	//   t2 = t - g(t) = ½t(3 - xt²)
+
+	u := newDecimal(prec)
+	v := newDecimal(prec)
+	xf, _ := x.Float64()
+	sqi := newDecimal(prec)
+	sqi.SetPrec(17).SetFloat64(1 / math.Sqrt(xf))
+	for prec := prec + _DW; sqi.prec < prec; {
+		sqi.prec *= 2
+		u.prec = sqi.prec
+		v.prec = sqi.prec
+		u.Mul(sqi, sqi) //   u = sqi²
+		u.Mul(x, u)     //     = x.sqi²
+		v.Sub(three, u) //   v = 3 - x.sqi²
+		u.Mul(sqi, v)   //   u = sqi(3 - x.sqi²)
+		sqi.Quo(u, two) // sqi = ½sqi(3 - x.sqi²)
+	}
+	// sqi = 1/√x
+
+	// x/√x = √x
+	z.Mul(x, sqi)
+
+	// re-attach halved exponent
+	return z.SetMantExp(z, b/2)
+}
+
+// newDecimal returns a new *Decimal with space for twice the given
+// precision.
+func newDecimal(prec2 uint32) *Decimal {
+	z := new(Decimal)
+	// dec.make ensures the slice length is > 0
+	z.mant = z.mant.make(int(prec2/_DW) * 2)
+	return z
+}