diff --git a/matrix/mat64/gsvd.go b/matrix/mat64/gsvd.go
index 7eb98da3..e3064848 100644
--- a/matrix/mat64/gsvd.go
+++ b/matrix/mat64/gsvd.go
@@ -203,9 +203,9 @@ func (gsvd *GSVD) ValuesB(s []float64) []float64 {
 	return s
 }
 
-// ZeroRFromGSVD extracts the matrix [ 0 R ] from the singular value decomposition, storing
-// the result in-place into the receiver. [ 0 R ] is size (k+l)×c.
-func (m *Dense) ZeroRFromGSVD(gsvd *GSVD) {
+// ZeroRTo extracts the matrix [ 0 R ] from the singular value decomposition, storing
+// the result in-place into dst. [ 0 R ] is size (k+l)×c.
+func (gsvd *GSVD) ZeroRTo(dst *Dense) {
 	if gsvd.kind == 0 {
 		panic("gsvd: no decomposition computed")
 	}
@@ -214,13 +214,13 @@ func (m *Dense) ZeroRFromGSVD(gsvd *GSVD) {
 	k := gsvd.k
 	l := gsvd.l
 	h := min(k+l, r)
-	m.reuseAsZeroed(k+l, c)
+	dst.reuseAsZeroed(k+l, c)
 	a := Dense{
 		mat:     gsvd.a,
 		capRows: r,
 		capCols: c,
 	}
-	m.Slice(0, h, c-k-l, c).(*Dense).
+	dst.Slice(0, h, c-k-l, c).(*Dense).
 		Copy(a.Slice(0, h, c-k-l, c))
 	if r < k+l {
 		b := Dense{
@@ -228,32 +228,32 @@ func (m *Dense) ZeroRFromGSVD(gsvd *GSVD) {
 			capRows: gsvd.p,
 			capCols: c,
 		}
-		m.Slice(r, k+l, c+r-k-l, c).(*Dense).
+		dst.Slice(r, k+l, c+r-k-l, c).(*Dense).
 			Copy(b.Slice(r-k, l, c+r-k-l, c))
 	}
 }
 
-// SigmaAFromGSVD extracts the matrix Σ₁ from the singular value decomposition, storing
-// the result in-place into the receiver. Σ₁ is size r×(k+l).
-func (m *Dense) SigmaAFromGSVD(gsvd *GSVD) {
+// SigmaATo extracts the matrix Σ₁ from the singular value decomposition, storing
+// the result in-place into dst. Σ₁ is size r×(k+l).
+func (gsvd *GSVD) SigmaATo(dst *Dense) {
 	if gsvd.kind == 0 {
 		panic("gsvd: no decomposition computed")
 	}
 	r := gsvd.r
 	k := gsvd.k
 	l := gsvd.l
-	m.reuseAsZeroed(r, k+l)
+	dst.reuseAsZeroed(r, k+l)
 	for i := 0; i < k; i++ {
-		m.set(i, i, 1)
+		dst.set(i, i, 1)
 	}
 	for i := k; i < min(r, k+l); i++ {
-		m.set(i, i, gsvd.s1[i])
+		dst.set(i, i, gsvd.s1[i])
 	}
 }
 
-// SigmaBFromGSVD extracts the matrix Σ₂ from the singular value decomposition, storing
-// the result in-place into the receiver. Σ₂ is size p×(k+l).
-func (m *Dense) SigmaBFromGSVD(gsvd *GSVD) {
+// SigmaBTo extracts the matrix Σ₂ from the singular value decomposition, storing
+// the result in-place into dst. Σ₂ is size p×(k+l).
+func (gsvd *GSVD) SigmaBTo(dst *Dense) {
 	if gsvd.kind == 0 {
 		panic("gsvd: no decomposition computed")
 	}
@@ -261,65 +261,65 @@ func (m *Dense) SigmaBFromGSVD(gsvd *GSVD) {
 	p := gsvd.p
 	k := gsvd.k
 	l := gsvd.l
-	m.reuseAsZeroed(p, k+l)
+	dst.reuseAsZeroed(p, k+l)
 	for i := 0; i < min(l, r-k); i++ {
-		m.set(i, i+k, gsvd.s2[k+i])
+		dst.set(i, i+k, gsvd.s2[k+i])
 	}
 	for i := r - k; i < l; i++ {
-		m.set(i, i+k, 1)
+		dst.set(i, i+k, 1)
 	}
 }
 
-// UFromGSVD extracts the matrix U from the singular value decomposition, storing
-// the result in-place into the receiver. U is size r×r.
-func (m *Dense) UFromGSVD(gsvd *GSVD) {
+// UTo extracts the matrix U from the singular value decomposition, storing
+// the result in-place into dst. U is size r×r.
+func (gsvd *GSVD) UTo(dst *Dense) {
 	if gsvd.kind&matrix.GSVDU == 0 {
 		panic("mat64: improper GSVD kind")
 	}
 	r := gsvd.u.Rows
 	c := gsvd.u.Cols
-	m.reuseAs(r, c)
+	dst.reuseAs(r, c)
 
 	tmp := &Dense{
 		mat:     gsvd.u,
 		capRows: r,
 		capCols: c,
 	}
-	m.Copy(tmp)
+	dst.Copy(tmp)
 }
 
-// VFromGSVD extracts the matrix V from the singular value decomposition, storing
-// the result in-place into the receiver. V is size p×p.
-func (m *Dense) VFromGSVD(gsvd *GSVD) {
+// VTo extracts the matrix V from the singular value decomposition, storing
+// the result in-place into dst. V is size p×p.
+func (gsvd *GSVD) VTo(dst *Dense) {
 	if gsvd.kind&matrix.GSVDV == 0 {
 		panic("mat64: improper GSVD kind")
 	}
 	r := gsvd.v.Rows
 	c := gsvd.v.Cols
-	m.reuseAs(r, c)
+	dst.reuseAs(r, c)
 
 	tmp := &Dense{
 		mat:     gsvd.v,
 		capRows: r,
 		capCols: c,
 	}
-	m.Copy(tmp)
+	dst.Copy(tmp)
 }
 
-// QFromGSVD extracts the matrix Q from the singular value decomposition, storing
-// the result in-place into the receiver. Q is size c×c.
-func (m *Dense) QFromGSVD(gsvd *GSVD) {
+// QTo extracts the matrix Q from the singular value decomposition, storing
+// the result in-place into dst. Q is size c×c.
+func (gsvd *GSVD) QTo(dst *Dense) {
 	if gsvd.kind&matrix.GSVDQ == 0 {
 		panic("mat64: improper GSVD kind")
 	}
 	r := gsvd.q.Rows
 	c := gsvd.q.Cols
-	m.reuseAs(r, c)
+	dst.reuseAs(r, c)
 
 	tmp := &Dense{
 		mat:     gsvd.q,
 		capRows: r,
 		capCols: c,
 	}
-	m.Copy(tmp)
+	dst.Copy(tmp)
 }
diff --git a/matrix/mat64/gsvd_example_test.go b/matrix/mat64/gsvd_example_test.go
index d85a820e..5f605642 100644
--- a/matrix/mat64/gsvd_example_test.go
+++ b/matrix/mat64/gsvd_example_test.go
@@ -26,8 +26,8 @@ func ExampleGSVD() {
 	}
 
 	var u, v mat64.Dense
-	u.UFromGSVD(&gsvd)
-	v.VFromGSVD(&gsvd)
+	gsvd.UTo(&u)
+	gsvd.VTo(&v)
 
 	s1 := gsvd.ValuesA(nil)
 	s2 := gsvd.ValuesB(nil)
@@ -38,8 +38,8 @@ func ExampleGSVD() {
 		s2, mat64.Formatted(&v, mat64.Prefix("\t    "), mat64.Excerpt(2)))
 
 	var zeroR, q mat64.Dense
-	zeroR.ZeroRFromGSVD(&gsvd)
-	q.QFromGSVD(&gsvd)
+	gsvd.ZeroRTo(&zeroR)
+	gsvd.QTo(&q)
 	q.Mul(&zeroR, &q)
 	fmt.Printf("\nCommon basis vectors\n\n\tQ^T = %.4f\n",
 		mat64.Formatted(q.T(), mat64.Prefix("\t      ")))
diff --git a/matrix/mat64/gsvd_test.go b/matrix/mat64/gsvd_test.go
index bb5532c1..0e4d8274 100644
--- a/matrix/mat64/gsvd_test.go
+++ b/matrix/mat64/gsvd_test.go
@@ -107,12 +107,12 @@ func TestGSVD(t *testing.T) {
 
 func extractGSVD(gsvd *GSVD) (c, s []float64, s1, s2, zR, u, v, q *Dense) {
 	var s1m, s2m, zeroR, um, vm, qm Dense
-	s1m.SigmaAFromGSVD(gsvd)
-	s2m.SigmaBFromGSVD(gsvd)
-	zeroR.ZeroRFromGSVD(gsvd)
-	um.UFromGSVD(gsvd)
-	vm.VFromGSVD(gsvd)
-	qm.QFromGSVD(gsvd)
+	gsvd.SigmaATo(&s1m)
+	gsvd.SigmaBTo(&s2m)
+	gsvd.ZeroRTo(&zeroR)
+	gsvd.UTo(&um)
+	gsvd.VTo(&vm)
+	gsvd.QTo(&qm)
 	c = gsvd.ValuesA(nil)
 	s = gsvd.ValuesB(nil)
 	return c, s, &s1m, &s2m, &zeroR, &um, &vm, &qm
diff --git a/matrix/mat64/hogsvd.go b/matrix/mat64/hogsvd.go
index 641d8bdd..d9ea0b5d 100644
--- a/matrix/mat64/hogsvd.go
+++ b/matrix/mat64/hogsvd.go
@@ -141,11 +141,11 @@ func (gsvd *HOGSVD) Len() int {
 	return gsvd.n
 }
 
-// UFromHOGSVD extracts the matrix U_n from the singular value decomposition, storing
-// the result in-place into the receiver. U_n is size r×c.
+// UTo extracts the matrix U_n from the singular value decomposition, storing
+// the result in-place into dst. U_n is size r×c.
 //
-// UFromHOGSVD will panic if the receiver does not contain a successful factorization.
-func (m *Dense) UFromHOGSVD(gsvd *HOGSVD, n int) {
+// UTo will panic if the receiver does not contain a successful factorization.
+func (gsvd *HOGSVD) UTo(dst *Dense, n int) {
 	if gsvd.n == 0 {
 		panic("hogsvd: unsuccessful factorization")
 	}
@@ -153,10 +153,10 @@ func (m *Dense) UFromHOGSVD(gsvd *HOGSVD, n int) {
 		panic("hogsvd: invalid index")
 	}
 
-	m.reuseAs(gsvd.b[n].Dims())
-	m.Copy(&gsvd.b[n])
+	dst.reuseAs(gsvd.b[n].Dims())
+	dst.Copy(&gsvd.b[n])
 	for j, f := range gsvd.Values(nil, n) {
-		v := m.ColView(j)
+		v := dst.ColView(j)
 		v.ScaleVec(1/f, v)
 	}
 }
@@ -188,13 +188,14 @@ func (gsvd *HOGSVD) Values(s []float64, n int) []float64 {
 	return s
 }
 
-// VFromHOGSVD extracts the matrix V from the singular value decomposition, storing
-// the result in-place into the receiver. V is size c×c.
+// VTo extracts the matrix V from the singular value decomposition, storing
+// the result in-place into dst. V is size c×c.
 //
-// VFromHOGSVD will panic if the receiver does not contain a successful factorization.
-func (m *Dense) VFromHOGSVD(gsvd *HOGSVD) {
+// VTo will panic if the receiver does not contain a successful factorization.
+func (gsvd *HOGSVD) VTo(dst *Dense) {
 	if gsvd.n == 0 {
 		panic("hogsvd: unsuccessful factorization")
 	}
-	*m = *DenseCopyOf(gsvd.v)
+	dst.reuseAs(gsvd.v.Dims())
+	dst.Copy(gsvd.v)
 }
diff --git a/matrix/mat64/hogsvd_example_test.go b/matrix/mat64/hogsvd_example_test.go
index e785a2ea..8fb2f00a 100644
--- a/matrix/mat64/hogsvd_example_test.go
+++ b/matrix/mat64/hogsvd_example_test.go
@@ -25,14 +25,14 @@ func ExampleHOGSVD() {
 
 	for i, n := range []string{"Africa", "Asia", "Latin America/Caribbean", "Oceania"} {
 		var u mat64.Dense
-		u.UFromHOGSVD(&gsvd, i)
+		gsvd.UTo(&u, i)
 		s := gsvd.Values(nil, i)
 		fmt.Printf("%s\n\ts_%d = %.4f\n\n\tU_%[2]d = %.4[4]f\n",
 			n, i, s, mat64.Formatted(&u, mat64.Prefix("\t      ")))
 	}
 
 	var v mat64.Dense
-	v.VFromHOGSVD(&gsvd)
+	gsvd.VTo(&v)
 	fmt.Printf("\nCommon basis vectors\n\n\tV^T = %.4f",
 		mat64.Formatted(v.T(), mat64.Prefix("\t      ")))
 
diff --git a/matrix/mat64/hogsvd_test.go b/matrix/mat64/hogsvd_test.go
index 5af166e1..320fa2bc 100644
--- a/matrix/mat64/hogsvd_test.go
+++ b/matrix/mat64/hogsvd_test.go
@@ -80,10 +80,10 @@ func extractHOGSVD(gsvd *HOGSVD) (u []*Dense, s [][]float64, v *Dense) {
 	s = make([][]float64, gsvd.Len())
 	for i := 0; i < gsvd.Len(); i++ {
 		u[i] = &Dense{}
-		u[i].UFromHOGSVD(gsvd, i)
+		gsvd.UTo(u[i], i)
 		s[i] = gsvd.Values(nil, i)
 	}
 	v = &Dense{}
-	v.VFromHOGSVD(gsvd)
+	gsvd.VTo(v)
 	return u, s, v
 }
diff --git a/matrix/mat64/lq.go b/matrix/mat64/lq.go
index 5fe9766c..f8a04732 100644
--- a/matrix/mat64/lq.go
+++ b/matrix/mat64/lq.go
@@ -36,7 +36,7 @@ func (lq *LQ) updateCond() {
 //
 // The LQ decomposition is a factorization of the matrix A such that A = L * Q.
 // The matrix Q is an orthonormal n×n matrix, and L is an m×n upper triangular matrix.
-// L and Q can be extracted from the LFromLQ and QFromLQ methods on Dense.
+// L and Q can be extracted from the LTo and QTo methods.
 func (lq *LQ) Factorize(a Matrix) {
 	m, n := a.Dims()
 	if m > n {
@@ -58,10 +58,10 @@ func (lq *LQ) Factorize(a Matrix) {
 // TODO(btracey): Add in the "Reduced" forms for extracting the m×m orthogonal
 // and upper triangular matrices.
 
-// LFromLQ extracts the m×n lower trapezoidal matrix from a LQ decomposition.
-func (m *Dense) LFromLQ(lq *LQ) {
+// LTo extracts the m×n lower trapezoidal matrix from a LQ decomposition.
+func (lq *LQ) LTo(dst *Dense) {
 	r, c := lq.lq.Dims()
-	m.reuseAs(r, c)
+	dst.reuseAs(r, c)
 
 	// Disguise the LQ as a lower triangular
 	t := &TriDense{
@@ -74,25 +74,25 @@ func (m *Dense) LFromLQ(lq *LQ) {
 		},
 		cap: lq.lq.capCols,
 	}
-	m.Copy(t)
+	dst.Copy(t)
 
 	if r == c {
 		return
 	}
 	// Zero right of the triangular.
 	for i := 0; i < r; i++ {
-		zero(m.mat.Data[i*m.mat.Stride+r : i*m.mat.Stride+c])
+		zero(dst.mat.Data[i*dst.mat.Stride+r : i*dst.mat.Stride+c])
 	}
 }
 
-// QFromLQ extracts the n×n orthonormal matrix Q from an LQ decomposition.
-func (m *Dense) QFromLQ(lq *LQ) {
+// QTo extracts the n×n orthonormal matrix Q from an LQ decomposition.
+func (lq *LQ) QTo(dst *Dense) {
 	r, c := lq.lq.Dims()
-	m.reuseAs(c, c)
+	dst.reuseAs(c, c)
 
 	// Set Q = I.
 	for i := 0; i < c; i++ {
-		v := m.mat.Data[i*m.mat.Stride : i*m.mat.Stride+c]
+		v := dst.mat.Data[i*dst.mat.Stride : i*dst.mat.Stride+c]
 		zero(v)
 		v[i] = 1
 	}
@@ -127,10 +127,10 @@ func (m *Dense) QFromLQ(lq *LQ) {
 
 		// Compute the multiplication matrix.
 		blas64.Ger(-lq.tau[i], v, v, h)
-		qCopy.Copy(m)
+		qCopy.Copy(dst)
 		blas64.Gemm(blas.NoTrans, blas.NoTrans,
 			1, h, qCopy.mat,
-			0, m.mat)
+			0, dst.mat)
 	}
 }
 
diff --git a/matrix/mat64/lq_test.go b/matrix/mat64/lq_test.go
index 05478ced..657b8b48 100644
--- a/matrix/mat64/lq_test.go
+++ b/matrix/mat64/lq_test.go
@@ -27,16 +27,16 @@ func TestLQ(t *testing.T) {
 		var want Dense
 		want.Clone(a)
 
-		lq := &LQ{}
+		var lq LQ
 		lq.Factorize(a)
 		var l, q Dense
-		q.QFromLQ(lq)
+		lq.QTo(&q)
 
 		if !isOrthonormal(&q, 1e-10) {
 			t.Errorf("Q is not orthonormal: m = %v, n = %v", m, n)
 		}
 
-		l.LFromLQ(lq)
+		lq.LTo(&l)
 
 		var got Dense
 		got.Mul(&l, &q)
diff --git a/matrix/mat64/lu.go b/matrix/mat64/lu.go
index 69f2daf0..a059a5a7 100644
--- a/matrix/mat64/lu.go
+++ b/matrix/mat64/lu.go
@@ -51,7 +51,7 @@ func (lu *LU) updateCond(norm float64) {
 // The LU factorization is computed with pivoting, and so really the decomposition
 // is a PLU decomposition where P is a permutation matrix. The individual matrix
 // factors can be extracted from the factorization using the Permutation method
-// on Dense, and the LFrom and UFrom methods on TriDense.
+// on Dense, and the LU LTo and UTo methods.
 func (lu *LU) Factorize(a Matrix) {
 	r, c := a.Dims()
 	if r != c {
@@ -204,30 +204,30 @@ func (lu *LU) RankOne(orig *LU, alpha float64, x, y *Vector) {
 	lu.updateCond(-1)
 }
 
-// LFromLU extracts the lower triangular matrix from an LU factorization.
-func (t *TriDense) LFromLU(lu *LU) {
+// LTo extracts the lower triangular matrix from an LU factorization.
+func (lu *LU) LTo(dst *TriDense) {
 	_, n := lu.lu.Dims()
-	t.reuseAs(n, false)
+	dst.reuseAs(n, false)
 	// Extract the lower triangular elements.
 	for i := 0; i < n; i++ {
 		for j := 0; j < i; j++ {
-			t.mat.Data[i*t.mat.Stride+j] = lu.lu.mat.Data[i*lu.lu.mat.Stride+j]
+			dst.mat.Data[i*dst.mat.Stride+j] = lu.lu.mat.Data[i*lu.lu.mat.Stride+j]
 		}
 	}
 	// Set ones on the diagonal.
 	for i := 0; i < n; i++ {
-		t.mat.Data[i*t.mat.Stride+i] = 1
+		dst.mat.Data[i*dst.mat.Stride+i] = 1
 	}
 }
 
-// UFromLU extracts the upper triangular matrix from an LU factorization.
-func (t *TriDense) UFromLU(lu *LU) {
+// UTo extracts the upper triangular matrix from an LU factorization.
+func (lu *LU) UTo(dst *TriDense) {
 	_, n := lu.lu.Dims()
-	t.reuseAs(n, true)
+	dst.reuseAs(n, true)
 	// Extract the upper triangular elements.
 	for i := 0; i < n; i++ {
 		for j := i; j < n; j++ {
-			t.mat.Data[i*t.mat.Stride+j] = lu.lu.mat.Data[i*lu.lu.mat.Stride+j]
+			dst.mat.Data[i*dst.mat.Stride+j] = lu.lu.mat.Data[i*lu.lu.mat.Stride+j]
 		}
 	}
 }
diff --git a/matrix/mat64/lu_test.go b/matrix/mat64/lu_test.go
index ed7dd6df..6468715e 100644
--- a/matrix/mat64/lu_test.go
+++ b/matrix/mat64/lu_test.go
@@ -20,12 +20,12 @@ func TestLUD(t *testing.T) {
 		var want Dense
 		want.Clone(a)
 
-		lu := &LU{}
+		var lu LU
 		lu.Factorize(a)
 
 		var l, u TriDense
-		l.LFromLU(lu)
-		u.UFromLU(lu)
+		lu.LTo(&l)
+		lu.UTo(&u)
 		var p Dense
 		pivot := lu.Pivot(nil)
 		p.Permutation(n, pivot)
@@ -93,8 +93,8 @@ func TestLURankOne(t *testing.T) {
 // luReconstruct reconstructs the original A matrix from an LU decomposition.
 func luReconstruct(lu *LU) *Dense {
 	var L, U TriDense
-	L.LFromLU(lu)
-	U.UFromLU(lu)
+	lu.LTo(&L)
+	lu.UTo(&U)
 	var P Dense
 	pivot := lu.Pivot(nil)
 	P.Permutation(len(pivot), pivot)
diff --git a/matrix/mat64/qr.go b/matrix/mat64/qr.go
index 15b3a8ea..770bd3b3 100644
--- a/matrix/mat64/qr.go
+++ b/matrix/mat64/qr.go
@@ -37,7 +37,7 @@ func (qr *QR) updateCond() {
 //
 // The QR decomposition is a factorization of the matrix A such that A = Q * R.
 // The matrix Q is an orthonormal m×m matrix, and R is an m×n upper triangular matrix.
-// Q and R can be extracted from the QFromQR and RFromQR methods on Dense.
+// Q and R can be extracted using the QTo and RTo methods.
 func (qr *QR) Factorize(a Matrix) {
 	m, n := a.Dims()
 	if m < n {
@@ -60,10 +60,10 @@ func (qr *QR) Factorize(a Matrix) {
 // TODO(btracey): Add in the "Reduced" forms for extracting the n×n orthogonal
 // and upper triangular matrices.
 
-// RFromQR extracts the m×n upper trapezoidal matrix from a QR decomposition.
-func (m *Dense) RFromQR(qr *QR) {
+// RTo extracts the m×n upper trapezoidal matrix from a QR decomposition.
+func (qr *QR) RTo(dst *Dense) {
 	r, c := qr.qr.Dims()
-	m.reuseAs(r, c)
+	dst.reuseAs(r, c)
 
 	// Disguise the QR as an upper triangular
 	t := &TriDense{
@@ -76,29 +76,29 @@ func (m *Dense) RFromQR(qr *QR) {
 		},
 		cap: qr.qr.capCols,
 	}
-	m.Copy(t)
+	dst.Copy(t)
 
 	// Zero below the triangular.
 	for i := r; i < c; i++ {
-		zero(m.mat.Data[i*m.mat.Stride : i*m.mat.Stride+c])
+		zero(dst.mat.Data[i*dst.mat.Stride : i*dst.mat.Stride+c])
 	}
 }
 
-// QFromQR extracts the m×m orthonormal matrix Q from a QR decomposition.
-func (m *Dense) QFromQR(qr *QR) {
+// QTo extracts the m×m orthonormal matrix Q from a QR decomposition.
+func (qr *QR) QTo(dst *Dense) {
 	r, _ := qr.qr.Dims()
-	m.reuseAsZeroed(r, r)
+	dst.reuseAsZeroed(r, r)
 
 	// Set Q = I.
 	for i := 0; i < r*r; i += r + 1 {
-		m.mat.Data[i] = 1
+		dst.mat.Data[i] = 1
 	}
 
 	// Construct Q from the elementary reflectors.
 	work := make([]float64, 1)
-	lapack64.Ormqr(blas.Left, blas.NoTrans, qr.qr.mat, qr.tau, m.mat, work, -1)
+	lapack64.Ormqr(blas.Left, blas.NoTrans, qr.qr.mat, qr.tau, dst.mat, work, -1)
 	work = make([]float64, int(work[0]))
-	lapack64.Ormqr(blas.Left, blas.NoTrans, qr.qr.mat, qr.tau, m.mat, work, len(work))
+	lapack64.Ormqr(blas.Left, blas.NoTrans, qr.qr.mat, qr.tau, dst.mat, work, len(work))
 }
 
 // SolveQR finds a minimum-norm solution to a system of linear equations defined
diff --git a/matrix/mat64/qr_test.go b/matrix/mat64/qr_test.go
index 82e9cb25..61407137 100644
--- a/matrix/mat64/qr_test.go
+++ b/matrix/mat64/qr_test.go
@@ -30,16 +30,16 @@ func TestQR(t *testing.T) {
 		var want Dense
 		want.Clone(a)
 
-		qr := &QR{}
+		var qr QR
 		qr.Factorize(a)
 		var q, r Dense
-		q.QFromQR(qr)
+		qr.QTo(&q)
 
 		if !isOrthonormal(&q, 1e-10) {
 			t.Errorf("Q is not orthonormal: m = %v, n = %v", m, n)
 		}
 
-		r.RFromQR(qr)
+		qr.RTo(&r)
 
 		var got Dense
 		got.Mul(&q, &r)
diff --git a/matrix/mat64/svd.go b/matrix/mat64/svd.go
index 87498f73..bd8f4992 100644
--- a/matrix/mat64/svd.go
+++ b/matrix/mat64/svd.go
@@ -138,42 +138,42 @@ func (svd *SVD) Values(s []float64) []float64 {
 	return s
 }
 
-// UFromSVD extracts the matrix U from the singular value decomposition, storing
-// the result in-place into the receiver. U is size m×m if svd.Kind() == SVDFull,
-// of size m×min(m,n) if svd.Kind() == SVDThin, and UFromSVD panics otherwise.
-func (m *Dense) UFromSVD(svd *SVD) {
+// UTo extracts the matrix U from the singular value decomposition, storing
+// the result in-place into dst. U is size m×m if svd.Kind() == SVDFull,
+// of size m×min(m,n) if svd.Kind() == SVDThin, and UTo panics otherwise.
+func (svd *SVD) UTo(dst *Dense) {
 	kind := svd.kind
 	if kind != matrix.SVDFull && kind != matrix.SVDThin {
 		panic("mat64: improper SVD kind")
 	}
 	r := svd.u.Rows
 	c := svd.u.Cols
-	m.reuseAs(r, c)
+	dst.reuseAs(r, c)
 
 	tmp := &Dense{
 		mat:     svd.u,
 		capRows: r,
 		capCols: c,
 	}
-	m.Copy(tmp)
+	dst.Copy(tmp)
 }
 
-// VFromSVD extracts the matrix V from the singular value decomposition, storing
-// the result in-place into the receiver. V is size n×n if svd.Kind() == SVDFull,
-// of size n×min(m,n) if svd.Kind() == SVDThin, and VFromSVD panics otherwise.
-func (m *Dense) VFromSVD(svd *SVD) {
+// VTo extracts the matrix V from the singular value decomposition, storing
+// the result in-place into dst. V is size n×n if svd.Kind() == SVDFull,
+// of size n×min(m,n) if svd.Kind() == SVDThin, and VTo panics otherwise.
+func (svd *SVD) VTo(dst *Dense) {
 	kind := svd.kind
 	if kind != matrix.SVDFull && kind != matrix.SVDThin {
 		panic("mat64: improper SVD kind")
 	}
 	r := svd.vt.Rows
 	c := svd.vt.Cols
-	m.reuseAs(c, r)
+	dst.reuseAs(c, r)
 
 	tmp := &Dense{
 		mat:     svd.vt,
 		capRows: r,
 		capCols: c,
 	}
-	m.Copy(tmp.T())
+	dst.Copy(tmp.T())
 }
diff --git a/matrix/mat64/svd_test.go b/matrix/mat64/svd_test.go
index fb342486..7f1e1b96 100644
--- a/matrix/mat64/svd_test.go
+++ b/matrix/mat64/svd_test.go
@@ -170,8 +170,8 @@ func TestSVD(t *testing.T) {
 
 func extractSVD(svd *SVD) (s []float64, u, v *Dense) {
 	var um, vm Dense
-	um.UFromSVD(svd)
-	vm.VFromSVD(svd)
+	svd.UTo(&um)
+	svd.VTo(&vm)
 	s = svd.Values(nil)
 	return s, &um, &vm
 }