2017-06-06 Thomas Koenig <tkoenig@gcc.gnu.org>

	PR fortran/80975
	* m4/matmul_internal.m4:  Move zeroing before early return.
	* generated/matmul_c10.c: Regenerated.
	* generated/matmul_c16.c: Regenerated.
	* generated/matmul_c4.c: Regenerated.
	* generated/matmul_c8.c: Regenerated.
	* generated/matmul_i1.c: Regenerated.
	* generated/matmul_i16.c: Regenerated.
	* generated/matmul_i2.c: Regenerated.
	* generated/matmul_i4.c: Regenerated.
	* generated/matmul_i8.c: Regenerated.
	* generated/matmul_r10.c: Regenerated.
	* generated/matmul_r16.c: Regenerated.
	* generated/matmul_r4.c: Regenerated.
	* generated/matmul_r8.c: Regenerated.
	* generated/matmulavx128_c10.c: Regenerated.
	* generated/matmulavx128_c16.c: Regenerated.
	* generated/matmulavx128_c4.c: Regenerated.
	* generated/matmulavx128_c8.c: Regenerated.
	* generated/matmulavx128_i1.c: Regenerated.
	* generated/matmulavx128_i16.c: Regenerated.
	* generated/matmulavx128_i2.c: Regenerated.
	* generated/matmulavx128_i4.c: Regenerated.
	* generated/matmulavx128_i8.c: Regenerated.
	* generated/matmulavx128_r10.c: Regenerated.
	* generated/matmulavx128_r16.c: Regenerated.
	* generated/matmulavx128_r4.c: Regenerated.
	* generated/matmulavx128_r8.c: Regenerated.

2017-06-06  Thomas Koenig  <tkoenig@gcc.gnu.org>

	PR fortran/80975
	* gfortran.dg/matmul_16.f90: New test.
	* gfortran.dg/inline_matmul_18.f90: New test.



git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@248932 138bc75d-0d04-0410-961f-82ee72b054a4

28 files changed, 491 insertions, 460 deletions

diff --git a/libgfortran/ChangeLog b/libgfortran/ChangeLog
index e763232744d..1f5a0264172 100644
--- a/libgfortran/ChangeLog
+++ b/libgfortran/ChangeLog
@@ -1,3 +1,34 @@
+2017-06-06  Thomas Koenig  <tkoenig@gcc.gnu.org>
+
+	PR fortran/80975
+	* m4/matmul_internal.m4:  Move zeroing before early return.
+	* generated/matmul_c10.c: Regenerated.
+        * generated/matmul_c16.c: Regenerated.
+        * generated/matmul_c4.c: Regenerated.
+        * generated/matmul_c8.c: Regenerated.
+        * generated/matmul_i1.c: Regenerated.
+        * generated/matmul_i16.c: Regenerated.
+        * generated/matmul_i2.c: Regenerated.
+        * generated/matmul_i4.c: Regenerated.
+        * generated/matmul_i8.c: Regenerated.
+        * generated/matmul_r10.c: Regenerated.
+        * generated/matmul_r16.c: Regenerated.
+        * generated/matmul_r4.c: Regenerated.
+        * generated/matmul_r8.c: Regenerated.
+        * generated/matmulavx128_c10.c: Regenerated.
+        * generated/matmulavx128_c16.c: Regenerated.
+        * generated/matmulavx128_c4.c: Regenerated.
+        * generated/matmulavx128_c8.c: Regenerated.
+        * generated/matmulavx128_i1.c: Regenerated.
+        * generated/matmulavx128_i16.c: Regenerated.
+        * generated/matmulavx128_i2.c: Regenerated.
+        * generated/matmulavx128_i4.c: Regenerated.
+        * generated/matmulavx128_i8.c: Regenerated.
+        * generated/matmulavx128_r10.c: Regenerated.
+        * generated/matmulavx128_r16.c: Regenerated.
+        * generated/matmulavx128_r4.c: Regenerated.
+        * generated/matmulavx128_r8.c: Regenerated.
+
 2017-05-29  Jerry DeLisle  <jvdelisle@gcc.gnu.org>
 
 	PR libgfortran/53029
diff --git a/libgfortran/generated/matmul_c10.c b/libgfortran/generated/matmul_c10.c
index 54e2714668d..4e156ba2ee2 100644
--- a/libgfortran/generated/matmul_c10.c
+++ b/libgfortran/generated/matmul_c10.c
@@ -307,6 +307,11 @@ matmul_c10_avx (gfc_array_c10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_c10_avx (gfc_array_c10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_c10_avx2 (gfc_array_c10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_c10_avx2 (gfc_array_c10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_c10_avx512f (gfc_array_c10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_c10_avx512f (gfc_array_c10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_c10_vanilla (gfc_array_c10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_c10_vanilla (gfc_array_c10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_c10 (gfc_array_c10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_c10 (gfc_array_c10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_c16.c b/libgfortran/generated/matmul_c16.c
index cd8aacd9c68..162e564f260 100644
--- a/libgfortran/generated/matmul_c16.c
+++ b/libgfortran/generated/matmul_c16.c
@@ -307,6 +307,11 @@ matmul_c16_avx (gfc_array_c16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_c16_avx (gfc_array_c16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_c16_avx2 (gfc_array_c16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_c16_avx2 (gfc_array_c16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_c16_avx512f (gfc_array_c16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_c16_avx512f (gfc_array_c16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_c16_vanilla (gfc_array_c16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_c16_vanilla (gfc_array_c16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_c16 (gfc_array_c16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_c16 (gfc_array_c16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_c4.c b/libgfortran/generated/matmul_c4.c
index ead22efabb2..43a0288f2cb 100644
--- a/libgfortran/generated/matmul_c4.c
+++ b/libgfortran/generated/matmul_c4.c
@@ -307,6 +307,11 @@ matmul_c4_avx (gfc_array_c4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_c4_avx (gfc_array_c4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_c4_avx2 (gfc_array_c4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_c4_avx2 (gfc_array_c4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_c4_avx512f (gfc_array_c4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_c4_avx512f (gfc_array_c4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_c4_vanilla (gfc_array_c4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_c4_vanilla (gfc_array_c4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_c4 (gfc_array_c4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_c4 (gfc_array_c4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_c8.c b/libgfortran/generated/matmul_c8.c
index a52e4bd7f0a..6efdca10baf 100644
--- a/libgfortran/generated/matmul_c8.c
+++ b/libgfortran/generated/matmul_c8.c
@@ -307,6 +307,11 @@ matmul_c8_avx (gfc_array_c8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_c8_avx (gfc_array_c8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_c8_avx2 (gfc_array_c8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_c8_avx2 (gfc_array_c8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_c8_avx512f (gfc_array_c8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_c8_avx512f (gfc_array_c8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_c8_vanilla (gfc_array_c8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_c8_vanilla (gfc_array_c8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_c8 (gfc_array_c8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_c8 (gfc_array_c8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_i1.c b/libgfortran/generated/matmul_i1.c
index dfd47e176e4..a658c456663 100644
--- a/libgfortran/generated/matmul_i1.c
+++ b/libgfortran/generated/matmul_i1.c
@@ -307,6 +307,11 @@ matmul_i1_avx (gfc_array_i1 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_i1_avx (gfc_array_i1 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_1));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_i1_avx2 (gfc_array_i1 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_i1_avx2 (gfc_array_i1 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_1));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_i1_avx512f (gfc_array_i1 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_i1_avx512f (gfc_array_i1 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_1));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_i1_vanilla (gfc_array_i1 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_i1_vanilla (gfc_array_i1 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_1));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_i1 (gfc_array_i1 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_i1 (gfc_array_i1 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_1));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_i16.c b/libgfortran/generated/matmul_i16.c
index a7bdcb5c67d..447112be70f 100644
--- a/libgfortran/generated/matmul_i16.c
+++ b/libgfortran/generated/matmul_i16.c
@@ -307,6 +307,11 @@ matmul_i16_avx (gfc_array_i16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_i16_avx (gfc_array_i16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_i16_avx2 (gfc_array_i16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_i16_avx2 (gfc_array_i16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_i16_avx512f (gfc_array_i16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_i16_avx512f (gfc_array_i16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_i16_vanilla (gfc_array_i16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_i16_vanilla (gfc_array_i16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_i16 (gfc_array_i16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_i16 (gfc_array_i16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_i2.c b/libgfortran/generated/matmul_i2.c
index d541fa3fe73..4e2165131f5 100644
--- a/libgfortran/generated/matmul_i2.c
+++ b/libgfortran/generated/matmul_i2.c
@@ -307,6 +307,11 @@ matmul_i2_avx (gfc_array_i2 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_i2_avx (gfc_array_i2 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_2));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_i2_avx2 (gfc_array_i2 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_i2_avx2 (gfc_array_i2 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_2));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_i2_avx512f (gfc_array_i2 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_i2_avx512f (gfc_array_i2 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_2));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_i2_vanilla (gfc_array_i2 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_i2_vanilla (gfc_array_i2 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_2));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_i2 (gfc_array_i2 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_i2 (gfc_array_i2 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_2));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_i4.c b/libgfortran/generated/matmul_i4.c
index f8f0cdb96cc..5331e945926 100644
--- a/libgfortran/generated/matmul_i4.c
+++ b/libgfortran/generated/matmul_i4.c
@@ -307,6 +307,11 @@ matmul_i4_avx (gfc_array_i4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_i4_avx (gfc_array_i4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_i4_avx2 (gfc_array_i4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_i4_avx2 (gfc_array_i4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_i4_avx512f (gfc_array_i4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_i4_avx512f (gfc_array_i4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_i4_vanilla (gfc_array_i4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_i4_vanilla (gfc_array_i4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_i4 (gfc_array_i4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_i4 (gfc_array_i4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_i8.c b/libgfortran/generated/matmul_i8.c
index 2aac1d76822..d150e699f2f 100644
--- a/libgfortran/generated/matmul_i8.c
+++ b/libgfortran/generated/matmul_i8.c
@@ -307,6 +307,11 @@ matmul_i8_avx (gfc_array_i8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_i8_avx (gfc_array_i8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_i8_avx2 (gfc_array_i8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_i8_avx2 (gfc_array_i8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_i8_avx512f (gfc_array_i8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_i8_avx512f (gfc_array_i8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_i8_vanilla (gfc_array_i8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_i8_vanilla (gfc_array_i8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_i8 (gfc_array_i8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_i8 (gfc_array_i8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_r10.c b/libgfortran/generated/matmul_r10.c
index 448c96bb99f..5407615fc37 100644
--- a/libgfortran/generated/matmul_r10.c
+++ b/libgfortran/generated/matmul_r10.c
@@ -307,6 +307,11 @@ matmul_r10_avx (gfc_array_r10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_r10_avx (gfc_array_r10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_r10_avx2 (gfc_array_r10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_r10_avx2 (gfc_array_r10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_r10_avx512f (gfc_array_r10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_r10_avx512f (gfc_array_r10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_r10_vanilla (gfc_array_r10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_r10_vanilla (gfc_array_r10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_r10 (gfc_array_r10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_r10 (gfc_array_r10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_r16.c b/libgfortran/generated/matmul_r16.c
index 57a47dcd846..1254ec11fa5 100644
--- a/libgfortran/generated/matmul_r16.c
+++ b/libgfortran/generated/matmul_r16.c
@@ -307,6 +307,11 @@ matmul_r16_avx (gfc_array_r16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_r16_avx (gfc_array_r16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_r16_avx2 (gfc_array_r16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_r16_avx2 (gfc_array_r16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_r16_avx512f (gfc_array_r16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_r16_avx512f (gfc_array_r16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_r16_vanilla (gfc_array_r16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_r16_vanilla (gfc_array_r16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_r16 (gfc_array_r16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_r16 (gfc_array_r16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_r4.c b/libgfortran/generated/matmul_r4.c
index 52eea532a9b..5d13e6713d6 100644
--- a/libgfortran/generated/matmul_r4.c
+++ b/libgfortran/generated/matmul_r4.c
@@ -307,6 +307,11 @@ matmul_r4_avx (gfc_array_r4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_r4_avx (gfc_array_r4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_r4_avx2 (gfc_array_r4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_r4_avx2 (gfc_array_r4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_r4_avx512f (gfc_array_r4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_r4_avx512f (gfc_array_r4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_r4_vanilla (gfc_array_r4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_r4_vanilla (gfc_array_r4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_r4 (gfc_array_r4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_r4 (gfc_array_r4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmul_r8.c b/libgfortran/generated/matmul_r8.c
index 074697dbf8b..e835ed17169 100644
--- a/libgfortran/generated/matmul_r8.c
+++ b/libgfortran/generated/matmul_r8.c
@@ -307,6 +307,11 @@ matmul_r8_avx (gfc_array_r8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -319,11 +324,6 @@ matmul_r8_avx (gfc_array_r8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -859,6 +859,11 @@ matmul_r8_avx2 (gfc_array_r8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -871,11 +876,6 @@ matmul_r8_avx2 (gfc_array_r8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1411,6 +1411,11 @@ matmul_r8_avx512f (gfc_array_r8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1423,11 +1428,6 @@ matmul_r8_avx512f (gfc_array_r8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -1977,6 +1977,11 @@ matmul_r8_vanilla (gfc_array_r8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -1989,11 +1994,6 @@ matmul_r8_vanilla (gfc_array_r8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -2603,6 +2603,11 @@ matmul_r8 (gfc_array_r8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -2615,11 +2620,6 @@ matmul_r8 (gfc_array_r8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_c10.c b/libgfortran/generated/matmulavx128_c10.c
index 53cdecbd4e8..ac8a013b864 100644
--- a/libgfortran/generated/matmulavx128_c10.c
+++ b/libgfortran/generated/matmulavx128_c10.c
@@ -272,6 +272,11 @@ matmul_c10_avx128_fma3 (gfc_array_c10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_c10_avx128_fma3 (gfc_array_c10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_c10_avx128_fma4 (gfc_array_c10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_c10_avx128_fma4 (gfc_array_c10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_c16.c b/libgfortran/generated/matmulavx128_c16.c
index e7657a098a5..6af99bec434 100644
--- a/libgfortran/generated/matmulavx128_c16.c
+++ b/libgfortran/generated/matmulavx128_c16.c
@@ -272,6 +272,11 @@ matmul_c16_avx128_fma3 (gfc_array_c16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_c16_avx128_fma3 (gfc_array_c16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_c16_avx128_fma4 (gfc_array_c16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_c16_avx128_fma4 (gfc_array_c16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_c4.c b/libgfortran/generated/matmulavx128_c4.c
index 950f1eb49de..0e358be89c3 100644
--- a/libgfortran/generated/matmulavx128_c4.c
+++ b/libgfortran/generated/matmulavx128_c4.c
@@ -272,6 +272,11 @@ matmul_c4_avx128_fma3 (gfc_array_c4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_c4_avx128_fma3 (gfc_array_c4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_c4_avx128_fma4 (gfc_array_c4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_c4_avx128_fma4 (gfc_array_c4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_c8.c b/libgfortran/generated/matmulavx128_c8.c
index a41c160c993..c4416b17562 100644
--- a/libgfortran/generated/matmulavx128_c8.c
+++ b/libgfortran/generated/matmulavx128_c8.c
@@ -272,6 +272,11 @@ matmul_c8_avx128_fma3 (gfc_array_c8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_c8_avx128_fma3 (gfc_array_c8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_c8_avx128_fma4 (gfc_array_c8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_c8_avx128_fma4 (gfc_array_c8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_COMPLEX_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_COMPLEX_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_i1.c b/libgfortran/generated/matmulavx128_i1.c
index e1871578ea2..011de3ab5c3 100644
--- a/libgfortran/generated/matmulavx128_i1.c
+++ b/libgfortran/generated/matmulavx128_i1.c
@@ -272,6 +272,11 @@ matmul_i1_avx128_fma3 (gfc_array_i1 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_i1_avx128_fma3 (gfc_array_i1 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_1));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_i1_avx128_fma4 (gfc_array_i1 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_i1_avx128_fma4 (gfc_array_i1 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_1));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_1)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_i16.c b/libgfortran/generated/matmulavx128_i16.c
index 1a7b733181a..3a8b48a670f 100644
--- a/libgfortran/generated/matmulavx128_i16.c
+++ b/libgfortran/generated/matmulavx128_i16.c
@@ -272,6 +272,11 @@ matmul_i16_avx128_fma3 (gfc_array_i16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_i16_avx128_fma3 (gfc_array_i16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_i16_avx128_fma4 (gfc_array_i16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_i16_avx128_fma4 (gfc_array_i16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_i2.c b/libgfortran/generated/matmulavx128_i2.c
index a095c5872eb..738f64222db 100644
--- a/libgfortran/generated/matmulavx128_i2.c
+++ b/libgfortran/generated/matmulavx128_i2.c
@@ -272,6 +272,11 @@ matmul_i2_avx128_fma3 (gfc_array_i2 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_i2_avx128_fma3 (gfc_array_i2 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_2));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_i2_avx128_fma4 (gfc_array_i2 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_i2_avx128_fma4 (gfc_array_i2 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_2));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_2)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_i4.c b/libgfortran/generated/matmulavx128_i4.c
index a01c56f7138..6c3bb7cd876 100644
--- a/libgfortran/generated/matmulavx128_i4.c
+++ b/libgfortran/generated/matmulavx128_i4.c
@@ -272,6 +272,11 @@ matmul_i4_avx128_fma3 (gfc_array_i4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_i4_avx128_fma3 (gfc_array_i4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_i4_avx128_fma4 (gfc_array_i4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_i4_avx128_fma4 (gfc_array_i4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_i8.c b/libgfortran/generated/matmulavx128_i8.c
index bc78ffe2779..58bdf2ca242 100644
--- a/libgfortran/generated/matmulavx128_i8.c
+++ b/libgfortran/generated/matmulavx128_i8.c
@@ -272,6 +272,11 @@ matmul_i8_avx128_fma3 (gfc_array_i8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_i8_avx128_fma3 (gfc_array_i8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_i8_avx128_fma4 (gfc_array_i8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_i8_avx128_fma4 (gfc_array_i8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_INTEGER_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_INTEGER_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_r10.c b/libgfortran/generated/matmulavx128_r10.c
index 943678d1b51..72277faada3 100644
--- a/libgfortran/generated/matmulavx128_r10.c
+++ b/libgfortran/generated/matmulavx128_r10.c
@@ -272,6 +272,11 @@ matmul_r10_avx128_fma3 (gfc_array_r10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_r10_avx128_fma3 (gfc_array_r10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_r10_avx128_fma4 (gfc_array_r10 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_10)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_r10_avx128_fma4 (gfc_array_r10 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_10));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_10)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_r16.c b/libgfortran/generated/matmulavx128_r16.c
index 3d5738bb6f5..a6e11785299 100644
--- a/libgfortran/generated/matmulavx128_r16.c
+++ b/libgfortran/generated/matmulavx128_r16.c
@@ -272,6 +272,11 @@ matmul_r16_avx128_fma3 (gfc_array_r16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_r16_avx128_fma3 (gfc_array_r16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_r16_avx128_fma4 (gfc_array_r16 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_16)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_r16_avx128_fma4 (gfc_array_r16 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_16));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_16)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_r4.c b/libgfortran/generated/matmulavx128_r4.c
index 6c6da3994c9..b981053606e 100644
--- a/libgfortran/generated/matmulavx128_r4.c
+++ b/libgfortran/generated/matmulavx128_r4.c
@@ -272,6 +272,11 @@ matmul_r4_avx128_fma3 (gfc_array_r4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_r4_avx128_fma3 (gfc_array_r4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_r4_avx128_fma4 (gfc_array_r4 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_4)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_r4_avx128_fma4 (gfc_array_r4 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_4));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_4)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/generated/matmulavx128_r8.c b/libgfortran/generated/matmulavx128_r8.c
index d628200e8e3..6392b7b1cb5 100644
--- a/libgfortran/generated/matmulavx128_r8.c
+++ b/libgfortran/generated/matmulavx128_r8.c
@@ -272,6 +272,11 @@ matmul_r8_avx128_fma3 (gfc_array_r8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -284,11 +289,6 @@ matmul_r8_avx128_fma3 (gfc_array_r8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
@@ -825,6 +825,11 @@ matmul_r8_avx128_fma4 (gfc_array_r8 * const restrict retarray,
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = (GFC_REAL_8)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -837,11 +842,6 @@ matmul_r8_avx128_fma4 (gfc_array_r8 * const restrict retarray,
 
       t1 = malloc (t1_dim * sizeof(GFC_REAL_8));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = (GFC_REAL_8)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)
diff --git a/libgfortran/m4/matmul_internal.m4 b/libgfortran/m4/matmul_internal.m4
index fdc93e74663..e20f9224b61 100644
--- a/libgfortran/m4/matmul_internal.m4
+++ b/libgfortran/m4/matmul_internal.m4
@@ -223,6 +223,11 @@ sinclude(`matmul_asm_'rtype_code`.m4')dnl
       b_offset = 1 + b_dim1;
       b -= b_offset;
 
+      /* Empty c first.  */
+      for (j=1; j<=n; j++)
+	for (i=1; i<=m; i++)
+	  c[i + j * c_dim1] = ('rtype_name`)0;
+
       /* Early exit if possible */
       if (m == 0 || n == 0 || k == 0)
 	return;
@@ -235,11 +240,6 @@ sinclude(`matmul_asm_'rtype_code`.m4')dnl
 
       t1 = malloc (t1_dim * sizeof('rtype_name`));
 
-      /* Empty c first.  */
-      for (j=1; j<=n; j++)
-	for (i=1; i<=m; i++)
-	  c[i + j * c_dim1] = ('rtype_name`)0;
-
       /* Start turning the crank. */
       i1 = n;
       for (jj = 1; jj <= i1; jj += 512)