diff --git a/lib_blackfin/memcpy.S b/lib_blackfin/memcpy.S
new file mode 100644
index 0000000..a80fe3d
--- /dev/null
+++ b/lib_blackfin/memcpy.S
@@ -0,0 +1,130 @@
+/*
+ * File:         arch/blackfin/lib/memcpy.S
+ * Based on:
+ * Author:
+ *
+ * Created:
+ * Description:  internal version of memcpy(), issued by the compiler
+ *               to copy blocks of data around.
+ *               This is really memmove() - it has to be able to deal with
+ *               possible overlaps, because that ambiguity is when the compiler
+ *               gives up and calls a function. We have our own, internal version
+ *               so that we get something we trust, even if the user has redefined
+ *               the normal symbol.
+ * Rev:          $Id: memcpy.S 2775 2007-02-21 13:58:44Z hennerich $
+ *
+ * Modified:
+ *               Copyright 2004-2006 Analog Devices Inc.
+ *
+ * Bugs:         Enter bugs at http://blackfin.uclinux.org/
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, see the file COPYING, or write
+ * to the Free Software Foundation, Inc.,
+ * 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
+ */
+
+
+
+.align 2
+
+.globl _memcpy_ASM;
+_memcpy_ASM:
+	CC = R2 <=  0;	/* length not positive?*/
+	IF CC JUMP  .L_P1L2147483647;	/* Nothing to do */
+
+	P0 = R0 ;	/* dst*/
+	P1 = R1 ;	/* src*/
+	P2 = R2 ;	/* length */
+
+	/* check for overlapping data */
+	CC = R1 < R0;	/* src < dst */
+	IF !CC JUMP .Lno_overlap;
+	R3 = R1 + R2;
+	CC = R0 < R3;	/* and dst < src+len */
+	IF CC JUMP .Lhas_overlap;
+
+.Lno_overlap:
+	/* Check for aligned data.*/
+
+	R3 = R1 | R0;
+	R0 = 0x3;
+	R3 = R3 & R0;
+	CC = R3;	/* low bits set on either address? */
+	IF CC JUMP .Lnot_aligned;
+
+	/* Both addresses are word-aligned, so we can copy
+	at least part of the data using word copies.*/
+	P2 = P2 >> 2;
+	CC = P2 <= 2;
+	IF !CC JUMP .Lmore_than_seven;
+	/* less than eight bytes... */
+	P2 = R2;
+	LSETUP(.Lthree_start, .Lthree_end) LC0=P2;
+	R0 = R1;	/* setup src address for return */
+.Lthree_start:
+	R3 = B[P1++] (X);
+.Lthree_end:
+	B[P0++] = R3;
+
+	RTS;
+
+.Lmore_than_seven:
+	/* There's at least eight bytes to copy. */
+	P2 += -1;	/* because we unroll one iteration */
+	LSETUP(.Lword_loop, .Lword_loop) LC0=P2;
+	R0 = R1;
+	I1 = P1;
+	R3 = [I1++];
+.Lword_loop:
+	MNOP || [P0++] = R3 || R3 = [I1++];
+
+	[P0++] = R3;
+	/* Any remaining bytes to copy? */
+	R3 = 0x3;
+	R3 = R2 & R3;
+	CC = R3 == 0;
+	P1 = I1;	/* in case there's something left, */
+	IF !CC JUMP .Lbytes_left;
+	RTS;
+.Lbytes_left:	P2 = R3;
+.Lnot_aligned:
+	/* From here, we're copying byte-by-byte. */
+	LSETUP (.Lbyte_start , .Lbyte_end) LC0=P2;
+	R0 = R1;	/* Save src address for return */
+.Lbyte_start:
+	R1 = B[P1++] (X);
+.Lbyte_end:
+	B[P0++] = R1;
+
+.L_P1L2147483647:
+	RTS;
+
+.Lhas_overlap:
+/* Need to reverse the copying, because the
+ * dst would clobber the src.
+ * Don't bother to work out alignment for
+ * the reverse case.
+ */
+	R0 = R1;	/* save src for later. */
+	P0 = P0 + P2;
+	P0 += -1;
+	P1 = P1 + P2;
+	P1 += -1;
+	LSETUP(.Lover_start, .Lover_end) LC0=P2;
+.Lover_start:
+	R1 = B[P1--] (X);
+.Lover_end:
+	B[P0--] = R1;
+
+	RTS;