[Blackfin] arch: fix bug - make memcpy return the dest addr.

	/* Check for aligned data.*/

	R3 = R1 | R0;

	R0 = 0x3;

	R3 = R3 & R0;

	R1 = 0x3;

	R3 = R3 & R1;

	CC = R3;	/* low bits set on either address? */

	IF CC JUMP .Lnot_aligned;

	/* 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:

	/* There's at least eight bytes to copy. */

	P2 += -1;	/* because we unroll one iteration */

	LSETUP(.Lword_loops, .Lword_loope) LC0=P2;

	R0 = R1;

	I1 = P1;

	R3 = [I1++];

#if ANOMALY_05000202

.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:

	 * 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;