pi64bmAddrBuf1 = (Int64 *)&bm[i][0];                     //  &bm[0][i]
            #pragma UNROLL(1)
            for (k = 0, m = 0; k < (NO_OF_STATES>>1); k++, m += 2)
            {

                ui32packedStates = _amem4(pi32softIdealOp++);

                i64TmpVar = _amem8(pi64bmAddrBuf1++);
                path1 = _loll(i64TmpVar);
                path2 = _hill(i64TmpVar);

                i64TmpVar = _ddotp4(i32_decoderIP01, ui32packedStates);
                i32Tmp1 = _loll(i64TmpVar);
                i32Tmp2 = _hill(i64TmpVar);

                path1  += i32Tmp1;
                path2  += i32Tmp2;

                if (path1 > path2)
                {

                    *pui8bestState1 = (m);
                    _amem4(pi32bmAddrBuf2++) = path1;

                }
                else
                {

                    *pui8bestState1 = (m+1);
                    _amem4(pi32bmAddrBuf2++) = path2;

                }

                pui8bestState1 += MAX_DECODER_OP_LENGTH;

            }