.align_code 4;
biq:                                                             		// Inner loop computations are SIMD, using both X and Y comp blocks
           word1: fr3=r0*r14; fr8=r8+r3; r5:4=q[j0+=4];    r17:16=q[k0+=4];;  // r3=a2*w(n-2), r8=x(n), r4=w(n-1), r16=a1 (S1,S2), r17=a1 (S3,S4)
           word2: fr7=r1*r15; fr9=r9+r7; q[j1+=4]=r5:4;;     					// r7=a2*w(n-2), r9=x(n), r5=w(n-1)
           word3: fr3=r4*r16; fr8=r8+r3;                   r15:14=q[k0+=4];;  // r3=a1*w(n-1), r8=x(n)+a2*w(n-2), store new w(n-2), r14=b2 (S1,S2), r15=b2 (S3,S4)
           word4: fr7=r5*r17; fr9=r9+r7;;     								// r7=a1*w(n-1), r9=x(n)+a2*w(n-2), store new w(n-2)
        
           word5: fr3=r0*r14; fr12=r8+r3;                  r17:16=q[k0+=4];;  // r3=b2*w(n-2), r12=new w(n), r0=next w(n-2), r16=b1 (S1,S2), r17=b1 (S3,S4)
           word6: fr7=r1*r15; fr13=r9+r7; r1:0=q[j0+=4];;    					// r7=b2*w(n-2), r13=new w(n), r1=next w(n-2)
           word7: fr3=r4*r16; fr8=r12+r3; q[j1+=4]=r13:12; r15:14=q[k0+=4];;  // r3=b1*w(n-1), r8=w(n)+b2*w(n-2), store new w(n-1), r14=next a2 (S1,S2), r15=next a2 (S3,S4)
.align_code 4;
	   if NLC1E, jump biq;
           word8: fr7=r5*r17; fr9=r13+r7;;   									// r7=b1*w(n-1), r9=w(n)+b2*w(n-2), store new w(n-1)





.align_code 4;
_xcorrs_main_loop:
	xytr15:0=xcorrs(r7:4,  thr3:0)(ext); r11:8=q[j4+=4];r19:16=q[k4+=4];;
	xytr15:0=xcorrs(r15:12,  thr3:0)(ext);;
	xytr15:0=xcorrs(r11:8,  thr3:0)(ext);r7:4=q[j4+=4];r15:12=q[k4+=4];;
	xytr15:0=xcorrs(r19:16,  thr3:0)(ext);;
	
	xytr15:0=xcorrs(r7:4,  thr3:0)(ext); r11:8=q[j4+=4];r19:16=q[k4+=4];;
	xytr15:0=xcorrs(r15:12,  thr3:0)(ext);;
	xytr15:0=xcorrs(r11:8,  thr3:0)(ext);r7:4=q[j4+=4];r15:12=q[k4+=4];;
.align_code 4;
	if NLC1E, jump _xcorrs_main_loop;	
	xytr15:0=xcorrs(r19:16,  thr3:0)(ext);;