*===============================================================================
*
*	TEXAS INSTRUMENTS, INC.
*
*	IIRCAS4
*
*	Revision Data: 05/14/97
*
*	USAGE 	This routine is C Callable and can be called as:
*
*		void iircas4(int n, short *c, int *d, int *y)
*
*		n	= the number of cascaded biquads
*		c	= array containing -a1, -a2, b1, b2 biquad coefs
*		d	= array of the delayed states within biquads
*		y	= inputs y[0] and y[1] (also outputs)
*
*	C CODE 	This is the C equivalent of the assembly code without
*		restrictions.  Note that the assembly code is hand optimized and
*		restrictions may apply.
*
*	void iircas4(int n, short *c, int *d, int *y)
*	{
*	int  k0, k1, i;
*	   for (i = 0; i < n; i++) {
*	      k0   = c[4*i+1]*(d[2*i+1] >> 16) + c[4*i+0]*(d[2*i+0]>> 16) +y[0];
*	      y[0] = c[4*i+3]*(d[2*i+1] >> 16) + c[4*i+2]*(d[2*i+0]>> 16) + k0;
*	      d[2*i+1] = k0;
*	      k1   = c[4*i+1]*(d[2*i+0] >> 16) + c[4*i+0]*(k0 >> 16) + y[1];
*	      y[1] = c[4*i+3]*(d[2*i+0] >> 16) + c[4*i+2]*(k0 >> 16) + k1;
*	      d[2*i+0] = k1;
*	      }
*	}
*	
*	DESCRIPTION
*		The iircas4 performs a cascaded biquad iir filter with
*		the direct form II structure (4-multiplies.)  It performs two
*		samples at a time.  Coefficients are stored in the order -a1,
*		-a2, b1, b2 for each successive biquad located in the c array.
*		Both outputs are stored back to the location of the inputs y[0] 
*		and y[1].  The inputs and outputs are 32 bit values while the 
*		coefficients are 16 bit values.
*		
*	TECHNIQUES
*		The loop is written so that one biquad for each of the two
*		inputs is completed every time through the loop.  There is a 
*		an extra priming delay for the second input so that the biquads
*		new delayed state k0 is calculated based on the first input
*		(i.e. the second input is being processed by the biquad
*		preceeding the biquad which is processing the first input.)
*
*	MEMORY NOTE
*		The d and c array pointers must be placed on opposite word
*		boundaries to avoid memory hits (i.e. one must start on an even
*		word boundary while the other starts on an odd word boundary.)
*
*	CYCLES	4*N + 16  (Note: the iircas4 cycle count is for two inputs.)
*		for N = 10 -> 56 cycles or 280 nsec
*
*===============================================================================
	.global _iircas4
	.text

_iircas4:
	STW	.D2	B10,	*B15--		; push B10 on stack
||	MV	.L1X	B6,	A3		; copy y pointer

*** BEGIN Benchmark Timing ***
B_START:

	B	.S2	LOOP			; for
||	LDW	.D1	*+A3[1],	B10	; y1 = y[1]
||	MV	.S1X	B11,	A9		; store B11 in A9

	ADD	.L1	2,	A4,	A1	; n+2
||	MVK	.S2	3,	B0		; setup priming count
||	MVK	.S1	2,	A2		; setup priming count
||	LDW	.D1	*+A3[0],	A7	; y0 = y[0]

LOOP:

  [!A1]	MVK	.S1	1,	A2
||	MPYH	.M2	B11,	B5,	B9	; g1 = (d0 >> 16) * b2
||	MV	.L2X	A0,	B8		;* copy a1, a2
||	MV	.S2	B2,	B11		;* copy d0
||	ADD	.L1	A5,	A8,	A8	;* h0 = a1d0 + a2d1
||	MPYH	.M1X	B6,	A4,	A8	;* b2d1 = (d1 >> 16) * b2
||	LDW	.D1	*A6++,	B2		;*** d0 = d[2*i+0]
||	LDW	.D2	*B4++,	A0		;*** a1 = c[4*i], a2 = c[4*i+1]

  [A1]	SUB	.S1	A1,	1,	A1	; i++
||[!B0]	ADD	.S2	B9,	B10,	B10	; h1 = j1 + y1
||[!A2]	ADD	.L1	A8,	A7,	A7	;* k0 = h0 + y0
||	MV	.L2X	A4,	B5		;* copy b1, b2
||	MPYH	.M2	B11,	B8,	B1	;* e1 = (d0 >> 16) * a2
||	MPYHL	.M1X	B2,	A0,	A5	;** a1d0 = (d0 >> 16) * a1
||	LDW	.D2	*B4++,	A4		;*** b1 = c[4*i+0],b2 = c[4*i+1]
||	LDW	.D1	*A6++,	B6		;*** d1 = d[2*i+1]	

  [A1]	B	.S1	LOOP			; for
||	ADD	.L2	B10,	B1,	B8	; k1 = h1 + e1
||	ADD	.D2	B7,	B9,	B9	; m1 = f1 + g1
||	MPYHL	.M2X	A7,	B8,	B9	;* j1 = (k0 >> 16) * a1
||[!A2]	STW	.D1	A7,	*-A6[5]		;* d[2*i+1] = k0 
||	ADD	.L1	A5,	A8,	A8	;* m0 = b1d0 + b2d1
||	MPYH	.M1X	B6,	A0,	A8	;** a2d1 = (d1 >> 16) * a2

  [B0]	SUB	.S2	B0,	1,	B0	; decrement priming count
||[A2]	SUB	.L1	A2,	1,	A2	; decrement priming count
||[!B0]	STW	.D1	B8,	*-A6[8]		; d[2*i+0] = k1		
||[!B0]	ADD	.L2	B9,	B8,	B10	; y1 = m1 + k1	
||	MPYHL	.M2X	A7,	B5,	B7	;* f1 = (k0 >> 16) * b1
||[!A2]	ADD	.S1	A8,	A7,	A7	;* y0 = m0 + k0
||	MPYHL	.M1X	B2,	A4,	A5	;** b1d0 = (d0 >> 16) * b1
;end of LOOP

	B	.S2	B3			; return
||	LDW	.D2	*++B15,	B10		; pop B10 off stack
||	STW	.D1	A7,	*+A3[0]		; y[0] = y0

	STW	.D1	B10,	*+A3[1]		; y[1] = y1
||	MV	.S2X	A9,	B11		; restore B11

B_END:
*** END Benchmark Timing ***

	NOP	4