VSBM documentation

John Strawn

9/1/87

The goal is to get (A-B)*C.  A "brute-force" implementation would
be something like


     move      x:(r1),a       ; A
     move      x:(r5),x0      ; B
     sub       x0,a           ; B-A
     tfr       a,y0           ; NB allows parallel move
     move      x:(r2),y1      ; C
     mpyr      y1,y0
     move      a,x:(r6)

This is wasteful, because the tfr, mpyr, and sub will require at least
3 operations in even the tightest inner loop. Therefore, we implement
this as A*C-B*C, for which the outline is:

     move      x:(r1),x0	; A
     move      x:(r5),y0	; B	
     move      x:(r2),y1	; C
     mpy  x0,y1,a
     macr -y0,y1,a
     move      a,x:(r6)