C & Assembly to Perform 16x16 Pixel-Block Motion Estimation

char *btpr; /* pointer to start row of 16x16 pixel block being compressed */

char *cptr; /* pointer to start row of 16x16 pixel reference block */

val = 0;

for (i=0; i<16; i++) {

for (j=0; j<16; j++) {

data = (*(bptr++)- *(cptr++));

if (data<0){val -= data;}

else {val += data;}

}

/* Fast out after this row if best match has been exceeded */

 if (val > best_value) break;

/* Update pointer to next row */

 bptr += (rm->width - 16);

/* Update pointer  to next row */

  cptr += (cm->width - 16);

}

Example 2. Inner-Core Absolute Differences for 16x16 Pixel Block

__asm {

   movdqu xmm0, [m1]

   movdqu xmm1, [m2]

   movdqa xmm2, xmm0

   psubusb xmm0, xmm1

   psubusb xmm1, xmm2

   por xmm0, xmm1

   movdqa xmm1, xmm0

   punpcklbw xmm0, xmm6

   punpcklbw xmm1, xmm6

   movdqa xmm3, xmm1

   pshufd xmm1, xmm0, 238

   pshufd xmm3, xmm0, 68

 paddw xmm1, xmm3

movdqa xmm4, xmm1

pshufd xmm4, xmm4, 78

   paddw xmm1, xmm4

}