user manual

Use 3DNow!™ PAVGUSB for MPEG-2 Motion Compensation 123
22007E/0—November 1999 AMD Athlon™ Processor x86 Code Optimization
;;
;; DESTROYS MM0,MM1,MM2,MM3,MM4
PXOR MM0, MM0 ; 0 | 0
MOVQ MM1, MM6 ; w | z
MOVQ MM4, MM5 ; y | x
PUNPCKHDQ MM1, MM1 ; w | w
MOVQ MM3, MM6 ; w | z
MOVQ MM2, MM5 ; y | x
PFSUBR MM3, MM0 ; -w | -z
PFSUBR MM2, MM0 ; -y | -x
PUNPCKLDQ MM3, MM6 ; z | -z
PFCMPGT MM4, MM1 ; y>w?FFFFFFFF:0 | x>w?FFFFFFFF:0
MOVQ MM0, QWORD PTR [ABOVE_RIGHT] ; ABOVE | RIGHT
PFCMPGT MM3, MM1 ; z>w?FFFFFFFF:0 | -z>w>FFFFFFFF:0
PFCMPGT MM2, MM1 ; -y>w?FFFFFFFF:0 | -x>w?FFFFFFFF:0
MOVQ MM1, QWORD PTR [BEHIND_BEFORE] ; BEHIND | BEFORE
PAND MM4, MM0 ; y > w ? ABOVE:0 | x > w ? RIGHT:0
MOVQ MM0, QWORD PTR [BELOW_LEFT] ; BELOW | LEFT
PAND MM3, MM1 ; z > w ? BEHIND:0 | -z > w ? BEFORE:0
PAND MM2, MM0 ; -y > w ? BELOW:0 | -x > w ? LEFT:0
POR MM2, MM4 ; BELOW,ABOVE | LEFT,RIGHT
POR MM2, MM3 ; BELOW,ABOVE,BEHIND | LEFT,RIGHT,BEFORE
MOVQ MM1, MM2 ; BELOW,ABOVE,BEHIND | LEFT,RIGHT,BEFORE
PUNPCKHDQ MM2, MM2 ; BELOW,ABOVE,BEHIND | BELOW,ABOVE,BEHIND
POR MM2, MM1 ; zclip, yclip, xclip = clip code
Use 3DNow!™ PAVGUSB for MPEG-2 Motion Compensation
Use the 3DNow! PAVGUSB instruction for MPEG-2 motion
compensation. The PAVGUSB instruction produces the rounded
averages of the eight unsigned 8-bit integer values in the source
operand (a MMX register or a 64-bit memory location) and the
eight corresponding unsigned 8-bit integer values in the
destination operand (a MMX register). The PAVGUSB
instruction is extremely useful in DVD (MPEG-2) decoding
where motion compensation performs a lot of byte averaging
between and within macroblocks. The PAVGUSB instruction
helps speed up these operations. In addition, PAVGUSB can
free up some registers and make unrolling the averaging loops
possible.
The following code fragment uses original MMX code to
perform averaging between the source macroblock and
destination macroblock: