PB Displacement filter really slow with big images

Question

hello;
     I am trying to make a Pixel bender displacement filter that will handle displacements bigger than 128 pixels, for flash. Right now flash only uses one 8 bit color channel to represent a displacement,  which can not be bigger than 128 pixels. I am trying to make a pixel bender kernel that will take 3 color channels( 24 bit) as a displacement, which should be able to handle displacements as big as 8388608. That is obviously bigger than I need.
     The problem i am finding is that when the image or the displacement maps are too big it runs really slow. With images the size 3072x768,  it runs at 3 frames per second on an intel based mac.
     The question is , does anybody know how to optimise this code?
the code:
<languageVersion : 1.0;>
kernel DisplacementMap
<
namespace : "dood";
vendor : "m0ose";
version : 1;
description : "Displacement by bitmap map 16 bit. X and Y are seperate bitmaps  ";
>
parameter   float       scale
    <
        minValue        : -10.0;
        maxValue        :  10.0;
        defaultValue    :  1.0;
    >;
parameter float offsetX
    <
        minValue        : -1000.0;
        maxValue        :  1000.0;
        defaultValue    : 0.0;
    >;
parameter float offsetY
    <
            minValue        : -1000.0;
        maxValue        :  1000.0;
        defaultValue    : 0.0;
    >;
parameter float zoom
    <
            minValue        : 0.1;
        maxValue        :  10.0;
        defaultValue    : 1.0;
    >;
input image4 src;
input image4 mapX;
input image4 mapY;
output pixel4 dst;
void evaluatePixel()
    float2 pos = outCoord();
    float4 amountX = sampleLinear(mapX , pos);
    float4 amountY = sampleLinear(mapY , pos);
    float x = (amountX.b * 256.0 - 128.0) ; //red
    x = x + (amountX.g * 256.0 - 128.0) * 128.0;//alpha
    x = x * scale;
    x = x + pos.x;
    x = x + offsetX ;
    x = x / zoom;
    float y =  (amountY.b * 256.0 - 128.0) ;
    y = y + (amountY.g * 256.0 - 128.0) * 128.0;
    y = y * scale;
    y = y + pos.y;
    y = y + offsetY ;
    y = y / zoom;
    float2 displacedPos = float2( x , y );
    dst = sampleLinear(src, displacedPos);
thanx

Answer 1

I am adding a shader to the filter attribute of a Flex Video Player
code: something like this
[Embed(source="displ24test.pbj" , mimeType="application/octet-stream")]
protected var displacer:Class;
var shader:Shader;
public function init():void
shader = new Shader( new displacer() );
shader.data.src.input = src;
shader.data.mapX.input = mapX;
shader.data.mapY.input = mapY;
shader.data.scale.value = [ scale ];
shader.data.offsetX.value = [ offsetX ];
shader.data.offsetY.value = [ offsetY ];
shader.data.zoom.value = [ zoom];
_mov.filters = [ new ShaderFilter( shader) ];
then in flex:
<s:VideoPlayer x="10" y="10" source="df_retro5_1k_sound.mov"  id="_mov"  autoPlay="true"/>