/*
 * Portions Copyright (C) 2003 Sun Microsystems, Inc.
 * All rights reserved.
 */

/*
 *
 * COPYRIGHT NVIDIA CORPORATION 2003. ALL RIGHTS RESERVED.
 * BY ACCESSING OR USING THIS SOFTWARE, YOU AGREE TO:
 *
 *  1) ACKNOWLEDGE NVIDIA'S EXCLUSIVE OWNERSHIP OF ALL RIGHTS
 *     IN AND TO THE SOFTWARE;
 *
 *  2) NOT MAKE OR DISTRIBUTE COPIES OF THE SOFTWARE WITHOUT
 *     INCLUDING THIS NOTICE AND AGREEMENT;
 *
 *  3) ACKNOWLEDGE THAT TO THE MAXIMUM EXTENT PERMITTED BY
 *     APPLICABLE LAW, THIS SOFTWARE IS PROVIDED *AS IS* AND
 *     THAT NVIDIA AND ITS SUPPLIERS DISCLAIM ALL WARRANTIES,
 *     EITHER EXPRESS OR IMPLIED, INCLUDING, BUT NOT LIMITED
 *     TO, IMPLIED WARRANTIES OF MERCHANTABILITY  AND FITNESS
 *     FOR A PARTICULAR PURPOSE.
 *
 * IN NO EVENT SHALL NVIDIA OR ITS SUPPLIERS BE LIABLE FOR ANY
 * SPECIAL, INCIDENTAL, INDIRECT, OR CONSEQUENTIAL DAMAGES
 * WHATSOEVER (INCLUDING, WITHOUT LIMITATION, DAMAGES FOR LOSS
 * OF BUSINESS PROFITS, BUSINESS INTERRUPTION, LOSS OF BUSINESS
 * INFORMATION, OR ANY OTHER PECUNIARY LOSS), INCLUDING ATTORNEYS'
 * FEES, RELATING TO THE USE OF OR INABILITY TO USE THIS SOFTWARE,
 * EVEN IF NVIDIA HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGES.
 *
 */

package demos.proceduralTexturePhysics;

import com.sun.opengl.util.FileUtil;
import com.sun.opengl.util.texture.Texture;
import com.sun.opengl.util.texture.TextureData;
import com.sun.opengl.util.texture.TextureIO;
import demos.util.Cubemap;
import gleem.linalg.Mat4f;
import gleem.linalg.Rotf;
import java.io.IOException;
import java.nio.ByteBuffer;
import java.util.ArrayList;
import java.util.Iterator;
import java.util.List;
import javax.media.opengl.GL;
import javax.media.opengl.GL2;
import javax.media.opengl.GLAutoDrawable;
import javax.media.opengl.GLCapabilities;
import javax.media.opengl.GLDrawableFactory;
import javax.media.opengl.GLEventListener;
import javax.media.opengl.GLException;
import javax.media.opengl.GLPbuffer;
import javax.media.opengl.glu.GLU;
import javax.media.opengl.util.BufferUtil;



/**
 * Auxiliary Water simulation class used by ProceduralTexturePhysics
 * main loop. Demonstration by NVidia Corporation.
 *
 * <P>
 *
 * Ported to Java and ARB_fragment_program by Kenneth Russell
 */

public class Water {
  // Note: this class is organized differently than most of the demos
  // due to the fact that it is used for two purposes: when the
  // pbuffer's context is current it is used to update the cellular
  // automata, and when the parent drawable's context is current it is
  // used to render the water geometry (with the parent drawable's GL
  // object).

  private GLU glu = new GLU();

  // Rendering modes
  public static final int CA_FULLSCREEN_REFLECT   = 0;
  public static final int CA_FULLSCREEN_FORCE     = 1;
  public static final int CA_FULLSCREEN_HEIGHT    = 2;
  public static final int CA_FULLSCREEN_NORMALMAP = 3;
  public static final int CA_TILED_THREE_WINDOWS  = 4;
  public static final int CA_DO_NOT_RENDER        = 5;

  private int[] initialMapDimensions = new int[2];
  private TextureData initialMapData;

  private String tmpSpinFilename;
  private String tmpDropletFilename;
  private String tmpCubeMapFilenamePrefix;
  private String tmpCubeMapFilenameSuffix;

  private GLPbuffer pbuffer;
  private Rotf cameraOrientation = new Rotf();

  // Dynamic texture names
  private static final int CA_TEXTURE_FORCE_INTERMEDIATE = 0;
  private static final int CA_TEXTURE_FORCE_TARGET       = 1;
  private static final int CA_TEXTURE_VELOCITY_SOURCE    = 2;
  private static final int CA_TEXTURE_VELOCITY_TARGET    = 3;
  private static final int CA_TEXTURE_HEIGHT_SOURCE      = 4;
  private static final int CA_TEXTURE_HEIGHT_TARGET      = 5;
  private static final int CA_TEXTURE_NORMAL_MAP         = 6;
  private static final int CA_NUM_DYNAMIC_TEXTURES       = 7;
    
  // List names
  private static final int CA_FRAGMENT_PROGRAM_EQ_WEIGHT_COMBINE     = 0;
  private static final int CA_FRAGMENT_PROGRAM_NEIGHBOR_FORCE_CALC_1 = 1;
  private static final int CA_FRAGMENT_PROGRAM_NEIGHBOR_FORCE_CALC_2 = 2;
  private static final int CA_FRAGMENT_PROGRAM_APPLY_FORCE           = 3;
  private static final int CA_FRAGMENT_PROGRAM_APPLY_VELOCITY        = 4;
  private static final int CA_FRAGMENT_PROGRAM_CREATE_NORMAL_MAP     = 5;
  private static final int CA_FRAGMENT_PROGRAM_REFLECT               = 6;
  private static final int CA_DRAW_SCREEN_QUAD                       = 7;
  private static final int CA_NUM_LISTS                              = 8;

  // Static textures
  private Texture initialMapTex;
  private Texture spinTex;
  private Texture dropletTex;
  private Texture cubemap;

  private Texture[] dynamicTextures = new Texture[CA_NUM_DYNAMIC_TEXTURES];
    
  private int       texHeightInput;                 // current input height texture ID.
  private int       texHeightOutput;                // current output height texture ID.
  private int       texVelocityInput;               // current input velocity texture ID.
  private int       texVelocityOutput;              // current output velocity texture ID.
  private int       texForceStepOne;                // intermediate force computation result texture ID.
  private int       texForceOutput;                 // current output force texture ID.

  private int[]     displayListIDs = new int[CA_NUM_LISTS];
    
  private int       vertexProgramID;                // one vertex program is used to choose the texcoord offset

  private int       flipState;                      // used to flip target texture configurations.

  private boolean   wrap;                           // CA can either wrap its borders, or clamp (clamp by default)  
  private boolean   reset = true;                   // are we resetting this frame? (user hit reset).
  private boolean   singleStep;                     // animation step on keypress.
  private boolean   animate = true;                 // continuous animation.
  private boolean   slow = true;                    // run slow.
  private boolean   wireframe;                      // render in wireframe mode
  private boolean   applyInteriorBoundaries = true; // enable / disable "boundary" image drawing.
  private boolean   spinLogo = true;                // draw spinning logo.
  private boolean   createNormalMap = true;         // enable / disable normal map creation.

  private float     perTexelWidth;                  // width of a texel (percentage of texture)
  private float     perTexelHeight;                 // height of a texel

  private float     blurDist = 0.5f;                // distance over which to blur.
  private boolean   mustUpdateBlurOffsets;          // flag indicating blurDist was set last tick

  private float     normalSTScale = 0.8f;           // scale of normals in normal map.
  private float     bumpScale = 0.25f;              // scale of bumps in water.

  private float     dropletFrequency = 0.175f;      // frequency at which droplets are drawn in water...

  private int       slowDelay = 1;                  // amount (milliseconds) to delay when running slow.
  private int       skipInterval;                   // frames to skip simulation.
  private int       skipCount;                      // frame count for skipping rendering

  private int       angle;                          // angle in degrees for spinning logo

  private List/*<Droplet>*/ droplets = new ArrayList/*<Droplet>*/();             // array of droplets

  private int       renderMode; 

  // Constant memory locations
  private static final int CV_UV_OFFSET_TO_USE =  0;

  private static final int CV_UV_T0_NO_OFFSET  =  1;
  private static final int CV_UV_T0_TYPE1      =  2;
  private static final int CV_UV_T0_TYPE2      =  3;
  private static final int CV_UV_T0_TYPE3      =  4;
  private static final int CV_UV_T0_TYPE4      =  5;

  private static final int CV_UV_T1_NO_OFFSET  =  6;
  private static final int CV_UV_T1_TYPE1      =  7;
  private static final int CV_UV_T1_TYPE2      =  8;
  private static final int CV_UV_T1_TYPE3      =  9;
  private static final int CV_UV_T1_TYPE4      = 10;

  private static final int CV_UV_T2_NO_OFFSET  = 11;
  private static final int CV_UV_T2_TYPE1      = 12;
  private static final int CV_UV_T2_TYPE2      = 13;
  private static final int CV_UV_T2_TYPE3      = 14;
  private static final int CV_UV_T2_TYPE4      = 15;

  private static final int CV_UV_T3_NO_OFFSET  = 16;
  private static final int CV_UV_T3_TYPE1      = 17;
  private static final int CV_UV_T3_TYPE2      = 18;
  private static final int CV_UV_T3_TYPE3      = 19;
  private static final int CV_UV_T3_TYPE4      = 20;

  private static final int CV_CONSTS_1         = 21;

  public void initialize(String initialMapFilename,
                         String spinFilename,
                         String dropletFilename,
                         String cubeMapFilenamePrefix,
                         String cubeMapFilenameSuffix,
                         GLAutoDrawable parentWindow) {
    loadInitialTexture(initialMapFilename);
    tmpSpinFilename           = spinFilename;
    tmpDropletFilename        = dropletFilename;
    tmpCubeMapFilenamePrefix  = cubeMapFilenamePrefix;
    tmpCubeMapFilenameSuffix  = cubeMapFilenameSuffix;
    
    // create the pbuffer.  Will use this as an offscreen rendering buffer.
    // it allows rendering a texture larger than our window.
    GLCapabilities caps = new GLCapabilities();
    caps.setDoubleBuffered(false);
    if (!GLDrawableFactory.getFactory().canCreateGLPbuffer()) {
      throw new GLException("Pbuffers not supported with this graphics card");
    }
    pbuffer = GLDrawableFactory.getFactory().createGLPbuffer(caps,
                                                             null,
                                                             initialMapDimensions[0],
                                                             initialMapDimensions[1],
                                                             parentWindow.getContext());
    pbuffer.addGLEventListener(new Listener());
  }

  public void destroy() {
    if (pbuffer != null) {
      pbuffer.destroy();
      pbuffer = null;
    }
    reset = true;
  }

  public void tick() { 
    pbuffer.display();
  }

  public void draw(GL2 gl, Rotf cameraOrientation) {
    this.cameraOrientation.set(cameraOrientation);

    if (skipCount >= skipInterval && renderMode != CA_DO_NOT_RENDER) {
      skipCount = 0;
      // Display the results of the rendering to texture
      if (wireframe) {
        gl.glPolygonMode(GL2.GL_FRONT_AND_BACK, GL2.GL_LINE);
           
        // chances are the texture will be all dark, so lets not use a texture
        gl.glDisable(GL2.GL_TEXTURE_2D);
      } else {
        gl.glPolygonMode(GL2.GL_FRONT_AND_BACK, GL2.GL_FILL);
            			
        gl.glActiveTexture(GL2.GL_TEXTURE0);
        gl.glEnable(GL2.GL_TEXTURE_2D);
      }

      switch (renderMode) {
        case CA_FULLSCREEN_REFLECT: {
          // include bump scale...
          Mat4f bscale = new Mat4f();
          bscale.makeIdent();
          bscale.set(0, 0, bumpScale);
          bscale.set(1, 1, bumpScale);
          Mat4f rot = new Mat4f();
          rot.makeIdent();
          rot.setRotation(cameraOrientation);
          Mat4f matRot = rot.mul(bscale);

          gl.glCallList(displayListIDs[CA_FRAGMENT_PROGRAM_REFLECT]);

          // Draw quad over full display
          gl.glActiveTexture(GL2.GL_TEXTURE0);
          dynamicTextures[CA_TEXTURE_NORMAL_MAP].bind();
          dynamicTextures[CA_TEXTURE_NORMAL_MAP].disable();
          gl.glActiveTexture(GL2.GL_TEXTURE3);
          cubemap.bind();
          cubemap.enable();

          gl.glColor4f(1, 1, 1, 1);
          gl.glBegin(GL2.GL_QUADS);
                
          gl.glMultiTexCoord2f(GL2.GL_TEXTURE0, 0,0);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE1, matRot.get(0,0), matRot.get(0,1), matRot.get(0,2),  1);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE2, matRot.get(1,0), matRot.get(1,1), matRot.get(1,2),  1);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE3, matRot.get(2,0), matRot.get(2,1), matRot.get(2,2),  1);
          gl.glVertex2f(-1,-1);
                
          gl.glMultiTexCoord2f(GL2.GL_TEXTURE0, 1,0);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE1, matRot.get(0,0), matRot.get(0,1), matRot.get(0,2), -1);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE2, matRot.get(1,0), matRot.get(1,1), matRot.get(1,2),  1);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE3, matRot.get(2,0), matRot.get(2,1), matRot.get(2,2),  1);
          gl.glVertex2f( 1,-1);
                
          gl.glMultiTexCoord2f(GL2.GL_TEXTURE0, 1,1);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE1, matRot.get(0,0), matRot.get(0,1), matRot.get(0,2), -1);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE2, matRot.get(1,0), matRot.get(1,1), matRot.get(1,2), -1);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE3, matRot.get(2,0), matRot.get(2,1), matRot.get(2,2),  1);
          gl.glVertex2f( 1, 1);
                
          gl.glMultiTexCoord2f(GL2.GL_TEXTURE0, 0,1);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE1, matRot.get(0,0), matRot.get(0,1), matRot.get(0,2),  1);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE2, matRot.get(1,0), matRot.get(1,1), matRot.get(1,2), -1);
          gl.glMultiTexCoord4f(GL2.GL_TEXTURE3, matRot.get(2,0), matRot.get(2,1), matRot.get(2,2),  1);
          gl.glVertex2f(-1, 1);
                
          gl.glEnd();
    
          cubemap.disable();
          gl.glDisable(GL2.GL_FRAGMENT_PROGRAM);
                
          break;
        }

        case CA_FULLSCREEN_NORMALMAP: {
          // Draw quad over full display
          gl.glActiveTexture(GL2.GL_TEXTURE0);
          dynamicTextures[CA_TEXTURE_NORMAL_MAP].bind();
                
          gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);
          break;
        }

        case CA_FULLSCREEN_HEIGHT: {
          // Draw quad over full display
          gl.glActiveTexture(GL2.GL_TEXTURE0);
          gl.glBindTexture(GL2.GL_TEXTURE_2D, texHeightOutput);
                
          gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);
          break;
        }

        case CA_FULLSCREEN_FORCE: {
          // Draw quad over full display
          gl.glActiveTexture(GL2.GL_TEXTURE0);
          dynamicTextures[CA_TEXTURE_FORCE_TARGET].bind();
			                 
          gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);
          break;
        }

        case CA_TILED_THREE_WINDOWS: {
          // Draw quad over full display
          // lower left
          gl.glActiveTexture(GL2.GL_TEXTURE0);
          dynamicTextures[CA_TEXTURE_FORCE_TARGET].bind();
          gl.glMatrixMode(GL2.GL_MODELVIEW);
          gl.glPushMatrix();
			                 
          gl.glTranslatef(-0.5f, -0.5f, 0);
          gl.glScalef(0.5f, 0.5f, 1);
          gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);
          gl.glPopMatrix();

          // lower right
          gl.glBindTexture(GL2.GL_TEXTURE_2D, texVelocityOutput);
          gl.glPushMatrix();
			                 
          gl.glTranslatef(0.5f, -0.5f, 0);
          gl.glScalef(0.5f, 0.5f, 1);
          gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);
          gl.glPopMatrix();

          // upper left
          dynamicTextures[CA_TEXTURE_NORMAL_MAP].bind();
          gl.glMatrixMode(GL2.GL_MODELVIEW);
          gl.glPushMatrix();
			                 
          gl.glTranslatef(-0.5f, 0.5f, 0);
          gl.glScalef(0.5f, 0.5f, 1);
          gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);
          gl.glPopMatrix();

          // upper right
          gl.glBindTexture(GL2.GL_TEXTURE_2D, texHeightOutput);
          gl.glMatrixMode(GL2.GL_MODELVIEW);
          gl.glPushMatrix();
			                 
          gl.glTranslatef(0.5f, 0.5f, 0);
          gl.glScalef(0.5f, 0.5f, 1);
          gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);
          gl.glPopMatrix();
			    
          break;
        }
      }
    } else {
      // skip rendering this frame
      skipCount++;
    }
  }

  public void singleStep()                               { singleStep  = true;                 }
  public void enableAnimation(boolean enable)            { animate     = enable;               }
  public void enableSlowAnimation(boolean enable)        { slow        = enable;               }
  public void reset()                                    { reset       = true;                 }
  public void setRenderMode(int mode)                    { renderMode  = mode;                 }
    
  public void enableWireframe(boolean enable)            { wireframe   = enable;               }
  public void enableBorderWrapping(boolean enable)       { wrap        = enable;               }
    
  public void enableBoundaryApplication(boolean enable)  { applyInteriorBoundaries = enable;   }
  public void enableSpinningLogo(boolean enable)         { spinLogo    = enable;               }

  public void  setBlurDistance(float distance)           { blurDist    = distance;
                                                           mustUpdateBlurOffsets = true;       }
  public float getBlurDistance()                         { return blurDist;                    }

  public void  setBumpScale(float scale)                 { bumpScale   = scale;                }
  public float getBumpScale()                            { return bumpScale;                   }

  public void  setDropFrequency(float frequency)         { dropletFrequency = frequency;       }
  public float getDropFrequency()                        { return dropletFrequency;            }

  public static class Droplet {
    private float rX;
    private float rY;
    private float rScale;

    Droplet(float rX, float rY, float rScale) {
      this.rX     = rX;
      this.rY     = rY;
      this.rScale = rScale;
    }
    
    float rX()     { return rX;     }
    float rY()     { return rY;     }
    float rScale() { return rScale; }
  }

  public synchronized void addDroplet(Droplet drop) {
    droplets.add(drop);    
  }

  //----------------------------------------------------------------------
  // Internals only below this point
  //

  class Listener implements GLEventListener {

    public void init(GLAutoDrawable drawable) {
      GL2 gl = drawable.getGL().getGL2();

      initOpenGL(gl);
    }

    public void display(GLAutoDrawable drawable) {

      GL2 gl = drawable.getGL().getGL2();
      if (mustUpdateBlurOffsets) {
        updateBlurVertOffset(gl);
        mustUpdateBlurOffsets = false;
      }
      
      // Take a single step in the cellular automaton

      // Disable culling
      gl.glDisable(GL2.GL_CULL_FACE);

      if (reset) {
        reset = false;
        flipState = 0;
      }

      if (animate) {
        // Update the textures for one step of the simulation
        doSingleTimeStep(gl);
      } else if (singleStep) {
        doSingleTimeStep(gl);
        singleStep = false;
      }
	
      // Force rendering to pbuffer to complete
      gl.glFlush();

      if (slow && (slowDelay > 0) ) {
        try {
          Thread.sleep(slowDelay);
        } catch (InterruptedException e) {
        }
      }
    }

    public void reshape(GLAutoDrawable drawable, int x, int y, int width, int height) {}

    // Unused routines
    public void displayChanged(GLAutoDrawable drawable, boolean modeChanged, boolean deviceChanged) {}
  }

  // We need to load the initial texture file early to get the width
  // and height for the pbuffer
  private void loadInitialTexture(String initialMapFilename) {
    try {
      initialMapData = TextureIO.newTextureData(getClass().getClassLoader().getResourceAsStream(initialMapFilename),
                                                false,
                                                FileUtil.getFileSuffix(initialMapFilename));
    } catch (IOException e) {
      throw new GLException(e);
    }
    initialMapDimensions[0] = initialMapData.getWidth();
    initialMapDimensions[1] = initialMapData.getHeight();
  }

  private void initOpenGL(GL2 gl) {
    try {
      loadTextures(gl, tmpSpinFilename, tmpDropletFilename, tmpCubeMapFilenamePrefix, tmpCubeMapFilenameSuffix);
    } catch (IOException e) {
      throw new GLException(e);
    }
    tmpSpinFilename           = null;
    tmpDropletFilename        = null;
    tmpCubeMapFilenamePrefix  = null;
    tmpCubeMapFilenameSuffix  = null;

    gl.glMatrixMode(GL2.GL_MODELVIEW);
    gl.glLoadIdentity();
    gl.glMatrixMode(GL2.GL_PROJECTION);
    gl.glLoadIdentity();
    glu.gluOrtho2D(-1, 1, -1, 1);
    
    gl.glClearColor(0, 0, 0, 0);
    gl.glDisable(GL2.GL_LIGHTING);
    gl.glDisable(GL2.GL_DEPTH_TEST);
      
    createAndWriteUVOffsets(gl, initialMapDimensions[0], initialMapDimensions[1]);

    checkExtension(gl, "GL_vertex_program");
    checkExtension(gl, "GL_fragment_program");
    checkExtension(gl, "GL_multitexture");

    ///////////////////////////////////////////////////////////////////////////
    // UV Offset Vertex Program
    ///////////////////////////////////////////////////////////////////////////

    int[] tmpInt = new int[1];
    gl.glGenPrograms(1, tmpInt, 0);
    vertexProgramID = tmpInt[0];
    gl.glBindProgram(GL2.GL_VERTEX_PROGRAM, vertexProgramID);

    String programBuffer = 
"!!ARBvp1.0\n" +
"# Constant memory location declarations (must match those in Java sources)\n" +
"# CV_UV_OFFSET_TO_USE = 0\n" +
"\n" +
"# CV_UV_T0_NO_OFFSET  = 1\n" +
"# CV_UV_T0_TYPE1      = 2\n" +
"# CV_UV_T0_TYPE2      = 3\n" +
"# CV_UV_T0_TYPE3      = 4\n" +
"# CV_UV_T0_TYPE4      = 5\n" +
"\n" +
"# CV_UV_T1_NO_OFFSET  = 6\n" +
"# CV_UV_T1_TYPE1      = 7\n" +
"# CV_UV_T1_TYPE2      = 8\n" +
"# CV_UV_T1_TYPE3      = 9\n" +
"# CV_UV_T1_TYPE4      = 10\n" +
"\n" +
"# CV_UV_T2_NO_OFFSET  = 11\n" +
"# CV_UV_T2_TYPE1      = 12\n" +
"# CV_UV_T2_TYPE2      = 13\n" +
"# CV_UV_T2_TYPE3      = 14\n" +
"# CV_UV_T2_TYPE4      = 15\n" +
"\n" +
"# CV_UV_T3_NO_OFFSET  = 16\n" +
"# CV_UV_T3_TYPE1      = 17\n" +
"# CV_UV_T3_TYPE2      = 18\n" +
"# CV_UV_T3_TYPE3      = 19\n" +
"# CV_UV_T3_TYPE4      = 20\n" +
"\n" +
"# CV_CONSTS_1         = 21\n" +
"\n" +
"# Parameters\n" +
"PARAM mvp [4]       = { state.matrix.mvp };     # modelview projection matrix\n" +
"PARAM uvOffsetToUse = program.env[0];\n" +
"PARAM uvOffsets[20] = { program.env[1..20] };\n" +
"\n" +
"# Addresses\n" +
"ADDRESS addr;\n" +
"\n" +
"# Per vertex inputs\n" +
"ATTRIB iPos         = vertex.position;          #position\n" +
"\n" +
"# Outputs\n" +
"OUTPUT oPos         = result.position;          #position\n" +
"\n" +
"# Transform vertex-position to clip-space\n" +
"DP4 oPos.x, iPos, mvp[0];\n" +
"DP4 oPos.y, iPos, mvp[1];\n" +
"DP4 oPos.z, iPos, mvp[2];\n" +
"DP4 oPos.w, iPos, mvp[3];\n" +
"\n" +
"# Read which set of offsets to use\n" +
"ARL addr.x, uvOffsetToUse.x;\n" +
"\n" +
"#    c[CV_CONSTS_1] = c[28]\n" +
"#    x = 0\n" +
"#    y = 0.5\n" +
"#    z = 1\n" +
"#    w = 2.0f\n" +
"\n" +
"#    Put a scale factor into r0 so the sample points\n" +
"#    can be moved farther from the texel being written\n" +
"#    MOV R0, c[28].z;\n" +
"\n" +
"# Add the offsets to the input texture\n" +
"# coordinate, creating 4 sets of independent\n" +
"# texture coordinates.\n" +
"ADD result.texcoord[0], uvOffsets[addr.x     ], vertex.texcoord[0];\n" +
"ADD result.texcoord[1], uvOffsets[addr.x + 5 ], vertex.texcoord[0];\n" +
"ADD result.texcoord[2], uvOffsets[addr.x + 10], vertex.texcoord[0];\n" +
"ADD result.texcoord[3], uvOffsets[addr.x + 15], vertex.texcoord[0];\n" +
"\n" +
"END\n";

    // set up constants (not currently used in the vertex program, though)
    float[] rCVConsts = new float[] { 0, 0.5f, 1.0f, 2.0f };
    gl.glProgramEnvParameter4fv(GL2.GL_VERTEX_PROGRAM, CV_CONSTS_1, rCVConsts, 0);

    loadProgram(gl, GL2.GL_VERTEX_PROGRAM, programBuffer);

    ///////////////////////////////////////////////////////////////////////////
    // fragment program setup for equal weight combination of texels
    ///////////////////////////////////////////////////////////////////////////
    displayListIDs[CA_FRAGMENT_PROGRAM_EQ_WEIGHT_COMBINE] = gl.glGenLists(1);
    initEqWeightCombine_PostMult(gl, displayListIDs[CA_FRAGMENT_PROGRAM_EQ_WEIGHT_COMBINE]);

    ///////////////////////////////////////////////////////////////////////////
    // fragment program setup for computing force from neighbors (step 1)
    ///////////////////////////////////////////////////////////////////////////
    displayListIDs[CA_FRAGMENT_PROGRAM_NEIGHBOR_FORCE_CALC_1] = gl.glGenLists(1);
    initNeighborForceCalcStep1(gl, displayListIDs[CA_FRAGMENT_PROGRAM_NEIGHBOR_FORCE_CALC_1]);

    ///////////////////////////////////////////////////////////////////////////
    // fragment program setup for computing force from neighbors (step 2)
    ///////////////////////////////////////////////////////////////////////////
    displayListIDs[CA_FRAGMENT_PROGRAM_NEIGHBOR_FORCE_CALC_2] = gl.glGenLists(1);
    initNeighborForceCalcStep2(gl, displayListIDs[CA_FRAGMENT_PROGRAM_NEIGHBOR_FORCE_CALC_2]);

    ///////////////////////////////////////////////////////////////////////////
    // fragment program setup to apply force
    ///////////////////////////////////////////////////////////////////////////
    displayListIDs[CA_FRAGMENT_PROGRAM_APPLY_FORCE] = gl.glGenLists(1);
    initApplyForce(gl, displayListIDs[CA_FRAGMENT_PROGRAM_APPLY_FORCE]);

    ///////////////////////////////////////////////////////////////////////////
    // fragment program setup to apply velocity
    ///////////////////////////////////////////////////////////////////////////
    displayListIDs[CA_FRAGMENT_PROGRAM_APPLY_VELOCITY] = gl.glGenLists(1);
    initApplyVelocity(gl, displayListIDs[CA_FRAGMENT_PROGRAM_APPLY_VELOCITY]);

    ///////////////////////////////////////////////////////////////////////////
    // fragment program setup to create a normal map
    ///////////////////////////////////////////////////////////////////////////
    displayListIDs[CA_FRAGMENT_PROGRAM_CREATE_NORMAL_MAP] = gl.glGenLists(1);
    initCreateNormalMap(gl, displayListIDs[CA_FRAGMENT_PROGRAM_CREATE_NORMAL_MAP]);

    ///////////////////////////////////////////////////////////////////////////
    // fragment program setup for dot product reflection
    ///////////////////////////////////////////////////////////////////////////
    displayListIDs[CA_FRAGMENT_PROGRAM_REFLECT] = gl.glGenLists(1);
    initDotProductReflect(gl, displayListIDs[CA_FRAGMENT_PROGRAM_REFLECT]);

    ///////////////////////////////////////////////////////////////////////////
    // display list to render a single screen space quad.
    ///////////////////////////////////////////////////////////////////////////
    displayListIDs[CA_DRAW_SCREEN_QUAD] = gl.glGenLists(1);
    gl.glNewList(displayListIDs[CA_DRAW_SCREEN_QUAD], GL2.GL_COMPILE);
    gl.glColor4f(1, 1, 1, 1);
    gl.glBegin(GL2.GL_TRIANGLE_STRIP);
    gl.glTexCoord2f(0, 1); gl.glVertex2f(-1,  1);
    gl.glTexCoord2f(0, 0); gl.glVertex2f(-1, -1);
    gl.glTexCoord2f(1, 1); gl.glVertex2f( 1,  1);
    gl.glTexCoord2f(1, 0); gl.glVertex2f( 1, -1);
    gl.glEnd();
    gl.glEndList();
  }

  private void checkExtension(GL gl, String extensionName) {
    if (!gl.isExtensionAvailable(extensionName)) {
      throw new GLException("Unable to initialize " + extensionName + " OpenGL extension");
    }
  }

  private void doSingleTimeStep(GL2 gl) {
    int temp;

    // Swap texture source & target indices & pointers
    //  0 = start from initial loaded texture
    //  1/2 = flip flop back and forth between targets & sources

    switch (flipState) {
    case 0:
      texHeightInput    = dynamicTextures[CA_TEXTURE_HEIGHT_SOURCE].getTextureObject();    // initial height map.
      texHeightOutput   = dynamicTextures[CA_TEXTURE_HEIGHT_TARGET].getTextureObject();    // next height map.

      texVelocityInput  = dynamicTextures[CA_TEXTURE_VELOCITY_SOURCE].getTextureObject();  // initial velocity.
      texVelocityOutput = dynamicTextures[CA_TEXTURE_VELOCITY_TARGET].getTextureObject();  // next velocity.

      // Clear initial velocity texture to 0x80 == gray
      gl.glClearColor(0.5f, 0.5f, 0.5f, 1.0f);
      gl.glClear(GL2.GL_COLOR_BUFFER_BIT);

      // Now we need to copy the resulting pixels into the intermediate force field texture
      gl.glActiveTexture(GL2.GL_TEXTURE0);
      gl.glBindTexture(GL2.GL_TEXTURE_2D, texVelocityInput);

      // use CopyTexSubImage for speed (even though we copy all of it) since we pre-allocated the texture
      gl.glCopyTexSubImage2D(GL2.GL_TEXTURE_2D, 0, 0, 0, 0, 0, initialMapDimensions[0], initialMapDimensions[1]);

      break;  
        
    case 1:
      temp              = texHeightInput;
      texHeightInput    = texHeightOutput;
      texHeightOutput   = temp;

      temp              = texVelocityInput;
      texVelocityInput  = texVelocityOutput;
      texVelocityOutput = temp;

      break;

    case 2:
      temp              = texHeightInput;
      texHeightInput    = texHeightOutput;
      texHeightOutput   = temp;

      temp              = texVelocityInput;
      texVelocityInput  = texVelocityOutput;
      texVelocityOutput = temp;
      break;
    }
	
    // even if wireframe mode, render to texture as solid
    gl.glPolygonMode(GL2.GL_FRONT_AND_BACK, GL2.GL_FILL);
	
    /////////////////////////////////////////////////////////////
    //  Render first 3 components of force from three neighbors
    //  Offsets selected are 1 center texel for center height
    //    and 3 of the 4 nearest neighbors.  Texture selected
    //    is same for all stages as we're turning height difference
    //    of nearest neightbor texels into a force value.

    gl.glCallList(displayListIDs[CA_FRAGMENT_PROGRAM_NEIGHBOR_FORCE_CALC_1]);

    // set current source texture for stage 0 texture
    for (int i = 0; i < 4; i++)
      {
        gl.glActiveTexture(GL2.GL_TEXTURE0 + i);
        gl.glBindTexture(GL2.GL_TEXTURE_2D, texHeightInput);
        gl.glEnable(GL2.GL_TEXTURE_2D);
      }

    int wrapMode = wrap ? GL2.GL_REPEAT : GL2.GL_CLAMP_TO_EDGE;
    gl.glTexParameteri(GL2.GL_TEXTURE_2D, GL2.GL_TEXTURE_WRAP_S, wrapMode);
    gl.glTexParameteri(GL2.GL_TEXTURE_2D, GL2.GL_TEXTURE_WRAP_T, wrapMode);

    // disable blending
    gl.glDisable(GL2.GL_BLEND);

    // render using offset 1 (type 1 -- center + 3 of 4 nearest neighbors).
    gl.glProgramEnvParameter4f(GL2.GL_VERTEX_PROGRAM, CV_UV_OFFSET_TO_USE, 1, 0, 0, 0);

    // bind the vertex program to be used for this step and the next one.
    gl.glBindProgram(GL2.GL_VERTEX_PROGRAM, vertexProgramID);
    gl.glEnable(GL2.GL_VERTEX_PROGRAM);

    // render a screen quad. with texture coords doing difference of nearby texels for force calc.
    gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);

    gl.glDisable(GL2.GL_FRAGMENT_PROGRAM);

    // Now we need to copy the resulting pixels into the intermediate force field texture
    gl.glActiveTexture(GL2.GL_TEXTURE2);
    dynamicTextures[CA_TEXTURE_FORCE_INTERMEDIATE].bind();    

    // use CopyTexSubImage for speed (even though we copy all of it) since we pre-allocated the texture
    gl.glCopyTexSubImage2D(GL2.GL_TEXTURE_2D, 0, 0, 0, 0, 0, initialMapDimensions[0], initialMapDimensions[1]);

    ////////////////////////////////////////////////////////////////
    // Now add in last component of force for the 4th neighbor
    //  that we didn't have enough texture lookups to do in the 
    //  first pass

    gl.glCallList(displayListIDs[CA_FRAGMENT_PROGRAM_NEIGHBOR_FORCE_CALC_2]);
    
    // Cannot use additive blending as the force contribution might
    //   be negative and would have to subtract from the dest.
    // We must instead use an additional texture as target and read
    //   the previous partial 3-neighbor result into the pixel shader
    //   for possible subtraction

    // Alphablend must be false

    //; t0 = center  (same as last phase)
    //; t1 = 2nd axis final point (same as last phase)
    //; t2 = previous partial result texture sampled at center (result of last phase copied to texture)
    //; t3 = not used (disable now)

    gl.glTexParameterf(GL2.GL_TEXTURE_2D, GL2.GL_TEXTURE_WRAP_S, wrapMode);
    gl.glTexParameterf(GL2.GL_TEXTURE_2D, GL2.GL_TEXTURE_WRAP_T, wrapMode);

    gl.glActiveTexture(GL2.GL_TEXTURE3);
    gl.glDisable(GL2.GL_TEXTURE_2D);

    // vertex program already bound.
    // render using offset 2 (type 2 -- final nearest neighbor plus center of previous result).
    gl.glProgramEnvParameter4f(GL2.GL_VERTEX_PROGRAM, CV_UV_OFFSET_TO_USE, 2, 0, 0, 0);

    // render a screen quad
    gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);

    gl.glDisable(GL2.GL_FRAGMENT_PROGRAM);

    // Now we need to copy the resulting pixels into the intermediate force field texture
    gl.glActiveTexture(GL2.GL_TEXTURE1);
    dynamicTextures[CA_TEXTURE_FORCE_TARGET].bind();

    // use CopyTexSubImage for speed (even though we copy all of it) since we pre-allocated the texture
    gl.glCopyTexSubImage2D(GL2.GL_TEXTURE_2D, 0, 0, 0, 0, 0, initialMapDimensions[0], initialMapDimensions[1]);

    /////////////////////////////////////////////////////////////////
    // Apply the force with a scale factor to reduce it's magnitude.
    // Add this to the current texture representing the water height.
    
    gl.glCallList(displayListIDs[CA_FRAGMENT_PROGRAM_APPLY_FORCE]);

    // use offsets of zero
    gl.glProgramEnvParameter4f(GL2.GL_VERTEX_PROGRAM, CV_UV_OFFSET_TO_USE, 0, 0, 0, 0);

    // bind the vertex program to be used for this step and the next one.

    gl.glActiveTexture(GL2.GL_TEXTURE0);
    gl.glBindTexture(GL2.GL_TEXTURE_2D, texVelocityInput);
    gl.glActiveTexture(GL2.GL_TEXTURE1);
    dynamicTextures[CA_TEXTURE_FORCE_TARGET].bind();    
    gl.glActiveTexture(GL2.GL_TEXTURE2);
    gl.glDisable(GL2.GL_TEXTURE_2D);
    gl.glActiveTexture(GL2.GL_TEXTURE3);
    gl.glDisable(GL2.GL_TEXTURE_2D);

    // Draw the quad to add in force.
    gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);

    gl.glDisable(GL2.GL_FRAGMENT_PROGRAM);

    ///////////////////////////////////////////////////////////////////
    // With velocity texture selected, render new excitation droplets
    //   at random freq.

    float randomFrequency = (float) Math.random();

    if (dropletFrequency > randomFrequency) {
      // a drop falls - decide where
      Droplet drop = new Droplet(2 * ((float)Math.random() - 0.5f),
                                 2 * ((float)Math.random() - 0.5f),
                                 0.02f +  0.1f * ((float)Math.random()));
      addDroplet(drop);
    }

    //  Now draw the droplets:
    if (!droplets.isEmpty()) {
      drawDroplets(gl);
      droplets.clear();
    }

    // Now we need to copy the resulting pixels into the velocity texture
    gl.glActiveTexture(GL2.GL_TEXTURE1);
    gl.glBindTexture(GL2.GL_TEXTURE_2D, texVelocityOutput);

    // use CopyTexSubImage for speed (even though we copy all of it) since we pre-allocated the texture
    gl.glCopyTexSubImage2D(GL2.GL_TEXTURE_2D, 0, 0, 0, 0, 0, initialMapDimensions[0], initialMapDimensions[1]);

    //////////////////////////////////////////////////////////////////////
    // Apply velocity to position
    gl.glCallList(displayListIDs[CA_FRAGMENT_PROGRAM_APPLY_VELOCITY]);
    gl.glEnable(GL2.GL_VERTEX_PROGRAM);

    gl.glActiveTexture(GL2.GL_TEXTURE0);
    gl.glBindTexture(GL2.GL_TEXTURE_2D, texHeightInput);
    gl.glActiveTexture(GL2.GL_TEXTURE1); // velocity output already bound
    gl.glEnable(GL2.GL_TEXTURE_2D);

    // use offsets of zero
    gl.glProgramEnvParameter4f(GL2.GL_VERTEX_PROGRAM, CV_UV_OFFSET_TO_USE, 0, 0, 0, 0);

    // Draw the quad to add in force.
    gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);

    gl.glDisable(GL2.GL_FRAGMENT_PROGRAM);

    // Now we need to copy the resulting pixels into the input height texture
    gl.glActiveTexture(GL2.GL_TEXTURE0);
    gl.glBindTexture(GL2.GL_TEXTURE_2D, texHeightInput);
    
    // use CopyTexSubImage for speed (even though we copy all of it) since we pre-allocated the texture
    gl.glCopyTexSubImage2D(GL2.GL_TEXTURE_2D, 0, 0, 0, 0, 0, initialMapDimensions[0], initialMapDimensions[1]);

    ///////////////////////////////////////////////////////////////////
    //  blur positions to smooth noise & generaly dampen things
    //  degree of blur is controlled by magnitude of 4 neighbor texel
    //   offsets with bilinear on
    
    for (int i = 1; i < 4; i++) {
      gl.glActiveTexture(GL2.GL_TEXTURE0 + i);
      gl.glBindTexture(GL2.GL_TEXTURE_2D, texHeightInput);
      gl.glEnable(GL2.GL_TEXTURE_2D);
    }

    // use offsets of 3
    gl.glProgramEnvParameter4f(GL2.GL_VERTEX_PROGRAM, CV_UV_OFFSET_TO_USE, 3, 0, 0, 0);

    gl.glCallList(displayListIDs[CA_FRAGMENT_PROGRAM_EQ_WEIGHT_COMBINE]);

    gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);
    gl.glDisable(GL2.GL_FRAGMENT_PROGRAM);

    // Draw the logo in the water.
    if (applyInteriorBoundaries) {
      gl.glDisable(GL2.GL_VERTEX_PROGRAM);
      drawInteriorBoundaryObjects(gl);
    }

    // Now we need to copy the resulting pixels into the velocity texture
    gl.glActiveTexture(GL2.GL_TEXTURE0);
    gl.glBindTexture(GL2.GL_TEXTURE_2D, texHeightOutput);

    // use CopyTexSubImage for speed (even though we copy all of it) since we pre-allocated the texture
    gl.glCopyTexSubImage2D(GL2.GL_TEXTURE_2D, 0, 0, 0, 0, 0, initialMapDimensions[0], initialMapDimensions[1]);
      
    ///////////////////////////////////////////////////////////////////
    // If selected, create a normal map from the height
      
    if (createNormalMap) {
      createNormalMap(gl);
    }
      
    ///////////////////////////////////////////////////////////
    // Flip the state variable for the next round of rendering
    switch (flipState) {
    case 0:
      flipState = 1;
      break;
    case 1:
      flipState = 2;
      break;
    case 2:
      flipState = 1;
      break;
    }
  }

  private void createNormalMap(GL2 gl) {
    // use the height output on all four texture stages
    for (int i = 0; i < 4; i++) {
      gl.glActiveTexture(GL2.GL_TEXTURE0 + i);
      gl.glBindTexture(GL2.GL_TEXTURE_2D, texHeightOutput);
      gl.glEnable(GL2.GL_TEXTURE_2D);
    }

    // Set constants for red & green scale factors (also essential color masks)
    // Red mask first
    float[] pixMasks = new float[] { normalSTScale, 0.0f, 0.0f, 0.0f };

    gl.glProgramEnvParameter4fv(GL2.GL_FRAGMENT_PROGRAM, 0, pixMasks, 0);

    // Now green mask & scale:
    pixMasks[0] = 0.0f;
    pixMasks[1] = normalSTScale;
    gl.glProgramEnvParameter4fv(GL2.GL_FRAGMENT_PROGRAM, 1, pixMasks, 0);

    gl.glCallList(displayListIDs[CA_FRAGMENT_PROGRAM_CREATE_NORMAL_MAP]);

    // set vp offsets to nearest neighbors
    gl.glProgramEnvParameter4f(GL2.GL_VERTEX_PROGRAM, CV_UV_OFFSET_TO_USE, 4, 0, 0, 0);
    gl.glEnable(GL2.GL_VERTEX_PROGRAM);
    
    gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);

    gl.glDisable(GL2.GL_FRAGMENT_PROGRAM);

    // Now we need to copy the resulting pixels into the normal map
    gl.glActiveTexture(GL2.GL_TEXTURE0);
    dynamicTextures[CA_TEXTURE_NORMAL_MAP].bind();
    
    // use CopyTexSubImage for speed (even though we copy all of it) since we pre-allocated the texture
    gl.glCopyTexSubImage2D(GL2.GL_TEXTURE_2D, 0, 0, 0, 0, 0, initialMapDimensions[0], initialMapDimensions[1]);
  }

  private void drawInteriorBoundaryObjects(GL2 gl) {
    
    gl.glActiveTexture(GL2.GL_TEXTURE0);
    initialMapTex.bind();
    initialMapTex.enable();

    gl.glEnable(GL2.GL_ALPHA_TEST);

    // disable other texture units.
    for (int i = 1; i < 4; i++) {
      gl.glActiveTexture(GL2.GL_TEXTURE0 + i);
      gl.glDisable(GL2.GL_TEXTURE_2D);
    }
    
    gl.glBlendFunc(GL2.GL_SRC_ALPHA, GL2.GL_ONE_MINUS_SRC_ALPHA);
    gl.glEnable(GL2.GL_BLEND);

    gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);

    if (spinLogo) {
      gl.glActiveTexture(GL2.GL_TEXTURE0);
      spinTex.bind();
      gl.glMatrixMode(GL2.GL_MODELVIEW);
      gl.glPushMatrix();
      gl.glRotatef(angle, 0, 0, 1);
      angle += 1;

      gl.glCallList(displayListIDs[CA_DRAW_SCREEN_QUAD]);

      gl.glPopMatrix();
    }

    gl.glDisable(GL2.GL_ALPHA_TEST);
    gl.glDisable(GL2.GL_BLEND);
  }

  private void loadTextures(GL gl,
                            String spinFilename,
                            String dropletFilename,
                            String cubeMapFilenamePrefix,
                            String cubeMapFilenameSuffix) throws IOException {
    if (initialMapData == null) {
      throw new GLException("Must call loadInitialTexture ahead of time");
    }

    initialMapTex = TextureIO.newTexture(initialMapData);
    spinTex       = TextureIO.newTexture(getClass().getClassLoader().getResourceAsStream(spinFilename), false,
                                         FileUtil.getFileSuffix(spinFilename));
    dropletTex    = TextureIO.newTexture(getClass().getClassLoader().getResourceAsStream(dropletFilename), false,
                                         FileUtil.getFileSuffix(dropletFilename));

    // load the cubemap texture
    cubemap = Cubemap.loadFromStreams(getClass().getClassLoader(),
                                      cubeMapFilenamePrefix,
                                      cubeMapFilenameSuffix,
                                      true);

    // now create dummy intermediate textures from the initial map texture
    for (int i = 0; i < CA_NUM_DYNAMIC_TEXTURES; i++) {
      dynamicTextures[i] = TextureIO.newTexture(initialMapData);
    }

    initialMapData = null;

    texHeightInput    = initialMapTex.getTextureObject();                               // initial height map.
    texHeightOutput   = dynamicTextures[CA_TEXTURE_HEIGHT_TARGET].getTextureObject();   // next height map.
    
    texVelocityInput  = dynamicTextures[CA_TEXTURE_VELOCITY_SOURCE].getTextureObject(); // initial velocity.
    texVelocityOutput = dynamicTextures[CA_TEXTURE_VELOCITY_TARGET].getTextureObject(); // next velocity.
  }

  private void createAndWriteUVOffsets(GL2 gl, int width, int height) {
    // This sets vertex shader constants used to displace the
    //  source texture over several additive samples.  This is 
    //  used to accumulate neighboring texel information that we
    //  need to run the game - the 8 surrounding texels, and the 
    //  single source texel which will either spawn or die in the 
    //  next generation.
    // Label the texels as follows, for a source texel "e" that
    //  we want to compute for the next generation:
    //
    //          abc
    //          def
    //          ghi:

    // first the easy one: no offsets for sampling center
    //  occupied or unoccupied
    // Use index offset value 0.0 to access these in the 
    //  vertex shader.
    
    perTexelWidth  = 1.0f / width;
    perTexelHeight = 1.0f / height;

    // Offset set 0 : center texel sampling
    float[] noOffsetX = new float[] { 0, 0, 0, 0 };
    float[] noOffsetY = new float[] { 0, 0, 0, 0 };

    // Offset set 1:  For use with neighbor force pixel shader 1
    //  samples center with 0, +u, -u, and +v,
    //  ie the 'e','d', 'f', and 'h' texels
    float dist = 1.5f;
    float[] type1OffsetX = new float[] { 0.0f, -dist * perTexelWidth,  dist * perTexelWidth,   dist * perTexelWidth  };
    float[] type1OffsetY = new float[] { 0.0f,  dist * perTexelHeight, dist * perTexelHeight, -dist * perTexelHeight };

    // Offset set 2:  for use with neighbor force pixel shader 2
    //  samples center with 0, and -v texels 
    //  ie the 'e' and 'b' texels
    // This completes a pattern of sampling center texel and it's
    //   4 nearest neighbors to run the height-based water simulation
    // 3rd must be 0 0 to sample texel center from partial result
    //   texture.

    float[] type2OffsetX = new float[] { 0.0f, -dist * perTexelWidth,  0.0f, 0.0f   };
    float[] type2OffsetY = new float[] { 0.0f, -dist * perTexelHeight, 0.0f, 0.0f   };
        
    // type 3 offsets
    updateBlurVertOffset(gl);

    /////////////////////////////////////////////////////////////
    // Nearest neighbor offsets:

    float[] type4OffsetX = new float[] { -perTexelWidth,   perTexelWidth,   0.0f,              0.0f   };
    float[] type4OffsetY = new float[] { 0.0f,             0.0f,            -perTexelHeight,   perTexelHeight };

    // write all these offsets to constant memory
    for (int i = 0; i < 4; ++i) {
      float noOffset[]    = { noOffsetX[i],    noOffsetY[i],    0.0f, 0.0f };
      float type1Offset[] = { type1OffsetX[i], type1OffsetY[i], 0.0f, 0.0f };
      float type2Offset[] = { type2OffsetX[i], type2OffsetY[i], 0.0f, 0.0f };
      float type4Offset[] = { type4OffsetX[i], type4OffsetY[i], 0.0f, 0.0f };

      gl.glProgramEnvParameter4fv(GL2.GL_VERTEX_PROGRAM, CV_UV_T0_NO_OFFSET + 5 * i, noOffset, 0);
      gl.glProgramEnvParameter4fv(GL2.GL_VERTEX_PROGRAM, CV_UV_T0_TYPE1     + 5 * i, type1Offset, 0);
      gl.glProgramEnvParameter4fv(GL2.GL_VERTEX_PROGRAM, CV_UV_T0_TYPE2     + 5 * i, type2Offset, 0);
      gl.glProgramEnvParameter4fv(GL2.GL_VERTEX_PROGRAM, CV_UV_T0_TYPE4     + 5 * i, type4Offset, 0);
    }
  }

  private void updateBlurVertOffset(GL2 gl) {
    float[] type3OffsetX = new float[] { -perTexelWidth * 0.5f, 
                                         perTexelWidth, 
                                         perTexelWidth * 0.5f, 
                                         -perTexelWidth 
    };
    float[] type3OffsetY = new float[] { perTexelHeight,
                                         perTexelHeight * 0.5f,
                                         -perTexelHeight,
                                         -perTexelHeight * 0.5f 
    };
    float[] offsets = new float[] { 0, 0, 0, 0 };

    for (int i = 0; i < 4; ++i) {
      offsets[0] = blurDist * ( type3OffsetX[i]);
      offsets[1] = blurDist * ( type3OffsetY[i]);
      gl.glProgramEnvParameter4fv(GL2.GL_VERTEX_PROGRAM, CV_UV_T0_TYPE3 + 5 * i, offsets, 0);
    }
  }

  private synchronized void drawDroplets(GL2 gl) {
    gl.glDisable(GL2.GL_FRAGMENT_PROGRAM);
    gl.glDisable(GL2.GL_VERTEX_PROGRAM);

    gl.glActiveTexture(GL2.GL_TEXTURE0);
    dropletTex.bind();
    dropletTex.enable();

    gl.glActiveTexture(GL2.GL_TEXTURE1);
    gl.glDisable(GL2.GL_TEXTURE_2D);

    gl.glBlendFunc(GL2.GL_ONE, GL2.GL_ONE);
    gl.glEnable(GL2.GL_BLEND);

    gl.glBegin(GL2.GL_QUADS);
    gl.glColor4f(1, 1, 1, 1);
    for (Iterator iter = droplets.iterator(); iter.hasNext(); ) {
      Droplet droplet = (Droplet) iter.next();
      // coords in [-1,1] range

      // Draw a single quad to the texture render target
      // The quad is textured with the initial droplet texture, and
      //   covers some small portion of the render target
      // Draw the droplet
       
      gl.glTexCoord2f(0, 0); gl.glVertex2f(droplet.rX() - droplet.rScale(), droplet.rY() - droplet.rScale());
      gl.glTexCoord2f(1, 0); gl.glVertex2f(droplet.rX() + droplet.rScale(), droplet.rY() - droplet.rScale());
      gl.glTexCoord2f(1, 1); gl.glVertex2f(droplet.rX() + droplet.rScale(), droplet.rY() + droplet.rScale());
      gl.glTexCoord2f(0, 1); gl.glVertex2f(droplet.rX() - droplet.rScale(), droplet.rY() + droplet.rScale());          
    }
    gl.glEnd();

    gl.glDisable(GL2.GL_BLEND);
  }

  //----------------------------------------------------------------------
  // Inlined register combiner and texture shader programs
  // (don't want to port nvparse as it's a dead-end; we'll focus on Cg instead)

  private void initEqWeightCombine_PostMult(GL2 gl, int displayListID) {
    // Take samples of all four texture inputs and average them,
    // adding on a bias
    //
    // Original register combiner program:
    //
    // Stage 0
    // rgb
    // {
    //   discard = half_bias(tex0);
    //   discard = half_bias(tex1);
    //   spare0 = sum();
    //   scale_by_one_half();
    // }
    // Stage 1
    // rgb
    // {
    //   discard = half_bias(tex2);
    //   discard = half_bias(tex3);
    //   spare1 = sum();
    //   scale_by_one_half();
    // }
    // Stage 2
    // rgb
    // {
    //   discard = spare0;
    //   discard = spare1;
    //   spare0 = sum();
    //   scale_by_one_half();
    // }
    // Stage 3
    // rgb
    // {
    //   discard = const0;
    //   discard = spare0;
    //   spare0 = sum();
    // }

    float[] const0 = new float[] { 0.5f, 0.5f, 0.5f, 1.0f };

    int[] tmpInt = new int[1];
    gl.glGenPrograms(1, tmpInt, 0);
    int fragProg = tmpInt[0];
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);

    String program =
"!!ARBfp1.0\n" +
"PARAM const0  = program.env[0];\n" +
"PARAM oneQtr  = { 0.25, 0.25, 0.25, 0.25 };\n" +
"PARAM two     = { 2.0, 2.0, 2.0, 2.0 };\n" +
"TEMP texSamp0, texSamp1, texSamp2, texSamp3;\n" +
"TEMP spare0, spare1;\n" +
"\n" +
"TEX texSamp0, fragment.texcoord[0], texture[0], 2D;\n" +
"TEX texSamp1, fragment.texcoord[1], texture[1], 2D;\n" +
"TEX texSamp2, fragment.texcoord[2], texture[2], 2D;\n" +
"TEX texSamp3, fragment.texcoord[3], texture[3], 2D;\n" +
"ADD spare0, texSamp0, texSamp1;\n" +
"ADD spare1, texSamp2, texSamp3;\n" +
"ADD spare0, spare0, spare1;\n" +
"SUB spare0, spare0, two;\n" +
"MAD result.color, oneQtr, spare0, const0;\n" +
"\n" +
"END\n";

    loadProgram(gl, GL2.GL_FRAGMENT_PROGRAM, program);

    gl.glNewList(displayListID, GL2.GL_COMPILE);
    gl.glProgramEnvParameter4fv(GL2.GL_FRAGMENT_PROGRAM, 0, const0, 0);
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);
    gl.glEnable(GL2.GL_FRAGMENT_PROGRAM);
    gl.glEndList();
  }

  private void initNeighborForceCalcStep1(GL2 gl, int displayListID) {
    // Step one in the nearest-neighbor force calculation for height-based water
    // simulation.  NeighborForceCalc2 is the second step.
    //
    // This step takes the center point and three neighboring points, and computes
    // the texel difference as the "force" acting to pull the center texel.
    // 
    // The amount to which the computed force is applied to the texel is controlled
    // in a separate shader.

    //  get colors from all 4 texture stages
    //  tex0 = center texel
    //  tex1 = 1st neighbor
    //  tex2 = 2nd neighbor - same axis as 1st neighbor point
    //       so force for that axis == t1 - t0 + t2 - t0
    //  tex3 = 3rd neighbor on other axis

    // Original register combiner program:
    //
    // Stage 0
    // rgb
    // {
    //   //s0 = t1 - t0;
    //   discard = -tex0;
    //   discard = tex1;
    //   spare0 = sum();  
    // }
    // Stage 1
    // rgb
    // {
    //   //s1 = t2 - t0;
    //   discard = -tex0;
    //   discard = tex2;
    //   spare1 = sum();  
    // }
    // Stage 2
    // // 'force' for 1st axis
    // rgb 
    // {
    //   //s0 = s0 + s1 = t1 - t0 + t2 - t0;
    //   discard = spare0;
    //   discard = spare1;
    //   spare0 = sum();  
    // }
    // Stage 3
    // // one more point for 2nd axis
    // rgb
    // {
    //   //s1 = t3 - t0;
    //   discard = -tex0;
    //   discard = tex3;
    //   spare1 = sum();  
    // }
    // Stage 4
    // rgb
    // {
    //   //s0 = s0 + s1 = t3 - t0 + t2 - t0 + t1 - t0;
    //   discard = spare0;
    //   discard = spare1;
    //   spare0 = sum();  
    // }
    // Stage 5
    // // Now add in a force to gently pull the center texel's 
    // //  value to 0.5.  The strength of this is controlled by
    // //  the PCN_EQ_REST_FAC  - restoration factor
    // // Without this, the simulation will fade to zero or fly
    // //  away to saturate at 1.0
    // rgb 
    // {
    //   //s1 = 0.5 - t0;  
    //   discard = -tex0;
    //   discard = const0;
    //   spare1 = sum();  
    // }
    // Stage 6
    // {
    //   rgb
    //   {
    //     discard = spare1 * const0;
    //     discard = spare0;
    //     spare0 = sum();
    //   }
    // }
    // Stage 7
    // rgb
    // {
    //   discard = spare0;
    //   discard = const0;
    //   spare0 = sum();
    // }

    float[] const0 = new float[] { 0.5f, 0.5f, 0.5f, 1.0f };

    int[] tmpInt = new int[1];
    gl.glGenPrograms(1, tmpInt, 0);
    int fragProg = tmpInt[0];
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);

    String program =
"!!ARBfp1.0\n" +
"PARAM const0 = program.env[0];\n" +
"PARAM three                 = {  3,     3,     3,    1.0 };\n" +
"TEMP texSamp0, texSamp1, texSamp2, texSamp3;\n" +
"TEMP spare0, spare1;\n" +
"\n" +
"TEX texSamp0, fragment.texcoord[0], texture[0], 2D;\n" +
"TEX texSamp1, fragment.texcoord[1], texture[1], 2D;\n" +
"TEX texSamp2, fragment.texcoord[2], texture[2], 2D;\n" +
"TEX texSamp3, fragment.texcoord[3], texture[3], 2D;\n" +
"ADD spare0, texSamp1, texSamp2;\n" +
"MAD spare1, const0, const0, const0;\n" +
"ADD spare0, texSamp3, spare0;\n" +
"ADD spare0, spare1, spare0;\n" +
"ADD spare1, three, const0;\n" +
"MAD result.color, -spare1, texSamp0, spare0;\n" +

// Faster version which hardcodes in value of const0:
//"ADD spare0, texSamp1, texSamp2;\n" +
//"ADD spare1, texSamp3, pointSevenFive;\n" +
//"ADD spare0, spare0, spare1;\n" +
//"MAD result.color, minusThreePointFive, texSamp0, spare0;\n" +

// Straightforward port:
//"SUB spare0, texSamp1, texSamp0;\n" +
//"SUB spare1, texSamp2, texSamp0;\n" +
//"ADD spare0, spare0, spare1;\n" +
//"SUB spare1, texSamp3, texSamp0;\n" +
//"ADD spare0, spare0, spare1;\n" +
//"SUB spare1, const0, texSamp0;\n" +
//"MAD spare0, const0, spare1, spare0;\n" +
//"ADD result.color, spare0, const0;\n" +

"\n" +
"END\n";

    loadProgram(gl, GL2.GL_FRAGMENT_PROGRAM, program);

    gl.glNewList(displayListID, GL2.GL_COMPILE);
    gl.glProgramEnvParameter4fv(GL2.GL_FRAGMENT_PROGRAM, 0, const0, 0);
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);
    gl.glEnable(GL2.GL_FRAGMENT_PROGRAM);
    gl.glEndList();
  }

  private void initNeighborForceCalcStep2(GL2 gl, int displayListID) {
    // 2nd step of force calc for render-to-texture
    // water simulation.
    //
    // Adds the 4th & final neighbor point to the 
    // force calc..
    //
    // Bias and scale the values so 0 force is 0.5, 
    // full negative force is 0.0, and full pos is
    // 1.0
    //
    // tex0    Center texel
    // tex1    2nd axis neighbor point
    // tex2    previous partial force amount
    // Result from t1 - t0 is added to this t2
    //  partial result & output

    // Original register combiner program:
    //
    // Stage 0
    // last element of neighbor force
    // rgb
    // {
    //   discard = -tex0;
    //   discard = tex1;
    //   spare0 = sum();
    // }
    // Stage 1
    // add with previous partial force amount
    // rgb
    // {
    //   discard = spare0;
    //   discard = tex2;
    //   spare0 = sum();
    // }

    int[] tmpInt = new int[1];
    gl.glGenPrograms(1, tmpInt, 0);
    int fragProg = tmpInt[0];
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);

    String program =
"!!ARBfp1.0\n" +
"PARAM const0 = program.env[0];\n" +
"TEMP texSamp0, texSamp1, texSamp2;\n" +
"TEMP spare0;\n" +
"\n" +
"TEX texSamp0, fragment.texcoord[0], texture[0], 2D;\n" +
"TEX texSamp1, fragment.texcoord[1], texture[1], 2D;\n" +
"TEX texSamp2, fragment.texcoord[2], texture[2], 2D;\n" +
"SUB spare0, texSamp1, texSamp0;\n" +
"ADD result.color, spare0, texSamp2;\n" +
"\n" +
"END\n";

    loadProgram(gl, GL2.GL_FRAGMENT_PROGRAM, program);

    gl.glNewList(displayListID, GL2.GL_COMPILE);
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);
    gl.glEnable(GL2.GL_FRAGMENT_PROGRAM);
    gl.glEndList();
  }

  private void initApplyForce(GL2 gl, int displayListID) {
    // This shader samples t1, biases its value to a signed number, and applies this
    // value multiplied by a scale factor to the t0 sample.
    //
    // This is used to apply a "force" texture value to a "velocity" state texture
    // for nearest-neighbor height-based water simulations.  The output pixel is
    // the new "velocity" value to replace the t0 sample in rendering to a new 
    // texture which will replace the texture selected into t0.
    //
    // A nearly identical shader using a different scaling constant is used to
    // apply the "velocity" value to a "height" texture at each texel.
    //
    // t1 comes in the range [0,1] but needs to hold signed values, so a value of
    // 0.5 in t1 represents zero force.  This is biased to a signed value in 
    // computing the new velocity.
    //
    // tex0 = previous velocity
    // tex1 = force
    //
    // Bias the force so that 0.5 input = no change in t0 value
    //  and 0.0 input means -0.5 * scale change in t0 value
    //
    // New velocity = force * scale + previous velocity

    // Original register combiner program:
    //
    // Stage 0
    // rgb
    // {
    //   discard = expand(tex1) * const0;
    //   discard = expand(tex0);
    //   spare0 = sum();
    //   scale_by_one_half();
    // }
    // Stage 1
    // rgb
    // {
    //   discard = spare0;
    //   discard = const1;
    //   spare0 = sum();
    // }

    float[] const0 = new float[] { 0.25f, 0.25f, 0.25f, 1.0f };
    float[] const1 = new float[] { 0.5f,  0.5f,  0.5f,  1.0f };

    int[] tmpInt = new int[1];
    gl.glGenPrograms(1, tmpInt, 0);
    int fragProg = tmpInt[0];
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);

    String program =
"!!ARBfp1.0\n" +
"PARAM const0 = program.env[0];\n" +
"PARAM const1 = program.env[1];\n" +
"PARAM one     = { 1.0, 1.0, 1.0, 0.0 };\n" +
"PARAM oneHalf = { 0.5, 0.5, 0.5, 1.0 };\n" +
"PARAM two     = { 2.0, 2.0, 2.0, 1.0 };\n" +
"TEMP texSamp0, texSamp1;\n" +
"TEMP spare0, spare1;\n" +
"\n" +
"TEX texSamp0, fragment.texcoord[0], texture[0], 2D;\n" +
"TEX texSamp1, fragment.texcoord[1], texture[1], 2D;\n" +
"MAD spare0, two, texSamp1, -one;\n" +
"MAD spare1, two, texSamp0, -one;\n" +
"MAD spare0, spare0, const0, spare1;\n" +
"MAD result.color, oneHalf, spare0, const1;\n" +
"\n" +
"END\n";

    loadProgram(gl, GL2.GL_FRAGMENT_PROGRAM, program);

    gl.glNewList(displayListID, GL2.GL_COMPILE);
    gl.glProgramEnvParameter4fv(GL2.GL_FRAGMENT_PROGRAM, 0, const0, 0);
    gl.glProgramEnvParameter4fv(GL2.GL_FRAGMENT_PROGRAM, 1, const1, 0);
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);
    gl.glEnable(GL2.GL_FRAGMENT_PROGRAM);
    gl.glEndList();
  }

  private void initApplyVelocity(GL2 gl, int displayListID) {
    // This shader samples t1, biases its value to a signed number, and applies this
    // value multiplied by a scale factor to the t0 sample.
    //
    // This is used to apply a "velocity" texture value to a "height" state texture
    // for nearest-neighbor height-based water simulations.  The output pixel is
    // the new "height" value to replace the t0 sample in rendering to a new 
    // texture which will replace the texture selected into t0.
    //
    // A nearly identical shader using a different scaling constant is used to
    // apply the "force" value to the "velocity" texture at each texel.
    //
    // t1 comes in the range [0,1] but needs to hold signed values, so a value of
    // 0.5 in t1 represents zero velocity.  This is biased to a signed value in 
    // computing the new position.                       
    //
    // tex0 = height field
    // tex1 = velocity          
    //
    // Bias the force/velocity to a signed value so we can subtract from
    //   the t0 position sample.
    //
    // New height = velocity * scale factor + old height

    // Original register combiner program:
    //
    // Stage 0
    // rgb
    // {
    //   discard = expand(tex1) * const0;
    //   discard = expand(tex0);
    //   spare0 = sum();
    //   scale_by_one_half();
    // }
    // Stage 1
    // rgb
    // {
    //   discard = spare0;
    //   discard = const0;
    //   spare0 = sum();
    // }
    // }

    float[] const0 = new float[] { 0.5f,  0.5f,  0.5f,  1.0f };

    int[] tmpInt = new int[1];
    gl.glGenPrograms(1, tmpInt, 0);
    int fragProg = tmpInt[0];
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);

    String program =
"!!ARBfp1.0\n" +
"PARAM const0 = program.env[0];\n" +
"PARAM one     = { 1.0, 1.0, 1.0, 0.0 };\n" +
"PARAM oneHalf = { 0.5, 0.5, 0.5, 1.0 };\n" +
"PARAM two     = { 2.0, 2.0, 2.0, 1.0 };\n" +
"TEMP texSamp0, texSamp1;\n" +
"TEMP spare0, spare1;\n" +
"\n" +
"TEX texSamp0, fragment.texcoord[0], texture[0], 2D;\n" +
"TEX texSamp1, fragment.texcoord[1], texture[1], 2D;\n" +
"MAD spare0, two, texSamp1, -one;\n" +
"MAD spare1, two, texSamp0, -one;\n" +
"MAD spare0, spare0, const0, spare1;\n" +
"MAD result.color, oneHalf, spare0, const0;\n" +
"\n" +
"END\n";

    loadProgram(gl, GL2.GL_FRAGMENT_PROGRAM, program);

    gl.glNewList(displayListID, GL2.GL_COMPILE);
    gl.glProgramEnvParameter4fv(GL2.GL_FRAGMENT_PROGRAM, 0, const0, 0);
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);
    gl.glEnable(GL2.GL_FRAGMENT_PROGRAM);
    gl.glEndList();
  }

  private void initCreateNormalMap(GL2 gl, int displayListID) {
    // Neighbor-differencing for RGB normal map creation.  Scale factors for s and t
    // axis components are set in program code.
    // This does a crude 1-s^2-t^2 calculation for the blue component in order to
    // approximately normalize the RGB normal map vector.  For s^2+t^2 close to 1.0,
    // this is a close approximation to blue = sqrt(1 - s^2 - t^2) which would give a
    // normalized vector.
    // An additional pass with a dependent texture lookup (alpha-red or green-blue)
    // could be used to produce an exactly normalized normal.

    // colors from all 4 texture stages
    // tex0 = -s,  0
    // tex1 = +s,  0
    // tex2 =  0, +t
    // tex3 =  0, -t

    // Original register combiner program:
    //
    // Stage 0
    // rgb
    // {
    //   // (t0 - t1)*4  : 4 for higher scale
    //   discard = -tex1;
    //   discard = tex0;
    //   spare0 = sum();
    //   scale_by_four();
    // }
    // Stage 1
    // rgb
    // {
    //   // (t3 - t2)*4 : 4 for higher scale
    //   discard = -tex2;
    //   discard = tex3;
    //   spare1 = sum();
    //   scale_by_four();
    // }
    // Stage 2
    // Define const0 in the third general combiner as RGBA = (scale, 0, 0, 0)
    //  Where scale [0,1] is applied to reduce the magnitude
    //  of the s axis component of the normal.
    // Define const1 in the third combiner similarly to affect the t axis component
    // define these by "ramboing" them in the C++ code that uses this combiner script.
    // Note: these variables have been renamed to "redMask" and "greenMask" in
    // the fragment program below.
    // rgb
    // {
    //   // see comment about consts above!
    //   // t0 = s result in red only
    //   discard = spare0 * const0;
    //   discard = spare1 * const1;
    //   spare0 = sum();
    // }
    // Stage 3
    // rgb
    // {
    //   tex1 = spare0 * spare0;
    //   scale_by_two();
    // }
    // Stage 4
    // const0 = (1, 1, 0, 0);
    // rgb
    // {
    //   spare1 = unsigned_invert(tex1) . const0;
    //   scale_by_one_half();
    // }
    // Stage 5
    // const0 = (0.5, 0.5, 0, 0);
    // rgb
    // {
    //   discard = spare0;
    //   discard = const0;
    //   spare0 = sum();
    // }
    // Stage 6
    // const0 = (0, 0, 1, 1);
    // rgb 
    // {
    //   discard = spare1 * const0;
    //   discard = spare0;
    //   spare0 = sum();
    // }


    float[] const0 = new float[] { 0.5f,  0.5f,  0.5f,  1.0f };

    int[] tmpInt = new int[1];
    gl.glGenPrograms(1, tmpInt, 0);
    int fragProg = tmpInt[0];
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);

    String program =
"!!ARBfp1.0\n" +
"PARAM redMask   = program.env[0];\n" +
"PARAM greenMask = program.env[1];\n" +
"PARAM const0    = { 1.0, 1.0, 0.0, 0.0 };\n" +
"PARAM const1    = { 0.5, 0.5, 0.0, 0.0 };\n" +
"PARAM const2    = { 0.0, 0.0, 1.0, 1.0 };\n" +
"PARAM one     = { 1.0, 1.0, 1.0, 0.0 };\n" +
"PARAM oneHalf = { 0.5, 0.5, 0.5, 1.0 };\n" +
"PARAM two     = { 2.0, 2.0, 2.0, 1.0 };\n" +
"PARAM four    = { 4.0, 4.0, 4.0, 1.0 };\n" +
"TEMP texSamp0, texSamp1, texSamp2, texSamp3;\n" +
"TEMP spare0, spare1, spare2;\n" +
"\n" +
"TEX texSamp0, fragment.texcoord[0], texture[0], 2D;\n" +
"TEX texSamp1, fragment.texcoord[1], texture[1], 2D;\n" +
"TEX texSamp2, fragment.texcoord[2], texture[2], 2D;\n" +
"TEX texSamp3, fragment.texcoord[3], texture[3], 2D;\n" +
"SUB spare0, texSamp0, texSamp1;\n" +
"MUL spare0, spare0, four;\n" +
"SUB spare1, texSamp3, texSamp2;\n" +
"MUL spare1, spare1, four;\n" +
"MUL spare0, spare0, redMask;\n" +
"MAD spare0, greenMask, spare1, spare0;\n" +
"MUL_SAT spare2, spare0, spare0;\n" +
"SUB spare2, one, spare2;\n" +
"DP3 spare1, spare2, const0;\n" +
"ADD spare0, spare0, const1;\n" +
"MAD result.color, const2, spare1, spare0;\n" +
"\n" +
"END\n";

    loadProgram(gl, GL2.GL_FRAGMENT_PROGRAM, program);

    gl.glNewList(displayListID, GL2.GL_COMPILE);
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);
    gl.glEnable(GL2.GL_FRAGMENT_PROGRAM);
    gl.glEndList();
  }

  private void initDotProductReflect(GL2 gl, int displayListID) {
    // Pseudocode for this operation, derived from the NVidia
    // texture_shader.txt documentation at
    // http://oss.sgi.com/projects/ogl-sample/registry/NV/texture_shader.txt

    // TEX texSamp0, fragment.texcoord[0], texture[0], 2D;
    // MAD texSamp0, two, texSamp0, minusOne;
    // TEMP dotPP = texSamp0 . texcoord[1];
    // TEMP dotP  = texSamp0 . texcoord[2];
    // TEMP dotC  = texSamp0 . texcoord[3];
    // TEMP R, N, E;
    // N = [dotPP, dotP, dotC];
    // ooNLength = N dot N;
    // RCP ooNLength, ooNLength;
    // E = [texcoord[1].w, texcoord[2].w, texcoord[3].w];
    // nDotE = N dot E;
    // MUL R, nDotE, N;
    // MUL R, R, two;
    // MUL R, R, ooNLength;
    // SUB R, R, E;
    // TEX result.color, R, texture[3], CUBE;

    // This fragment program is pretty length-sensitive; making it too
    // big causes the frame rate to be cut in half on my machine
    // (Quadro FX Go700) due to sync-to-vertical-refresh. The program
    // below is more optimized in its use of temporaries. Some of the
    // scaling operations on the first component of the normal vector
    // (before subtracting off the E vector) don't appear to make much
    // of a visual difference so they are skipped as well.

    int[] tmpInt = new int[1];
    gl.glGenPrograms(1, tmpInt, 0);
    int fragProg = tmpInt[0];
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);

    String program =
"!!ARBfp1.0\n" +
"PARAM minusOne = { -1.0, -1.0, -1.0, 0.0 };\n" +
"PARAM two      = {  2.0,  2.0,  2.0, 0.0 };\n" +
"TEMP texSamp0, R, N, E;\n" +
"\n" +
"TEX texSamp0, fragment.texcoord[0], texture[0], 2D;\n" +
"MAD texSamp0, two, texSamp0, minusOne;\n" +
"DP3 N.x,   texSamp0, fragment.texcoord[1];\n" +
"DP3 N.y,   texSamp0, fragment.texcoord[2];\n" +
"DP3 N.z,   texSamp0, fragment.texcoord[3];\n" +
"MOV E.x, fragment.texcoord[1].w;\n" +
"MOV E.y, fragment.texcoord[2].w;\n" +
"MOV E.z, fragment.texcoord[3].w;\n" +
"MUL N, N, two;\n" +
"SUB R, N, E;\n" +
"TEX result.color, R, texture[3], CUBE;\n" +
"\n" +
"END";

    loadProgram(gl, GL2.GL_FRAGMENT_PROGRAM, program);

    gl.glNewList(displayListID, GL2.GL_COMPILE);
    gl.glBindProgram(GL2.GL_FRAGMENT_PROGRAM, fragProg);
    gl.glEnable(GL2.GL_FRAGMENT_PROGRAM);
    gl.glEndList();
  }

  private void loadProgram(GL2 gl,
                           int target,
                           String programBuffer) {

    gl.glProgramString(target, GL2.GL_PROGRAM_FORMAT_ASCII, programBuffer.length(), programBuffer);

    int[] errPos = new int[1];
    gl.glGetIntegerv(GL2.GL_PROGRAM_ERROR_POSITION, errPos, 0);
    if (errPos[0] >= 0) {
      String kind = "Program";
      if (target == GL2.GL_VERTEX_PROGRAM) {
        kind = "Vertex program";
      } else if (target == GL2.GL_FRAGMENT_PROGRAM) {
        kind = "Fragment program";
      }
      System.out.println(kind + " failed to load:");
      String errMsg = gl.glGetString(GL2.GL_PROGRAM_ERROR_STRING);
      if (errMsg == null) {
        System.out.println("[No error message available]");
      } else {
        System.out.println("Error message: \"" + errMsg + "\"");
      }
      System.out.println("Error occurred at position " + errPos[0] + " in program:");
      int endPos = errPos[0];
      while (endPos < programBuffer.length() && programBuffer.charAt(endPos) != '\n') {
        ++endPos;
      }
      System.out.println(programBuffer.substring(errPos[0], endPos));
      throw new GLException("Error loading " + kind);
    } else {
      if (target == GL2.GL_FRAGMENT_PROGRAM) {
        int[] isNative = new int[1];
        gl.glGetProgramiv(GL2.GL_FRAGMENT_PROGRAM,
                             GL2.GL_PROGRAM_UNDER_NATIVE_LIMITS,
                             isNative, 0);
        if (isNative[0] != 1) {
          System.out.println("WARNING: fragment program is over native resource limits");
          Thread.dumpStack();
        }
      }
    }
  }
}