User Interface

Scene Menu

• Reset Camera - Restores the camera to its initial position for the current scene.  Very useful if you get lost.
• Load Spheres - Loads the spheres model.  Simply 7 spheres floating in space (one of which is a mirror ball).
• Load Loft - Loads the loft model.  A small single room house with an upstairs and downstairs.  A diffuse only model with 23 thousand triangles and two point lights.
• Load Lotus - Loads the lotus room model.  A more interesting space with multiple objects and several large mirrors.  Includes 21 thousand triangles and 3 point lights.  (Thanks to Hector Yee for providing the model.)
• Quit - Exits the sample application.  You can also exit by simply closing the window.

Ratio Menu

The ratio is the "render mismatch ratio"[1], defined as the number of pixels in the image divided by the number of eye rays computed (shaded) per frame.  Thus a ratio of 32 means that ray tracer is only doing enough work to compute 1/32nd of a complete image each frame.  The rest of the data comes from the cache of prior shading results maintained by the render cache.  The actual ratio being used is displayed at the bottom of the options panel and may be higher if the ray tracer is unable to shade enough rays within the frame time.  (Very common in the more complicated scenes or shadings.)

Mode Menu

• Ray Trace Only Mode - This mode does not actually use the render cache at all.  Instead to achieve an interactive frame rate, the ray tracer is used to compute and update 1/16th of the pixels each frame.  Once set, a pixel retains its color until it is updated again 16 frames later.  (Updating every pixel, every frame would generally produce frame rates too low to be interactive.)  This mode is single threaded, but since it avoids the overhead of the render cache computations, the thread can devote all its time to tracing rays.  Despite this, the effective interactive performance is generally worse than when using the render cache.  Note: many of the other controls do not apply in this mode since the render cache is not being used.
• Single Processor Mode - In this mode, the render cache and the ray tracer share time on a single thread with most of the time going to the render cache computations.  Despite the decreased amount of time spent tracing rays, interactive performance is generally better than the pure raytracing  mode.
• Dual Processor Mode - In this mode the render cache and the ray tracer run on separate threads and communicate asychronously.  This is the way the render cache is really designed to be run, but works best on dual processor systems, where it allows full utilization of both processors.

Help Menu

• About - Brings up information box about the sample application
• Camera Controls - Brings up an information box with a brief reminder of some of the camera control keys.

Image Display and Camera Controls

• 8 or I - move camera forward
• 5 or K - move camera backward
• 4 or J - move camera to your left
• 6 or L - move camera to your right
• 9 or O - move camera up
• 3 or > - move camera down
• 0 or Spacebar - toggle rotation mode on/off.  In rotation mode the camera will turn based on the current mouse position relative to the center of the image. 
• Minus - zoom out camera field of view
• Plus - zoom in camera field of view
  

Framerate Display

The framerate line below the image display panel shows the current framerate (in frames per second).

Movement Scale

Exposure Controls

• Minus - Darken the image by reducing the effective exposure time.
• Auto - Automatically guess at an appropriate exposure setting for the current image.  May need to be fine-tuned with the lighten and darken buttons
• Plus - Lighten the image by increasing the effective exposure time
  

Shading Mode Controls

• Diffuse - Diffuse only shading
• +Phong - Diffuse shading plus phong highlights (depending on object's material)
• +Shadows - Diffuse shading plus phong highlights and shadow from light sources
• +Reflections - Diffuse plus phong, with shadows and reflections on mirror surfaces

Display Mode Controls

• Raw Points - Show the "raw" projected points for this frame without any filtering (except that the depth cull is still applied if active)
• Prefilter Only - Show the image after point projection and the 7x7 prefilter, but before the normal smaller 3x3 filter.  This image is typically quite blurry.  Only works if the prefilter is active, otherwise it just produces a black image.
• Normal Image - Show the image after point projection, prefilter (if active) and the 3x3 interpolation filter have been used.  This is the image that would normally be shown to the user.
• Priority - Show the priority image which is constructed to guide the sampling.  Highest priority (white) is given to pixels when no valid points map into their 3x3 neighborhood.  Lowest priority (black) is given to pixels which had a recently computed point map to them
• Requests - Shows the subset of image locations where the render cache chooses to ask the ray tracer to compute a new sample (i.e. shade an eye ray).  Locations should be spread somewhat uniformly over the image plane while also more concentrated in regions of sparse or changing data.
• Occupancy - Shows a false-color visualization of which pixels had valid sample mapped to them (green) or were empy (blue).  Note that the depth cull filter (when active) incorrectly  invalidates some pixels near depth discontinuities.
• Pred. Occupancy - When prediction is active, this mode shows the estimated occupancy in the predicted (future) image plane.  White for predicted occupied and black for regions predicted to lack samples.  Note this image has 1/16th as many pixels as the normal image, so it is magnified for display.
• Pred. Requests - Shows the locations in the predicted image plane where samples are requested because a lack of data is predicted for that location. Allows data to be precomputed just before it is needed and provide at least sparse data in new regions as they become visible.

Options controls

• Use Prediction - enable/disable predictive sampling.  When enabled, uses a predicted camera to detect soon to be visible regions that have no shading results (points) cached and requests a sparse set of samples to provide at least some data in such regions before they become visible.
• Clear First - enable/disable clearing of the image to black between frames.  Sometime due to a lack of nearby cached shading results, the render cache is unable to compute a color for a pixel.  In this case the pixel can either be cleared to black (enabled) or left the same color it had in the prior frame (disabled).  Leaving it with its old color is usually less visually distracting, but clearing to black helps when visualizing on the render cache works.
• 7x7 Prefilter - enable/disable use of the prefilter stage.  The prefilter has a larger filter kernel and is thus able to fill in larger gaps between points (cached shading results) than the regular interpolation filter with its 3x3 filter.  Because the regular interpolation overwrites any pixels that it can, effectively only pixels that cannot be computed by the regular interpolation are affected.
• Freeze Points - Freeze the state of the point cloud (shading result cache).  When frozen, points are neither added nor removed from the cache and aging of the points is disabled.  Very useful for visualizing how the render cache works.  You can  freeze the point cloud from particular viewpoint and then look at it from other positions and visualization modes.
• Rate Display - Show the current "render mismatch" ratio.  This ratio is defined as the number of pixels divided by the number of new samples (shaded eye rays) computed by the underlying renderer (ray tracer in this case).  A 512x512 image contains 262144 pixels.  If the ray tracer is able to compute 8192 eye rays per frame, then the mismatch ratio is 32.  If the ray tracer can only compute compute 2000 eye rays per frame then the ratio is 131.