SourceGen: Visualizations

Back to index

Overview

Programs are generally a combination of code and data. Sometimes the data is graphical in nature, e.g. a bitmap used as a font or game sprite. Being able to see the data in graphic form can make it easier to determine the purpose of associated code.

While modern systems use GIF, JPEG, and PNG to hold 2D bitmaps, graphical elements embedded in 6502 applications are almost always in a platform-specific form. For this reason, the task of generating images from data is performed by extension scripts. Some scripts for common formats are included in the SourceGen runtime directory. If these don't do what you need, you can write your own scripts and include them in your project.

The project file doesn't store the converted graphics. Instead, the project file holds a string that identifies the converter, and a list of parameters that are passed to the converter. Images are generated when the project is first opened, and updated when certain things change in the project.

Visualizations are not included in generated assembly output. They may be included in HTML exports.

Because visualizations are associated with a specific file offset, they will become "hidden" if the offset isn't at the start of a line, e.g. it's in the middle of a multi-byte instruction or data item. The editors will try to prevent you from doing this.

Bitmaps will always be scaled up as much as possible to make them easy to see. This means that small shapes and large shapes may appears to be the same size when displayed as thumbnails in the code list.

The role of a visualization generator is to take a collection of input parameters and generate graphical data. It's most useful for graphical sources like bitmaps, but it's not limited to that. You could, for example, write a script that generates random flowers, and use it to make your source listings more cheerful.

Visualizations and Visualization Sets

Visualizations are essentially decorative: they do not affect the assembled output, and do not change how code is analyzed. They are contained in sets that are placed at arbitrary offsets. Each set can contain multiple items. For example, if a file has data for 10 bitmaps, you can place a visualization near each, or create a single visualization set with all 10 items and put it at the start of the file. You can display a visualization near the data or near the instructions that perform the drawing. Or both.

To create a visualization set, select a code or data line, and use Actions > Create/Edit Visualization Set. To edit a visualization set, select it and use the same menu item, or just double-click on it. This opens the Visualization Set Editor window.

The visualization set editor shows a list of visualizations associated with the selected file offset. You can create a new visualization, edit or remove an existing entry, or rearrange them. If you select New Bitmap or edit an existing bitmap entry, the Bitmap Visualization Editor window will open. Similarly, if you select New Bitmap Animation or edit an existing bitmap animation, the Bitmap Animation Editor will open.

Visualization Editor

The combo box at the top of the screen lists every visualization generator defined by an active extension script. Select the one that is appropriate for the data you're trying to visualize. Every visualizer may have different parameters, so as you select different entries the set of input parameters below the preview window may change.

There are two categories of visualization generator: bitmap, and wireframe. Bitmaps are simple 2D images, but wireframes are 2D or 3D meshes that can be viewed from different angles. When you select a wireframe generator, additional view controls will be added at the bottom. (See below.)

The "tag" is a unique string that will be shown in the display list. This is not a label, and may contain any characters you want (but leading and trailing whitespace will be trimmed). The only requirement is that it be unique across all visualizations (bitmaps, animations, etc).

The preview window shows the visualizer output. The generated image is expanded to fill the window, so small bitmaps will be shown with very large pixels. If you resize the editor window, the preview window will expand, which can make it easier to see detail on larger images. If the generator fails, the preview window will show a red 'X', and an error message will appear below it.

Parameters may be numeric or boolean. The latter use a simple checkbox, the former a text entry field that accepts decimal and hexadecimal values. The range of allowable values is shown to the right of the entry field. If you enter an invalid value, the parameter description will turn red.

The Export button at the top right can be used to save a copy of the bitmap or wireframe rendering with the current parameters.

Wireframe View Controls

The wireframe generator may offer the choice of perspective vs. orthographic projection, and whether or not to enable backface culling. These are declared in the visualization generator script, but implemented in the viewer. If the generator doesn't declare them, the default is to render with a perspective projection and without culling.

The viewer allows you to rotate the image about the X, Y, and Z axes. The viewer provides a left-handed coordinate system, with +X toward the right, +Y toward the top of the screen, and +Z going into the screen. The object will be placed a short distance down the Z axis and scaled to fit the window. Positive rotations cause a counter-clockwise rotation when the axis about which rotations are performed points toward the viewer. The rotations are performed with a matrix using Euler angles, and are subject to gimbal lock (e.g. if you set Y to 90 degrees, X and Z rotate about the same axis).

If you check the Animated box, you can add a simple spin. Choose the number of degrees to rotate per frame, how many frames to generate before resetting, and the delay between each frame. Clicking the Auto button will automatically select the number of frames needed to display the animation in an unbroken loop (useful for animated GIFs). Click the Test Animation button to see it in action.

Bitmap Animation Editor

Bitmap animations allow you to create a simple animation from a collection of other visualizations. This can be useful when a program stores animated graphics as a series of frames.

The "tag" is a unique string that will be shown in the display list. The same rules apply as for bitmap visualizations.

The list at the top left holds all visualizations. Select items on the left and use the Add button to add them to the list on the right, which has the set that is included in the animation. You can reorder the list with the up/down buttons. Adding the same frame multiple times is allowed.

The frame delay field lets you specify how long each frame is shown on screen, in milliseconds. Some animation formats may use a different time resolution; for example, animated GIFs use units of 1/100th of a second. The closest value will be used. Note also that some viewers (notably web browsers) will cap the update rate.

When you have one or more frames in the animation list, you can preview the result in the window at the bottom. The actual appearance may be slightly different, especially if the frames are different sizes. For example, the preview window scales individual frames, but animated GIFs will be scaled to the size of the largest frame.

Scripts Included with SourceGen

A number of visualization generation scripts are included with SourceGen, in the platform-specific runtime data directories.

Most generators will take the file offset, bitmap width, and bitmap height as parameters. Offsets are handled as they are elsewhere, i.e. always in hexadecimal, with a leading '+'. Some less-common parameters include:

• Column stride - number of bytes used to hold a column. This is uncommon, but could be used if (say) a pair of bitmaps was stored with interleaved bytes. If you set this to zero the visualizer will default to no interleave (col_stride = 1).
• Row stride - number of bytes between the start of each row. This is used when a row has padding on the end, e.g. a bitmap that's 7 bytes wide might be padded to 8 for easy indexing, or when bitmap data is interleaved. If you set this to zero the visualizer will default to no padding (row_stride = width * column_stride).
• Cell stride - for multi-bitmap data like a font or sprite sheet, this determines the number of bytes between the start of one item and the next. If set to zero a "dense" arrangement is assumed (cell_stride = row_stride * item_height).

Remember that this is a disassembler, not an image converter. The results do not need to be perfectly accurate to be useful when disassembling code.

Apple II : Apple/VisHiRes and Apple/VisShapeTable

There is no standard format for small hi-res bitmaps, but certain arrangements are common. The VisHiRes script defines four generators:

• Hi-Res Bitmap - converts an MxN row-major bitmap.
• Hi-Res Sprite Sheet - converts a series of bitmaps and renders them in a grid. Useful for games that use cell animation. The generated bitmap has a 1-pixel transparent gap between elements.
• Hi-Res Bitmap Font - a simplified version of the Sprite Sheet, intended for the common 7x8 monochrome fonts. Most fonts have 96 or 128 glyphs, though some drop the last character. (This also works for Apple /// fonts, but currently ignores the high bit in each byte.)
• Hi-Res Screen Image - used for 8KiB screen images. The data is linearized and converted to a 280x192 bitmap. Because images are relatively large, the generator does not require them to be contiguous in the file, i.e. two halves of the image can be in different parts of the file so long as they end up contiguous in memory.

Widths are specified in bytes, not pixels. Each byte represents 7 pixels (with some hand-waving).

In addition to offset, dimensions, and stride values, the bitmap converter has a checkbox for monochrome or color, and two checkboxes that affect the color. The first causes the first byte to be treated as being in an odd column rather than an even one, which affects green vs. purple and orange vs. blue. The second flips the high bits on every byte, switching green vs. orange and purple vs. blue. Neither has any effect on black & white bitmaps.

The converter generates one output pixel for every source pixel, so half-pixel shifts are not represented.

The VisShapeTable script renders Applesoft shape tables, which can have multiple vector shapes. There are two generators:

• Shape Table - full table. Set the offset to the first byte of the table, and set the image index to the desired shape index. (Note: indices here start at zero, but in Applesoft programs they start at 1.)
• Shape Table Shape - single shape. Set the offset to the first byte of the shape definition.

Atari 2600 : Atari/Vis2600

The Atari 2600 graphics system has registers that determine the appearance of a sprite or playfield on a single row. The register values are typically changed as the screen is drawn to get different data on successive rows. The visualization generator doesn't attempt to emulate this behavior, but works well for data stored in a straightforward fashion.

• Sprite - basic 1xN sprite, converted to an image 8 pixels wide. Square pixels are assumed.
• Playfield - assumes PF0,PF1,PF2 are stored in that order, multiple entries following each other. Specify the number of 3-byte entries as the height. Since most playfields aren't the full height of the screen, it will tend to look squashed. Use the "row thickness" feature to repeat each row N times to adjust the proportions. The "Reflected" checkbox determines whether the right-side image is repeated as-is or flipped.

Atari Arcade : Atari/VisAVG and Atari/VisDVG

Different versions of Atari's Analog Vector Graphics and Digital Vector Graphics were used in several games, notably Asteroids, Battlezone, Tempest, and Star Wars. The commands drove a vector display monitor. SourceGen visualizes them as 2D wireframes, which isn't a perfect fit since they can describe points as well as lines, but works fine for annotating a disassembly.

The visualizer takes two arguments: the offset of the start of the commands to visualize, and the base address of vector RAM. The latter is needed to convert vector JMP/JSR commands into offsets.

Commodore 64 : Commodore/VisC64

The Commodore 64 has a 64-byte sprite format defined by the hardware. It comes in two basic varieties:

• High-resolution sprite - 24x21 monochrome. Pixels are either colored or transparent.
• Multi-color sprite - 12x21 3-color. The width of each pixel is doubled to make it 24x21.

Sprites can be doubled in width and/or height.

Colors come from a hardware-defined palette of 16:

0.  black 
1.  white 
2.  red 
3.  cyan 
4.  purple 
5.  green 
6.  blue 
7.  yellow 
8.  orange 
9.  brown 
10.  light red 
11.  dark grey 
12.  grey 
13.  light green 
14.  light blue 
15.  light grey 

Bear in mind that the editor scales images to their maximum size, so a sprite that is doubled in both width and height will look exactly like a sprite that is not doubled at all.

If multiple sprites are defined consecutively, at 64-byte intervals, they can be combined into a sprite sheet. A single set of parameters for color and scaling is applied to all sprites.

The visualizer also handles high-resolution and multi-color fonts, which are displayed in a grid.

Nintendo Entertainment System : Nintendo/VisNES

NES PPU pattern tables hold 8x8 tiles with 2 bits of color per pixel. Converting the full collection to a reference bitmap is straightforward. A few color palette options are offered.

Sprites and backgrounds are formed from collections of tiles. In some cases this is straightfoward, in others it's not. A visualization generator that renders a "tile grid" is available for simpler cases.

Back to index