Bitmap Memory Debugger
Version 1.10
June 25, 2003
Last Update Feb 18, 2006
Copyright 2003-2006 Octopus Motor Games
Lars Norpchen
bmd@NOSPAM_theycamefromhollywood.com

.
 
 



The area on the left are the programmer's controls for describing the data to be examined. The picture on the right is the data leeched from the debugged application

The window / process pull down lists all open application windows. Pick the application you wish to debug. If your application doesn't show up, for example, it was launched after BMD opened, hit the rescan button. The application is picked by window name and the process ID is reacquired on each refresh, so stopping and restarting the debugged application does not cause BMD to lose which application is being debugged (as long as it's window title remains consistent). However, it can't distinguish between multiple instances of the same application, so it's recommended only one instance be debugger / run.

Image type field picks how BMD should convert the data to a bitmap as well as setting the data field size. There are three palettes available for byte data based on what's most appropriate for the developer. See the notes below on what the options mean. The pad byte is inserted between each data field, to allow for reading a given field in an array of structures.

Image data area sets where the data to be read can be found within the debugged application's memory space. The address shown is the base address used for both the image data and the dimensions (if dims are in address+offset mode). Usually you want to copy and paste the base address from your debugger-it will take hex or decimal input (it figures it out depending on a leading 0x ) The offset is the location of the image data as offset from the base address (for accessing image data within a structure or class) Address+offset mode is used when the memory's location is direct, statically declared, or embedded within the structure directly. *(address+offset) is used when the memory's location is referenced via a pointer in a structure. The net address being viewed is shown in hex.

The Dims area describes the image size. Direct Entry mode lets the developer specify the image size directly and statically. Address+offset mode is used to pull the dimensions from a structure within the debugged application dynamically, with their bit size specified. In Address+offset mode, you set the offset from the base address (as specified in the image data region), and the read-from-memory dimensions are shown in the width, height and (optionally) pitch fields. See the examples below for more detail. The program has a built in limit of 4000 for width or height for sanity checking.

Pitch is used when the row length is greater than the image width (as some video memory buffers require). Pitch cannot be smaller than width..

Refresh controls when the display is refreshed. Paused suspends updates allowing the developer to enter the data into the fields without having BMD try to update it's image with half valid data. This is the default state on startup. "Refresh now" will refresh the display and all dynamic data on demand. Auto refresh will cause BMD to automatically update itself at a given interval (in milliseconds). It's not recommended to set the interval very low on slow machines when viewing large amounts of data as this can really bog down the machine. Auto update while the debugged application is running is very useful for watching the data change in real time.

Vert flip will display the image upside down (a vertical mirror flip, not a 180' rotation). This is useful for some image formats which store data bottom to top.

The Zoom buttons allow the image to be scaled (the scale factor is displayed in the window's title bar). It will not allow a zoomed image to be greater than 1500 pixels tall or wide. If a data read is larger than 1500 pixels, it will automatically be scaled down until it fits. (zoom levels smaller than 1:1 will skip pixels, not interpolate)

Display Modes Notes:

• 8 bit grayscale : maps bytes from black (0) to white (255)
• 8 bit RRGGGBBB: converts bytes to an RGB color, with blue being the least sig bits and red the msb. While BMD is mostly designed to show unsigned data, this mode shows the sign bit as the high red bit, so negative values will have a strong red tint, positive show little to no red.
• 8 bit banded: same as 8 bit grayscale, except alternate values are inverted to show contrast in the LSB.
• 16 bit 555: unsigned words in the xRRRRRGGGGGBBBBB color format
• 16 bit 565: unsigned words in the RRRRRGGGGGGBBBBB color format
• 24 bit: Reads bytes in RGB order and displays in true color mode.
• 24bit BGR: same as RGB but in reverse order.
• 32 bit RGBA. 8 bit Red, green and blue followed by an 8 bit alpha channel (transparency).
• 32 bit BGRA: same as RGBA but reverse order for the RGB bytes
• 24bit BGR: same as RGB but in reverse order.
• Float grayscale: displays 32bit floats as grayscale, automatically setting the range. Min (black) and max (white) values found are displayed at the top. Invalid floats (NAN, +/- inf, etc.) are shown as red pixels
• Float VST: specialized mode for looking at VST audio data. Values from -1.0 to 1.0 are displayed in grayscale, valid floating point values outside that range are shown as warnings in yellow. Denormal floats are shown in magenta and invalid floats are in red. Invalid pixels and warning counts are listed at the top of the bitmap.

The total number of data fields read is pitch * height, whereas the size of each data field is set by the image type plus pad bytes. So the total bytes read is (pitch * height * (data byte size + pad bytes)). The memory request may fail if the read size exceeds the real size of the available data

The aInv button will invert the alpha channel, which is useful if your alpha channel represents opacity instead of transparency.


File Menu

Load and Save Config will load / save a configuration so you don't have to remember or recalculate offsets, dims, etc. for structures and classes used frequently. Config files are stored as .ini files for convenient editing, should you feel the need.

Export to TGA will save the current image, at its current zoom level, as a 24 bit TGA file.

The 9 most recently saved and loaded configurations are listed under the file commands for quick access.

And then there's Exit...


Tools Menu

Offset Finder
This is a tool to hunt for offset values automatically, since it can be hard to calculate offsets manually when padding, vtables, inheritance, etc. come into play. Set up the base address of your structure in the main window. Then open the offset finder (make sure it shows the proper structure base address in the window). Using the debugger, find the value of the structure field you want. Enter the value to hunt for, it's byte size, and a reasonable upper limit to the search range. Hit start to start the search at the base address. The resulting offset will be displayed if it's found, or it will say not found. Hit find again to find the next instance of the hunt value (find again will gray out if it hits the end of the search range). Copy the found offset to the desired offset field in the main window.