moyix / hilbert_kcov Goto Github PK

SDL-based hilbert curve visualizer for kernel coverage data.

It looks like this:

Input is a file with lines like:

0x80124415  34
0x80124419  4
0x80194415  1
0x80194416  257

Representing the number of times each kernel address was seen. Currently all kernel addresses must be in the range 0x80000000-0xFFFFFFFF (the default for 32-bit Windows systems).

Because the hilbert curve for a 2GB range of memory is quite large, the viewer just shows a view into it, initially centered on a random address. The colors (greyscale, going from black to white), represent how rare a given address is, with white being the most rare. Black means the address was never seen (yes, this is inconsistent, but worked better than having the view be mostly white all the time). Specifically the color is just 255 * (1 - frequency).

Inside the viewer, you can use the following keyboard commands:

• Arrow keys - pan around the viewport
• q - quit the viewer
• r - recenter the view on a random (nonzero count) address

You can click on any part of the viewport to display the actual address and count.

Building

On OS X, assuming you have SDL 1.2 installed to /Library/Frameworks/ you should be able to just type make.

On Linux it should compile with:

g++ hilbert.cpp -o hilbert -lSDL -lbsd

On Windows, you're on your own.

Bugs

The screen starts off blank until you use the arrow keys or press "r" to move. I have no idea why and very little interest in figuring it out.