Is there a tool to examine the difference of two callgrind/valgrind profiles?

I'm trying to understand how a program's performance changes when run with different options—I have a factor of 1.8 I'd like to explain. I've been doing profiling with [valgrind]'s callgrind tool and using kcachegrind to visualize the results. But I'd really like to look at differences. I found the cg_diff program but it doesn't work with callgrind profiles, only with cachegrind profiles, which are too expensive for me to gather. Does anybody know of tools for visualizing the difference between two profiles?