Where did you find those values? I'm pretty new to the world of optimizations, but I have heard about a qemu feature that runs the cpu from your device in your qemu session instead of emulating one. There's something about it in /scratchbox/doc/cpu-transparency.howto

Now, if you were to use the n8x0's processor to do the compile... that would take a long time... but it might get around the segfaults.