> BTW, I tried setting the precision control field for the FPU, and while > that did change the intermediate results (as inspected by gdb), the final > calculation of minPos always came out to 0.0. This would be another option I would be happy with, if it could be made to work.