Targets with no floating-point unit

For embedded targets that have limited or no support for floating-point instructions, you can use fixed-point arithmetic blocks. This trade-off has a very limited impact on computational accuracy.