Currently FP and vector instructions use a simple 'pipe_class_default' which models a single instruction with a fixed latency of 2 cycles.

This issue proposes extending this to model a representative FP/Vector pipeline.