Optimization Techniques for Verification of Out-of-Order Execution Machines
Figure 3
(a) Scatterplot compares the flush-machine (FM) approach with standard flushing using 23 ooo pipelined machine models
(b) Scatterplot compares the flush-machine (FM) approach with standard GFP-based commitment using 23 ooo pipelined machine models
(c) Scatterplot compares the use of both collapsed flushing (CF) and flush-machine (FM) with standard flushing based on 23 ooo pipelined machine models
(d) Scatterplot compares the use of both collapsed flushing (CF) and flush-machine (FM) with GFP-based commitment on 23 ooo pipelined machine models
(e) Scatterplot compares the use of both collapsed flushing (CF) and flush-machine (FM) with FM alone on 23 ooo pipelined machine models
(f) Plot shows the increase in verification time when using flushing (F), commitment (C), flush-machine (FM), and collapsed flushing (CF) and FM, when the size of the reorder buffer is increased. All the models have two execution units with a 2-cycle latency each