josvth / hermes-optimization Goto Github PK

Currently you need to filter the pass selection in post-processing twice to get to the actual 'used' result:

• once for down selection of overlapping passes
• once for passes that result in zero link time

The latter is due to the 'soft' way of using passes. If a pass results into no link-time, the pass is ignored in the throughput & latency & energy calculation. It is however considered in the pointing calculation.

In post-processing the link time has to be re-evaluated which is time consuming.

• Use a dict / *arg / **kwarg based way of passing results of models