When I implement a design with the multithreading activated and then again
not activated for MAP and PAR, the results are different. with the multithreading I can confirm that the results are different because the timing scores are different.
I had thought that using the multithreading would just speed up the implementation but not change the result.
Is it expected behavior that the results are not the same?
For MAP I used -mt 2 and PAR with -mt 4.
This is an expected result if you are using multithreading for PAR, A drop in performance by about 1% should be expected.
When the Router algorithms works in parallel, there will be multiple nets being routed at the same time while keeping the resources free, which could lead to a slight drop in performance. Please note that the main objective behind multi-threading is to reduce run time.
Multi-threading should be used if you are still in the design process but once you have your final design, you can run PAR without multi-threading for extra performance gain and to improve the slacks.
Multi-threading for the MAPPER should give roughly the same results as when it is run without multi-threading.
Was this Answer Record helpful?