Report 9, Murat Bozkurt

In this week, we mainly worked on improving our last parallel algorithm.

In our algorithm, we firstly levelized gates. After that, we divided gates into clusters by making all of them independent from each other. Then, we simulate and gates, after that we simulate latches for all cycles.

The most time consuming part was the optimization part of our code. To do this, we analyzed our code to be able to decide where we could use shared memory instead of global memory. After we made use of shared memory in some of our code, we decreased time to run to nearly 60% of the recent week’s result.

In the next week, we are planning to make further optimizations to make our result better than the sequential part.

Murat Bozkurt

Advertisement