At the moment, there is a huge memory overhead as we keep a full columns for each selectors for each traces, i.e. if we have N relations columns and M selectors, we will keep M * (N + M) columns of L rows, L being the domain size.
It is not efficient at all, as we only have to keep a value 0 or 1 for M columns for each of the M traces.
At the moment, there is a huge memory overhead as we keep a full columns for each selectors for each traces, i.e. if we have N relations columns and M selectors, we will keep M * (N + M) columns of L rows, L being the domain size. It is not efficient at all, as we only have to keep a value 0 or 1 for M columns for each of the M traces.