Backward DP

Sometimes you want dp[u] to represent paths from $u$ to a fixed destination, not from a source to $u$. Reverse all edges in the graph and run DP in the reversed graph. If the original graph has edge $u \to v$, the reversed graph has edge $v \to u$. Compute topological order on the reversed graph, initialize dp[destination] = 1, and propagate.

Now dp[u] in the reversed graph represents paths from $u$ to the destination in the original graph. This is useful for problems like counting paths that can reach a specific node.