Memory Layout

Floyd-Warshall uses a $2D$ array dist[i][j] of size $n \times n$. For $n = 1000$, this is $1$ million entries ($4$-$8$ MB for integers). This fits in cache on modern machines. Initialize the array carefully: set diagonal to $0$, edges to weights, rest to infinity (use a large constant like $10^9$).

Proper initialization is critical. If you mess up the initial state, the algorithm computes garbage. Memory layout affects performance on large inputs.