Good afternoon!
Should the matrices in the benchmarks be tested with the same density (as it can be controlled), or is it possible to test with different? E.g., matrix A with N=N1 and D=D1 runs >1 minute and Matrix B with N=N2 and D=D2 runs >3 minutes, where N2>N1 and D1!=D2 (in particular, D1>D2, e.g., B has more 0).
Is it fine, if the upper bound for each connection between two nodes were added (e.g., % MAX_DIST) or should it be not upper bounded?
Best regards,
Dmitry Maklakov