How does it work?

This can indeed be more efficient in debug build mode, but at the second level of optimization we get an identical assembler in both cases.

I'm glad you enjoyed it!

ПАнимаю

Added a message for fixed and small tests.

You can print the input data println!("{:?}", degrees), testing ends at the first assertion failed.

I am very happy to read this)

Perhaps I am not well versed in the mechanics of spoilers.

If I understand correctly, you suggested the following algorithm:

for i in range(num_of_vertices):
    for j in in range(i+1, num_of_vertices):
        e = randint(0, 1)
        matrix[i][j] = e
        matrix[j][i] = e

I tried it and got a huge performance drop.

I changed the tests according to your advice but left the graph generation the same because your algorithm runs much longer (a total of 4-6 seconds difference).

Specified in @test.it.