jkrs2's blog

By jkrs2, history, 4 months ago, In English

It seems that codeforces compiles and runs submissions in a windows environment. This can produce some surprises to anyone who develops solutions under Linux.

(I have not checked the details of anything written here....)


CPU time and clock

A linux developer will believe that the clock function will return a good approximation of the processor time used by a program. In the codeforces environment this is occasionally untrue.

Reference: https://codeforces.com/blog/entry/85677

The solution is code along these lines:

long double cpu_time() {
#if defined(_WIN32) || defined(_WIN64)
    FILETIME creation_ft, exit_ft, kernel_ft, user_ft;
    GetProcessTimes(GetCurrentProcess(), &creation_ft, &exit_ft, &kernel_ft, &user_ft);

    auto extract_time = [](FILETIME ft) {
        return 1e-7L * (ft.dwLowDateTime | uint64_t(ft.dwHighDateTime) << 32);

    return extract_time(user_ft) + extract_time(kernel_ft);
    return (long double) clock() / CLOCKS_PER_SEC;


Random numbers

A linux developer will read that the rand() function uses the same code as the random() function, which produces a random number generator with a huge period of roughly $$$16^{2^{31}}$$$. RAND_MAX has value $$$2^{31}-1$$$.

In the code forces environment rand() produces a cycle of random numbers unknown period (though longer than 32768). RAND_MAX has value 32767 or $$$2^{15}-1$$$.


Full text and comments »

  • Vote: I like it
  • +27
  • Vote: I do not like it