#include #include using namespace std; /* TIME LIMIT EXCEEDED? Just try again in a minute. Testing wisdom: endl should only be used to FLUSH output every so often -- \n should be used otherwise. Interference: Though hi-resolution timer ticks 1 billion times/sec, the scheduler returns every 15623 or 15624 microseconds, with variance in the 3rd position (the 6) detected in later tests. The finest measure appears to be 1/64 of a second (possibly unrelated: float precision is 15.6 for 52-bit floats, 15.9 for 53 bits. Also, system RTC should be ticking at 18.2 ticks/s). Stabilization to no significant difference may be caused by optimization / smart cache. Perhaps the best tests are done when the cache is helping someone else instead. (CORRECTED text from micro- to nanoseconds; the math appears to work out the same) */ const int ITERS=25500000; std::ostream nullcharbuf(0); // null buffer // std::wostream nullwcharbuf(0); // wide, fyi long long test_endl() { int i=0; auto start = std ::chrono ::high_resolution_clock ::now(); while(i // microseconds (endTime-start) .count(); } long long test_newline() { int i=0; char newline='\n'; auto start = std ::chrono ::high_resolution_clock ::now(); while(i // microseconds (endTime-start) .count(); } int main() { int numerator = chrono ::high_resolution_clock ::period ::num; long denominator = chrono ::high_resolution_clock ::period ::den; cout << " Clock: " << numerator << "/" << denominator << " (nanoseconds)\n"; long long e = test_endl(); long long nl = test_newline(); // 15623 and 15624 are the smallest possible microseconds we get cout << " Tests: " << ITERS << "\n"; cout << "newline: " << nl << "ns (" << nl/15623000 << "/64 seconds)" << endl; cout << " endl: " << e << "ns (" << e/15623000 << "/64 seconds)\n\n"; cout << "*** Nanoseconds actually arrive in 1/64s chunks (could overlap if too close)\n"; return 0; }

Newlines faster than << endl