In yesderday's contest, my submission for D got TLE in the 29 th case (17495055) where I used cin/cout with ios_base :: sync_with_stdio (0), which of course took more than 2 seconds to execute. :O
But today I re-submitted that same code, but changed the cin, couts into scanf and printf (17504450), and got AC! The weird part is that scanf/printf code took only 826 ms to execute, which is more than 2 times faster than ios_base :: sync_with_stdio (0)!
I knew ios_base :: sync_with_stdio (0) ensures almost same speed as scanf and printf. My question is that, does scanf/printf actually ensures significantly faster performance than ios_base :: sync_with_stdio (0) ?
no ios,scanf 17216504 685ms
using ios 17216440 327ms
scanf 17216479 93ms
using ios and
'\n'
instead of endl 17510489 233msOk. I'm NOT going to use cin/cout now :(
I had the same problem, I think there should be a hint like: Input is huge, Fast I/O needed or something like that. Or big test cases on pretest, to have TLE on the competition and not later.
Here are some bechmarks http://codeforces.net/blog/entry/562 (unfortunately only in Russian). I think it not so hard to understand them using google translate.
It's a bug in MinGW GCC build used by codeforces. Cin/cout should of course be just as fast as printf/scanf.
I wonder when it will be fixed, either by MinGW authors or by Mike. I mean, Microsoft distributes MSVC 2015 as free standalone package, why not to install it on codeforces servers?