Let's look on this code example:
#include <iostream>
int main() {
std::ios_base::sync_with_stdio(false);
int n;
std::cin >> n;
for (int i = 0; i < n; ++i) {
int buf;
std::cin >> buf;
}
}
Performance of this code sample on input like this:
10000000
0
1
...
9999999
on my machine:
g++-5 -O2 -std=c++11:
./a.out < input.txt 0.86s user 0.07s system 98% cpu 0.942 total
clang-700.0.72 -O2 -std=c++11:
./a.out < input.txt 38.69s user 0.21s system 99% cpu 39.248 total
After some profiling I found that libc++ doesn't disable synchronization at all.
Then I look at their code and found this: https://github.com/llvm-mirror/libcxx/blob/6a85e8a355be05b9efa8408f875014e6b47cef3b/src/ios.cpp#L458
So my question, is it by design or bug?
sync_with_stdio
is just an advisory function, as far as I can tell.
ios.members.static/2 says:
Effects: If any input or output operation has occurred using the standard streams prior to the call, the effect is implementation-defined. Otherwise, called with a false argument, it allows the standard streams to operate independently of the standard C streams.
"allows ... to operate independently" is the key, IMHO.
There's no requirement on an implementation to actually do so, but rather, a prohibition on doing so unless this call has been made.
So, yes, this is conforming behavior.