之前没有注意过这个问题, 转自: http://book.51cto.com/art/200901/106880.htm
问：有人告诉我不能在printf中使用%lf。为什么printf()用%f输出double型，而scanf却用%lf呢？
答：printf的%f说明符的确既可以输出float型又可以输出double型。 根据"默认参数提升"规则（在printf这样的函数的可变参数列表中 ，不论作用域内有没有原型，都适用这一规则）float型会被提升为double型。因此printf()只会看到双精度数。参见问题15.2。
对于scanf，情况就完全不同了，它接受指针，这里没有类似的类型提升。（通过指针）向float存储和向double存储大不一样，因此，scanf区别%f和%lf。
下表列出了printf和scanf对于各种格式说明符可以接受的参数类型。
|
格式 |
printf |
scanf |
|
%c |
int |
char * |
|
%d, %i |
int |
int * |
|
%o, %u, %x |
unsigned int |
unsigned int * |
（续）
|
格式 |
printf |
scanf |
|
%ld, %li |
long int |
long int * |
|
%lo, %lu, %lx |
unsinged long int |
unsigned long int * |
|
%hd, %hi |
int |
short int * |
|
%ho, %hu, %hx |
unsigned int |
unsigned short int * |
|
%e, %f, %g |
double |
float * |
|
%le, %lf, %lg |
n/a |
double * |
|
%s |
char * |
char * |
|
%[...] |
n/a |
char * |
|
%p |
void |
void ** |
|
%n |
int * |
int * |
|
%% |
none |
none |
（严格地讲，%lf在printf下是未定义的，但是很多系统可能会接受它。要确保可移植性，就要坚持使用%f。）
另外, 关于才c++标准输入输出同步, 如果是用了ios::sync_with_stdio(false)的话是可以达到近似scanf的速度的。
static void sync_with_stdio();
Synchronizes the C++ streams with the standard I/O system. The first time this function is called, it resets the predefined streams (cin, cout, cerr, clog) to use astdiobuf object rather than a filebufobject. After that, you can mix I/O using these streams with I/O using stdin, stdout, and stderr. Expect some performance decrease because there is buffering both in the stream class and in the standard I/O file system.
After the call to sync_with_stdio, the ios::stdio bit is set for all affected predefined stream objects, and cout is set to unit buffered mode.