放上一道用来测试快读效率的题:A+B Problem (IO)
题目描述
输入 4 ∗ 1 0 6 4*10^6 4∗106 个整数 a , b a,b a,b,输出它们的和
1. 直接用 cin、cout,毫无疑问,超时(>1s):
#include <bits/stdc++.h>
using namespace std;
signed main(){
int m=(int)2e6;
while(m--){
int a,b;cin>>a>>b;
cout<<a+b<<'\n';
}
return 0;
}
2. cin、cout 关闭同步/解除绑定(380ms):
#include <bits/stdc++.h>
using namespace std;
signed main(){
ios::sync_with_stdio(false);
cin.tie(0);cout.tie(0);
int m=(int)2e6;
while(m--){
int a,b;cin>>a>>b;
cout<<a+b<<'\n';
}
return 0;
}
3. scanf、printf (410ms):
#include <bits/stdc++.h>
using namespace std;
signed main(){
int m=(int)2e6;
while(m--){
int a,b;scanf("%d%d",&a,&b);
printf("%d\n",a+b);
}
return 0;
}
4. 普通快读、快输(215ms):
#include <cstdio>
using namespace std;
namespace IO{
#define re register
#define tpl template<typename T>
#define tpa template<typename T,typename... Args>
tpl inline void read(T &x){
x=0;re bool ff=false;re char ch=getchar();
while(ch<'0'||ch>'9'){ff|=(ch=='-');ch=getchar();}
while(ch>='0'&&ch<='9'){x=(x<<1)+(x<<3)+(ch^48);ch=getchar();}
x=ff?~x+1:x;
}
tpa inline void read(T& x,Args&... args){read(x);read(args...);}
inline void readc(char &c){
do{c=getchar();}while(!((c>='a'&&c<='z')||(c>='A'&&c<='Z')));
}
tpl inline void write(T x){
static T st[45];T top=0;
if(x<0)x=~x+1,putchar('-');
do{st[top++]=x%10;}while(x/=10);
while(top)putchar(st[--top]^48);
}
tpl inline void write(T x,const char c){write(x);putchar(c);}
inline void writes(const char *c){while(*c)putchar(*c++);putchar('\n');}
}
using namespace IO;
signed main(){
int m=(int)2e6;
while(m--){
int a,b;read(a,b);
write(a+b,'\n');
}
return 0;
}
5. fread 优化快读、快输(scanf 风格)(120ms):
#include <cstdio>
using namespace std;
namespace IO{
#define re register
#define getchar() (p1==p2&&(p2=(p1=ibuf)+fread(ibuf,1,buf_size,stdin),p1==p2)?EOF:*p1++)
#define putchar(x) ((p3==obuf+buf_size)&&(fwrite(obuf,p3-obuf,1,stdout),p3=obuf),*p3++=x)
#define tpl template<typename T>
#define tpa template<typename T,typename... Args>
typedef long long ll;
const ll buf_size=1<<21;
char ibuf[buf_size],*p1=ibuf,*p2=ibuf;
char obuf[buf_size],*p3=obuf;
tpl inline void read(T &x){
x=0;re bool ff=false;re char ch=getchar();
while(ch<'0'||ch>'9'){ff|=(ch=='-');ch=getchar();}
while(ch>='0'&&ch<='9'){x=(x<<1)+(x<<3)+(ch^48);ch=getchar();}
x=ff?~x+1:x;
}
tpa inline void read(T& x,Args&... args){read(x);read(args...);}
inline void readc(char &c){
do{c=getchar();}while(!((c>='a'&&c<='z')||(c>='A'&&c<='Z')));
}
tpl inline void write(T x){
static T st[45];T top=0;
if(x<0)x=~x+1,putchar('-');
do{st[top++]=x%10;}while(x/=10);
while(top)putchar(st[--top]^48);
}
tpl inline void write(T x,const char c){write(x);putchar(c);}
inline void writes(const char *c){while(*c)putchar(*c++);putchar('\n');}
}
using namespace IO;
signed main(){
int m=(int)2e6;
while(m--){
int a,b;read(a,b);
write(a+b,'\n');
}
fwrite(obuf,1,p3-obuf,stdout);
return 0;
}
6. fread 优化快读、快输(cin 风格)(96ms):
#include <bits/stdc++.h>
using namespace std;
namespace SOL {
#ifdef LOCAL
FILE*Fin(fopen("test.in","r")),*Fout(fopen("test.out","w"));
#else
FILE*Fin(stdin),*Fout(stdout);
#endif
namespace IO {
class qistream {
static const size_t SIZE=1<<16,BLOCK=32;
FILE*fp;
char buf[SIZE];
int p;
public:
qistream(FILE*_fp=stdin):fp(_fp),p(0) {
fread(buf+p,1,SIZE-p,fp);
} void flush() {
memmove(buf,buf+p,SIZE-p),fread(buf+SIZE-p,1,p,fp),p=0;
} template<class T>qistream&operator>>(T&x) {
x=0;
p+BLOCK>=SIZE?flush():void();
bool flag=false;
for(; !isdigit(buf[p]); ++p)flag=buf[p]=='-';
for(; isdigit(buf[p]); ++p)x=x*10+buf[p]-'0';
x=flag?-x:x;
return*this;
} char getch() {
return buf[p++];
} qistream&operator>>(char*str) {
char ch=getch();
while(ch<=' ')ch=getch();
for(int i=0; ch>' '; ++i,ch=getch())str[i]=ch;
return*this;
}
} qcin(Fin);
class qostream {
static const size_t SIZE=1<<16,BLOCK=32;
FILE*fp;
char buf[SIZE];
int p;
public:
qostream(FILE*_fp=stdout):fp(_fp),p(0) {}~qostream() {
fwrite(buf,1,p,fp);
} void flush() {
fwrite(buf,1,p,fp),p=0;
} template<class T>qostream&operator<<(T x) {
int len=0;
p+BLOCK>=SIZE?flush():void();
x<0?(x=-x,buf[p++]='-'):0;
do buf[p+len]=x%10+'0',x/=10,++len;
while(x);
for(int i=0,j=len-1; i<j; ++i,--j)swap(buf[p+i],buf[p+j]);
p+=len;
return*this;
} void putch(char ch) {
buf[p++]=ch;
} qostream&operator<<(const char*str) {
for(int i=0; str[i]; ++i)putch(str[i]);
return*this;
}
} qcout(Fout);
} using namespace IO;
} using namespace SOL;
signed main() {
int m=(int)2e6;
while(m--) {
int a,b;
qcin>>a>>b;
(qcout<<a+b).putch('\n');
}
return 0;
}
6. mmap 优化快读 + fread 优化快输(110ms):
mmap 需要加 #include <sys/mman.h> 头文件,考 CSP 时不能用!!!
#include <bits/stdc++.h>
#include <sys/mman.h>
using namespace std;
namespace IO{
#define putchar(x) ((p3==obuf+buf_size)&&(fwrite(obuf,p3-obuf,1,stdout),p3=obuf),*p3++=x)
#define tpl template<typename T>
#define tpa template<typename T,typename... Args>
const char*mbuf=(char*)mmap(nullptr,1<<28,PROT_READ,MAP_PRIVATE,fileno(stdin),0);
typedef long long ll;
const ll buf_size=1<<21;
char obuf[buf_size],*p3=obuf;
tpl inline void read(T& x){
x=0;T f=1;
while(*mbuf<'0'||*mbuf>'9'){if(*mbuf=='-')f=-1;mbuf++;}
while(*mbuf>='0'&&*mbuf<='9')x=x*10+*mbuf++-'0'; x*=f;
}
tpa inline void read(T& x,Args&... args){read(x);read(args...);}
inline void readc(char &c){
do{c=getchar();}while(!((c>='a'&&c<='z')||(c>='A'&&c<='Z')));
}
tpl inline void write(T x){
static T st[45];T top=0;
if(x<0)x=~x+1,putchar('-');
do{st[top++]=x%10;}while(x/=10);
while(top)putchar(st[--top]^48);
}
tpl inline void write(T x,const char c){write(x);putchar(c);}
inline void writes(const char *c){while(*c)putchar(*c++);putchar('\n');}
}
using namespace IO;
signed main(){
int m=(int)2e6;
while(m--){
int a,b;read(a,b);
write(a+b,'\n');
}
fwrite(obuf,1,p3-obuf,stdout);
return 0;
}
补充:
- 在 Linux 中可以用 getchar_unlocked() 来代替 getchar(),或者像下面这样(不过差别不大):
#define getchar() stdin->_IO_read_ptr<stdin->_IO_read_end?*stdin->_IO_read_ptr++:__uflow(stdin)
#define putchar(x) stdout->_IO_write_ptr<stdout->_IO_write_end?*stdout->_IO_write_ptr++=x:__overflow(stdout,x)
- CF不支持 #include <sys/mman.h> 头文件,所以用不了 mmap
- 读写文件时如果把 cin、cout 关闭同步/解除绑定了,一定要在最后刷新缓冲区(cout<<endl;)
至于为什么没有 mmap 实现快输,是因为我太蒻力(悲),如果有哪位大佬会请撅我