C++ 对TXT 的串并行读写

任务说明：有36篇文档，现在要读入，并统计词频，字典长度25，希望能够比较串并行读写操作的时间差距。

1. 串行读入并统计词频

   // LoadDocsInUbuntu.cpp
   
   //
   
   #include <iostream>
   
   #include <stdio.h>
   
   #include <vector>
   
   using namespace std;
   
   int main()
   
   {
   
       char filename[100];
   
       size_t d;
   
       FILE *fileptr;
   
       int word;
   
       vector< vector<int> > corpus;
   
       printf("load data ...\n");
   
       for (d = 1; d < 37; d++){
   
           sprintf(filename, "..//data/doc_%d.txt", d);
   
           fileptr = fopen(filename, "r");
   
           vector<int> doc;
   
           int ff[25] = { 0 };
   
           while (fscanf(fileptr, "%d", &word) != EOF)
   
           {
   
               ff[word - 1] = ff[word - 1] + 1;
   
               doc.push_back(word);
   
           }
   
           corpus.push_back(doc);
   
           fclose(fileptr);
   
           sprintf(filename, "..//result/freqUbuntuSerial_%d.txt", d);
   
           fileptr = fopen(filename, "w");
   
           for (int f = 0; f < 25; f++)
   
           {    
   
               fprintf(fileptr, "%d ", ff[f]);
   
           }
   
           fclose(fileptr);
   
       }
   
       cout <<"corpus.size()="<< corpus.size() << endl;
   
       return 0;
   
   }
   

2. 这里讨论并行有三种思路：一，按照文档序号进行分组读入统计等操作；二，在文档内按单词数目分组进行统计；三，将统计与读写操作并行处理。

   针对第一种思路，使用openmp做多线程处理：

　　　　

// LoadDocsByOpenMP.cpp
//
#include <omp.h>
#include <iostream>
#include <stdio.h>
#include <vector>
#include <stdlib.h>
#include <time.h>
#include <string>
using namespace std;

int main()
{
	char filename[100],resultname[100];
	int d;
	FILE *fileptr[360];
	int word;
	int ff[360][25] = { 0 };
	//vector< vector<int> > corpus;
	clock_t start,finish;
	int f[360]={0};

	start=clock();
	printf("load data ...\n");
#pragma omp parallel for num_threads(4)
	for (d = 1; d < 361; d++){
		printf("Hello world, I am %d, docs index %d.\n",omp_get_thread_num(),d);
		sprintf(filename, "..//data/doc_%d.txt", d);
		fileptr[d-1] = fopen(filename, "r");
		//int ff[25]={0};
		////vector<int> doc;

		while (fscanf(fileptr[d-1], "%d", &word) != EOF)
		{
			ff[d-1][word - 1] = ff[d-1][word - 1] + 1;
			//ff[word-1]=ff[word-1]+1;
		//	//doc.push_back(word);
		}
		////corpus.push_back(doc);
		fclose(fileptr[d-1]);
		sprintf(resultname, "..//result/freqByOpenMP_%d.txt", d);//Be CAREFUL!For the name "filename" has been used before, we must name the string differently here.
		fileptr[d-1] = fopen(resultname, "w");
		for (f[d-1] = 0; f[d-1] < 25; f[d-1]++)
		{

			fprintf(fileptr[d-1], "%d ", ff[f[d-1]]);
		}
		fclose(fileptr[d-1]);
	}

	//cout <<"corpus.size()="<< corpus.size() << endl;
	finish=clock();
	cout<<"time cost : "<< (double)(finish-start)/ CLOCKS_PER_SEC<<endl;
	return 0;
}

但初步比较openmp对串行读取的速度并没有太多提升，反而是当进程数多于系统物理核数的时候，程序时间会加长。

另外两种实现思路在后续学习中继续实现。