-
Notifications
You must be signed in to change notification settings - Fork 5
/
parSort.cpp
191 lines (165 loc) · 4.38 KB
/
parSort.cpp
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
#include "parSort.h"
#include <math.h>
#include <iostream>
#include <omp.h>
using namespace std;
/**********************************************************
* sample sort (returns rank)
/**********************************************************/
void sampleSort(float * a, int *index, int n) {
int i, j, k, p, start, end, avg, split, * core, * coreCount, * newIndex;
float * splitters, * newA;
int * rank = (int *)malloc(sizeof(int) * n);
#pragma omp parallel shared(a, n, p, avg, splitters, core, coreCount, rank, index) private(i, j, k, start, end, split, newA, newIndex)
{
/* determine section of array to sort */
p = omp_get_num_threads();
avg = (int)floor((float)n / (float)p);
start = (int)omp_get_thread_num() * avg;
if (omp_get_thread_num() == (p-1)) {
end = n;
splitters = (float *)malloc(sizeof(float) * p * p);
}
else {
end = (omp_get_thread_num() + 1) * avg;
}
#pragma omp barrier
//printf("tid %d :: start = %d, end = %d\n", omp_get_thread_num(), start, end);
/* conduct sequential sort on "owned" portion */
seqShellSort((a + start), (index + start), (end - start));
/* determine personal splitters and save to shared location */
split = (int)floor((float)(end - start) / (float)p);
j = (omp_get_thread_num() + 1) * p - 1;
for(i=(end-1); i>start; i-=split) {
*(splitters + j) = *(a + i);
j--;
}
#pragma omp barrier
if(omp_get_thread_num() == 0) {
seqShellSort(splitters, p*p);
/*printf("splitters = [%f,", *(splitters + 0));
for(i=1; i<(p*p - 1); i++) {
printf("%f, ", *(splitters + i));
}
printf("%f]\n", *(splitters + (p*p) - 1));*/
j = 0;
for(i=(p-1); i<(p*p); i+=p){
*(splitters + j) = *(splitters + i);
j++;
}
/*printf("splitters = [%f,", *(splitters + 0));
for(i=1; i<(p-1); i++) {
printf("%f, ", *(splitters + i));
}
printf("%f]\n", *(splitters + p - 1));*/
core = (int *) malloc(sizeof(int)*n);
coreCount = (int *) malloc(sizeof(int) * p);
for(i=0; i<p; i++) {
*(coreCount + i) = 0;
}
}
#pragma omp barrier
#pragma omp for
for(i=0; i<n; i++) {
j = 0;
while(*(a + i) > *(splitters + j)) {
j++;
}
*(core + i) = j;
}
/*if(omp_get_thread_num() == 0) {
printf("core = [%d, ", *(core + 0));
for(i=1; i<(n-1); i++) {
printf("%d, ", *(core + i));
}
printf("%d]\n", *(core + n - 1));
}*/
#pragma omp barrier
newA = (float *)malloc(sizeof(float) * 2 * avg);
newIndex = (int *)malloc(sizeof(float) * 2 * avg);
j = 0;
k = omp_get_thread_num();
for(i=0; i<n; i++) {
if(*(core + i) == k) {
*(newA + j) = *(a + i);
*(newIndex + j) = *(index + i);
j++;
}
}
*(coreCount + k) = j;
/*#pragma omp barrier
if(omp_get_thread_num() == 0) {
printf("coreCount = [%d, ", *(coreCount + 0));
for(i=1; i<(p-1); i++) {
printf("%d, ", *(coreCount + i));
}
printf("%d]\n", *(coreCount + p - 1));
}*/
#pragma omp barrier
seqShellSort(newA, newIndex, j);
start = 0;
for(i=0; i<k; i++) {
start += *(coreCount + i);
}
#pragma omp barrier
/* each process copies its data items back to the full shared array */
for(i=0; i<j; i++) {
*(a + start + i) = *(newA + i);
*(index + start + i) = *(newIndex + i);
}
free(newA);
free(newIndex);
/** Determine Rank given INDEX array */
#pragma omp barrier
#pragma omp for
for(i = 0; i<n; i++) {
*(rank + *(index + i)) = i;
}
}// END parallel region
#pragma omp parallel for
for(i = 0; i < n; i++) {
*(index + i) = *(rank + i);
}
}
/**********************************************************
* sequential Shell Sorts
/**********************************************************/
void seqShellSort(float *a, int *index, int n){
//printf("tid = %d, n = %d\n", omp_get_thread_num(), n);
int inc, i, j, k, tempI;
float tempF;
inc = n/2;
while (inc > 0) {
for(i=inc; i<n; i++) {
tempF = *(a + i);
tempI = *(index + i);
j = i;
while((j >= inc) && (*(a + j - inc) > tempF)) {
*(a+j) = *(a + j - inc);
*(index + j) = *(index + j - inc);
j = j - inc;
}
*(a + j) = tempF;
*(index + j) = tempI;
}
inc = (int)floor((inc / 2.2) + 0.5);
}
}
void seqShellSort(float *a, int n){
//printf("tid = %d, n = %d\n", omp_get_thread_num(), n);
int inc, i, j, k;
float tempF;
inc = n/2;
while (inc > 0) {
for(i=inc; i<n; i++) {
tempF = *(a + i);
j = i;
while((j >= inc) && (*(a + j - inc) > tempF)) {
*(a+j) = *(a + j - inc);
j = j - inc;
}
*(a + j) = tempF;
}
inc = (int)floor((inc / 2.2) + 0.5);
}
}