Partial Threaded Sorting in C - c

Question

I'm trying to do a partial sort with a threads,
my current output it
27 12 21 48 15 28 82 69 35 91
13 82 33 35 46 5 35 28 87 95
0 10 20 22 23 30 52 80 86 96
3 8 42 53 67 70 70 71 75 79
5 8 8 18 41 43 70 79 86 88
10 51 56 60 65 84 87 91 94 99
23 25 38 39 40 44 51 56 69 75
20 21 25 29 29 38 66 71 73 96
33 50 9 6 13 27 97 21 70 22
3 4 6 6 7 15 34 59 63 70
As you can see I am getting it partially sorted I want my output to be this (no merging at the end)
12 15 21 27 28 35 48 69 82 91
5 13 28 33 35 35 46 82 87 95
0 10 20 22 23 30 52 80 86 96
3 8 42 53 67 70 70 71 75 79
5 8 8 18 41 43 70 79 86 88
10 51 56 60 65 84 87 91 94 99
23 25 38 39 40 44 51 56 69 75
20 21 25 29 29 38 66 71 73 96
6 9 13 21 22 27 33 50 70 97
3 4 6 6 7 15 34 59 63 70
I can get the right output if instead of using a struct I use &array[i] and manually input the length
This is the code I have so far:
#include <stdio.h>
#include <unistd.h>
#include <fcntl.h>
#include <stdlib.h>
#include <pthread.h>
int cmpfunc(const void *a, const void *b) {
return (*(int*)a - *(int*)b);
}
struct values {
int *arrayptr;
int length;
};
void *thread_fn(void *a) {
struct values *start = a;
qsort(start->arrayptr, start->length, sizeof(int), cmpfunc);
return (void*)a;
}
int main(int argc, const char *argv[]) {
FILE *fp = fopen(argv[3], "r");
FILE *fp1 = fopen("numS1.dat", "w+");
//amount of threads
int threadAmount = atoi(argv[1]);
//size of input
int numberAmount = atoi(argv[2]);
//multidimensional array
int array[threadAmount][numberAmount / threadAmount];
for (int i = 0; i < threadAmount; i++)
for (int j = 0; j < numberAmount / threadAmount; j++)
fscanf(fp, "%d", &array[i][j]);
pthread_t threadid[threadAmount];
for (int i = 0; i < threadAmount; ++i) {
struct values a = { array[i], numberAmount / threadAmount };
pthread_create(&threadid[i], NULL, thread_fn, &a);
}
for (int i = 0; i < threadAmount; ++i)
pthread_join(threadid[i], NULL);
for (int i = 0; i < threadAmount; i++) {
if (i != 0)
fprintf(fp1, "\n");
for (int j = 0; j < numberAmount / threadAmount; j++)
fprintf(fp1 ,"%d ", array[i][j]);
}
return 0;
}
Do you know where I am going wrong?
I think its the struct but everything I see online does what I'm doing.

Answer 1

You are passing a pointer to automatic storage to newly created threads: the struct values object becomes invalid as soon as the calling scope is exited, thus it cannot be reliably accessed by the new thread. You should allocate the struct values and pass the pointer to the allocated object as a parameter to pthread_create:
for (int i = 0; i < threadAmount; ++i) {
struct values *a = malloc(sizeof(*a));
a->arrayptr = array[i];
a->length = numberAmount / threadAmount;
pthread_create(&threadid[i], NULL, thread_fn, a);
}
The structure can be freed by the thread function before exiting.
Notes:
the way you split the array into chunks only works if the length is a multiple of the number of threads.
the comparison function does not work for large int values, you should use this instead:
int cmpfunc(const void *a, const void *b) {
return (*(int*)b < *(int*)a) - (*(int*)a < *(int*)b);
}