Taylor series using OpenMP - c

Question

ex = 1 + x + x2/2! + x3/3! + x4/4! + x5/5! +...
I have converted a Taylor series of ex (above) into a program of OpenMp.
All the codes are written below.
When I run the code through Oracle Ubuntu it works.
It is giving me e^0=1,e^1=2.718,e^2=7.389056
But when I run it on Ubuntu (not virtually), then it doesn't work right.
It is giving me e^0=nan,e^1=0.40..,e^2=4.780.
And output is totally random as in its not exact as I mentioned above.
I need help.
#include <math.h>
#include <pthread.h>
#include <stdlib.h>
long double x, fact[150], pwr[150], s[1];
int i, term;
void *Power(void *temp) {
int k;
for (k = 0; k < 150; k++) {
pwr[k] = pow(x, k);
//printf("%.2Lf\n", pwr[k]);
}
return pwr;
}
void *Fact(void *temp) {
long double f;
int j;
fact[0] = 1.0;
for (term = 1; term < 150; term++) {
f = 1.0;
for (j = term; j > 0; j--)
f = f * j;
fact[term] = f;
//printf("%.2Lf\n", fact[term]);
}
return fact;
}
void *Exp(void *temp) {
int t;
s[0] = 0;
for (t = 0; t < 150; t++)
s[0] = s[0] + (pwr[t] / fact[t]);
return s;
}
int main(void) {
pthread_t thread1, thread2, thread3;
long double **sum;
printf("Exponential [PROMPT] Enter the value of x (between 0 to 100) (for calculating exp(x)):");
scanf("%Lf", &x);
printf("\nExponential [INFO] Threads creating.....\n");
pthread_create(&thread1, NULL, Power, NULL); //calling power function
pthread_create(&thread2, NULL, Fact, NULL); //calling factorial function
printf("Exponential [INFO] Threads created\n");
pthread_join(thread1, NULL);
pthread_join(thread2, NULL);
printf("Exponential [INFO] Master thread and terminated threads are joining\n");
printf("Exponential [INFO] Result collected in Master thread\n");
pthread_create(&thread3, NULL, Exp, NULL);
pthread_join(thread3, sum);
printf("\neXPONENTIAL [INFO] Value of exp(%.2Lf) is : %Lf\n\n", x, s[0]);
exit(1);
}
The above code is originally for ex using threads which works.
#include <math.h>
#include <omp.h>
#include <stdio.h>
#include <stdlib.h>
int main(void) {
long double x, f, fact[150], pwr[150], s[1];
int i, term, k, j, t;
long double sum;
printf("Exponential [PROMPT] Enter the value of x (between 0 to 100) (for calculating exp(x)):");
scanf("%Lf", &x);
#pragma omp parallel num_threads(10)
#pragma omp for
for (k = 0; k < 150; k++) {
for (int h = 0; h <= k; h++) {
if (h == 0)
x = 1;
else
pwr[k] = pow(x, k);
}
#pragma omp for
for (term = 1; term < 150; term++) {
f = 1.0;
for (j = term; j > 0; j--)
f = f * j;
fact[term] = f;
}
#pragma omp for
for (t = 0; t < 150; t++)
s[0] = s[0] + (pwr[t] / fact[t]);
printf("\neXPONENTIAL [INFO] Value of exp(%.2Lf) is : %Lf\n\n", x, s[0]);
exit(1);
}
And the code above is a conversion of the previous code to an OpenMP.

Answer 1

for (t = 0; t < 150; t++)
s[0] = s[0] + (pwr[t] / fact[t]);
This code, when parallelized, will overwrite the same variable concurrently with partial calculation results. This can only work when the threads are coordinated somehow. Fortunately, openmp has a dedicated directive reduce for calculating sums, so you can fix this easily.
In the pthread version of the code, one thread does this calculation, so no problem there.