надо сравнивать
$ time ./a.out
и
$ time LD_PRELOAD=/usr/local/lib/libjemalloc.so ./a.out

/* gcc memtest.c -lrt */
#include <stdio.h>
#include <unistd.h>
#include <pthread.h>
#include <stdlib.h>
#include <string.h>

int block_size =  256;
int array_size =  1024*1024;
void* thread (void *v)
{
        long i;
        int **array = malloc(array_size * sizeof(int*));
        for(i = 0;  i < array_size; i++){
                int *s = malloc(block_size);
                int j;
                for(j = 0; j <block_size/sizeof(int); j++)
                        s[j] = j;
                array[i] = s;
                if(array_size >= 10 && !(i%(array_size/10))){
                        float percent = (float)i/array_size*100.0;
                        printf("\t thread %x: %f%%%% creation done\n", (unsigned)pthread_self(), percent);
                }
        }
        for(i = 0;  i < array_size; i++){
                int *s = array[i];
                int j;
                for(j = 0; j <block_size/sizeof(int); j++)
                        if(s[j] != j)
                                printf("\t MEMERROR in thread #%x\n", (unsigned)pthread_self());
                free(s);
                if(array_size >= 10 && !(i%(array_size/10))){
                        float percent = (float)i/array_size*100.0;
                        printf("\t thread %x: %f%%%% deletion done\n", (unsigned)pthread_self(), percent);
                }
        }
        printf("\tfinished thread #%x\n", (unsigned)pthread_self());
        free(array);
        return NULL;
}
int main (int argc, char** argv)
{
        int thread_count = 5;
        if(argc > 1)
                block_size = atoi(argv[1]);
        if(argc > 2)
                array_size = atoi(argv[2]);
        if(argc > 3)
                thread_count = atoi(argv[3]);
        printf("malloc performance tester\nUsage:  %s [block_size] [array_size] [thread_count]\n", argv[0]);
        printf("going to allocate array of blocks in each of the %d threads [block_size = %d,"
                        "array_size = %d] \n", thread_count, block_size, array_size);
        pthread_t pids[thread_count];
        memset(pids, 0, sizeof(pids));
        int i;
        for(i = 0; i < thread_count; i++)
                if (-1 == pthread_create(pids + i, NULL, &thread, NULL)){
                        perror("creation of the thread failed");
                        exit(-1);
                }
        for(i = 0; i < thread_count; i++){
                pthread_join(pids[i], NULL);
                printf("joined thread #%x\n", (unsigned)pids[i]);
        }
        printf("finished main thread #%x\n", (unsigned)pthread_self());
        return 0;
}