Examples

Using FTI_Snapshot
Using FTI Checkpoint

---

Using FTI_Snapshot

⬆️ Top

#include <stdlib.h>
#include <fti.h>

int main(int argc, char** argv){
    MPI_Init(&argc, &argv);
    char* path = "config.fti"; //config file path
    FTI_Init(path, MPI_COMM_WORLD);
    int world_rank, world_size; //FTI_COMM rank & size
    MPI_Comm_rank(FTI_COMM_WORLD, &world_rank);
    MPI_Comm_size(FTI_COMM_WORLD, &world_size);

    int *array = malloc(sizeof(int) * world_size);
    int number = world_rank;
    int i = 0;
    //adding variables to protect
    FTI_Protect(1, &i, 1, FTI_INTG);
    FTI_Protect(2, &number, 1, FTI_INTG);
    for (; i < 100; i++) {
        FTI_Snapshot();
        MPI_Allgather(&number, 1, MPI_INT, array,
		      1, MPI_INT, FTI_COMM_WORLD);
	number += 1;
    }
    free(array);
    FTI_Finalize();
    MPI_Finalize();
    return 0;
}

DESCRIPTION

FTI_Snapshot() makes a checkpoint by given time and also recovers data after a failure, thus makes the code shorter. Checkpoints intervals can be set in configuration file (see: ckpt_L1 - ckpt_L4).

Using FTI_Checkpoint

⬆️ Top

#include <stdlib.h>
#include <fti.h>
#define ITER_CHECK 10

int main(int argc, char** argv){
    MPI_Init(&argc, &argv);
    char* path = "config.fti"; //config file path
    FTI_Init(path, MPI_COMM_WORLD);
    int world_rank, world_size; //FTI_COMM rank & size
    MPI_Comm_rank(FTI_COMM_WORLD, &world_rank);
    MPI_Comm_size(FTI_COMM_WORLD, &world_size);

    int *array = malloc(sizeof(int) * world_size);
    int number = world_rank;
    int i = 0;
    //adding variables to protect
    FTI_Protect(1, &i, 1, FTI_INTG);
    FTI_Protect(2, &number, 1, FTI_INTG);
    if (FTI_Status() != 0) {
        FTI_Recover();
    }
    for (; i < 100; i++) {
        if (i % ITER_CHECK == 0) {
            FTI_Checkpoint(i / ITER_CHECK + 1, 2);
        }
        MPI_Allgather(&number, 1, MPI_INT, array,
		      1, MPI_INT, FTI_COMM_WORLD);
	number += 1;
    }
    free(array);
    FTI_Finalize();
    MPI_Finalize();
    return 0;
}

DESCRIPTION

FTI_Checkpoint() allows to checkpoint at precise application intervals. Note that when using FTI_Checkpoint(), ckpt_L1, ckpt_L2, ckpt_L3 and ckpt_L4 are not taken into account.