Related
I'm exploring MPI in C++ and I wanted to parallelize the creation of a picture of the Mandelbrot set. I'm using the ppm format. Each processor builds its part and sends it back to the main process that receives it as MPI_CHAR. This is the code:
#include "mpi.h"
#include <iostream>
#include <string>
#include <fstream>
#include <complex>
using namespace std;
int mandelbrot(int x, int y, int width, int height, int max) {
complex<float> point((float) (y - height/2.0) * 4.0/width, (float) (x - width/2.0) * 4.0/width);
complex<float> z(0, 0);
unsigned int iteration = 0;
while (abs(z) < 4 && iteration < max) {
z = z * z + point;
iteration++;
}
return iteration;
}
int main(int argc, char **argv) {
int numprocs;
int myid;
int buff_size = 404270; // 200x200
char buff[buff_size];
int i;
MPI_Status stat;
MPI_Init(&argc,&argv);
MPI_Comm_size(MPI_COMM_WORLD,&numprocs);
MPI_Comm_rank(MPI_COMM_WORLD,&myid);
int width = 200, height = 200, max_iter = 1000;
if (myid == 0) {
ofstream image("mandel.ppm");
image << "P3\n" << width << " " << height << " 255\n";
for(i=1; i < numprocs; i++) {
MPI_Probe(i, 0, MPI_COMM_WORLD, &stat);
int length;
MPI_Get_count(&stat, MPI_CHAR, &length);
MPI_Recv(buff, length, MPI_CHAR, i, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
image << buff;
}
} else {
stringstream ss;
// proc rank: 1, 2, ..., n
int part = height/(numprocs-1), start = (myid - 1) * part, end = part * myid;
printf("%d -> %d\n", start, end);
for (int row = start; row < end; row++) {
for (int col = 0; col < width; col++) {
int iteration = mandelbrot(row, col, width, height, max_iter);
if (row == start) ss << 255 << ' ' << 255 << ' ' << 255 << "\n";
else if (iteration < max_iter) ss << iteration * 255 << ' ' << iteration * 20 << ' ' << iteration * 5 << "\n";
else ss << 0 << ' ' << 0 << ' ' << 0 << "\n";
}
}
printf("\n sizeof = %d\n", ss.str().length());
MPI_Send(ss.str().c_str(), ss.str().length(), MPI_CHAR, 0, 0, MPI_COMM_WORLD);
}
MPI_Finalize();
return 0;
}
Code compilation:
$ mpic++ -std=c++0x mpi.mandel.cpp -o mpi.mandel
Running with 3 processes (process main + process rank 1 and 2)
$ mpirun -np 3 ./mpi.mandel
Resulting ppm pictures when running with 3, 4, and 5 process:
It seems that the point-to-point communication of sending-receiving is mixing the results when more than 3 processes try to send the MPI_CHAR elements to the main process. How can avoid this behavior?
It works when creating the buffer buff with the same length as the receiving message:
.
.
for (int i=1; i < numprocs; i++) {
MPI_Probe(i, 0, MPI_COMM_WORLD, &stat);
int length;
MPI_Get_count(&stat, MPI_CHAR, &length);
printf("\nfrom %d <<-- %d (stat.source=%d) Receiving %d chars\n", myid, i, stat.MPI_SOURCE, length);
char buff[length + 1];
MPI_Recv(buff, length, MPI_CHAR, i, 0, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
buff[length] = '\0';
image << buff;
}
.
.
Thus, we don't need anymore the declaration at the beginning int buff_size = 404270; neither char buff[buff_size];
I would like to use a shared memory between processes. I tried MPI_Win_allocate_shared but it gives me a strange error when I execute the program:
Assertion failed in file ./src/mpid/ch3/include/mpid_rma_shm.h at line 592: local_target_rank >= 0
internal ABORT
Here's my source:
# include <stdlib.h>
# include <stdio.h>
# include <time.h>
# include "mpi.h"
int main ( int argc, char *argv[] );
void pt(int t[], int s);
int main ( int argc, char *argv[] )
{
int rank, size, shared_elem = 0, i;
MPI_Init ( &argc, &argv );
MPI_Comm_rank ( MPI_COMM_WORLD, &rank );
MPI_Comm_size ( MPI_COMM_WORLD, &size );
MPI_Win win;
int *shared;
if (rank == 0) shared_elem = size;
MPI_Win_allocate_shared(shared_elem*sizeof(int), sizeof(int), MPI_INFO_NULL, MPI_COMM_WORLD, &shared, &win);
if(rank==0)
{
MPI_Win_lock(MPI_LOCK_EXCLUSIVE, 0, MPI_MODE_NOCHECK, win);
for(i = 0; i < size; i++)
{
shared[i] = -1;
}
MPI_Win_unlock(0,win);
}
MPI_Barrier(MPI_COMM_WORLD);
int *local = (int *)malloc( size * sizeof(int) );
MPI_Win_lock(MPI_LOCK_SHARED, 0, 0, win);
for(i = 0; i < 10; i++)
{
MPI_Get(&(local[i]), 1, MPI_INT, 0, i,1, MPI_INT, win);
}
printf("processus %d (avant): ", rank);
pt(local,size);
MPI_Win_unlock(0,win);
MPI_Win_lock(MPI_LOCK_EXCLUSIVE, 0, 0, win);
MPI_Put(&rank, 1, MPI_INT, 0, rank, 1, MPI_INT, win);
MPI_Win_unlock(0,win);
MPI_Win_lock(MPI_LOCK_SHARED, 0, 0, win);
for(i = 0; i < 10; i++)
{
MPI_Get(&(local[i]), 1, MPI_INT, 0, i,1, MPI_INT, win);
}
printf("processus %d (apres): ", rank);
pt(local,size);
MPI_Win_unlock(0,win);
MPI_Win_free(&win);
MPI_Free_mem(shared);
MPI_Free_mem(local);
MPI_Finalize ( );
return 0;
}
void pt(int t[],int s)
{
int i = 0;
while(i < s)
{
printf("%d ",t[i]);
i++;
}
printf("\n");
}
I get the following result:
processus 0 (avant): -1 -1 -1 -1 -1 -1 -1 -1 -1 -1
processus 0 (apres): 0 -1 -1 -1 -1 -1 -1 -1 -1 -1
processus 4 (avant): 0 -1 -1 -1 -1 -1 -1 -1 -1 -1
processus 4 (apres): 0 -1 -1 -1 4 -1 -1 -1 -1 -1
Assertion failed in file ./src/mpid/ch3/include/mpid_rma_shm.h at line 592: local_target_rank >= 0
internal ABORT - process 5
Assertion failed in file ./src/mpid/ch3/include/mpid_rma_shm.h at line 592: local_target_rank >= 0
internal ABORT - process 6
Assertion failed in file ./src/mpid/ch3/include/mpid_rma_shm.h at line 592: local_target_rank >= 0
internal ABORT - process 9
Can someone please help me figure out what's going wrong & what that error means ? Thanks a lot.
MPI_Win_allocate_shared is a departure from the very abstract nature of MPI. It exposes the underlying memory organisation and allows the programs to bypass the expensive (and often confusing) MPI RMA operations and utilise the shared memory directly on systems that have such. While MPI typically deals with distributed-memory environments where ranks do not share the physical memory address space, a typical HPC system nowadays consists of many interconnected shared-memory nodes. Thus, it is possible for ranks that execute on the same node to attach to shared memory segments and communicate by sharing data instead of message passing.
MPI provides a communicator split operation that allows one to create subgroups of ranks such that the ranks in each subgroup are able to share memory:
MPI_Comm_split_type(comm, MPI_COMM_TYPE_SHARED, key, info, &newcomm);
On a typical cluster, this essentially groups the ranks by the nodes they execute on. Once the split is done, a shared-memory window allocation can be executed over the ranks in each newcomm. Note that for a multi-node cluster job this will result in several independent newcomm communicators and thus several shared memory windows. Ranks on one node won't (and shouldn't) be able to see the shared memory windows on other nodes.
In that regard, MPI_Win_allocate_shared is a platform-independent wrapper around the OS-specific mechanisms for shared memory allocation.
There are several problems with this code and the usage. Some of these are mentioned in #Hristolliev's answer.
you have to run all the processes in the same node to have a intranode communicator or use "communicator split shared".
you need to run this code with at least 10 processes.
Third, local should be deallocated with free().
you should get the shared pointer from a query.
you should deallocate shared (I think this is taken care by Win_free)
This is the resulting code:
# include <stdlib.h>
# include <stdio.h>
# include <time.h>
# include "mpi.h"
int main ( int argc, char *argv[] );
void pt(int t[], int s);
int main ( int argc, char *argv[] )
{
int rank, size, shared_elem = 0, i;
MPI_Init ( &argc, &argv );
MPI_Comm_rank ( MPI_COMM_WORLD, &rank );
MPI_Comm_size ( MPI_COMM_WORLD, &size );
MPI_Win win;
int *shared;
// if (rank == 0) shared_elem = size;
// MPI_Win_allocate_shared(shared_elem*sizeof(int), sizeof(int), MPI_INFO_NULL, MPI_COMM_WORLD, &shared, &win);
if (rank == 0)
{
MPI_Win_allocate_shared(size, sizeof(int), MPI_INFO_NULL,
MPI_COMM_WORLD, &shared, &win);
}
else
{
int disp_unit;
MPI_Aint ssize;
MPI_Win_allocate_shared(0, sizeof(int), MPI_INFO_NULL,
MPI_COMM_WORLD, &shared, &win);
MPI_Win_shared_query(win, 0, &ssize, &disp_unit, &shared);
}
if(rank==0)
{
MPI_Win_lock(MPI_LOCK_EXCLUSIVE, 0, MPI_MODE_NOCHECK, win);
for(i = 0; i < size; i++)
{
shared[i] = -1;
}
MPI_Win_unlock(0,win);
}
MPI_Barrier(MPI_COMM_WORLD);
int *local = (int *)malloc( size * sizeof(int) );
MPI_Win_lock(MPI_LOCK_SHARED, 0, 0, win);
for(i = 0; i < 10; i++)
{
MPI_Get(&(local[i]), 1, MPI_INT, 0, i,1, MPI_INT, win);
}
printf("processus %d (avant): ", rank);
pt(local,size);
MPI_Win_unlock(0,win);
MPI_Win_lock(MPI_LOCK_EXCLUSIVE, 0, 0, win);
MPI_Put(&rank, 1, MPI_INT, 0, rank, 1, MPI_INT, win);
MPI_Win_unlock(0,win);
MPI_Win_lock(MPI_LOCK_SHARED, 0, 0, win);
for(i = 0; i < 10; i++)
{
MPI_Get(&(local[i]), 1, MPI_INT, 0, i,1, MPI_INT, win);
}
printf("processus %d (apres): ", rank);
pt(local,size);
MPI_Win_unlock(0,win);
MPI_Win_free(&win);
// MPI_Free_mem(shared);
free(local);
// MPI_Free_mem(local);
MPI_Finalize ( );
return 0;
}
void pt(int t[],int s)
{
int i = 0;
while(i < s)
{
printf("%d ",t[i]);
i++;
}
printf("\n");
}
So I have a some code where I am using MPI_Bcast to send information from the root node to all nodes, but instead I want to get my P0 to send chunks of the array to individual processes.
How do I do this with MPI_Send and MPI_Receive?
I've never used them before and I don't know if I need to loop my MPI_Receive to effectively send everything or what.
I've put giant caps lock comments in the code where I need to replace my MPI_Bcast(), sorry in advance for the waterfall of code.
Code:
#include "mpi.h"
#include <stdio.h>
#include <math.h>
#define MAXSIZE 10000000
int add(int *A, int low, int high)
{
int res = 0, i;
for(i=low; i<=high; i++)
res += A[i];
return(res);
}
int main(argc,argv)
int argc;
char *argv[];
{
int myid, numprocs, x;
int data[MAXSIZE];
int i, low, high, myres, res;
double elapsed_time;
MPI_Init(&argc,&argv);
MPI_Comm_size(MPI_COMM_WORLD,&numprocs);
MPI_Comm_rank(MPI_COMM_WORLD,&myid);
if (myid == 0)
{
for(i=0; i<MAXSIZE; i++)
data[i]=1;
}
/* star the timer */
elapsed_time = -MPI_Wtime();
//THIS IS WHERE I GET CONFUSED ABOUT MPI_SEND AND MPI_RECIEVE!!!
MPI_Bcast(data, MAXSIZE, MPI_INT, 0, MPI_COMM_WORLD);
x = MAXSIZE/numprocs;
low = myid * x;
high = low + x - 1;
if (myid == numprocs - 1)
high = MAXSIZE-1;
myres = add(data, low, high);
printf("I got %d from %d\n", myres, myid);
MPI_Reduce(&myres, &res, 1, MPI_INT, MPI_SUM, 0, MPI_COMM_WORLD);
/* stop the timer*/
elapsed_time += MPI_Wtime();
if (myid == 0)
printf("The sum is %d, time taken = %f.\n", res,elapsed_time);
MPI_Barrier(MPI_COMM_WORLD);
printf("The sum is %d at process %d.\n", res,myid);
MPI_Finalize();
return 0;
}
You need MPI_Scatter. A good intro is here: http://mpitutorial.com/tutorials/mpi-scatter-gather-and-allgather/
I think in your code it could look like this:
elements_per_proc = MAXSIZE/numprocs;
// Create a buffer that will hold a chunk of the global array
int *data_chunk = malloc(sizeof(int) * elements_per_proc);
MPI_Scatter(data, elements_per_proc, MPI_INT, data_chunk,
elements_per_proc, MPI_INT, 0, MPI_COMM_WORLD);
If you really want use MPI_Send and MPI_Recv, then you can use something like this:
int x = MAXSIZE / numprocs;
int *procData = new int[x];
if (rank == 0) {
for (int i = 1; i < num; i++) {
MPI_Send(data + i*x, x, MPI_INT, i, 0, MPI_COMM_WORLD);
}
} else {
MPI_Recv(procData, x, MPI_INT, 0, 0, MPI_COMM_WORLD, &status);
}
I'm trying to divide the string str_data1 and send it to the slave processors in MPI_COMM_WORLD, but I am getting an error on the slaves.
The error looks something like this:
2
2
3
�E0� �E0�� �E0O�
#include <stdio.h>
#include <string.h>
#include "mpi.h"
int main(int argc, char* argv[]) {
int rank;
int p;
MPI_Status status;
int msg_size = 0;
MPI_Init(&argc, &argv);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &p);
char *str_data1 = "Samsung Galaxy Tab S 10.5 Wi-Fi, Tablet PC Android";
int len1 = strlen(str_data1), i;
char *str1[len1];
char *a[len1];
if (rank == 0) {
char *ds = strdup(str_data1);
int n = 0;
a[n] = strtok(ds, " ,");
while (a[n] && n < len1) {
a[++n] = strtok(NULL, " ,");
}
int chunk = n / p;
int str_size = chunk;
for (i = 1; i < p; i++) {
if (i == p - 1) {
str_size = n - chunk * i;
}
MPI_Send(&str_size, 1, MPI_INT, i, 0, MPI_COMM_WORLD);
MPI_Send(&a, str_size + 1, MPI_CHAR, i, 0, MPI_COMM_WORLD);
}
} else {
MPI_Recv(&msg_size, 1, MPI_INT, 0, 0, MPI_COMM_WORLD, &status);
char messagsg_size];
printf(" \n %d ", msg_size);
MPI_Recv(&message, msg_size + 1, MPI_CHAR, 0, 0, MPI_COMM_WORLD,
&status);
printf(" \n %s ", message);
}
MPI_Finalize();
return 0;
}
does anyone have any clues what im doing wrong? Thanks.
Some newer compilers will give you nice warning about these sorts of things. It's been very nice ever since Clang added this. If you were to use that compiler, you'd see this:
$ mpic++ asdf.c
clang: warning: treating 'c' input as 'c++' when in C++ mode, this behavior is deprecated
asdf.c:17:23: warning: conversion from string literal to 'char *' is deprecated [-Wc++11-compat-deprecated-writable-strings]
char *str_data1 = "Samsung Galaxy Tab S 10.5 Wi-Fi, Tablet PC Android";
^
asdf.c:39:22: warning: argument type 'char *(*)[len1]' doesn't match specified 'MPI' type tag that requires 'char *' [-Wtype-safety]
MPI_Send(&a, str_size + 1, MPI_CHAR, i, 0, MPI_COMM_WORLD);
^~ ~~~~~~~~
asdf.c:47:18: warning: argument type 'char (*)[msg_size]' doesn't match specified 'MPI' type tag that requires 'char *' [-Wtype-safety]
MPI_Recv(&message, msg_size + 1, MPI_CHAR, 0, 0, MPI_COMM_WORLD,
^~~~~~~~ ~~~~~~~~
3 warnings generated.
That shows you that you're using the wrong types for your character arrays. You should be using either just character arrays:
char a[len1];
or character pointers:
char *a;
Either way, you need to do some code rework to make that work correctly. Specifically, this section:
char *ds = strdup(str_data1);
int n = 0;
a[n] = strtok(ds, " ,");
while (a[n] && n < len1) {
a[++n] = strtok(NULL, " ,");
}
I don't think that string tokenizer is doing what you think it is since it would just be overwriting itself over and over.
Thanks Wesly.. i tired with structure.. looks like its working..
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include<mpi.h>
struct tokan {
char buff[30];
} t[50];
int main(int argc, char* argv[]) {
int rank;
int p;
MPI_Status status;
int msg_size = 0, i = 0, j = 0, msg_size1 = 0;
MPI_Init(&argc, &argv);
MPI_Comm_rank(MPI_COMM_WORLD, &rank);
MPI_Comm_size(MPI_COMM_WORLD, &p);
int n = 0, m = 0, k = 0, N = 0, M = 0;
char *str1, *str2, *str_data2, *str_data1;
if (rank == 0) {
str_data1 =
"processes in this communicator will now abort, and potentially your MPI job";
str_data2 =
"The behavior is undefined if lhs or rhs are not pointers to null-terminated byte strings";
str1 = (char *) malloc(strlen(str_data1) * sizeof(char));
str2 = (char *) malloc(strlen(str_data2) * sizeof(char));
strcpy(str1, str_data1);
strcpy(str2, str_data2);
int len;
M = strlen(str2);
N = strlen(str1);
char *ptr;
k = 0;
char *ds = strdup(str_data1);
ptr = strtok(ds, " ,");
while (ptr != NULL) {
++n;
len = strlen(ptr);
for (j = 0; j < len; j++) {
t[k].buff[j] = *(ptr + j);
}
//printf(" %s ", t[k].buff);
k++;
ptr = strtok(NULL, " ,");
}
int chunk = n / p;
int str_size = chunk, cnt = 0;
j = chunk;
for (i = 1; i < p; i++) {
if (i == p - 1) {
str_size = n - chunk * i;
}
MPI_Send(&str_size, 1, MPI_INT, i, 0, MPI_COMM_WORLD);
for (cnt = 0; cnt < str_size; cnt++) {
/*printf("process 0: %c %d %d %d %d %d %d\n", t[j].buff, str_size, n,
chunk, cnt, j, l);*/
MPI_Send(t[j].buff, 100, MPI_CHAR, i, 0, MPI_COMM_WORLD);
j++;
}
}
str_size = chunk;
for (i = 1; i < p; i++) {
MPI_Send(&M, 1, MPI_INT, i, 0, MPI_COMM_WORLD);
//printf("process 0: %s %d %d %d %d\n", str2, n, chunk, cnt, j);
MPI_Send(str2, M, MPI_CHAR, i, 0, MPI_COMM_WORLD);
}
} else {
MPI_Recv(&msg_size, 1, MPI_INT, 0, 0, MPI_COMM_WORLD, &status);
int k;
for (k = 0; k < msg_size; k++) {
MPI_Recv(t[k].buff, 100, MPI_CHAR, 0, 0, MPI_COMM_WORLD, &status);
printf(" \nstr1: %s %d %d %d\n", t[k].buff, rank, msg_size, k);
}
printf("***************");
MPI_Recv(&M, 1, MPI_INT, 0, 0, MPI_COMM_WORLD, &status);
str2 = (char *) malloc((M + 1) * sizeof(char));
MPI_Recv(str2, M, MPI_CHAR, 0, 0, MPI_COMM_WORLD, &status);
str2[M] = '\0';
printf(" \nstr2: %s %d %d %d \n", str2, rank, k, M);
}
if (rank == 0)
free(str1);
free(str2);
MPI_Finalize();
return 0;
}
I have a very weird problem. I've written a con in mpi that one process should print something, but amazingly the code is terminated without any output. I can't understand where it's wrong...
PS: this code is supposed to multiply two matrices.
int main( int argc, char *argv[] )
{
int M = atoi(argv[1]);
// N = 2 ^ M
N = (unsigned int) pow (2.0, M); //you need to modify this code!
int my_rank, comm_sz,mt;
int rows,offset,extra,averow ,dest;
int i,j,k;
srand(time(0));
time_t t1, t2;
double dt; //t2-t1
double tavg=0.0;
//input array
A = (double*) malloc ( sizeof(double) * N * N );
B = (double*) malloc ( sizeof(double) * N * N );
C = (double*) malloc ( sizeof(double) * N * N );
//int r; for (r = 0; r < REP; r++)
//{
//fill in matrix A and B with random numbers
//t1
//t1 = time(0);
MPI_Init(&argc, &argv);
MPI_Comm_size(MPI_COMM_WORLD, &comm_sz);
MPI_Comm_rank(MPI_COMM_WORLD, &my_rank);
if (my_rank =0){
printf("mpi_mm has started with %d tasks.\n",comm_sz);
printf("Initializing arrays...\n");
fillmatrix(A,N);
fillmatrix(B,N);
averow = N/comm_sz;
extra = N%comm_sz;
offset = 0;
mt = 0;
MPI_Bcast(B,N*N,MPI_DOUBLE,0,MPI_COMM_WORLD);
for ( dest=1;dest<=comm_sz;dest++){
rows = (dest <=extra) ? averow+1 : averow;
MPI_Send(&offset,1,MPI_INT,dest,mt,MPI_COMM_WORLD);
MPI_Send(&rows,1,MPI_INT,dest,mt,MPI_COMM_WORLD);
MPI_Send(&A[offset*N],rows*N,MPI_DOUBLE,dest,mt,MPI_COMM_WORLD);
offset=offset+rows;}
mt = 1;
for (i=1; i<=comm_sz; i++){
MPI_Recv(&offset, 1, MPI_INT, i, mt, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
MPI_Recv(&rows, 1, MPI_INT, i, mt, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
MPI_Recv(&C[offset*N], rows*N, MPI_DOUBLE, i, mt, MPI_COMM_WORLD, MPI_STATUS_IGNORE);
printf("Received results from task %d\n",i);
}
/* Print results */
printf("******************************************************\n");
printf("Result Matrix:\n");
for (i=0; i<N; i++)
{
printf("\n");
for (j=0; j<N; j++)
printf("%6.2f ", C[i*N+j]);
}
printf("\n******************************************************\n");
printf ("Done.\n");
}
if(my_rank !=0){
MPI_Bcast(B,N*N,MPI_DOUBLE,0,MPI_COMM_WORLD);
mt = 0;
MPI_Recv(&offset,1,MPI_INT,0,mt,MPI_COMM_WORLD,MPI_STATUS_IGNORE);
MPI_Recv(&rows,1,MPI_INT,0,mt,MPI_COMM_WORLD,MPI_STATUS_IGNORE);
MPI_Recv(&A,rows*N,MPI_DOUBLE,0,mt,MPI_COMM_WORLD,MPI_STATUS_IGNORE);
for(i=0;i<N;i++)
for(j=0;j<rows;j++){
C[j*N+i] =0.0;
for(k=0;k<N;k++)
C[j*N+i] += A[j*N+k]*B[k*N+i];
}
mt = 1;
MPI_Send(&offset, 1, MPI_INT, 0, mt, MPI_COMM_WORLD);
MPI_Send(&rows, 1, MPI_INT, 0, mt, MPI_COMM_WORLD);
MPI_Send(&C, rows*N, MPI_DOUBLE, 0, mt, MPI_COMM_WORLD);
}
MPI_Finalize();
Found it.
You say
if (my_rank =0)
This should be
if (my_rank == 0)
;-)