I'll give it a go. Attached is the code.
Thanks!
-Mike
Tim S. Woodall wrote:
Hello Mike,
Mike Houston wrote:
When only sending a few messages, we get reasonably good IB performance,
~500MB/s (MVAPICH is 850MB/s). However, if I crank the number of
messages up, we drop to 3MB/s(!!!). This is with the OSU NBCL
mpi_bandwidth test. We are running Mellanox IB Gold 1.8 with 3.3.3
firmware on PCI-X (Couger) boards. Everything works with MVAPICH, but
we really need the thread support in OpenMPI.
Ideas? I noticed there are a plethora of runtime options configurable
for mvapi. Do I need to tweak these to get performacne up?
You might try running w/ the:
mpirun -mca mpi_leave_pinned 1
Which will cause mvapi port to maintain an mru cache of registrations,
rather than dynamically pinning/unpinning memory.
If this does not resolve the BW problems, try increasing the
resources allocated to each connection:
-mca btl_mvapi_rd_min 128
-mca btl_mvapi_rd_max 256
Also can you forward me a copy of the test code or a reference to it?
Thanks,
Tim
_______________________________________________
users mailing list
us...@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/users
/*
* Copyright (C) 2002-2003 the Network-Based Computing Laboratory
* (NBCL), The Ohio State University.
*/
#include "mpi.h"
#include <stdio.h>
#include <unistd.h>
#include <stdlib.h>
#include <math.h>
#include <assert.h>
#define MYBUFSIZE (4*1024*1028)
#define MAX_REQ_NUM 100000
char s_buf1[MYBUFSIZE];
char r_buf1[MYBUFSIZE];
MPI_Request request[MAX_REQ_NUM];
MPI_Status stat[MAX_REQ_NUM];
int main(int argc,char *argv[])
{
int myid, numprocs, i;
int size, loop, page_size;
char *s_buf, *r_buf;
double t_start=0.0, t_end=0.0, t=0.0;
MPI_Init(&argc,&argv);
MPI_Comm_size(MPI_COMM_WORLD,&numprocs);
MPI_Comm_rank(MPI_COMM_WORLD,&myid);
if ( argc < 3 ) {
fprintf(stderr, "Usage: bw loop msg_size\n");
MPI_Finalize();
return 0;
}
size=atoi(argv[2]);
loop = atoi(argv[1]);
if(size > MYBUFSIZE){
fprintf(stderr, "Maximum message size is %d\n",MYBUFSIZE);
MPI_Finalize();
return 0;
}
if(loop > MAX_REQ_NUM){
fprintf(stderr, "Maximum number of iterations is
%d\n",MAX_REQ_NUM);
MPI_Finalize();
return 0;
}
page_size = getpagesize();
s_buf = (char*)(((unsigned long)s_buf1 + (page_size -1))/page_size *
page_size);
r_buf = (char*)(((unsigned long)r_buf1 + (page_size -1))/page_size *
page_size);
assert( (s_buf != NULL) && (r_buf != NULL) );
for ( i=0; i<size; i++ ){
s_buf[i]='a';
r_buf[i]='b';
}
/*warmup */
if (myid == 0)
{
for ( i=0; i< loop; i++ ) {
MPI_Isend(s_buf, size, MPI_CHAR, 1, 100, MPI_COMM_WORLD, request+i);
}
MPI_Waitall(loop, request, stat);
MPI_Recv(r_buf, 4, MPI_CHAR, 1, 101, MPI_COMM_WORLD, &stat[0]);
}else{
for ( i=0; i< loop; i++ ) {
MPI_Irecv(r_buf, size, MPI_CHAR, 0, 100, MPI_COMM_WORLD, request+i);
}
MPI_Waitall(loop, request, stat);
MPI_Send(s_buf, 4, MPI_CHAR, 0, 101, MPI_COMM_WORLD);
}
MPI_Barrier(MPI_COMM_WORLD);
if (myid == 0)
{
t_start=MPI_Wtime();
for ( i=0; i< loop; i++ ) {
MPI_Isend(s_buf, size, MPI_CHAR, 1, 100, MPI_COMM_WORLD, request+i);
}
MPI_Waitall(loop, request, stat);
MPI_Recv(r_buf, 4, MPI_CHAR, 1, 101, MPI_COMM_WORLD, &stat[0]);
t_end=MPI_Wtime();
t = t_end - t_start;
}else{
for ( i=0; i< loop; i++ ) {
MPI_Irecv(r_buf, size, MPI_CHAR, 0, 100, MPI_COMM_WORLD, request+i);
}
MPI_Waitall(loop, request, stat);
MPI_Send(s_buf, 4, MPI_CHAR, 0, 101, MPI_COMM_WORLD);
}
if ( myid == 0 ) {
double tmp;
tmp = ((size*1.0)/1.0e6)*loop;
fprintf(stdout,"%d\t%f (MillionBytes/sec)\t%f(MegaBytes/sec)\n", size,
tmp/t, tmp/t/1.048576);
}
MPI_Barrier(MPI_COMM_WORLD);
MPI_Finalize();
return 0;
}
