Dr. Hanif Durad 2
Lecture Outline
Virtual Topologies
Topology Types
Creating a Cartesian Virtual Topology
Cartesian Example
Cartesian Mapping Functions
MPI_CART_RANK
MPI_CART_COORDS
MPI_CART_SHIFT
Cartesian Partitioning
IntroMPI.ppt
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Guidelines and tips
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
Virtual Topologies-Message Passing Interface Library-Lecture Slides, Slides of Microprocessor and Interfacing
This lecture is related to Message Passing Interface Library concepts. its helpful for Parallel Computing students. It was delivered by Dr. Hanif Durad at Pakistan Institute of Engineering and Applied Sciences, Islamabad (PIEAS). It includes: Virtual, Topologies, Cartesian, Mapping, Functions, Partitioning, Rationale, Graph, Boundaries, Cyclic
Typology: Slides
2011/2012
1 / 35
This page cannot be seen from the preview
Don't miss anything!
Related documents
Partial preview of the text
Download Virtual Topologies-Message Passing Interface Library-Lecture Slides and more Slides Microprocessor and Interfacing in PDF only on Docsity!
Dr. Hanif Durad
Lecture Outline
^
Virtual Topologies
^
Topology Types
^
Creating a Cartesian Virtual Topology
^
Cartesian Example
^
Cartesian Mapping Functions^
MPI_CART_RANK MPI_CART_COORDS MPI_CART_SHIFT
^
Cartesian Partitioning
IntroMPI.ppt
Virtual Topologies
^
Convenient process naming
^
Naming scheme to fit the communication pattern
^
Simplifies writing of code
^
Can allow MPI to optimize communications
^
Rationale: access to useful topology routines
Dr. Hanif Durad
Example - 2D Torus
Dr. Hanif Durad
Topology types
^
Cartesian topologies^
Each process is connected to its neighbors in a virtualgrid
Boundaries can be cyclic
^
Processes can be identified by Cartesian coordinates
^
Graph topologies^
General graphs
Will not be covered here
Dr. Hanif Durad
Arguments
Dr. Hanif Durad
comm_old
existing communicator
ndims
number of dimensions
periods
logical array indicating whether a dimension is cyclic(If TRUE, cyclic boundary conditions)
reorder
logical(If FALSE, rank preserved)(If TRUE, possible rank reordering)
comm_cart
new cartesian communicator
Cartesian Example
MPI_Comm vu;int dim[2], period[2], reorder;dim[0]=4; dim[1]=3;period[0]=TRUE; period[1]=FALSE;reorder=TRUE;MPI_Cart_create(MPI_COMM_WORLD,2,dim,period,reorder,&vu);
Cartesian Mapping Functions
^
Mapping ranks to process grid coordinates^
C:
^
int MPI_Cart_coords (MPI_Comm comm, int rank, intmaxdims, int *coords)
^
Fortran:^
INTEGER COMM,RANK,MAXDIMS,COORDS(*),IERROR CALLMPI_CART_COORDS(COMM,RANK,MAXDIMS,COORDS,IERROR)
Dr. Hanif Durad
Example using C
#include <stdio.h>#include <mpi.h>/* Run with 12 processes */void main (int argc, char *argv[]) {int rank;MPI_Comm vu;int dim[2], period[2], reorder;int coord[2] , id;int TRUE=1,FALSE=0;MPI_Init(&argc, &argv);MPI_Comm_rank(MPI_COMM_WORLD,&rank);dim[0]=4; dim[1]=3;period[0]=TRUE; period[1]=FALSE;reorder=TRUE;MPI_Cart_create(MPI_COMM_WORLD,2,dim,period,reorder,&vu);if(rank==5){MPI_Cart_coords(vu,rank,2,coord);printf("P:%d My coordinates are %d %d\n",rank,coord[0],coord[1]);} if(rank==0) {coord[0]=3; coord[1]=1;MPI_Cart_rank(vu,coord,&id);printf("The processor at position (%d, %d) has rank %d\n",coord[0],coord[1],id);} MPI_Finalize();}^
cart.c
Cartesian Mapping Functions
^
Computing ranks of neighboring processes^
C:
^
int MPI_Cart_shift (MPI_Comm comm, int direction, intdisp, int *rank_source, int *rank_dest)
^
Fortran:^
INTEGERCOMM,DIRECTION,DISP,RANK_SOURCE,RANK_DEST,IERROR CALLMPI_CART_SHIFT(COMM,DIRECTION,DISP,RANK_SOURCE,RANK_DEST,IERROR
MPI_Cart_shift
^
Does not actually shift data: returns the correct ranks for a shiftwhich can be used in subsequent communication calls
^
Arguments:
Dr. Hanif Durad
^
If shift off of the topology, MPI_Proc_null is returned
direction
dimension in which the shift should be made
disp
length of the shift in processor coordinates (+ or -)
rank_source
where calling process should receive a message from during theshift
rank_dest
where calling process should send a message to during the shift
Example using FORTRAN
cartshift.f
Groups and Communicators
^
How to partition a group of processes into sub-groups?
^
Group by color (different communicators).
^
Sort by key (new ranks in the sub-groups)
Dr. Hanif Durad
10a-MVP06-slides.pdf
Example using C (1/2)
#include <stdio.h>#include "mpi.h"#include <math.h>main(int argc, char* argv[]) {int
p; int^
my_rank; MPI_Comm my_row_comm;int^
my_row; int^
q; int^
test; int^
my_rank_in_row; MPI_Init(&argc, &argv);MPI_Comm_size(MPI_COMM_WORLD, &p);MPI_Comm_rank(MPI_COMM_WORLD, &my_rank);q = (int) sqrt((double) p);
comm_split.c
Try different nodes?
Pacheco
Example using C (2/2)
q = (int) sqrt((double) p);/* my_rank is rank in MPI_COMM_WORLD.* q*q = p /my_row = my_rank/q;MPI_Comm_split(MPI_COMM_WORLD, my_row, my_rank,&my_row_comm);/ Test the new communicators /MPI_Comm_rank(my_row_comm, &my_rank_in_row);if (my_rank_in_row == 0)test = my_row;elsetest = 0;MPI_Bcast(&test, 1, MPI_INT, 0, my_row_comm);printf("Process %d > my_row = %d, my_rank_in_row = %d, test = %d\n",my_rank, my_row, my_rank_in_row, test);MPI_Finalize();} / main */