Distributed Memory Programming With MPI: Peter Pacheco

An Introduction to Parallel Programming
Peter Pacheco
Chapter 3
Distributed Memory
Programming with
MPI
Copyright 2010, Elsevier Inc. All rights Reserved
Writing your first MPI program.

Using the common MPI functions.
The Trapezoidal Rule in MPI.
Collective communication.
MPI derived datatypes.
Performance evaluation of MPI programs.
Parallel sorting.
Safety in MPI programs.
# Chapter Subtitle
Roadmap
A distributed memory system
A shared memory system
Hello World!
(a classic)
Identifying MPI processes
Common practice to identify processes by

nonnegative integer ranks.
p processes are numbered 0, 1, 2, .. p-1
Our first MPI program
Set MPICH2 Path
Then add the following at

the end of your ~/.bashrc file and source it with the command
source ~/.bashrc (or log in again).
#---------------------------------------------------------------------# MPICH2 setup
export PATH=/opt/MPICH2/bin:$PATH
export MANPATH= =/opt/MPICH2/bin :$MANPATH
#----------------------------------------------------------------------
Some logging and visulaization help:
You can Link with the libraries -llmpe -lmpe to enable logging and the MPE
environment. Then run the program as usual and a log file will be produced. The
log file can be visualized using the jumpshot program that comes bundled with
MPICH2.
Compilation
wrapper script to compile
source file
mpicc -g -Wall -o mpi_hello mpi_hello.c

produce
debugging
information
create this executable file name

(as opposed to default a.out)
turns on all warnings
Setup Multiple Computer

Environment
Create a file called say machines
containing the list of machines:
athena.cs.siu.edu
oscarnode1.cs.siu.edu
.
oscarnode8.cs.siu.edu
Establish network environments
mpdboot n 9 f machines
mpdtrace
mpdallexit
10
Execution
mpiexec -n <number of processes> <executable>
mpiexec -n 1 ./mpi_hello
run with 1 process
run with 4 processes
11
Execution
Greetings from process 0 of 1 !

12
MPI Programs
Written in C.
Has main.
Uses stdio.h, string.h, etc.
Need to add mpi.h header file.

Identifiers defined by MPI start with
MPI_.
First letter following underscore is
uppercase.
For function names and MPI-defined types.

Helps to avoid confusion.
13
MPI Components
MPI_Init
Tells MPI to do all the necessary setup.
MPI_Finalize
Tells MPI were done, so clean up anything

allocated for this program.
14
Basic Outline
15
Communicators
A collection of processes that can send

messages to each other.
MPI_Init defines a communicator that
consists of all the processes created when
the program is started.
Called MPI_COMM_WORLD.
16
Communicators
number of processes in the communicator
my rank
(the process making this call)
17
#include mpi.h
#include <stdio.h>
#include <math.h>
#define MAXSIZE 1000
void main(int argc, char *argv)
{
int myid, numprocs;
int data[MAXSIZE], i, x, low, high, myresult, result;
char fn[255];
char *fp;
MPI_Init(&argc,&argv);
MPI_Comm_size(MPI_COMM_WORLD,&numprocs);
MPI_Comm_rank(MPI_COMM_WORLD,&myid);
if (myid == 0) { /* Open input file and initialize data */
strcpy(fn,getenv(HOME));
strcat(fn,/MPI/rand_data.txt);
if ((fp = fopen(fn,r)) == NULL) {
printf(Cant open the input file: %s\n\n, fn);
exit(1);
}
for(i = 0; i < MAXSIZE; i++) fscanf(fp,%d, &data[i]);
}
MPI_Bcast(data, MAXSIZE, MPI_INT, 0, MPI_COMM_WORLD); /* broadcast data */
x = n/nproc; /* Add my portion Of data */
low = myid * x;
high = low + x;
for(i = low; i < high; i++)
myresult += data[i];
printf(I got %d from %d\n, myresult, myid); /* Compute global sum */
MPI_Reduce(&myresult, &result, 1, MPI_INT, MPI_SUM, 0, MPI_COMM_WORLD);
if (myid == 0) printf(The sum is %d.\n, result);
MPI_Finalize();
}
Sample MPI program
18
SPMD
Single-Program Multiple-Data
We compile one program.
Process 0 does something different.
Receives messages and prints them while the

other processes do the work.
The if-else construct makes our program

SPMD.
19
Communication
20
Data types
21
Communication
22
Message matching
r
MPI_Send
src = q
MPI_Recv
dest = r
23
Receiving messages
A receiver can get a message without

knowing:
the amount of data in the message,

the sender of the message, ( MPI_ANY_SOURCE)
or the tag of the message. ( MPI_ANY_TAG)
24
status_p argument
How receiver finds out the sender, tag if they are not needed
by the receiver
MPI_Status*
MPI_Status* status;
status.MPI_SOURCE
status.MPI_TAG
MPI_SOURCE
MPI_TAG
MPI_ERROR
25
How much data am I receiving?
26
Issues with send and receive
Exact behavior is determined by the MPI

implementation.
MPI_Send (locally blocking with buffer copied to

internal storage or block starts transmission) may
behave differently with regard to buffer size,
cutoffs and blocking.
MPI_Recv always blocks until a matching
message is received.
Non-blocking MPI_Isend and MPI_Irecv,

immediate return.
Know your implementation; dont make assumptions!
27
TRAPEZOIDAL RULE IN MPI
28
The Trapezoidal Rule
29
The Trapezoidal Rule
30
One trapezoid
31
Pseudo-code for a serial

program
32
Parallelizing the Trapezoidal Rule

1. Partition problem solution into tasks.
2. Identify communication channels between
tasks.
3. Aggregate tasks into composite tasks.
4. Map composite tasks to cores.
33
Parallel pseudo-code
34
Tasks and communications for

Trapezoidal Rule
35
First version (1)
36
First version (2)
37
First version (3)
38
Dealing with I/O

Each process just
prints a message.
39
Running with 6 processes
unpredictable output
40
Input
Most MPI implementations only allow

process 0 in MPI_COMM_WORLD access
to stdin.
Process 0 must read the data (scanf) and
send to the other processes.
41
Function for reading user input
42
COLLECTIVE
COMMUNICATION
43
Tree-structured communication
1. In the first phase:
(a) Process 1 sends to 0, 3 sends to 2, 5 sends to 4, and
7 sends to 6.
(b) Processes 0, 2, 4, and 6 add in the received values.
(c) Processes 2 and 6 send their new values to
processes 0 and 4, respectively.
(d) Processes 0 and 4 add the received values into their
new values.
2. (a) Process 4 sends its newest value to process 0.
(b) Process 0 adds the received value to its newest
value.
44
A tree-structured global sum
45
An alternative tree-structured
global sum
46
MPI_Reduce
47
Predefined reduction operators

in MPI
48
Collective vs. Point-to-Point

Communications
All the processes in the communicator

must call the same collective function.
For example, a program that attempts to

match a call to MPI_Reduce on one
process with a call to MPI_Recv on
another process is erroneous, and, in all
likelihood, the program will hang or crash.
49

Communications
The arguments passed by each process to

an MPI collective communication must be
compatible.
For example, if one process passes in 0 as

the dest_process and another passes in 1,
then the outcome of a call to MPI_Reduce
is erroneous, and, once again, the
program is likely to hang or crash.
50

Communications
The output_data_p argument is only used

on dest_process.
However, all of the processes still need to

pass in an actual argument corresponding
to output_data_p, even if its just NULL.
51

Communications
Point-to-point communications are

matched on the basis of tags and
communicators.
Collective communications dont use tags.

Theyre matched solely on the basis of the
communicator and the order in which
theyre called.
52
Example (1)
Multiple calls to MPI_Reduce
53
Example (2)
Suppose that each process calls

MPI_Reduce with operator MPI_SUM, and
destination process 0.
54
Example (3)
The order of the calls will determine the

matching so the value stored in b will be
1+2+1 = 4, and the value stored in d will
be 2+1+2 = 5.
55
MPI_Allreduce
Useful in a situation in which all of the

processes need the result of a global sum
in order to complete some larger
computation.
56
A global sum followed

by distribution of the
result.
57
A butterfly-structured global sum.
58
Broadcast
Data belonging to a single process is sent

to all of the processes in the
communicator.
59
A tree-structured broadcast.
60
A version of Get_input that uses

MPI_Bcast
61
Data distributions
Compute a vector sum.
62
Serial implementation of vector

addition
63
Different partitions of a 12component vector among 3

processes
64
Partitioning options
Block partitioning
Cyclic partitioning
Assign blocks of consecutive components to

each process.
Assign components in a round robin fashion.
Block-cyclic partitioning
Use a cyclic distribution of blocks of

components.
65
Parallel implementation of
vector addition
66
Scatter
MPI_Scatter can be used in a function that

reads in an entire vector on process 0 but
only sends the needed components to
each of the other processes.
67
Reading and distributing a vector
68
Gather
Collect all of the components of the vector

onto process 0, and then process 0 can
process all of the components.
69
Print a distributed vector (1)
70
Print a distributed vector (2)
71
Allgather
Concatenates the contents of each

process send_buf_p and stores this in
each process recv_buf_p.
As usual, recv_count is the amount of data
being received from each process.
72
MPI DERIVED DATATYPES
73
Derived datatypes
Used to represent any collection of data items in

memory by storing both the types of the items
and their relative locations in memory.
The idea is that if a function that sends data
knows this information about a collection of data
items, it can collect the items from memory
before they are sent.
Similarly, a function that receives data can
distribute the items into their correct destinations
in memory when theyre received.
74
Derived datatypes
Formally, consists of a sequence of basic

MPI data types together with a
displacement for each of the data types.
Trapezoidal Rule example:
75
MPI_Type create_struct
Builds a derived datatype that consists of

individual elements that have different
basic types.
76
MPI_Get_address
Returns the address of the memory

location referenced by location_p.
The special type MPI_Aint is an integer
type that is big enough to store an address
on the system.
77
MPI_Type_commit
Allows the MPI implementation to optimize

its internal representation of the datatype
for use in communication functions.
78
MPI_Type_free
When were finished with our new type,

this frees any additional storage used.
79
Get input function with a derived

datatype (1)
80

datatype (2)
81

datatype (3)
82
PERFORMANCE EVALUATION
83
Elapsed parallel time
Returns the number of seconds that have

elapsed since some time in the past.
84
Average/least/most execution time spent by

individual process
int myrank, numprocs;
double mytime, maxtime, mintime, avgtime; /*variables used for gathering timing statistics*/
MPI_Comm_rank(MPI_COMM_WORLD, &myrank); MPI_Comm_size(MPI_COMM_WORLD, &numprocs);
MPI_Barrier(MPI_COMM_WORLD); /*synchronize all processes*/
mytime = MPI_Wtime(); /*get time just before work section */
work();
mytime = MPI_Wtime() - mytime; /*get time just after work section*/ /*compute max, min, and average timing
statistics*/
MPI_Reduce(&mytime, &maxtime, 1, MPI_DOUBLE,MPI_MAX, 0, MPI_COMM_WORLD);
MPI_Reduce(&mytime, &mintime, 1, MPI_DOUBLE, MPI_MIN, 0,MPI_COMM_WORLD);
MPI_Reduce(&mytime, &avgtime, 1, MPI_DOUBLE, MPI_SUM, 0,MPI_COMM_WORLD);
if (myrank == 0) {
avgtime /= numprocs;
printf("Min: %lf Max: %lf Avg: %lf\n", mintime, maxtime,avgtime);
}
85
Elapsed serial time
In this case, you dont need to link in the

MPI libraries.
Returns time in microseconds elapsed
from some point in the past.
86
Elapsed serial time
87
MPI_Barrier
Ensures that no process will return from

calling it until every process in the
communicator has started calling it.
88
MPI_Barrier
89
Run-times of serial and parallel

matrix-vector multiplication
(Seconds)
90
Speedup
91
Efficiency
92
Speedups of Parallel MatrixVector Multiplication
93
Efficiencies of Parallel MatrixVector Multiplication
94
Scalability
A program is scalable if the problem size

can be increased at a rate so that the
efficiency doesnt decrease as the number
of processes increase.
95
Scalability
Programs that can maintain a constant

efficiency without increasing the problem
size are sometimes said to be strongly
scalable.
Programs that can maintain a constant

efficiency if the problem size increases at
the same rate as the number of processes
are sometimes said to be weakly scalable.
96
A PARALLEL SORTING
ALGORITHM
97
Sorting
n keys and p = comm sz processes.

n/p keys assigned to each process.
No restrictions on which keys are assigned
to which processes.
When the algorithm terminates:
The keys assigned to each process should be

sorted in (say) increasing order.
If 0 q < r < p, then each key assigned to
process q should be less than or equal to
every key assigned to process r.
98
Serial bubble sort
99
Odd-even transposition sort
A sequence of phases.
Even phases, compare swaps:
Odd phases, compare swaps:
100
Example
Start: 5, 9, 4, 3
Even phase: compare-swap (5,9) and (4,3)
getting the list 5, 9, 3, 4
Odd phase: compare-swap (9,3)
Even phase: compare-swap (5,3) and (9,4)
Odd phase: compare-swap (5,4)
101
Serial odd-even transposition

sort
102
Communications among tasks in

odd-even sort
Tasks determining a[i] are labeled with a[i].
103
Parallel odd-even transposition

sort
104
Pseudo-code
105
Compute_partner
106
Safety in MPI programs
The MPI standard allows MPI_Send to

behave in two different ways:
it can simply copy the message into an MPI

managed buffer and return,
or it can block until the matching call to
MPI_Recv starts.
107
Many implementations of MPI set a

threshold at which the system switches
from buffering to blocking.
Relatively small messages will be buffered
by MPI_Send.
Larger messages, will cause it to block.
108
If the MPI_Send executed by each process

blocks, no process will be able to start
executing a call to MPI_Recv, and the
program will hang or deadlock.
Each process is blocked waiting for an

event that will never happen.
(see pseudo-code)
109
A program that relies on MPI provided

buffering is said to be unsafe.
Such a program may run without problems

for various sets of input, but it may hang or
crash with other sets.
110
MPI_Ssend
An alternative to MPI_Send defined by the

MPI standard.
The extra s stands for synchronous and
MPI_Ssend is guaranteed to block until the
matching receive starts.
111
Restructuring communication
112
MPI_Sendrecv
An alternative to scheduling the

communications ourselves.
Carries out a blocking send and a receive
in a single call.
The dest and the source can be the same
or different.
Especially useful because MPI schedules
the communications so that the program
wont hang or crash.
113
MPI_Sendrecv
114
Safe communication with five

processes
115
Parallel odd-even transposition sort
116
Run-times of parallel odd-even

sort
(times are in milliseconds)
117
Concluding Remarks (1)
MPI or the Message-Passing Interface is a

library of functions that can be called from
C, C++, or Fortran programs.
A communicator is a collection of
processes that can send messages to
each other.
Many parallel programs use the singleprogram multiple data or SPMD approach.
118
Most serial programs are deterministic: if

we run the same program with the same
input well get the same output.
Parallel programs often dont possess this
property.
Collective communications involve all the
processes in a communicator.
119
When we time parallel programs, were

usually interested in elapsed time or wall
clock time.
Speedup is the ratio of the serial run-time
to the parallel run-time.
Efficiency is the speedup divided by the
number of parallel processes.
120
If its possible to increase the problem size

(n) so that the efficiency doesnt decrease
as p is increased, a parallel program is
said to be scalable.
An MPI program is unsafe if its correct
behavior depends on the fact that
MPI_Send is buffering its input.
121

Distributed Memory Programming With MPI: Peter Pacheco

Uploaded by

Copyright:

Available Formats

Distributed Memory Programming With MPI: Peter Pacheco

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Distributed Memory Programming With MPI: Peter Pacheco

Uploaded by

Copyright:

Available Formats

An Introduction to Parallel Programming

Copyright 2010, Elsevier Inc. All rights Reserved

Writing your first MPI program.

Copyright 2010, Elsevier Inc. All rights Reserved

A distributed memory system

Copyright 2010, Elsevier Inc. All rights Reserved

A shared memory system

Copyright 2010, Elsevier Inc. All rights Reserved

Identifying MPI processes

Common practice to identify processes by

p processes are numbered 0, 1, 2, .. p-1

Copyright 2010, Elsevier Inc. All rights Reserved

Our first MPI program

Copyright 2010, Elsevier Inc. All rights Reserved

Set MPICH2 Path

Then add the following at

Some logging and visulaization help:

Copyright 2010, Elsevier Inc. All rights Reserved

mpicc -g -Wall -o mpi_hello mpi_hello.c

create this executable file name

turns on all warnings

Copyright 2010, Elsevier Inc. All rights Reserved

Setup Multiple Computer

Copyright 2010, Elsevier Inc. All rights Reserved

Greetings from process 0 of 1 !

Need to add mpi.h header file.

For function names and MPI-defined types.

Tells MPI to do all the necessary setup.

Tells MPI were done, so clean up anything

Copyright 2010, Elsevier Inc. All rights Reserved

Copyright 2010, Elsevier Inc. All rights Reserved

A collection of processes that can send

Copyright 2010, Elsevier Inc. All rights Reserved

number of processes in the communicator

Copyright 2010, Elsevier Inc. All rights Reserved

Sample MPI program

Copyright 2010, Elsevier Inc. All rights Reserved

Receives messages and prints them while the

The if-else construct makes our program

Copyright 2010, Elsevier Inc. All rights Reserved

Copyright 2010, Elsevier Inc. All rights Reserved

Copyright 2010, Elsevier Inc. All rights Reserved

Copyright 2010, Elsevier Inc. All rights Reserved

Copyright 2010, Elsevier Inc. All rights Reserved

A receiver can get a message without

the amount of data in the message,

Copyright 2010, Elsevier Inc. All rights Reserved

Copyright 2010, Elsevier Inc. All rights Reserved

How much data am I receiving?

Copyright 2010, Elsevier Inc. All rights Reserved

Issues with send and receive

Exact behavior is determined by the MPI

MPI_Send (locally blocking with buffer copied to

Non-blocking MPI_Isend and MPI_Irecv,

Copyright 2010, Elsevier Inc. All rights Reserved

TRAPEZOIDAL RULE IN MPI

Copyright 2010, Elsevier Inc. All rights Reserved

The Trapezoidal Rule

Copyright 2010, Elsevier Inc. All rights Reserved