- The
MPI_Recv()
requires the user to specify a
specific sender ID
and a
specific tag:
MPI_Recv(buff, count, datatype, source, tag, comm, request);
- If the programmer does not know
in advance
what values
source
and
tag
the message will be,
you can use the
wildcard values
to receive the message, and then
use the
status
variable to find out who the sender was:
MPI_Recv(buff, count, type, MPI_ANY_SOURCE, MPI_ANY_TAG, status);
- After the
MPI_Recv()
returns with a message, you can use
these fields to find the
source id of the sender
and
the tag of the message:
status.SOURCE = source id of sender status.TAG = tag of message
- NOTE:
- So far, we have assumed that
we know the
format of the data that is
sent/received.
- In order to use MPI_Recv(),
you
must know
the
type of data
sent !!!
- Usually, MPI programs exchange messages that have
a fixed format (e.g.,
sender sends a number to the receiver)
In more complicated program, this assumption is no longer true...
- So far, we have assumed that
we know the
format of the data that is
sent/received.
- What can we do if we receive a message
without knowing apriori
about its
format and
length
Part 1 discuss how to handle unknown format
-
The simplest solution would be to agree on
one format (or a few)
for every sender
This is a "cop out" solution...
- The general solution is to
use a header message description
to describe the data:
- The message consists of a fixed format
header part that given
the number and the
type of values in the message
- And a variable format body part that contains the actual values
Example:
// Message header int NDataField - number of data field field 1(name, type, length) - name, type and length of field 1 field 2(name, type, length) - name, type and length of field 2 ... field n(name, type, length) - name, type and length of field n // Content of message value of field 1 value of field 2 ... value of field n
- The message consists of a fixed format
header part that given
the number and the
type of values in the message
- There is still one more problem
that we need to solve:
- How much buffer space do we need to use to receive the message ???
- Part 2 discuss how to handle unknown length...
- If you do not know the
length
of the data,
then you
cannot use the
MPI_Recv()
function
(because you must specify a buffer variable
and the size of the buffer
may be
too small)
- To receive a message
of unknown length,
we must proceed in
3 steps:
- Wait for a message to arrive and
find out the SIZE of the message
(using MPI_Probe())
- Find the
length (number of bytes)
of the message.
- Allocate a buffer of sufficient size to receive the message (using MPI_Get_count())
- Wait for a message to arrive and
find out the SIZE of the message
(using MPI_Probe())
- Example:
MPI_Status status; int nbytes; MPI_Probe(MPI_ANY_SOURCE, MPI_ANY_TAG, MPI_COMM_WORLD, &status); // It will block the caller until a message is ready // Obtain message size... MPI_Get_count(&status, MPI_CHAR, &nbytes); // Allocate memory to receive data if ( nbytes != MPI_UNDEFINED ) buff = malloc( nbytes ); // Finally, receive the message with a correctly sized buffer... MPI_Recv(buff, nbytes, MPI_CHAR, status.SOURCE, status.TAG, MPI_COMM_WORLD, &status);
- Example Program:
(Demo above code)
- Prog file: click here
Demo instruction:
- ssh2 puma
- cd sunhome/teaching/web/355/Syllabus/92-MPI/Progs
- mpiCC -o Comm05 Comm05.C
- mpirun -np 4 ./Comm05
- The approach to send/receive data of
arbitrary format
uses
META DATA
along with
the actual data
- This approach is
similar
to
the
meta data
used in
Database Systems
- For an in dept study of meta data , please see: click here