程序代写代做代考 data structure compiler C Fortran c/c++ Message Passing Programming

程序代写 CS代考 / c++代做, compiler, C语言代写, data structure, Fortran代写

Message Passing Programming
Introduction to MPI

Reusing this material
This work is licensed under a Creative Commons Attribution- NonCommercial-ShareAlike 4.0 International License.
http://creativecommons.org/licenses/by-nc-sa/4.0/
This means you are free to copy and redistribute the material and adapt and build on the
material under the following terms: You must give appropriate credit, provide a link to the license and indicate if changes were made. If you adapt or build on the material you must distribute your work under the same license as the original.
Acknowledge EPCC as follows: 􏰀© EPCC, The University of Edinburgh, www.epcc.ed.ac.uk􏰁
Note that this presentation contains images owned by others. Please seek their permission before reusing these images.
3

What is MPI?
4

MPI Forum
• First message-passing interface standard.
• Sixty people from forty different organisations.
• Users and vendors represented, from the US and Europe. • Two-year process of proposals, meetings and review.
• Message Passing Interface document produced in 1993
论坛
5

Implementation
• MPI is a library of function/subroutine calls ii.
• MPI is not a language
• There is no such thing as an MPI compiler
• The C or Fortran compiler you invoke knows nothing about what MPI actually does
– only knows prototype/interface of the function/subroutine calls
hompzaupiar
6

Goals and Scope of MPI
• MPI’s prime goals are:
– To provide source-code portability. tobefasar – To allow efficient implementation.
• It also offers:
– A great deal of functionality.
– Support for heterogeneous parallel architectures.
各 一样的
。
7

Header files
• C/C++:
• Fortran 77:
#include
X. outdatedmstpopkwe
include ‘mpif.h’
• Fortran 90:
use mpi
• Fortran 2008:
use mpi_f08
8

MPI Function Format
• C:
error = MPI_Xxxxx(parameter, …);
Mzttss
• Fortran:
CALL MPI_XXXXX(parameter, …, IERROR)
– IERROR optional in 2008 version only, otherwise essential
MPI_Xxxxx(parameter, …);
retum
9

Handles
• MPI controls its own internal data structures.
• MPI releases `handles’ to allow programmers to refer to
掌控
these.
• C handles are of defined typedefs. 类型定义 • Fortran handles are INTEGERs.
10

Initialising MPI
• C:
• Fortran:
int MPI_Init(int *argc, char ***argv)
MPI_INIT(IERROR)
INTEGER IERROR
• Must be the first MPI procedure called.
– but multiple processes are already running before MPI_Init
11

MPI_Init
int main(int argc, char *argv[]) {
…
MPI_Init(&argc, &argv);
…
int main() {
…
MPI_Init(NULL, NULL);
…
program my_mpi_program integer :: ierror …
CALL MPI_INIT(IERROR)
12

MPI_COMM_WORLD
01
Communicators
MPI_COMM_WORLD
23 6
4 5
inheuegmedgs.IS knowrankwye㓛 )
13

Rank
• How do you identify different processes in a communicator?
MPI_Comm_rank(MPI_Comm comm, int *rank)
MPI_COMM_RANK(COMM, RANK, IERROR) INTEGER COMM, RANK, IERROR
• The rank is not the physical processor number. – numbering is always 0, 1, 2, …., N-1
14

MPI_Comm_rank
int rank; …
MPI_Comm_rank(MPI_COMM_WORLD, &rank); printf(􏰀Hell􏰂 f􏰃􏰂m 􏰃ank %d\n􏰁, 􏰃ank); …
integer :: ierror
integer :: rank
…
CALL MPI_COMM_RANK(MPI_COMM_WORLD, rank, ierror) write(*,*) 􏰄Hell􏰂 f􏰃􏰂m 􏰃ank 􏰄, 􏰃ank
…
15

Size
• How many processes are contained within a communicator?
MPI_Comm_size(MPI_Comm comm, int *size)
MPI_COMM_SIZE(COMM, SIZE, IERROR) INTEGER COMM, SIZE, IERROR
16

Exiting MPI
􏰅C: 􏰅Fortran:
int MPI_Finalize()
MPI_FINALIZE(IERROR)
INTEGER IERROR
􏰅Must be the last MPI procedure called.
17

What machine am I on?
• Can be useful on a cluster 串 一群
– e.g. to confirm mapping of processes to nodes/processors/cores
寸
。
藏
integer namelen
character*(MPI_MAX_PROCESSOR_NAME) :: procname
…
call MPI_GET_PROCESSOR_NAME(procname, namelen, ierror) write(*,*) ‘rank ‘, rank, ‘ is on machine ‘, procname(1:namelen)
ht.name
int namelen;
char procname[MPI_MAX_PROCESSOR_NAME];
…
MPI_Get_processor_name(procname, &namelen); printf(“rank %d is on machine %s\n”, rank, procname);
19

Summary
• Have covered some basic MPI calls – but no explicit message-passing yet
• Can still write useful programs
– e.g. a task farm of independent jobs
• Need to compile and launch parallel jobs – procedure is not specified by MPI
– next lecture gives machine-specific details
20

Summary
• Have covered some basic MPI calls – but no explicit message-passing yet
• Can still write useful programs
– e.g. a task farm of independent jobs
• Need to compile and launch parallel jobs – procedure is not specified by MPI
– next lecture gives machine-specific details
20

Related Posts