Cypress CSC-1200T Guia do Utilizador Página 33

  • Descarregar
  • Adicionar aos meus manuais
  • Imprimir
  • Página
    / 124
  • Índice
  • MARCADORES
  • Avaliado. / 5. Com base em avaliações de clientes
Vista de página 32
Chapter 4. Program development 33
The option -Xn or -X n is used to indicate how many processors you
want for your application. If you do not provide this option, the program
can be run on any number of processors using the mpprun command.
This kind of executable is called malleable.
Here is a typical example of generating and running a non-malleable
executable, which has to be run on a fixed number of PEs:
t3e% f90 -X 16 -o prog.x prog.f90
t3e% ./prog.x
Here we are using 16 processors for our application. However, we can
also generate and run a malleable executable prog.x:
t3e% f90 -o prog.x prog.f90
t3e% mpprun -n 16 ./prog.x
t3e% mpprun -n 8 ./prog.x
Here the number of processors was given to the mpprun command and
the option -X was omitted. The program was first run on 16 processors
and then on eight processors.
The same applies to the C/C++ compilers. Here is an example of com-
piling and running a C program:
t3e% cc -o prog.x prog.c
t3e% mpprun -n 8 ./prog.x
Here we used eight processors for running our malleable executable
prog.x.
The MPI, PVM, and SHMEM libraries are automatically linked to your
application, when needed. Therefore, you do not need to provide any
special flags to be able to use, e.g., MPI calls. The Cray scientific library
(Libsci) is also automatically linked to your program.
The option -O indicates the optimization level used in the compilation.
If you are running production jobs, you should always turn code opti-
mization on! The default is moderate optimization, but you can request
more aggressive optimization.
The Fortran 90 compiler is discussed in more detail in Chapter 5,and
the C/C++ compiler in Chapter 6.
4.3 Libsci — Cray scientific library
Libsci is a collection of various mathematical subroutines. Most of the
routines solve some specific problem of linear algebra, but there are a
few routines for fast Fourier transforms as well.
Libsci is Cray’s own package of subroutines for computational tasks. Lib-
sci is divided into following sublibraries: BLAS 1,2,3, LAPACK, BLAS_S,
Vista de página 32
1 2 ... 28 29 30 31 32 33 34 35 36 37 38 ... 123 124

Comentários a estes Manuais

Sem comentários