OpenMP¤òÍÑ¤¤¤¿ÊÂÎó·×»»¡Ê£²¡Ë

Size: px

Start display at page:

Download "OpenMP¤òÍÑ¤¤¤¿ÊÂÎó·×»»¡Ê£²¡Ë"

ああすいくのや
4 years ago
Views:

2 (schedule) (omp sections) (omp single, omp master) (barrier, critical, atomic)

3 program pi i m p l i c i t none integer, parameter : : SP = kind ( 1. 0 ) integer, parameter : : DP = selected real kind (2 precision ( 1. 0 SP ) ) integer, parameter : : n = integer : : i real (DP) : : x, dx, p real(dp) :: time0, time1, omp get wtime dx = 1.0 DP / r eal ( n, DP) p = 0.0 DP time0 = omp get wtime()!$omp parallel do default(none) private(i,x) shared(dx) reduction(+:p) do i = 1,n x = real ( i, DP) dx p = p DP / ( 1. 0 DP + x 2) dx time1 = omp get wtime() i x p r i n t, p print *, time1-time0 end program

4 a 11 a 12 a 13 a 14 a 15 a 16 a 17 a 18 a 22 a 23 a 24 a 25 a 26 a 27 a 28 a 33 a 34 a 35 a 36 a 37 a 38 a 44 a 45 a 46 a 47 a 48 a 55 a 56 a 57 a 58 a 66 a 67 a 68 a 77 a 78 a 88 x 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8

5 a 11 a 12 a 13 a 14 a 15 a 16 a 17 a 18 a 22 a 23 a 24 a 25 a 26 a 27 a 28 a 33 a 34 a 35 a 36 a 37 a 38 a 44 a 45 a 46 a 47 a 48 a 55 a 56 a 57 a 58 a 66 a 67 a 68 a 77 a 78 a 88 OpenMP! $omp parallel do schedule(static,2) x 1 x 2 x 3 x 4 x 5 x 6 x 7 x 8

6 schedule schedule(, )!$omp parallel do schedule(static, 4) static dynamic guided dynamic runtime OMP SCHEDULE

7 1 enshu-openmp2 mkdir enshu openmp2 cd enshu openmp2 2 /tmp/openmp2/schedule.f90! omp parallel do schedule(,) schedule static, dynamic, guided 3

8 program schedule i m p l i c i t none integer, parameter : : SP = kind ( 1. 0 ) integer, parameter : : DP = selected real kind (2 precision ( 1. 0 SP ) ) integer, parameter : : n = 2000 integer : : i, j real (DP), dimension ( n ) : : x, y real (DP), dimension ( n, n ) : : A real (DP) : : time0, time1, omp get wtime x ( : ) = 2.0 DP A ( :, : ) = 1.0 DP time0 = omp get wtime ( )! $omp parallel do schedule(,) default(none) private(i,j) shared(a,x,y) do i =1,n y ( i ) = 0.0 DP do j =i, n y ( i ) = y ( i ) + A( i, j ) x ( j )! $omp end parallel do time1 = omp get wtime ( ) p r i n t, time1 time0 end program

9 enshu.sh pjsub enshu. sh jobname.o???????? #!/bin/bash #PJM -N jobname #PJM -L rscgrp=small #PJM -L node=1 #PJM -L elapse=2:00 #PJM -j export OMP NUM THREADS=4. / a. out /tmp/openmp2/enshu.sh

10 ! $omp parallel do do i =1,n y ( i ) = 0.0 DP do j=1,n y ( i ) = y ( i ) + A( i, j ) x ( j )! $omp end parallel do fortran do j =1,n y(i) = 0.0 DP do i=1,n y ( i ) = y ( i ) + A( i, j ) x ( j ) y(:) = 0.0 DP do j =1,n do i=1,n y ( i ) = y ( i ) + A( i, j ) x ( j )

11 !$omp sections program do while (t < ) x y z end program!$omp sections section!$omp parallel!$omp sections!$omp section! x omp end section!$omp section! y!$omp section! z!$omp end sections!$omp end parallel

12 1 /tmp/openmp2/sierp.f90 2 do x(i) y(i) sections sections 3 /tmp/openmp2/jscript.sh gnuplot x(i), y(i) X gnuplot plot jobname.o??? exit X

13 program sierpinski i m p l i c i t none! do i =1,n call random number (myrand ( i ) ) time0 = omp get wtime ( ) do i =1,n 1 i f ( myrand ( i ) < 0.33 DP ) then x(i+1) = x(i) 0.5 DP DP y(i+1) = y(i) 0.5 DP else i f (myrand ( i ) > 0.66 DP ) then x(i+1) = x(i) 0.5 DP DP y(i+1) = y(i) 0.5 DP else x(i+1) = x(i) 0.5 DP y(i+1) = y(i) 0.5 DP DP end i f time1 = omp get wtime ( ) print *, time1-time0!do i = 1,n! print, x(i), y(i)! end program gnuplot

14 1/3 x (n+1) = 1 2 x(n) + 1, y (n+1) = 1 2 y(n). 1/3 x (n+1) = 1 2 x(n) 1, y (n+1) = 1 2 y(n). 1/3 x (n+1) = 1 2 x(n), y (n+1) = 1 2 y(n) + 1.

15 !$omp single!$omp parallel val = 1.0 DP! $omp do do j =1,n do i =1,n A( i, j ) = val! $omp! $omp end parallel!$omp parallel!$omp single val = 1.0 DP!$omp end single! $omp do do j =1,n do i =1,n A( i, j ) = val! $omp! $omp end parallel val

16 !$omp parallel val = 1.0 DP! $omp do do j =1,n do i =1,n A( i, j ) = val! $omp! $omp end parallel!$omp master single!$omp parallel!$omp single val = 1.0 DP!$omp end single! $omp do do j =1,n do i =1,n A( i, j ) = val! $omp! $omp end parallel master!$omp parallel!$omp master val = 1.0 DP!$omp end master! $omp do do j =1,n do i =1,n A( i, j ) = val! $omp! $omp end parallel!$omp master: val val barrier

17 !$omp barrier!$omp!$omp end single barrier!$omp nowait!$omp critical!$omp atomic real (DP) : : sval, pval real (DP), dimension ( n ) : : svec!$omp parallel shared(sval, svec) private(pval) pval!$omp critical svec ( : ) = pval svec ( : )!$omp end critical!$omp atomic sval = sval + pval omp end atomic! $omp end parallel

18 reduction 1 2 omp atomic 3 omp atomic omp critical, omp end critical program summation integer, parameter : : SP=kind ( 1. 0 ) integer, parameter : : DP=selected real kind (2 precision (1.0 SP ) ) integer, parameter : : n=1000 real (DP) : : sval, pval real (DP), dimension (n) : : svec svec ( : ) = 1.0 DP sval = 0.0 DP!$omp parallel shared(sval, svec) private(pval) pval=0.0 DP! $omp do do i =1,n pval = pval + svec ( i )! $omp!$omp atomic sval = sval + pval! $omp end parallel p r i n t, sval /tmp/openmp2/sum.f90 end program

19 1 2 3 result.txt yaguchi mail yaguchi < result.txt

20 omp barrier nowait barrier!$omp parallel shared(sval, svec) private(pval) pval =0.0 DP! $omp do do i =1,n pval = pval + svec ( i )! $omp nowait! $omp atomic sval = sval + pval! $omp end parallel nowait atomic critical

21 program sierpinski i m p l i c i t none! do i=1,n call random number(myrand(i)) time0 = omp get wtime ( ) do i =1,n 1 i f ( myrand ( i ) < 0.33 DP ) then x(i+1) = x(i) 0.5 DP DP y(i+1) = y(i) 0.5 DP else i f (myrand ( i ) > 0.66 DP ) then x(i+1) = x(i) 0.5 DP DP y(i+1) = y(i) 0.5 DP else x(i+1) = x(i) 0.5 DP y(i+1) = y(i) 0.5 DP DP end i f time1 = omp get wtime ( ) do print *, time1-time0!do i = 1,n! print, x(i), y(i)! end program

22 r (n+1) = ar (n) + c mod m a c m random numer() recursive doubling r (n+1) = a(ar (n 1) + c) + c mod m = a 2 r (n 1) + (a + 1)c mod m =: a r (n 1) + c mod m

23 II 1 2 a, c!$omp sections 3 do i =1,n call random number (myrand ( i ) ) integer, parameter : : a = 109 integer, parameter : : c = 1021 integer, parameter : : m = 2 15 integer : : tmp tmp = 15 do i =1,n tmp = mod( tmp a+c,m) myrand ( i ) = real (tmp,dp) / (m 1)

24 1 2 3 result.txt yaguchi mail yaguchi < result.txt

25 OpenMP (1), (2), (3) C OpenMP c.pdf. OpenMP - CPU Barbara Chapman, Gabriele Jost and Ruud van der Pas (Foreword by David J. Kuck). Using OpenMP Portable Shared Memory Parallel Programming, The MIT Press, 2007.

OpenMP¤òÍÑ¤¤¤¿ÊÂÎó·×»»¡Ê£±¡Ë

OpenMP¤òÍÑ¤¤¤¿ÊÂÎó·×»»¡Ê£±¡Ë 2012 5 24 scalar Open MP Hello World Do (omp do) (omp workshare) (shared, private) π (reduction) PU PU PU 2 16 OpenMP FORTRAN/C/C++ MPI OpenMP 1997 FORTRAN Ver. 1.0 API 1998 C/C++ Ver. 1.0 API 2000 FORTRAN