サンプル集

http://www.nag-j.co.jp/openMP/

https://web.kudpc.kyoto-u.ac.jp/Archives/PDF/NewsLetter/2003-6_openmp.pdf

http://exp.cs.kobe-u.ac.jp/wiki/comp_practice/index.php?plugin=attach&refer=4.OpenMP%A4%F2%CD%D1%A4%A4%A4%BF%CA%C2%CE%F3%B7%D7%BB%BB&openfile=openmp1_yaguchi_2012.pdf

http://www.cc.kyushu-u.ac.jp/scp/system/library/OpenMP/openmp0209.pdf

[2015年 5月 2日 土曜日 19:11:57 JST]

[~/OpenMP]

[am@aofd165]

$ cat compile.run.sh

#!/bin/sh

usage(){

cat <<EOF

usage : $0 prog_name

EOF

}

if [ $# -ne 1 ]; then

  Error in $0 : Wrong number of argument.

  usage

  exit 1

fi

prog=$1

src=${prog}.f90

exe=${prog}.exe

if [ ! -f $src ]; then

  echo Error in $0 : No such file, $src

  usage

  exit 1

fi

echo

cat $src

echo

ifort -openmp ${src} -o ${exe}

export OMP_NUM_THREADS=32

echo "OMP_NUM_THREADS= $OMP_NUM_THREADS"

./${exe}

[2015年 5月 2日 土曜日 19:13:00 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh hello

program helloOpenMP

  !$ use omp_lib

  implicit none

  print *, "START"

!$omp parallel

  print *, "Hello! N =", omp_get_num_threads(), " and I am ", omp_get_thread_num()

!$omp end parallel

  print *, "END"

end

OMP_NUM_THREADS= 32

START

Hello! N = 32 and I am 16

Hello! N = 32 and I am 19

Hello! N = 32 and I am 5

Hello! N = 32 and I am 17

Hello! N = 32 and I am 11

Hello! N = 32 and I am 1

Hello! N = 32 and I am 4

Hello! N = 32 and I am 10

Hello! N = 32 and I am 30

Hello! N = 32 and I am 2

Hello! N = 32 and I am 29

Hello! N = 32 and I am 8

Hello! N = 32 and I am 7

Hello! N = 32 and I am 6

Hello! N = 32 and I am 18

Hello! N = 32 and I am 14

Hello! N = 32 and I am 25

Hello! N = 32 and I am 31

Hello! N = 32 and I am 9

Hello! N = 32 and I am 28

Hello! N = 32 and I am 15

Hello! N = 32 and I am 13

Hello! N = 32 and I am 3

Hello! N = 32 and I am 24

Hello! N = 32 and I am 26

Hello! N = 32 and I am 20

Hello! N = 32 and I am 23

Hello! N = 32 and I am 12

Hello! N = 32 and I am 21

Hello! N = 32 and I am 22

Hello! N = 32 and I am 27

Hello! N = 32 and I am 0

END

[2015年 5月 2日 土曜日 19:13:04 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh simpleDo

program simpleDo

  implicit none

  integer,parameter :: N = 100

  integer i

  double precision x(N)

  do i=1, N

    x(i) = dble(i)/N

  end do

  print *, sum(x)

end program

OMP_NUM_THREADS= 32

50.5000000000000

[2015年 5月 2日 土曜日 20:20:43 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh simpleDo.2

program simpleDo2

  !$ use omp_lib

  implicit none

  integer,parameter :: im = 1000,jm=1000,km=100

  integer i,j,k

  double precision x(im,jm,km)

  !$ double precision st, en

  !$ st = omp_get_wtime()

  do k=1,km

    do j=1,jm

      do i=1,im

        x(i,j,k) = dble(1000*i+100*j+k)

      end do !i

    enddo !j

  enddo !k

  print *, sum(x)

  !$ en = omp_get_wtime()

  !$ print *, "Elapsed time :", en-st

end program

OMP_NUM_THREADS= 1

55060050000000.0

Elapsed time : 2.43593597412109

real 0m3.093s

user 0m2.665s

sys 0m0.428s

OMP_NUM_THREADS= 2

55060050000000.0

Elapsed time : 2.40606188774109

real 0m3.062s

user 0m2.644s

sys 0m0.418s

OMP_NUM_THREADS= 4

55060050000000.0

Elapsed time : 2.31341409683228

real 0m2.970s

user 0m2.540s

sys 0m0.430s

OMP_NUM_THREADS= 8

55060050000000.0

Elapsed time : 2.48756599426270

real 0m3.105s

user 0m2.696s

sys 0m0.409s

OMP_NUM_THREADS= 32

55060050000000.0

Elapsed time : 2.31031107902527

real 0m2.963s

user 0m2.544s

sys 0m0.419s

$ compile.run.sh doTest

program doTest

  implicit none

  integer,parameter :: N = 50

  integer i, a(N)

!$omp parallel

!$omp do

  do i=1,N

    a(i) = i

  end do

!$omp end do

!$omp end parallel

  print *, a

end program

OMP_NUM_THREADS= 32

1 2 3 4 5 6

7 8 9 10 11 12

13 14 15 16 17 18

19 20 21 22 23 24

25 26 27 28 29 30

31 32 33 34 35 36

37 38 39 40 41 42

43 44 45 46 47 48

49 50

[2015年 5月 2日 土曜日 19:16:04 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh kadaiTime

program kadaiTime

  !$ use omp_lib

  implicit none

  integer,parameter :: N = 1000*1000, M = 100

  integer i, j

  double precision,allocatable :: a(:)

  double precision x

  !$ double precision st, en

  allocate(a(N))

  !$ st = omp_get_wtime()

!$omp parallel private(x)

!$omp do

  do i=1,N

    x = 0

    do j=1,M

      x = x + log(dble(i+j))

    end do

    a(i) = x/M

  end do

!$omp end do

!$omp end parallel

  !$ en = omp_get_wtime()

  print *, nint(sum(a))

  !$ print *, "Elapsed time :", en-st

end program

OMP_NUM_THREADS= 2

12816056

Elapsed time : 0.736011981964111

[2015年 5月 2日 土曜日 19:16:21 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh kadaiTime

program kadaiTime

  !$ use omp_lib

  implicit none

  integer,parameter :: N = 1000*1000, M = 100

  integer i, j

  double precision,allocatable :: a(:)

  double precision x

  !$ double precision st, en

  allocate(a(N))

  !$ st = omp_get_wtime()

!$omp parallel private(x)

!$omp do

  do i=1,N

    x = 0

    do j=1,M

      x = x + log(dble(i+j))

    end do

    a(i) = x/M

  end do

!$omp end do

!$omp end parallel

  !$ en = omp_get_wtime()

  print *, nint(sum(a))

  !$ print *, "Elapsed time :", en-st

end program

OMP_NUM_THREADS= 32

12816056

Elapsed time : 0.251950979232788

[2015年 5月 2日 土曜日 19:21:38 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh kadaiSections

program kadaiSections

  implicit none

  integer, parameter :: N = 20000

  real,allocatable :: a(:), b(:)

  allocate( a(N), b(N) )

  call random_number(a)

  call random_number(b)

!$omp parallel

!$omp sections

!$omp section

  call sort(a)

!$omp section

  call sort(b)

!$omp end sections

!$omp end parallel

  print '("first 4 numbers in a : ",4f10.7)', a(1:4)

  print '("first 4 numbers in b : ",4f10.7)', b(1:4)

contains

  subroutine sort(x)

    real x(:), tmp

    integer i, j

    do i = 1, N - 1

      do j = i + 1, N

        if (x(i)>x(j)) then

          tmp = x(i)

          x(i) = x(j)

          x(j) = tmp

        end if

      end do

    end do

  end subroutine sort

end program

OMP_NUM_THREADS= 1

first 4 numbers in a : 0.0000004 0.0000211 0.0000626 0.0000764

first 4 numbers in b : 0.0000067 0.0000323 0.0001214 0.0001603

real 0m1.569s

user 0m1.565s

sys 0m0.004s

[2015年 5月 2日 土曜日 19:21:06 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh kadaiSections

program kadaiSections

  implicit none

  integer, parameter :: N = 20000

  real,allocatable :: a(:), b(:)

  allocate( a(N), b(N) )

  call random_number(a)

  call random_number(b)

!$omp parallel

!$omp sections

!$omp section

  call sort(a)

!$omp section

  call sort(b)

!$omp end sections

!$omp end parallel

  print '("first 4 numbers in a : ",4f10.7)', a(1:4)

  print '("first 4 numbers in b : ",4f10.7)', b(1:4)

contains

  subroutine sort(x)

    real x(:), tmp

    integer i, j

    do i = 1, N - 1

      do j = i + 1, N

        if (x(i)>x(j)) then

          tmp = x(i)

          x(i) = x(j)

          x(j) = tmp

        end if

      end do

    end do

  end subroutine sort

end program

OMP_NUM_THREADS= 3

first 4 numbers in a : 0.0000004 0.0000211 0.0000626 0.0000764

first 4 numbers in b : 0.0000067 0.0000323 0.0001214 0.0001603

real 0m1.012s

user 0m2.144s

sys 0m0.028s

[2015年 5月 2日 土曜日 19:34:23 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh kadaiDataSharing

program kadaiDataSharing

!$ use omp_lib

implicit none

integer i, j

integer,parameter :: N=1000000

integer,allocatable :: a(:), b(:)

!$ double precision st, en

allocate( a(N), b(N) )

!$ st = omp_get_wtime()

a = (/(i,i=1,N)/)

b = 0

!$omp parallel default(none), private(i,j), shared(a,b)

!$omp do

do i=1, N

j = N - i + 1

b(j) = a(i)

end do

!$omp end do

!$omp end parallel

!$ en = omp_get_wtime()

print *, sum(a), "=", sum(b)

!$ print *, "Elapsed time :", en-st

end program

OMP_NUM_THREADS= 1

1784293664 = 1784293664

Elapsed time : 8.723974227905273E-003

real 0m0.013s

user 0m0.006s

sys 0m0.007s

[2015年 5月 2日 土曜日 19:34:51 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh kadaiDataSharing

program kadaiDataSharing

!$ use omp_lib

implicit none

integer i, j

integer,parameter :: N=1000000

integer,allocatable :: a(:), b(:)

!$ double precision st, en

allocate( a(N), b(N) )

!$ st = omp_get_wtime()

a = (/(i,i=1,N)/)

b = 0

!$omp parallel default(none), private(i,j), shared(a,b)

!$omp do

do i=1, N

j = N - i + 1

b(j) = a(i)

end do

!$omp end do

!$omp end parallel

!$ en = omp_get_wtime()

print *, sum(a), "=", sum(b)

!$ print *, "Elapsed time :", en-st

end program

OMP_NUM_THREADS= 32

1784293664 = 1784293664

Elapsed time : 0.121893882751465

real 0m0.413s

user 0m6.048s

sys 0m0.010s

[2015年 5月 2日 土曜日 19:48:46 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh pi

! http://jp.xlsoft.com/documents/intel/compiler/527J-001.pdf

 program omp

 !$ use omp_lib

 integer num_steps

 real*8 step,x,pi,sum

 num_steps = 100000000

 step = 1.0D0 / dble(num_steps)

 sum = 0.0D0

 !$OMP PARALLEL PRIVATE(X)

 nthread = OMP_GET_NUM_THREADS()

 !$OMP DO REDUCTION(+:SUM)

 do i = 1, num_steps

 x = (dble(i)-0.5d0)*step

 sum = sum + 4.0D0 / (1.0D0 + x * x)

 end do

 !$OMP END DO

 !$OMP END PARALLEL

 pi = step * sum

 write (6,*) nthread,' Threads',' PI = ',pi

 end program omp

OMP_NUM_THREADS= 8

8 Threads PI = 3.14159265358981

real 0m0.608s

user 0m4.035s

sys 0m0.003s

[2015年 5月 2日 土曜日 19:56:41 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh pi

! http://jp.xlsoft.com/documents/intel/compiler/527J-001.pdf

 program omp

 !$ use omp_lib

 integer num_steps

 real*8 step,x,pi,sum

 num_steps = 100000000

 step = 1.0D0 / dble(num_steps)

 sum = 0.0D0

 !$OMP PARALLEL PRIVATE(X)

 nthread = OMP_GET_NUM_THREADS()

 !$OMP DO REDUCTION(+:SUM)

 do i = 1, num_steps

 x = (dble(i)-0.5d0)*step

 sum = sum + 4.0D0 / (1.0D0 + x * x)

 end do

 !$OMP END DO

 !$OMP END PARALLEL

 pi = step * sum

 write (6,*) nthread,' Threads',' PI = ',pi

 end program omp

OMP_NUM_THREADS= 32

32 Threads PI = 3.14159265358981

real 0m0.408s

user 0m9.673s

sys 0m0.184s

[2015年 6月 5日 金曜日 09:44:44 JST]

[~/OpenMP]

[am@aofd165]

$ cat omp_parallel_do.sh

#!/bin/bash

# 4 doループの並列化

# http://www.nag-j.co.jp/openMP/openMPDoDirective.html

#

exe=$(basename $0 .sh)

src=$(basename $0 .sh).f90

cat <<EOF > $src

program ${exe}

!$ use omp_lib

implicit none

integer,parameter :: N = 10

integer i

real a(N),sum

!\$OMP PARALLEL DO   &

!\$OMP PRIVATE ( i )

do i=1,N

  a(i) = float(i)

  print *, "Num_threads =", omp_get_num_threads(), ".I am ", omp_get_thread_num(),"a(i)=",a(i)

end do

!\$OMP END PARALLEL DO

print *,'a(i):'

print *, a

print *

do i=1,N

  sum=sum+a(i)

enddo

print *,'sum=',sum

end program ${exe}

EOF

ifort -o ${exe} -openmp ${src}

if [ $? -ne 0 ]; then

  echo

  echo COMPILE ERROR!

  echo

  exit 1

fi

nt=5

export OMP_NUM_THREADS=$nt

echo

echo "OMP_NUM_THREADS= "$nt

echo

./$exe

exit 0

$ omp_parallel_do.sh

OMP_NUM_THREADS= 5

Num_threads = 5 .I am 0 a(i)= 1.000000

Num_threads = 5 .I am 0 a(i)= 2.000000

Num_threads = 5 .I am 2 a(i)= 5.000000

Num_threads = 5 .I am 2 a(i)= 6.000000

Num_threads = 5 .I am 4 a(i)= 9.000000

Num_threads = 5 .I am 4 a(i)= 10.00000

Num_threads = 5 .I am 3 a(i)= 7.000000

Num_threads = 5 .I am 3 a(i)= 8.000000

Num_threads = 5 .I am 1 a(i)= 3.000000

Num_threads = 5 .I am 1 a(i)= 4.000000

a(i):

1.000000 2.000000 3.000000 4.000000 5.000000

6.000000 7.000000 8.000000 9.000000 10.00000

sum= 55.00000

$ compile.run.sh tama

!

!http://tama.green.gifu-u.ac.jp/~tama/Memo/openmp.html

!

implicit none

integer, parameter :: N=1000, M=1000, NN=10, MAX_THREAD=10

integer :: i, j, i1, j1

real, allocatable :: x(:,:), y(:,:)

!$use omp_lib

character(len=1024) :: omp_num_threads

integer, external :: omp_get_thread_num

integer :: i_omp_num_threads, i_thread

integer :: ifile

character(len=1024) :: filename

integer :: neach, istart, iend

call getenv("OMP_NUM_THREADS", omp_num_threads)

read(omp_num_threads, *) i_omp_num_threads

write(0,*) 'OMP_NUM_THREADS ', i_omp_num_threads

!allocate( x(N,M) )

!allocate( y(N,M) )

neach = N/i_omp_num_threads

!$omp parallel private(i_thread, istart, iend, filename, ifile, i, j, x, y )

i_thread = omp_get_thread_num()

write(*,*) 'thread num = ', i_thread

istart = neach * i_thread + 1

iend   = neach * (i_thread+1)

ifile = 10+i_thread

write(filename, '("OUT-", I1)') i_thread

allocate( x(neach,M) )

allocate( y(neach,M) )

do i=istart, iend

  do j=1, M

    x(i,j) = log(abs( sin(0.2*i*j)))

  end do

end do

open(unit=ifile, file=filename(1:len_trim(filename)))

if( istart < NN+1 ) istart = NN+1

if( iend > N-NN )   iend   = N-NN

do i=istart, iend

 do j=NN+1, M-NN

  y(i,j)  = 0

   do i1=-NN, NN

    do j1 = -NN, NN

     y(i, j) = y(i,j) + x(i+i1, j+j1)

    end do

   end do

   y(i,j) = y(i,j) / ((2*NN+1)*(2*NN+1))

 end do

 write(ifile, *) i, (y(i,j), j=NN+1, M-NN)

end do

close(ifile)

!$omp end parallel

end

OMP_NUM_THREADS= 32

OMP_NUM_THREADS 32

thread num = 14

thread num = 30

thread num = 21

thread num = 13

thread num = 2

thread num = 7

thread num = 9

thread num = 17

thread num = 1

thread num = 22

thread num = 11

thread num = 5

thread num = 20

thread num = 12

thread num = 8

thread num = 31

thread num = 3

thread num = 4

thread num = 10

thread num = 16

thread num = 24

thread num = 27

thread num = 26

thread num = 15

thread num = 23

thread num = 18

thread num = 25

thread num = 0

thread num = 28

thread num = 19

thread num = 6

thread num = 29

forrtl: 致命的なエラー (174): SIGSEGV、segmentation fault occurred

Image PC Routine Line Source

tama.exe 0000000000418875 Unknown Unknown Unknown

tama.exe 00000000004343B4 Unknown Unknown Unknown

tama.exe 0000000000404061 Unknown Unknown Unknown

libiomp5.so 00002B6F6E14F4F3 Unknown Unknown Unknown

$ export OMP_STACKSIZE=64000000

[2015年 5月 2日 土曜日 19:43:13 JST]

[~/OpenMP]

[am@aofd165]

$ ulimit -s unlimited

[2015年 5月 2日 土曜日 19:43:22 JST]

[~/OpenMP]

[am@aofd165]

$ compile.run.sh tama

!

!http://tama.green.gifu-u.ac.jp/~tama/Memo/openmp.html

!

implicit none

integer, parameter :: N=1000, M=1000, NN=10, MAX_THREAD=10

integer :: i, j, i1, j1

real, allocatable :: x(:,:), y(:,:)

!$use omp_lib

character(len=1024) :: omp_num_threads

integer, external :: omp_get_thread_num

integer :: i_omp_num_threads, i_thread

integer :: ifile

character(len=1024) :: filename

integer :: neach, istart, iend

call getenv("OMP_NUM_THREADS", omp_num_threads)

read(omp_num_threads, *) i_omp_num_threads

write(0,*) 'OMP_NUM_THREADS ', i_omp_num_threads

!allocate( x(N,M) )

!allocate( y(N,M) )

neach = N/i_omp_num_threads

!$omp parallel private(i_thread, istart, iend, filename, ifile, i, j, x, y )

i_thread = omp_get_thread_num()

write(*,*) 'thread num = ', i_thread

istart = neach * i_thread + 1

iend   = neach * (i_thread+1)

ifile = 10+i_thread

write(filename, '("OUT-", I1)') i_thread

allocate( x(neach,M) )

allocate( y(neach,M) )

do i=istart, iend

  do j=1, M

    x(i,j) = log(abs( sin(0.2*i*j)))

  end do

end do

open(unit=ifile, file=filename(1:len_trim(filename)))

if( istart < NN+1 ) istart = NN+1

if( iend > N-NN )   iend   = N-NN

do i=istart, iend

 do j=NN+1, M-NN

  y(i,j)  = 0

   do i1=-NN, NN

    do j1 = -NN, NN

     y(i, j) = y(i,j) + x(i+i1, j+j1)

    end do

   end do

   y(i,j) = y(i,j) / ((2*NN+1)*(2*NN+1))

 end do

 write(ifile, *) i, (y(i,j), j=NN+1, M-NN)

end do

close(ifile)

!$omp end parallel

end

OMP_NUM_THREADS= 32

OMP_NUM_THREADS 32

thread num = 10

thread num = 21

thread num = 23

thread num = 24

thread num = 27

thread num = 26

thread num = 1

thread num = 2

thread num = 4

thread num = 20

thread num = 13

thread num = 19

thread num = 16

thread num = 17

thread num = 14

thread num = 3

thread num = 29

thread num = 0

thread num = 30

forrtl: 致命的なエラー (174): SIGSEGV、segmentation fault occurred

Image PC Routine Line Source

tama.exe 0000000000493D6F Unknown Unknown Unknown

tama.exe 000000000046C94F Unknown Unknown Unknown

tama.exe 0000000000463E77 Unknown Unknown Unknown

tama.exe 0000000000434714 Unknown Unknown Unknown

tama.exe 0000000000404061 Unknown Unknown Unknown

libiomp5.so 00002AD4857B44F3 Unknown Unknown Unknown

thread num = 18

thread num = 22

thread num = 28

thread num = 7

thread num = 31

thread num = 15

thread num = 6

thread num = 12

thread num = 8

thread num = 25

thread num = 5

thread num = 11

thread num = 9

real 0m0.041s

user 0m0.163s

sys 0m0.226s