Page Menu
Home
c4science
Search
Configure Global Search
Log In
Files
F92113820
pi_hybrid.f90
No One
Temporary
Actions
Download File
Edit File
Delete File
View Transforms
Subscribe
Mute Notifications
Award Token
Subscribers
None
File Metadata
Details
File Info
Storage
Attached
Created
Sun, Nov 17, 11:44
Size
1 KB
Mime Type
text/x-c
Expires
Tue, Nov 19, 11:44 (2 d)
Engine
blob
Format
Raw Data
Handle
22377895
Attached To
rSCINTROPAR Introduction to parallelism
pi_hybrid.f90
View Options
! ==========================================================================
! This exercise is taken from the class Parallel Programming Workshop (MPI,
! OpenMP and Advanced Topics) at HLRS given by Rolf Rabenseifner
! ==========================================================================
program
pi_hybrid
use
mpi
implicit none
integer
,
parameter
::
n
=
10000000
double precision
::
omp_t1
,
omp_t2
double precision
::
dx
,
x
,
sm
,
pif
double precision
,
external
::
f
integer
::
i
integer
::
ierror
,
prank
,
psize
,
error_code
,
nthreads
,
nrank
integer
::
nlocal
,
istart
,
iend
!$ integer, external :: omp_get_max_threads
!$ integer, external :: omp_get_thread_num
integer
::
provided
call
MPI_init_thread
(
MPI_THREAD_MULTIPLE
,
provided
,
ierror
)
if
(
ierror
.ne.
MPI_SUCCESS
)
then
print
*
,
'Cannot continue with MPI not accepting multi-threaded communications'
call
MPI_ABORT
(
MPI_COMM_WORLD
,
error_code
,
ierror
)
end if
call
MPI_Comm_size
(
MPI_COMM_WORLD
,
psize
,
ierror
)
call
MPI_Comm_rank
(
MPI_COMM_WORLD
,
prank
,
ierror
)
! calculate pi = integral [0..1] 4 / (1 + x**2) dx */
omp_t1
=
MPI_Wtime
()
nlocal
=
n
/
psize
istart
=
1
+
nlocal
*
prank
iend
=
nlocal
*
(
prank
+
1
);
dx
=
1.
/
n
sm
=
0.0
;
!$omp parallel private(x)
!$ nthreads = omp_get_max_threads()
!$ nrank = omp_get_thread_num()
!$omp do reduction(+:sm)
do
i
=
istart
,
iend
x
=
(
1.
*
i
-
0.5
)
*
dx
sm
=
sm
+
f
(
x
)
enddo
!$omp end do
!$omp single
call
MPI_Allreduce
(
MPI_IN_PLACE
,
sm
,
1
,
MPI_DOUBLE
,
MPI_SUM
,
MPI_COMM_WORLD
,
ierror
);
!$omp end single nowait
!$omp end parallel
pif
=
dx
*
sm
omp_t2
=
MPI_Wtime
()
if
(
prank
.eq.
0
)
then
print
*
,
'computed pi ='
,
pif
print
*
,
'Running time = '
,
(
omp_t2
-
omp_t1
)
endif
call
MPI_Finalize
(
ierror
)
end program
pi_hybrid
double precision
function
f
(
a
)
implicit none
double precision
,
intent
(
in
)
::
a
f
=
4.
/
(
1.
+
(
a
**
2
))
end function
f
Event Timeline
Log In to Comment