I am trying to calculate a sum of a column vector using mpi.
I have read the following thread
Can MPI sendbuf and recvbuf be the same thing?
and found the use of MPI_IN_PLACE would be a good choice to minimize the memory usage for very large data.
However, in my trivial test the mpi_allreduce keeps reporting memory segmentation fault:
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
I could not figure out what went wrong.
Any help would be greatly appreciated.
Here's my code:
program mpitest
implicit none
include "mpif.h"
integer :: i, mprank, mpsize, mpierr
real(KIND=8), dimension(5) :: Qin
call MPI_INIT(mpierr)
call MPI_COMM_SIZE(MPI_COMM_WORLD,mpsize,mpierr)
call MPI_COMM_RANK(MPI_COMM_WORLD,mprank,mpierr)
write(*,*) mprank,'/',mpsize
do i = 1, 5
Qin(i) = (mprank+1)*i
write(*,*) Qin(i), mprank
enddo
call MPI_AllReduce(MPI_IN_PLACE,Qin,5,MPI_DOUBLE, MPI_SUM, MPI_COMM_WORLD)
if (mprank == 0) then
do i = 1, 5
write(*,*) Qin(i)
enddo
endif
call MPI_FINALIZE(mpierr)
end program mpitest
I'm learning Fortran with the book Fortran 90 for scientists and engineers by Brian Hahn. In chapter 9 about arrays, pages 131/132, he gives the following code as an example of dynamic arrays
Program Chap_9_Allocatable_Array
Implicit none
! Variables
Real, dimension(:), Allocatable :: X, OldX
Real A
Integer IO, N, i
! Body of Chap_9_Allocatable_Array
Allocate( X(0) ) !Size zero to sart with?
N = 0
Open(1, File = 'Data.txt')
Do
Read(1, *, IOStat = IO) A
If (IO < 0) Exit
N = N + 1
Allocate( OldX( Size(X) ) )
OldX = X !Entire array can be assigned
Deallocate( X )
Allocate( X(N) )
X = OldX
X(N) = A
Deallocate( OldX )
End do
Print *, (X(i), i = 1, N)
End program Chap_9_Allocatable_Array
I have implemented this program in Visual Studio Community 2019 with the Intel Visual Fortran Compiler. The purpose of this program as he explains is
The following program extract shows how to use allocatable arrays, as these beasts are called, to read an unknown amount of data, which unfortunately must be supplied one item per line because of the way READ works.
I found an interesting error. The file data.txt consists of 100 random numbers, 1 per row. When I try to run it, it just seems to stall for a couple of seconds and then the console simply prints the
Press any key to continue.
prompt, without an error message. I have inserted some debug prints and determined that the program runs the do cycle between 3 to 8 times before stopping. I have not been able to determine the reason. If I then change the data.txt file to only be 3 numbers long, the program runs as intended. With the debug prints, I have pinned the error to being the
Deallocate( X )
line. If I debug the program in Visual Studio I just get the following message:
Chap_9_Allocatable_Array.exe has triggered a breakpoint.
There have been a few minor errors in the book. Just in this example, the author seems to have forgotten to declare i, which caused a compile error. However, as I'm only beggining to understand arrays, I don't know what else to try. Any ideas?
I have the following code.
PROGRAM CTS
implicit none
!C driver for routine fourn
INTEGER NDAT,NDIM
PARAMETER(NDIM=1,NDAT=1024)
INTEGER i,idum,isign,j,k,l,nn(NDIM)
REAL data1(NDAT),data2(NDAT),ran1 ,x,dx
REAL,DIMENSION(:),ALLOCATABLE::F,F1
allocate(F(NDAT),F1(NDAT))
x=1.
dx = (200.-1.)/real(NDAT)
nn(1)=NDAT
do i=1,NDAT
F1(i) =atan(x-100)
x= x + dx
enddo
x=1.
x=1.
isign=1
call fo(F1,nn,1,isign)
open(1,file="zresult.dat",status="replace")
do i=1,NDAT
write(1,*)x,F1(i)*dx
x= x + dx
enddo
stop
END
!!!!!!!!!!!!!!!!!!!!!!!!!!
SUBROUTINE fo(data,nn,ndim,isign)
INTEGER isign,ndim,nn(ndim)
REAL data(*)
INTEGER i1,i2,i2rev,i3,i3rev,ibit,idim,ifp1,ifp2,ip1,ip2,ip3,k1,&
k2,n,nprev,nrem,ntot
REAL tempi,tempr
DOUBLE PRECISION theta,wi,wpi,wpr,wr,wtemp
ntot=1
do 11 idim=1,ndim
ntot=ntot*nn(idim)
11 continue
nprev=1
do 18 idim=1,ndim
n=nn(idim)
nrem=ntot/(n*nprev)
ip1=2*nprev
ip2=ip1*n
ip3=ip2*nrem
i2rev=1
do 14 i2=1,ip2,ip1
if(i2.lt.i2rev)then
do 13 i1=i2,i2+ip1-2,2
do 12 i3=i1,ip3,ip2
i3rev=i2rev+i3-i2
tempr=data(i3)
tempi=data(i3+1)
data(i3)=data(i3rev)
data(i3+1)=data(i3rev+1)
data(i3rev)=tempr
data(i3rev+1)=tempi
12 continue
13 continue
endif
ibit=ip2/2
1 if ((ibit.ge.ip1).and.(i2rev.gt.ibit)) then
i2rev=i2rev-ibit
ibit=ibit/2
goto 1
endif
i2rev=i2rev+ibit
14 continue
ifp1=ip1
2 if(ifp1.lt.ip2)then
ifp2=2*ifp1
theta=isign*6.28318530717959d0/(ifp2/ip1)
wpr=-2.d0*sin(0.5d0*theta)**2
wpi=sin(theta)
wr=1.d0
wi=0.d0
do 17 i3=1,ifp1,ip1
do 16 i1=i3,i3+ip1-2,2
do 15 i2=i1,ip3,ifp2
k1=i2
k2=k1+ifp1
tempr=sngl(wr)*data(k2)-sngl(wi)*data(k2+1)
tempi=sngl(wr)*data(k2+1)+sngl(wi)*data(k2)
data(k2)=data(k1)-tempr
data(k2+1)=data(k1+1)-tempi
data(k1)=data(k1)+tempr
data(k1+1)=data(k1+1)+tempi
15 continue
16 continue
wtemp=wr
wr=wr*wpr-wi*wpi+wr
wi=wi*wpr+wtemp*wpi+wi
17 continue
ifp1=ifp2
goto 2
endif
nprev=n*nprev
18 continue
return
END
!!!!!!!!!!!
The problem is If I do not allocate F1 and put REAL F1(NDAT), the code runs without any problem, but when I allocate F1 I will get the following error
I have tried all possibilities to understand what is happening -fcheck=all etc. it seems memory corruption.
*** Error in `./out': free(): invalid next size (normal): 0x088a7f20 ***
Program received signal SIGABRT: Process abort signal.
Backtrace for this error:
#0 0xB76BE133
#1 0xB76BE7D0
#2 0xB77C73FF
#3 0xB77C7424
#4 0xB74E4686
#5 0xB74E7AB2
#6 0xB751EFD2
#7 0xB75294C9
#8 0xB752A13C
#9 0xB7777607
#10 0xB776EECF
#11 0xB776EFB9
#12 0xB76BDA93
#13 0xB77D733B
#14 0xB74E9230
#15 0xB74E928C
#16 0xB76C09E7
#17 0x80496D4 in cts at z2.f90:33
Aborted (core dumped)
Could you please help me to find out where the problem is.
Thank you so much
If you move the END after the subroutine, put CONTAINS before the subroutine to make it internal the program, change assumed size array
data(*)
to assumed shape array
data(:)
(just using data(NDAT) would also help)
then you can compile your code as
gfortran-7 -Wall -Wno-unused-variable -fcheck=all memcorr.f90
and get clear message
> ./a.out
At line 63 of file memcorr.f90
Fortran runtime error: Index '1025' of dimension 1 of array 'data' above upper bound of 1024
That means your are accessing your array out of bounds.
Line 63 is:
data(i3)=data(i3rev)
so i3 or i3rev is too large (larger than NDAT). You must find out why and fix that.
The point is: use explicit interfaces, assumed shape arrays and all other Fortran 90 stuff that will help you find bugs.
The best thing is to use modules for all your subroutines and functions.
I'm currenlty learning modern Fortran, I use the TDM-GCC compiler on Windows 10 64bit.
After compiling correcly my code, I received this message when trying to run the output .exe
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
Backtrace for this error:
#0 ffffffffffffffff
#1 ffffffffffffffff
#2 ffffffffffffffff
#3 ffffffffffffffff
#4 ffffffffffffffff
#5 ffffffffffffffff
#6 ffffffffffffffff
#7 ffffffffffffffff
#8 ffffffffffffffff
#9 ffffffffffffffff
#10 ffffffffffffffff
#11 ffffffffffffffff
#12 ffffffffffffffff
#13 ffffffffffffffff
#14 ffffffffffffffff
#15 ffffffffffffffff
#16 ffffffffffffffff
#17 ffffffffffffffff
#18 ffffffffffffffff
I understood the problem may be caused by reading I/O files. Here's the code (it's not mine, it's from the course teacher, he told me it's a runtime error, but has no idea why I get the error while reading a file):
program golden_ratio
! experiments with the golden ratio iterative relation
implicit none
integer, parameter :: rk = kind(1.0d0)
real(rk) :: phi, phi_old
real(rk) :: phi_start, tol
integer :: i, max_iter
open(11,FILE='goldenfile.in',STATUS='old')
read(11,*) phi_start, tol, max_iter
close(11)
! how I wrote the input file
! 5.0
! 0.0001
! 1000
phi_old = phi_start
do i=1,max_iter
phi = 1.0d0/phi_old + 1.0d0
if (abs(phi - phi_old) < tol) exit
phi_old = phi
end do
open(12,FILE='goldenfile.out',STATUS='replace')
write(12,100) 'Start value:',phi_start
write(12,100) 'Tolerance:',tol
write(12,'(2(A," ",I11," "))') 'Ended at iteration:', i, 'of', max_iter
write(12,100) 'Final value:',phi
close(12)
print *, 'Output file created'
100 format(A," ",F13.10)
end program golden_ratio
So this problem occurs when I try to open the input file 'goldenfile.in' (which is present in the same directory, defined as I wrote in the commented section, obviously it doesn't contain the comment marks).
It also occurs when I forgo using the input file, and I try to overwrite the output file 'goldenfile.out' after having obtained it a first time. Using 'replace' or not doesn't change the situation.
The strange thing is that the compilation goes right.
This is my program to solve a problem of three dimensional cylindrical fin. But when i run this program in Fortran 6.2 this error shows up:
Error: The shapes of the array expressions do not conform. [T]"
Now i don't understand why this is happening. I need quick assistance. Anyone please help me.
PROGRAM CYLINDRICAL FIN
DIMENSION T(500,500,500), OLDT(500,500,500),ERR(500,500,500)
DIMENSION R(500),Y(500),Z(500)
REAL CC,H,DR,DY,DZ,A,D,RY,YZ,ZR,B,E,TA,RL,YL,ZL
+P,AC,QF,MF,Q,EF,EFF,QMAX,AS,LC,QT,QF1,QF2
INTEGER I,J,K,M,N,L,M1,N1,L1,M2,M4,M34,N2,N4,N34,L2,L4,L34
RL=0.5
YL=6.283
ZL=0.04
M=100
N=40
L=20
M2=((M/2)+1)
M4=((M/4)+1)
M34=((3*M/4)+1)
N2=((N/2)+1)
N4=((N/4)+1)
N34=((3*N/4)+1)
L2=((L/2)+1)
L4=((L/4)+1)
L34=((3*L/4)+1)
DR=RL/M
DY=YL/N
DZ=ZL/L
CC=400.0
H=10.0
TA=25
M1=M-1
N1=N-1
L1=L-1
************VARIABLES************
A=DR*DY*DZ
D=DR+DY+DZ
RY=DR*DY
YZ=DY*DZ
ZR=DZ*DR
E=RY+YZ+ZR
************VARIABLES FOR EFFICIENCY AND EFFECTIVENESS (CROSS-SECTION AREA,PERIMETER,M,SURFACE AREA OF FIN)************
AC=3.1416*DR**2
P=2*(3.1416*DR+DZ)
MF=((H*P)/(CC*AC))**(0.5)
AS=2*3.1416*DR*DZ+3.1416*DR**2
************************************** distance discritization ******************
R(1)=0.0
Y(1)=0.0
Z(1)=0.0
R(M+1)=RL
Y(N+1)=YL
Z(L+1)=ZL
DO I=2,M
R(I)=R(I-1)+DR
END DO
DO J=2,N
Y(J)=Y(J-1)+DY
END DO
DO K=2,L
Z(K)=Z(K-1)+DZ
END DO
DO I=1,M
DO J=1,N
DO K=1,L
T(I,J,K)=0.0
END DO
END DO
END DO
DO I=1,M
DO J=1,N
T(I,J,1)=400
END DO
END DO
ITER=0.0
READ(*,*) LAST
31 CONTINUE
ITER=ITER+1
*************************************** FORMULAS**********************************
DO I=2,M1
DO J=2,N1
DO K=2,L1
T(I,J,K)=((R*DR*T(I+1,J,K)*(YZ)**2.0)-((T(I+1,J,K)+T(I-1,J,K))*
+(R*YZ)**2.0)-((T(I,J+1,K)+T(I,J-1,K))*(ZR**2.0))-((T(I,J,K+1)+
+T(I,J,K-1))*(R*RY)**2.0))/((R*DR*(YZ)**2.0)-(2.0*(R*YZ)**2.0)-
+(2.0*(ZR)**2.0)-(2.0*(R*RY)**2.0))
END DO
END DO
END DO
*************************************** END OF ITERATIONG FORMULAS****************
DO I=1,M
DO J=1,N
DO K=1,L
OLDT(I,J,K)=T(I,J,K)
END DO
END DO
END DO
DO I=1,M
DO J=1,N
DO K=1,L
ERR(I,J,K)=T(I,J,K)-OLDT(I,J,K)
END DO
END DO
END DO
EMAX=0.0
EMAX=MAX(EMAX,ERR(I,J,K))
WRITE(*,*) ITER, EMAX
IF (ITER.LT.LAST) GOTO 31
WRITE(*,*) DR,A,B,E
END PROGRAM CYLINDRICAL FIN
In your long formula you have R without index. But R is a vector so that's why you get the error.
DO I=1,M
DO J=1,N
T(I,J,1)=400
END DO
END DO
can be written more swiftly as
T(1:M,1:N,1) = 400
also 400 is an integer so you better use 400.
I quickly converted the program to free-format and compiled it with gfortran. The problem line is:
T(I,J,K)=((R*DR*T(I+1,J,K)*(YZ)**2.0)-((T(I+1,J,K)+T(I-1,J,K))* &
(R*YZ)**2.0)-((T(I,J+1,K)+T(I,J-1,K))*(ZR**2.0))-((T(I,J,K+1)+ &
T(I,J,K-1))*(R*RY)**2.0))/((R*DR*(YZ)**2.0)-(2.0*(R*YZ)**2.0)- &
(2.0*(ZR)**2.0)-(2.0*(R*RY)**2.0))
The gfortran error message is:
"Error: Incompatible ranks 0 and 1 in assignment at (1)"
with the (1) designation the "T" on the LHS. Obviously the T(I,J,K) is a scaler of rank 0. Somewhere on the RHS you have an array of rank 1 -- identified by Azrael3000 as "R".
P.S. I recommend using "implicit none" and typing all of your variables. Implicit typing is pernicious, e.g., it allows typos in variable names to be accepted by the compiler.
Your problem comes from the array operation you are trying to perform.
If you simplify the line:
T(I,J,K)=((R*DR*T(I+1,J,K)*(YZ)**2.0)-((T(I+1,J,K)+T(I-1,J,K))*
+(R*YZ)**2.0)-((T(I,J+1,K)+T(I,J-1,K))*(ZR**2.0))-((T(I,J,K+1)+
+T(I,J,K-1))*(R*RY)**2.0))/((R*DR*(YZ)**2.0)-(2.0*(R*YZ)**2.0)-
+(2.0*(ZR)**2.0)-(2.0*(R*RY)**2.0))
You can end up with this other line that also shows the same error:
T(I,J,K)= T(I,J,K-1)*R
The shapes of the array expressions do not conform. [T]
So you can see that your problem comes from trying to assign an array R to a scalar T(I,J,K).