"Killed" message when trying to read huge file.

[Yeon__Jejoon]

I'm trying to read a file using the same script I asked a question before in 

https://community.intel.com/t5/Intel-Fortran-Compiler/How-can-I-read-a-file-with-varying-number-of-c...

This post. The input data looks like 

8055 3 2 5608 5176 0 0.900 0.814 1.714 2.000 -0.466
8444 3 3 5755 6479 3139 0 0.473 0.309 0.729 1.511 2.000 -0.391
6479 1 6 8444 6895 6710 144 8257 3374 0 0.309 0.623 0.743 0.515 0.650 0.636 3.726 0.000 1.142
8407 2 3 8760 2459 9228 0 0.751 0.811 0.412 2.474 0.000 1.133
5222 3 2 4801 7092 0 0.690 0.615 1.524 2.000 -0.437
1089 3 3 7284 4801 7092 0 0.392 0.900 0.359 1.651 2.000 -0.519

and I'm using a script like: 

Program test
implicit none
integer nn
parameter (nn = 50000)
...
integer,dimension(nn,nn) :: conn
integer,dimension(nn) :: atype,nbonds,atindex,atpointer
integer,dimension(nn) :: molnr,atid2,atype2,paint,atid
real,dimension(nn,nn) :: bos
real,dimension(nn) :: x,y,z,abo,nlp,q,aBO2,aBO3,aBO4,aBO5,aBO6
real,dimension(nn) :: aBO7,aBO8,aBO9,charge,charge2
real,dimension(nn) :: sBO2,sBO3,sBO4,sBO5,sBO6,sBO7,sBO8,sBO9

...
do i=1,nat
read (2,*)atid2(i),atype2(i),nbonds(i),(conn(i,j),j=1,nbonds(i)),molnr(i),(bos(i,j),j=1,nbonds(i)),abo(i),nlp(i),charge2(i)
enddo
...

Now, I need to read a huge number of lines (the number of line per each datablock is 400000 lines. ) So I increased "nn" to 450000. Then, I compiled this with: 

ifort test.f -mcmodel medium -o test_script

Then when I execute this script, then "killed" message appears. 

$ ./test_script

Killed 

The script reads the data and works perfectly with a small-sized file with the same format, and small number of "nn", like nn=50000. But when I increase "nn", it just shows "Killed" message. What would be the best way to escape from this situation? 

This is not my PC or laptop, this is the server cluster of my workplace. 

$ ifort -V
Intel(R) Fortran Intel(R) 64 Compiler for applications running on Intel(R) 64, Version 18.0.5.274 Build 20180823
Copyright (C) 1985-2018 Intel Corporation. All rights reserved.

$ uname -a
Linux login00 3.10.0-693.21.1.el7.x86_64 #1 SMP Wed Mar 7 19:03:37 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux

[Steve_Lionel]

There's probably some resource limit set by the cluster system managers that you exceeded. You'll need to ask them about it.

[Arjen_Markus]

You have two two-dimensional arrays, conn and bos. If you set the parameter nn to 450000 then each will occupy some 8 GB - giving a total of 16 GB. This may exceed the stack space. Try instead to allocate these arrays.

Another thing: compile the program with array bound checking to make sure that nbonds(i) does not exceed the value of nn.

[Yeon__Jejoon]

May I ask other ways to allocate arrays, and compile with array bound checking? 

[jimdempseyatthecove]

...

integer,allocatable,dimension(:,:) :: conn
integer,allocatable,dimension(:) :: atype,nbonds,atindex,atpointer
integer,allocatable,dimension(:) :: molnr,atid2,atype2,paint,atid
real,allocatable,dimension(:,:) :: bos
real,allocatable,dimension(nn) :: x,y,z,abo,nlp,q,aBO2,aBO3,aBO4,aBO5,aBO6
real,allocatable,dimension(nn) :: aBO7,aBO8,aBO9,charge,charge2
real,allocatable,dimension(nn) :: sBO2,sBO3,sBO4,sBO5,sBO6,sBO7,sBO8,sBO9

...
allocate(conn(nn,nn))
allocate(atype(nn),nbonds(nn),atindex(nn),atpointer(nn))
allocate(molnr(nn),atid2(nn),atype2(nn),paint(nn),atid(nn))
allocate(bos(nn,nn))
allocate(x(nn),y(nn),z(nn),abo(nn),nlp(nn),q(nn),aBO2(nn),aBO3(nn),aBO4(nn),aBO5(nn),aBO6(nn))
allocate(aBO7(nn),aBO8(nn),aBO9(nn),charge(nn),charge2(nn))
allocate(sBO2(nn),sBO3(nn),sBO4(nn),sBO5(nn),sBO6(nn),sBO7(nn),sBO8(nn),sBO9(nn))

Also, you will likely have to avoid array expressions and revert to explicit DO loop(s), as an array expression may generate a temporary array on stack.

Jim Dempsey

[jimdempseyatthecove]

I forgot to mention. Linux has a "feature" called the OOM-killer which is designed to kill runaway applications. See the link highlighted for additional information.

Jim Dempsey