The ABINIT Users Mailing List ( CLOSED )

[Gilles Zerah <gilles.zerah@cea.fr>]

Dear all,

There is a bug in v5.4.3 which probably produced the effect below.
It will be corrected in V5.4.4, but in case you need to test the v5.4.3, 
here is a patch.

run patch -p0 < prtrhomxmn.F90-patch.txt in your abinit directory to 
apply it (keep a copy of 15common/prtrhomxmn.F90 to
make sure things went well).

Gilles (and also Damien Caliste, Muriel Delaveau, Philippe Blaise)

> Dear Abinit users,
>
> I try to relax (ionmov=2) a large system with PAW pseudopotentials and 
> band-fft parallelisation (64 CPUs) with abinip. I find this message in the 
> log file (I just give the begining and the end because it is very long):
>
>
>  *** glibc detected *** 
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip: free(): invalid 
> next size (normal): 0x60000000122b0de0 ***
> *** glibc detected *** 
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip: free(): invalid 
> next size (normal): 0x60000000123f2d20 ***
> *** glibc detected *** 
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip: free(): invalid 
> next size (normal): 0x600000001238af70 ***
> *** glibc detected *** 
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip: free(): invalid 
> next size (normal): 0x6000000012229d90 ***
> ======= Backtrace: =========
> /lib/libc.so.6.1[0x2000000003f37b00]
> /lib/libc.so.6.1(__libc_free+0x3ef2eb8)[0x2000000003f3b8a0]
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip[0x40000000021ac360]
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip[0x4000000001f044e0]
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip[0x4000000001b6cfa0]
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip[0x40000000010d30c0]
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip[0x4000000000094c10]
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip[0x40000000000357b0]
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip[0x400000000000fa50]
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip[0x40000000000043c0]
> *** glibc detected *** 
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip: free(): invalid 
> next size (normal): 0x6000000012318b90 ***
> /lib/libc.so.6.1(__libc_start_main+0x3e46c38)[0x2000000003e8f630]
> /RQexec/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip[0x4000000000004140]
>
> [.......
> ........
> ........]
>
> 2000001c74f58000-2000001c75758000 rw-s 00000000 00:00 0
> 2000001c75758000-2000001cf03e0000 rw-s 00000000 00:00 0
> 2000001cf03e0000-2000001cf0be0000 rw-s 00000000 00:00 0
> 2000001cf0be0000-2000001d6b868000 rw-s 00000000 00:00 0
> 2000001d6b868000-2000001d6c068000 rw-s 00000000 00:00 0
> 2000001d6c068000-2000001de6cf0000 rw-s 00000000 00:00 0
> 2000001de6cf0000-2000001de74f0000 rw-s 00000000 00:00 0
> 2000001de74f0000-2000001e62178000 rw-s 00000000 00:00 0
> 2000001e62178000-2000001e62978000 rw-s 00000000 00:00 0
> 2000001e62978000-2000001edd600000 rw-s 00000000 00:00 0
> 2000001edd600000-2000001edde00000 rw-s 00000000 00:00 0
> 2000001edde00000-2000001edde04000 r--s 00000000 00:0f 104708             
> /dev/mmtimer
> 2000001edde04000-2000001edde0c000 rw-p 2000001edde04000 00:00 0
> 2000001eddf00000-2000001eddf24000 rw-p 2000001eddf00000 00:00 0
> 2000001eddf24000-2000001ede000000 ---p 2000001eddf24000 00:00 0
> 4000000000000000-40000000022cc000 r-xp 00000000 fc:08 55838240856        
> /RQMNT/exec3/gcote/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip
> 6000000000008000-6000000000150000 rw-p 022c8000 fc:08 55838240856        
> /RQMNT/exec3/gcote/cotemich/Abinit/abinit-5.5.0/tmp_fft/src/main/abinip
> 6000000000150000-6000000059e70000 rw-p 6000000000150000 00:00 0          
> [heap]
> 603fffffffffc000-6040000000000000 rw-p 603fffffffffc000 00:00 0
> 607ffffffe99c000-607ffffffea34000 rw-p 607ffffffe99c000 00:00 0          
> [stack]
> a000000000000000-a000000000020000 ---p 00000000 00:00 0                  
> [vdso]
>
>
> which is the last thing written in the log file and the first Broyden step 
> never starts...
> Any idea?
> I also give my input file hereafter:
>
>
> acell 14.2154 24.6252 28.2404 angstrom
> rprim  1.0 0.0 0.0
>        0.0 1.0 0.0
>        0.0 0.0 1.0
> vacwidth 10
> ntypat 4
> znucl 8 13 39 40
> natom 640
> typat
>  3 4 4 4 3 4 4 4 4 4 4 3 9*4
>  3 4 4 3 3 16*4 3 4 4 4 3 4 92*1
>  3 4 4 4 3 4 4 4 4 4 4 3 9*4
>  3 4 4 3 3 16*4 3 4 4 4 3 4 92*1
>  216*1 144*2
> xangst
>
> [.......
> ........
> ........]
>
> ecut 15
> nband 2640
> ngfft 96 192 192
> ngfftdg 144 256 288
> npband 8
> npfft 8
> pawecutdg 35
> nkpt 1
> kpt 0.0 0.0 0.0
> istwfk 1
> fftalg 401
> wfoptalg 4
> fft_opt_lob 2
> intxc 0
> iprcch 0
> nloalg 4
> nstep 1000
> toldfe 1.0d-8
> diemac 3.0
> diemix 0.3
> iscf 17
> ionmov 2
> ntime 200
> tolmxf 1.0d-4
>
>
> Thanks to all
>
> Francois
>
>   

--- src/15common/prtrhomxmn.F90 2007-09-10 17:15:40.000000000 +0200
+++ src/15common/prtrhomxmn.F90.new     2007-09-12 14:11:19.000000000 +0200
@@ -205,9 +205,9 @@
   mpi_enreg%paral_level=3
   call xcomm_init(mpi_enreg,spaceComm)
   if(mpi_enreg%mode_para=='b') spaceComm=mpi_enreg%comm_fft
-  
allocate(value_fft(2,2,nspden+2,mpi_enreg%nproc_fft),index_fft(2,2,nspden+2,mpi_enreg%nproc_fft))
+  
allocate(value_fft(2,2,nitems,mpi_enreg%nproc_fft),index_fft(2,2,nitems,mpi_enreg%nproc_fft))
   value_fft(:,:,:,:)=zero ; index_fft(:,:,:,:)=0
-  value_fft(:,:,:,mpi_enreg%me_fft)=value(:,:,:) ; 
index_fft(:,:,:,mpi_enreg%me_fft)=index(:,:,:)
+  value_fft(:,:,:,mpi_enreg%me_fft+1)=value(:,:,:) ; 
index_fft(:,:,:,mpi_enreg%me_fft+1)=index(:,:,:)
   call xsum_mpi(value_fft,spaceComm,ierr)
   call xsum_mpi(index_fft,spaceComm,ierr)
 
@@ -253,6 +253,7 @@
 
   deallocate(value_fft,index_fft)
 
+  mpi_enreg%paral_level=old_paral_level
  endif ! if (mpi_enreg%paral_fft == 1)
 
  call xmaster_init_fft(mpi_enreg,master)