The ABINIT Users Mailing List ( CLOSED )

["Zhenhua Zeng" <myid520@163.com>]

Dear All,

I want to switch 5.3.4 to 5.6.4 now.
Both mpi and abinit are compiled using infort10.1 as recommended by 5.6.4, 
and the compiling process seems ok.
however some error (just) occur when running in mutli-node + read wfk from 
disk model.
I will explain the errors below:

[1] I tested using both openmpi-1.2.8 and lam/mpi-7.1.3, mpi run ok
[2] abinit is configured as below: ./configure --disable-netcdf 
--disable-etsf-io  --disable-bigdft --disable-wannier90 FC=mpif90 
--enable-mpi="yes" (for openmpi)
[3] Abinip is test in following combination, and the states are given below:

state  multi-node multi-processor  multi-dataset  getwfk from disk 
fail            yes             yes                     yes                   
          yes
fail            yes             yes                     no                    
          yes
ok                      yes                     yes                           
  yes                             no
ok                      no                      yes                           
  yes                             yes
(fail           yes                     --                              --    
                          yes)

I.e. Abinip crash when it runs in mutli-node and read data from disk model at 
the same time.
Messages given bellow.

#############################  error in error file (for lam) 
#############################
n-1<22283> ssi:boot:base:linear: booting n0 (compute-1-17.local)
n-1<22283> ssi:boot:base:linear: booting n1 (compute-1-39.local)
n-1<22283> ssi:boot:base:linear: finished
1    #Zhenhua: pay attention this number
1    #Zhenhua: pay attention this number

################################ error in log file (for lam) 
############################
 hdr_check:  Density/Potential file is OK for restart of calculation
================================================================================
 ioarr: data read from disk file telphon_1o_DS1_DEN
 
================================================================================
 
     iter   Etot(hartree)      deltaE(h)  residm     vres2    diffor    maxfor
 
 getcut: wavevector=  0.0000  0.0000  0.0000  ngfft=  16  16  27
         ecut(hartree)=     10.000   => boxcut(ratio)=   2.11224
 
 ewald : nr and ng are    2 and   25
 vtorho : nnsclo_now=800, note that nnsclo,dbl_nnsclo,istep=  4 0  1
MPI_Recv: process in local group is dead (rank 0, comm 3)
Rank (0, MPI_COMM_WORLD): Call stack within LAM:
Rank (0, MPI_COMM_WORLD):  - MPI_Recv()
Rank (0, MPI_COMM_WORLD):  - MPI_Barrier()
Rank (0, MPI_COMM_WORLD):  - MPI_Barrier()
Rank (0, MPI_COMM_WORLD):  - main()

############ no error message in openmpi, however abinip stoped at the 
position below  #############
 hdr_check:  Density/Potential file is OK for restart of calculation
================================================================================
 ioarr: data read from disk file telphon_1o_DS1_DEN
 
================================================================================
 
     iter   Etot(hartree)      deltaE(h)  residm     vres2    diffor    maxfor
 
 getcut: wavevector=  0.0000  0.0000  0.0000  ngfft=  16  16  27
         ecut(hartree)=     10.000   => boxcut(ratio)=   2.11224
 
 ewald : nr and ng are    2 and   25
 vtorho : nnsclo_now=800, note that nnsclo,dbl_nnsclo,istep=  4 0  1
######################################################################################

I have tried alot, however the error is still on.
Did someone also encounter the similar error?
Your comments are greatly appreciated!
                                
Best Wishes
Zhenhua Zeng