[QE-users] Parallel computing of QE7.1 vc-relax crashes when using large number of processors

[Xin Jin]

Dear Quantum Espresso Forum,

I encountered a problem related to the parallel computing using QE7.1 
for vc-relax.

I was trying to perform a vc-relax for a 3*3*3 BCC tungsten super cell. 
The code works fine for non-parallel computing, also works fine for 
parallel computing if the number of processors is smaller than 10.

However, if the number of processors is larger than 10, I will get 
following MPI error:
/*** An error occurred in MPI_Comm_free//
//*** reported by process [3585895498,2]//
//*** on communicator MPI_COMM_WORLD//
//*** MPI_ERR_COMM: invalid communicator//
//*** MPI_ERRORS_ARE_FATAL (processes in this communicator will now abort,//
//***    and potentially your MPI job)/

For parallel computing, I am using /OpenMPI/3.1.4-gcccuda/. (In 
addition, it seems that If I use OpenMPI V4, the simulation speed will 
be much slower than that of V3.)

Another thing is that, if I decrease the size of the supper cell, for 
example to 2*2*2, then there is no problem in the parallel computing 
even if I use more than 30 processors.

Could you help me look at this problem, please?

The input for QE can be found below.

Thank you in advance!

Xin Jin

/&control//
//
//    calculation='vc-relax' //
//    restart_mode='from_scratch', //
//    prefix='W_relax', //
//    pseudo_dir="../../PP_files",//
//    outdir='./'//
//
// ///
////
//
// &system//
//    ibrav= 0, //
//    celldm(1)=5.972,//
//    nat=  54, //
//    ntyp= 1,//
//    ecutwfc = 50,//
//    ecutrho = 500,//
//    occupations='smearing', smearing='mp', degauss=0.06//
// ///
//
// &electrons//
//    diagonalization='david',//
//    conv_thr =  1.0d-8,//
//    mixing_beta = 0.5,//
// ///
////
// &ions//
// ///
//
// &cell//
//    press = 0.0,//
// ///
////
//ATOMIC_SPECIES//
// W  183.84 W.pbe-spn-kjpaw_psl.1.0.0.UPF//
////
//CELL_PARAMETERS {alat}//
//   3.0  0.0  0.0//
//   0.0  3.0  0.0//
//   0.0  0.0  3.0 //
////
//ATOMIC_POSITIONS {alat}//
//W 0.00000 0.00000 0.00000//
//W 0.50000 0.50000 0.50000//
//W 1.00000 0.00000 0.00000//
//W 1.50000 0.50000 0.50000//
//W 2.00000 0.00000 0.00000//
//W 2.50000 0.50000 0.50000//
//W 0.00000 1.00000 0.00000//
//W 0.50000 1.50000 0.50000//
//W 1.00000 1.00000 0.00000//
//W 1.50000 1.50000 0.50000//
//W 2.00000 1.00000 0.00000//
//W 2.50000 1.50000 0.50000//
//W 0.00000 2.00000 0.00000//
//W 0.50000 2.50000 0.50000//
//W 1.00000 2.00000 0.00000//
//W 1.50000 2.50000 0.50000//
//W 2.00000 2.00000 0.00000//
//W 2.50000 2.50000 0.50000//
//W 0.00000 0.00000 1.00000//
//W 0.50000 0.50000 1.50000//
//W 1.00000 0.00000 1.00000//
//W 1.50000 0.50000 1.50000//
//W 2.00000 0.00000 1.00000//
//W 2.50000 0.50000 1.50000//
//W 0.00000 1.00000 1.00000//
//W 0.50000 1.50000 1.50000//
//W 1.00000 1.00000 1.00000//
//W 1.50000 1.50000 1.50000//
//W 2.00000 1.00000 1.00000//
//W 2.50000 1.50000 1.50000//
//W 0.00000 2.00000 1.00000//
//W 0.50000 2.50000 1.50000//
//W 1.00000 2.00000 1.00000//
//W 1.50000 2.50000 1.50000//
//W 2.00000 2.00000 1.00000//
//W 2.50000 2.50000 1.50000//
//W 0.00000 0.00000 2.00000//
//W 0.50000 0.50000 2.50000//
//W 1.00000 0.00000 2.00000//
//W 1.50000 0.50000 2.50000//
//W 2.00000 0.00000 2.00000//
//W 2.50000 0.50000 2.50000//
//W 0.00000 1.00000 2.00000//
//W 0.50000 1.50000 2.50000//
//W 1.00000 1.00000 2.00000//
//W 1.50000 1.50000 2.50000//
//W 2.00000 1.00000 2.00000//
//W 2.50000 1.50000 2.50000//
//W 0.00000 2.00000 2.00000//
//W 0.50000 2.50000 2.50000//
//W 1.00000 2.00000 2.00000//
//W 1.50000 2.50000 2.50000//
//W 2.00000 2.00000 2.00000//
//W 2.50000 2.50000 2.50000//
//
//K_POINTS {automatic}//
//4 4 4 0 0 0//
/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20221028/87b8c648/attachment.html>