Package: Quantum-ESPRESSO
In this page:
Test 1
A Car-Parrinello molecular dynamics of a large system containing 1532 atoms were performed.
The code shows a good scalability up to 65536 cores.
Kernel: cp.x
Atomic Species/Number of Atoms: C/1164, H/320, N/32, O/16
Pseudopotential: Ultrasoft-Vanderbilt
XC Functional: PBE
Plane Wave Energy cut-off: 25Rydberg
Execution parameters
# @ bg_size = 128, 256, 512, 1024, 2048
cp.x parameters: -ntg 4 –ndiag 512 –nbgrp N
Where N = 1, 2, 4, 8, 16
Number of OpenMP threads: 8
Number of MPI tasks: 512, 1024, 2048, 4096, 8192
Test 2
A Car-Parrinello molecular dynamics of the CdSe1214 were performed (courtesy by Arrigo Calzolari).
The code shows a good scalability up to 65536 cores.
Kernel: cp.x
Input Dataset Name: CdSe1214
Atomic Species/Number of Atoms: Cd/610, Se/604
Pseudopotential: Ultrasoft-Vanderbilt
XC Functional: PBE
Plane Wave Energy cut-off: 25Rydberg
Execution parameters
# @ bg_size = 256, 1024, 2048
cp.x parameters: -ntg 4 –ndiag D –nbgrp N
Where D = 256, 1024, 512 and N = 4, 4, 16
Number of OpenMP threads: 8
Number of MPI tasks: 1024, 4096, 8192
Virtual Cores | Real Cores | MPI task | OpenMP | Band Groups | Task Groups | Ortho procs | Time / step |
8192 | 4096 | 1024 | 8 | 4 | 4 | 256 | 472 |
32768 | 16384 | 2048 | 8 | 4 | 4 | 1024 | 241 |
65536 | 32768 | 8192 | 8 | 16 | 4 | 512 | 149 |
Test 3
Geometry optimization of a medium size (120 atoms) FeO system using the PW kernel (courtesy of Ivan Girotto).
Kernel: pw.x
Input Dataset Name: FeO120
Atomic Species/Number of Atoms: Fe/120, O/120
Pseudopotential: Ultrasoft-Vanderbilt
XC Functional: PBE
MPI | MPI/node | threads | total cores | npool | ntg | ndiag | WTIME (s) |
512 | 4 | 4 | 2048 | 4 | 4 | 64 | 835 |
512 | 8 | 4 | 2048 | 4 | 2 | 64 | 1006 |
1024 | 16 | 2 | 2048 | 4 | 2 | 100 | 1260 |
256 | 4 | 8 | 2048 | 4 | 1 | 64 | 59280 |
1024 | 4 | 4 | 4096 | 4 | 4 | 256 | 596 |
1024 | 8 | 4 | 4096 | 4 | 8 | 256 | 715 |
1024 | 8 | 4 | 4096 | 4 | 4 | 81 | 753 |
1024 | 4 | 8 | 8192 | 4 | 4 | 256 | 575 |
best cases:
MPI | MPI/node | threads | total cores | npool | ntg | ndiag | WTIME (s) |
512 | 4 | 4 | 2048 | 4 | 4 | 64 | 835 |
1024 | 8 | 4 | 4096 | 4 | 4 | 81 | 735 |
1024 | 4 | 8 | 8192 | 4 | 4 | 256 | 575 |
© Copyright 2