GPU版本的VASP如何算CINEB呢

ddddnight

GPU版本的VASP只能单线程运行，想用GPU版本的VASP算CINEB，每次只能优化01文件夹里面的POSCAR，优化结束之后就结束了，请问如何进行GPU版本的VASP如何CINEB的计算呢
(base) [tg@tg TS]$ mpirun -np 1 vasp
running    1 mpi-ranks, on    1 nodes
each image running on    1 cores
distrk:  each k-point on    1 cores,    1 groups
distr:  one band on    1 cores,    1 groups
using from now: 01/INCAR  
OpenACC runtime initialized ...    1 GPUs detected
vasp.6.4.2 20Jul23 (build Apr 21 2024 18:57:01) complex                        
POSCAR found type information on POSCAR CuC H Cl
01/POSCAR found :  4 types and     164 ions
scaLAPACK will be used selectively (only on CPU)
-----------------------------------------------------------------------------
|                                                                             |
|               ----> ADVICE to this user running VASP <----                  |
|                                                                             |
|     You have a (more or less) 'large supercell' and for larger cells it     |
|     might be more efficient to use real-space projection operators.         |
|     Therefore, try LREAL= Auto in the INCAR file.                           |
|     Mind: For very accurate calculation, you might also keep the            |
|     reciprocal projection scheme (i.e. LREAL=.FALSE.).                      |
|                                                                             |
-----------------------------------------------------------------------------

LDA part: xc-table for Pade appr. of Perdew
POSCAR found type information on POSCAR CuC H Cl
00/POSCAR found :  4 types and     164 ions
POSCAR found type information on POSCAR CuC H Cl
13/POSCAR found :  4 types and     164 ions
found WAVECAR, reading the header
POSCAR, INCAR and KPOINTS ok, starting setup
FFT: planning ... GRIDC
FFT: planning ... GRID_SOFT
FFT: planning ... GRID
reading WAVECAR
the WAVECAR file was read successfully
initial charge from wavefunction
entering main loop
       N       E                     dE             d eps       ncg     rms          rms(c)
DAV:   1    -0.595332772825E+03   -0.59533E+03   -0.49363E-06  5152   0.238E-02    0.163E-02
RMM:   2    -0.595332773324E+03   -0.49919E-06   -0.15104E-05  4488   0.105E-02    0.182E-02
RMM:   3    -0.595332771557E+03    0.17669E-05   -0.61705E-06  4291   0.672E-03    0.145E-02
RMM:   4    -0.595332771661E+03   -0.10326E-06   -0.95492E-07  4174   0.166E-03
   1 F= -.59533277E+03 E0= -.59528551E+03  d E =-.595333E+03
quench:  g(F)=  0.000E+00 g(S)=  0.000E+00 dE (1.order)= 0.000E+00
reached required accuracy - stopping structural energy minimisation
writing wavefunctions
Warning: ieee_invalid is signaling
Warning: ieee_divide_by_zero is signaling
Warning: ieee_underflow is signaling
Warning: ieee_inexact is signaling
FORTRAN STOP


INCAR的内容：
System = Cu001PVCs

IMAGES=12
ICHAIN=0
POTIM=0
IOPT=1
LCLIMB = .TRUE.
SPRING = -5

PREC = Normal
ENCUT = 400
EDIFF = 1e-5
# ADDGRID= .TRUE.                   (Increase grid； helps GGA convergence)

NELM=100
NELMIN=4

ISMEAR = 1
SIGMA = 0.2
ALGO = FAST
IALGO = 48
ISIF = 2
IBRION = 3
EDIFFG = -0.01
NSW = 128

GGA = PE
ISPIN = 1

NWRITE = 0
LWAVE = .TRUE.
LCHARG = .FALSE.

ADDGRID = .TRUE.

NPAR = 8
# NSIM = 4