在xshell中使用FDS做MPI并行运算问题

[lizijiang]

在FDS用户手册中提出，运用下面的shell进行MPI并行运算。

代码： 全选

#!/bin/bash
#PBS -N job_name
#PBS -e <pwd>/job_name.err
#PBS -o <pwd>/job_name.log
#PBS -o <pwd>/job_name.log
#PBS -o <pwd>/job_name.log
export OMP_NUM_THREADS=2
export I_MPI_PIN_DOMAIN=omp
cd <pwd>
mpiexec -n 8 <full_path>/fds job_name.fds

我根据我需要的节点数写了这个shell，但运行的时候报错。

代码： 全选

#!/bin/bash
#PBS -N detailed_MPI22
#PBS -e /THL6/home/openfoam/FDS/FDSmodel/single_building/detailed_MPI22/detailed_MPI22.err
#PBS -o /THL6/home/openfoam/FDS/FDSmodel/single_building/detailed_MPI22/detailed_MPI22.log
#PBS -l nodes=11:ppn=2
#PBS -l walltime=24:0:0
export OPM_NUM_THREADS=8
export I_MPI_PIN_DOMAIN=omp
cd /openfoam/FDS/FDSmodel/single_building/detailed_MPI22
mpiexec -n 22 /THL6/home/openfoam/FDS/FDSmodel/single_building/detailed_MPI22/fds detailed_MPI22.fds

运算以后的报错
[openfoam@th-hpc1-ln0 ~]$ mpiexec -np 22 FDS_MPI.pbs
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)
[proxy:0:0@th-hpc1-ln0] HYD_spawn (../../../../../src/pm/i_hydra/libhydra/spawn/intel/hydra_spawn.c:128): execvp error on file FDS_MPI.pbs (No such file or directory)

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 14 PID 6040 RUNNING AT th-hpc1-ln0
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 20 PID 6046 RUNNING AT th-hpc1-ln0
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 21 PID 6047 RUNNING AT th-hpc1-ln0
= KILLED BY SIGNAL: 9 (Killed)
===================================================================================

我是远程连接超算的，请问有没有人知道我该如何修改或者应该如何在Xshell里份MPI数，才能对FDS进行并行计算。
本人是linux小白，完全不会shell这些命令，但是无奈运行fds要用到，请大家多多指导！

[astolia]

既然你要用linux系统，要写shell脚本，就稍微学一下shell基础吧。如果有其它编程语言基础的话学起来很快的，挤个半天时间出来都够了。

你执行的命令是 mpiexec -np 22 FDS_MPI.pbs，其中FDS_MPI.pbs是一个文件路径。在linux系统中，凡是不以/开头的文件路径都是相对路径，也就是相对于当前所在目录的路径。而根据你[openfoam@th-hpc1-ln0 ~]的提示，当前路径在当前用户的家目录~，也就是说，你是让mpiexec去处理当前用户的家目录下的FDS_MPI.pbs文件。但这个文件不存在，所以报错No such file or directory

弄清楚你把FDS_MPI.pbs放到哪里了，把路径写正确。另外，由于mpiexec是去执行FDS_MPI.pbs，也别忘了给FDS_MPI.pbs加上可执行权限

[lizijiang]

既然你要用linux系统，要写shell脚本，就稍微学一下shell基础吧。如果有其它编程语言基础的话学起来很快的，挤个半天时间出来都够了。

你执行的命令是 mpiexec -np 22 FDS_MPI.pbs，其中FDS_MPI.pbs是一个文件路径。在linux系统中，凡是不以/开头的文件路径都是相对路径，也就是相对于当前所在目录的路径。而根据你[openfoam@th-hpc1-ln0 ~]的提示，当前路径在当前用户的家目录~，也就是说，你是让mpiexec去处理当前用户的家目录下的FDS_MPI.pbs文件。但这个文件不存在，所以报错No such file or directory

弄清楚你把FDS_MPI.pbs放到哪里了，把路径写正确。另外，由于mpiexec是去执行FDS_MPI.pbs，也别忘了给FDS_MPI.pbs加上可执行权限

非常感谢，我修改看看

[lizijiang]

通过大家的帮助，我将shell文件修改为
#!/bin/bash
#PBS -N detailed_MPI22
#PBS -e /THL6/home/openfoam/FDS/FDSmodel/single_building/detailed_MPI22/detailed_MPI22.err
#PBS -o /THL6/home/openfoam/FDS/FDSmodel/single_building/detailed_MPI22/detailed_MPI22.log
#PBS -l nodes=6:ppn=2
#PBS -l walltime=24:0:0
export OPM_NUM_THREADS=8
export I_MPI_PIN_DOMAIN=omp
cd /THL6/home/openfoam/FDS/FDSmodel/single_building/detailed_MPI22
mpiexec -n 12 fds detailed_MPI22.fds

然后给予该shell文件权限，并且进行运行
chmod +x ./FDS_MPI.pbs
mpirun -np 12 FDS_MPI.pbs
成功运行了并行运算的FDS文件

注：pbs文件保存在了/THL6/home/openfoam，需要运行的fds保存在了/THL6/home/openfoam/FDS/FDSmodel/single_building/detailed_MPI22
希望可以帮助到后面运用fds进行mpi并行的小白，user guide上写的实在太简单