Cluster的使用

来源：互联网发布：里约中国男篮数据编辑：程序博客网时间：2024/05/17 04:36

Interactive Job on Comet

参考：http://www.sdsc.edu/support/user_guides/gordon.html，在http://www.sdsc.edu/support/user_guides/comet.html上没有.
Comet机器上compute队列的CPU配置是：
Processor Type Intel Xeon E5-2680v3
Sockets 2
Cores/socket 12
qsub -I -q shared -l nodes=1:ppn=16:native:noflash,walltime=00:30:00

Above may be a misslead command.

The follow are correct.

srun –partition=shared –nodes=1 –ntasks-per-node=6 -t 00:30:00 –pty –wait=0 /bin/bash
srun –partition=gpu-shared –nodes=1 –ntasks-per-node=6 –gres=gpu:1 -t 00:30:00 –pty –wait=0 /bin/bash

Interactive Job on Stampede

Would you like to know how much memory your code uses during execution? Try TACC’s remora tool:
moduleloadremora module help remora

Stampede

idev –help

Bridges

Bridges Regular Memory
interact -p GPU-shared -N 1 –gres=gpu:1

BlueWaters

2 CPUs with 8 Cores per Node, submit application for 2 nodes with 32 cores in XE Queue.
For XE node specification: #PBS -l nodes=1024:ppn=32:xe
For XK node specification: #PBS -l nodes=1024:ppn=16:xk
For both XE and XK: #PBS -l nodes=1024:ppn=32:xe+1024:ppn=16:xk
For XE/XK-non-specific (X-feature) node specification: #PBS -l nodes=1024:ppn=16:x
For XE large memory nodes: #PBS -l feature=xehimem
For XK large memory nodes: #PBS -l feature=xkhimem

qsub -I -l nodes=2:ppn=32:xe -l walltime=01:00:00

0 0