| 查看: 2959 | 回复: 3 | ||
[求助]
“天河二号”上计算cl-neb出错 已有1人参与
|
|
我在“天河二号”上计算cl-neb，刚开始我算的一个任务IMAGES设的是1，可以顺利算完，但我把IMAGES改成5后算两步不到一分钟就出错退出，我想是否跟我并行的核数有关，之前看小木虫上有前辈说到计算的cpu核心数需要是IMAGES的整数倍，（“天河二号”上一个节点是24核）于是我改成一个节点、并行20核，还有2个节点、并行40核，结果还是一样的出错退出计算，以下是slurm-174172.out里面的内容，投的几次都是一样的出错信息，求大神帮忙支招。 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 M_divide: can not subdivide 24 nodes by 5 Fatal error in MPI_Topo_test: Invalid communicator, error stack: MPI_Topo_test(125): MPI_Topo_test(MPI_COMM_NULL, topo_type=0x7fff3b557f04) failed MPI_Topo_test(76).: Null communicator Fatal error in MPI_Topo_test: Invalid communicator, error stack: MPI_Topo_test(125): MPI_Topo_test(MPI_COMM_NULL, topo_type=0x7fffc1f20c84) failed MPI_Topo_test(76).: Null communicator Fatal error in MPI_Topo_test: Invalid communicator, error stack: MPI_Topo_test(125): MPI_Topo_test(MPI_COMM_NULL, topo_type=0x7fff7ae64b04) failed MPI_Topo_test(76).: Null communicator Fatal error in MPI_Topo_test: Invalid communicator, error stack: MPI_Topo_test(125): MPI_Topo_test(MPI_COMM_NULL, topo_type=0x7fffe7806a04) failed MPI_Topo_test(76).: Null communicator yhrun: error: cn7699: tasks 20-23: Exited with exit code 1 yhrun: First task exited 60s ago yhrun: tasks 0-19: running yhrun: tasks 20-23: exited abnormally yhrun: Terminating job step 174172.0 slurmd[cn7699]: *** STEP 174172.0 KILLED AT 2016-04-01T20:45:58 WITH SIGNAL 9 *** yhrun: Job step aborted: Waiting up to 2 seconds for job step to finish. slurmd[cn7699]: *** STEP 174172.0 KILLED AT 2016-04-01T20:45:58 WITH SIGNAL 9 *** yhrun: error: cn7699: tasks 0-19: Killed |
回复此楼» 猜你喜欢
这样的filecode谁见过
已经有5人回复
-
国基金的申报应该改成非等额制，评价高的钱多评价低的钱少，但是增加资助率
已经有4人回复
-
关于代码变化问题，想知道的进来
已经有13人回复
-
售SCI-T0P文章，我:8O.5.5.1.O.54,科目齐全,可+急
已经有4人回复
-
售SCI-T0P文章，我:8O.5.5.1.O.54,科目齐全,可+急
已经有6人回复
-
fileCode有新解读？
已经有16人回复
-
售SCI一区文章，我:8.O.55.1.O.54,科目齐全,可伽急
已经有5人回复
-
售SCI一区T0P文章，我:8.O.55.1.O54,科目全,可伽急
已经有4人回复
-
售SCI-T0P文章，我:8O.5.5.1.O.54,科目齐全,可+急
已经有4人回复
-
售SCI一区文章，我:8.O.55.1.O.54,科目齐全,可伽急
已经有3人回复
andavid2007
新虫 (正式写手)
- 应助: 1 (幼儿园)
- 金币: 7464
- 红花: 4
- 帖子: 598
- 在线: 141.1小时
- 虫号: 3982645
- 注册: 2015-07-21
- 专业: 理论和计算化学
2楼2016-04-01 22:45:05
3楼2016-04-02 12:27:58
冰凌sunshine
金虫 (小有名气)
- 应助: 1 (幼儿园)
- 金币: 1386.5
- 帖子: 86
- 在线: 187.7小时
- 虫号: 2840540
- 注册: 2013-12-01
- 性别: MM
- 专业: 凝聚态物性 II ：电子结构
4楼2018-01-17 11:49:01
100