| 查看: 825 | 回复: 8 | |||
| 当前主题已经存档。 | |||
| 【有奖交流】积极回复本帖子，参与交流，就有机会分得作者 dxp5380 的 3 个金币 | |||
| 当前只显示满足指定条件的回帖，点击这里查看本话题的所有回帖 | |||
dxp5380金虫 (著名写手)
|
[交流]
【求助】计算时出现的问题
|
||
|
用ABINIT 计算时出现一下问题，请虫友们帮忙找一下原因，该如何解决。初次接触ABINIT，望大家不吝赐教！谢谢！ -P-0000 kpgsph : BUG - -P-0000 The variables ikg, mkmem, and mpw must satisfy ikg<=(mkmem-1)*mpw, -P-0000 while the arguments of the routine are -P-0000 ikg = 291671, mkmem = 18, mpw = 16959 -P-0000 Probable cause: Known error in invars1 for parallel spin-polarized case. -P-0000 Temporary solution: Change the number of parallel processes. -P-0000 -P-0000 leave_new : decision taken to exit ... -P-0000 leave_new : synchronization done... -P-0000 leave_new : exiting... |
回复此楼» 猜你喜欢
269求调剂
已经有6人回复
-
297求调剂
已经有20人回复
-
297求调剂
已经有28人回复
-
267求调剂
已经有5人回复
-
考研调剂
已经有6人回复
-
314求调剂
已经有9人回复
-
285求调剂
已经有7人回复
-
一志愿矿大，材料工程专硕314分，0856可调都可以
已经有10人回复
-
还有化工二轮调剂的学校吗
已经有31人回复
-
一志愿211 0703化学 346分求调剂
已经有15人回复
netx_ray
木虫 (小有名气)
- 应助: 1 (幼儿园)
- 金币: 2358.1
- 红花: 2
- 帖子: 146
- 在线: 140小时
- 虫号: 133380
- 注册: 2005-12-14
- 性别: GG
- 专业: 凝聚态物性I:结构、力学和
★ ★ ★ ★ ★
zxzj05(金币+2,VIP+0):3Q! 鼓励讨论！ 10-13 13:53
dxp5380(金币+3,VIP+0):不太明白！谢谢你！ 10-13 14:33
zxzj05(金币+2,VIP+0):3Q! 鼓励讨论！ 10-13 13:53
dxp5380(金币+3,VIP+0):不太明白！谢谢你！ 10-13 14:33
|
ABINIT的maillist有人提到过这个问题，是getcell 和getxred 引起的 Dear Colleagues, I have identified a bug in kpgsph.F90 (or else, a bug in how it's called) which arises when getcell 1 is used to initiate a calculation given the optimized cell from a previous run. The problem arises in the calculation of the number of planewaves npw. At about line 390 is the line npw=ig, where ig was calculated in the above loops. The variable npw is then used later in several places involving the array kg_small, which was dimensioned (3,mpw*np_band). The problem is that in the getcell 1 case, it occurs occasionally to have ig and hence npw exceed mpw, in which case the program crashes due to array out of bounds. In the nested loops where ig is computed, there are multiple places where the array limits are checked for (see line 273 for example: if (ig <= mpw*np_band) then...). However, the value of ig itself is not controlled in this way, so later when npw=ig is set, it can be that npw is too big for the array. I have implemented the following work-around (wouldn't call it a bug fix because it seems too dirty): npw=ig is replaced by if (mpw*np_band > 0 .and. ig > mpw*np_band) then npw=mpw else npw=ig endif this works but I'm not sure it's quite the proper thing to do. To be honest I don't really understand the calculation of ig leading up to this point so there might be unintended consequences of this assignment. Please note that this is not a failure to use dilatmx in the second calculation, the code requires you in the getcell = 1 case to use a big enough dilatmx to account for the changes to acell that occured in the optimization. So in other words I use dilatmx in both the optimization and subsequent calculations using getcell 1. I would appreciate comments and clarifications. |
9楼2009-10-13 11:02:44
casjxm
铜虫 (正式写手)
- 应助: 19 (小学生)
- 金币: 737.1
- 散金: 20
- 帖子: 443
- 在线: 356.1小时
- 虫号: 445122
- 注册: 2007-10-27
- 性别: GG
- 专业: 凝聚态物性 II ：电子结构
2楼2009-08-29 14:52:13
dxp5380
金虫 (著名写手)
- 应助: 0 (幼儿园)
- 金币: 5577.7
- 散金: 15
- 帖子: 2329
- 在线: 81.2小时
- 虫号: 632008
- 注册: 2008-10-21
- 专业: 凝聚态物性I:结构、力学和
★ ★
fegg7502(金币+2,VIP+0):鼓励交流2 8-30 17:06
fegg7502(金币+2,VIP+0):鼓励交流2 8-30 17:06
|
我感觉不是cpu数量的原因。我连续优化过， ndtset 3 getcell -1 getxred -1 ngkpt1 8 8 8 ngkpt2 10 10 10 ngkpt3 12 12 12 一切正常。但在下面的计算中出现了问题： ndtset 4 getcell -1 getxred -1 strtarget1 strtarget2 strtarget3 strtarget4 (后面具体的数就不列了） 只算了一步，就出现了上面的问题。请问：在这种情况下的优化参数如何设？如optcell，ionmov，iscf等。另外还需要那些参数？谢谢！ |
4楼2009-08-29 16:19:23
zxzj05
荣誉版主 (著名写手)
- 1ST强帖: 1
- 应助: 0 (幼儿园)
- 贵宾: 1.242
- 金币: 5912.5
- 红花: 2
- 帖子: 1528
- 在线: 144.5小时
- 虫号: 406147
- 注册: 2007-06-18
- 性别: GG
- 专业: 金属功能材料
- 管辖: 第一性原理
6楼2009-09-16 21:34:19
100