计算化学公社

 找回密码 Forget password
 注册 Register

GROMACS-PLUMED自由能计算性能调优分享

查看数: 3420 | 评论数: 11 | 收藏 Add to favorites 11
关灯 | 提示:支持键盘翻页<-左 右->
    组图打开中,请稍候......
发布时间: 2024-5-13 14:38

正文摘要:

本帖最后由 Dempey 于 2024-5-14 08:55 编辑 GROMACS-PLUMED做metadynamics性能调优分享 前言       目前GROMACS计算自由能的方法有多种,据笔者所知,主流可以做metadynamics的方法其 ...

回复 Reply

fhh2626 发表于 Post on 2025-5-14 17:25:17
neocc 发表于 2025-5-13 10:02
请问gromacs24以上的版本直接在mdp里面写入colvar参数就可以直接使用吗?不需要重新编译吧

不需要重新编译
neocc 发表于 Post on 2025-5-13 10:02:04
fhh2626 发表于 2024-10-24 09:52
Colvars已经收编在Gromacs2024及未来的官方发布版中了,相比Plumed可以做更多的代码级别的优化

请问gromacs24以上的版本直接在mdp里面写入colvar参数就可以直接使用吗?不需要重新编译吧
MilesYYh 发表于 Post on 2025-5-10 21:53:28
补充:
“PLUMED is not compatible with the internal multi-threading implementation of GROMACS so you need to configure gromacs as
cmake -DGMX_THREAD_MPI=OFF and add -DGMX_MPI=ON if you want to use MPI.”
liuxdhs 发表于 Post on 2024-11-10 17:28:45
如果plumed是在cpu下进行计算,那gmx+plumed连用在没有gpu下的性能会表现如何。
plumed没有使用mpi编译,就只能使用-ntmpi=1 -ntomp=n了
Dempey 发表于 Post on 2024-10-25 23:03:25
16aDream 发表于 2024-10-24 10:55
NV热浴竟然也不能用update-gpu, 能简要说明一下吗

我也不知道是什么原理,知识有一次突然发现用NH热浴的update在CPU
  1. PP task will update and constrain coordinates on the CPU
复制代码
就想着试试能不能update gpu,一提交就报错了
  1. Inconsistency in user input:
  2. Update task on the GPU was required,
  3. but the following condition(s) were not satisfied:
  4. Nose-Hoover temperature coupling is not supported.
复制代码

16aDream 发表于 Post on 2024-10-24 10:55:50
本帖最后由 16aDream 于 2024-10-24 11:00 编辑
Dempey 发表于 2024-10-24 08:45
是的,PLUMED部分在CPU计算,而且GROMACS用NV热浴也不能update gpu

NV热浴竟然也不能用update-gpu, 能简要说明一下吗
fhh2626 发表于 Post on 2024-10-24 09:52:52
Colvars已经收编在Gromacs2024及未来的官方发布版中了,相比Plumed可以做更多的代码级别的优化
Dempey 发表于 Post on 2024-10-24 08:45:38
16aDream 发表于 2024-10-23 21:34
updata-gpu模式是与plumed的metad不兼容吗?

是的,PLUMED部分在CPU计算,而且GROMACS用NV热浴也不能update gpu
16aDream 发表于 Post on 2024-10-23 21:34:19
updata-gpu模式是与plumed的metad不兼容吗?
天青色 发表于 Post on 2024-9-3 01:53:49
老师您好,请问一下老师,我计算的下面的这个md.log文件中的日志记录正常吗?
为什么会有两行dE_term呢,分别代表什么呢?体系是一个RNA体系,一共有46万个原子,用的REST2方法做的计算。在300-1000度之间插值20个replicas.

之前算了一个300-1000之间插值5个replicas,算了15ns,没有一个副本交换,是不是我的代码有问题呢?谢谢老师和大家!

————————————
step 1000: timed with pme grid 72 72 72, coulomb cutoff 1.375: 549367.2 M-cycles
Replica exchange at step 1000 time 2.00000
Repl 0 <-> 1  dE_term = -0.000e+00 (kT)
dplumed =  7.619e+02  dE_Term =  7.619e+02 (kT)
Repl ex  0    1    2    3    4    5    6    7    8    9   10   11   12   13   14   15   16   17   18   19
Repl pr   .00       .00       .00       .00       .00       .00       .00       .00       .00       .00

202409030146074821..png (71.91 KB, 下载次数 Times of downloads: 80)

202409030146074821..png
yihanxu 发表于 Post on 2024-7-9 10:24:42
老师您好,metaMD适用于计算什么问题呢?运行MetaMD与运行MD是相互独立的吗?

手机版 Mobile version|北京科音自然科学研究中心 Beijing Kein Research Center for Natural Sciences|京公网安备 11010502035419号|计算化学公社 — 北京科音旗下高水平计算化学交流论坛 ( 京ICP备14038949号-1 )|网站地图

GMT+8, 2025-8-15 09:39 , Processed in 0.749491 second(s), 32 queries , Gzip On.

快速回复 返回顶部 返回列表 Return to list