计算化学公社»论坛首页 › 理论与计算化学 (Theoretical and Computational Chemistry) › 第一性原理 (First-principles) › 求助：不同编译方式的cp2k-2024.3计算杂化泛函速度和内 ...

求助：不同编译方式的cp2k-2024.3计算杂化泛函速度和内存消耗差异巨大的问题

查看数: 998 | 评论数: 3 | 收藏 Add to favorites 1

关灯 | 提示：支持键盘翻页<-左右->

帖子模式

cxtt

发布时间: 2024-12-7 22:58

正文摘要:

本帖最后由 cxtt 于 2024-12-7 23:00 编辑最近编译了多个版本的cp2k-2024.3，但每个版本都存在一些问题，尝试了很多编译器和mpi的组合仍无法解决，因此发帖求助一下。我编译的版本包含： [1] icc+intelmpi+in ...

回复 Reply

BangbooCat 发表于 Post on 2024-12-8 15:20:55

本帖最后由 BangbooCat 于 2024-12-8 15:23 编辑

我的配置：
i9 13900K 24核32线程
Ubuntu 22.04; GNU 11.4;openmpi-4.1.2；openblas-0.3.27;
CP2K version 2024.3
跑了一下你给的文件，第一步：

HFX_MEM_INFO| Est. max. program size before HFX [MiB]: 591
HFX_MEM_INFO| Number of cart. primitive ERI's calculated: 13033338590
HFX_MEM_INFO| Number of sph. ERI's calculated: 4561925105
HFX_MEM_INFO| Number of sph. ERI's stored in-core: 1647376146
HFX_MEM_INFO| Number of sph. ERI's stored on disk: 0
HFX_MEM_INFO| Number of sph. ERI's calculated on the fly: 0
HFX_MEM_INFO| Total memory consumption ERI's RAM [MiB]: 1523
HFX_MEM_INFO| Whereof max-vals [MiB]: 240
HFX_MEM_INFO| Total compression factor ERI's RAM: 8.25
HFX_MEM_INFO| Total memory consumption ERI's disk [MiB]: 0
HFX_MEM_INFO| Total compression factor ERI's disk: 0.00
HFX_MEM_INFO| Size of density/Fock matrix [MiB]: 16
HFX_MEM_INFO| Size of buffers [MiB]: 6
HFX_MEM_INFO| Number of periodic image cells considered: 27
HFX_MEM_INFO| Est. max. program size after HFX [MiB]: 865
1 OT DIIS 0.15E+00 43.2 0.04039601 -1057.1313965835 -1.06E+03
2 OT DIIS 0.15E+00 8.1 0.03345808 -1077.0228139642 -1.99E+01
3 OT DIIS 0.15E+00 8.6 0.01690949 -1098.2755824071 -2.13E+01
4 OT DIIS 0.15E+00 8.5 0.00978668 -1101.2981899075 -3.02E+00
5 OT DIIS 0.15E+00 8.2 0.00738061 -1102.6056959729 -1.31E+00
6 OT DIIS 0.15E+00 8.6 0.00468891 -1103.3545364870 -7.49E-01
7 OT DIIS 0.15E+00 8.2 0.00341403 -1103.6932368419 -3.39E-01
8 OT DIIS 0.15E+00 8.0 0.00265499 -1103.8433030392 -1.50E-01
9 OT DIIS 0.15E+00 8.1 0.00192820 -1103.9084616064 -6.52E-02
10 OT DIIS 0.15E+00 7.9 0.00114899 -1103.9362898468 -2.78E-02
11 OT DIIS 0.15E+00 8.5 0.00072371 -1103.9487195436 -1.24E-02
12 OT DIIS 0.15E+00 8.5 0.00043947 -1103.9542021360 -5.48E-03
13 OT DIIS 0.15E+00 8.1 0.00029312 -1103.9567641097 -2.56E-03
14 OT DIIS 0.15E+00 7.6 0.00021168 -1103.9577818879 -1.02E-03
15 OT DIIS 0.15E+00 7.9 0.00011461 -1103.9584436210 -6.62E-04
16 OT DIIS 0.15E+00 7.8 0.00008098 -1103.9587865001 -3.43E-04
17 OT DIIS 0.15E+00 8.7 0.00007366 -1103.9589886264 -2.02E-04
18 OT DIIS 0.15E+00 8.5 0.00005943 -1103.9591246385 -1.36E-04
19 OT DIIS 0.15E+00 8.0 0.00004775 -1103.9592543497 -1.30E-04
20 OT DIIS 0.15E+00 8.1 0.00004303 -1103.9593575462 -1.03E-04
Leaving inner SCF loop after reaching 20 steps.
Electronic density on regular grids: -511.9999999975 0.0000000025
Core density on regular grids: 511.9999999832 -0.0000000168
Total charge density on r-space grids: -0.0000000144
Total charge density g-space grids: -0.0000000144
Overlap energy of the core charge distribution: 0.00000227027007
Self energy of the core charge distribution: -2838.21791979120962
Core Hamiltonian energy: 827.21032830864033
Hartree energy: 1179.82649458010951
Exchange-correlation energy: -210.56777151025406
Hartree-Fock Exchange energy: -61.81094940571113
Dispersion energy: -0.39954199805786
Total energy: -1103.95935754621269
outer SCF iter = 1 RMS gradient = 0.43E-04 energy = -1103.9593575462

复制代码

后面稳定在 11-12 s/step，内存每步占用在1G左右

我觉得应该还是你编译过程存在问题

logzzz 发表于 Post on 2024-12-8 14:12:52

你可以试试不用openmpi,用mpich，我之前用openmpi也发现过类似的，但改成mpich就好了。

sobereva 发表于 Post on 2024-12-8 03:52:48

严格按本文说的软件环境和步骤安装，RockyLinux 9自带的gcc结合CP2K toolchain自动装的OpenMPI（或者结合自己装的OpenMPI 4.1.x），从8.1到目前2024最新版，什么问题都没有
CP2K第一性原理程序在Linux中的安装方法
http://sobereva.com/586（http://bbs.keinsci.com/thread-21608-1-1.html）

		自动登录 Automatic login	找回密码 Forget password
密码 Password			注册 Register