<p style="border:0; padding:0; margin:0; font-family:'Gulim'; font-size:10pt; cursor: text;"><p style="font-family: Gulim; font-size: 10pt; border: 0px; padding: 0px; margin: 0px; cursor: text;">Hi Mark, </p><p style="font-family: Gulim; font-size: 10pt; border: 0px; padding: 0px; margin: 0px; cursor: text;"><br></p><p style="font-family: Gulim; font-size: 10pt; border: 0px; padding: 0px; margin: 0px; cursor: text;">Thanks for your reply email.</p><div style="font-family: 돋움, arial; font-size: 12px; margin-top: 30px; margin-left: 0.8em; color: rgb(0, 102, 204); font-weight: bold;">--------- 원본 메일 ---------</div><blockquote style="font-family: Gulim; font-size: 12px; border-left-style: solid; border-left-width: 2px; margin: 0pt 0pt 0pt 0.8em; padding-left: 1em;"><div style="font-family:arial,돋움;line-height:1.5"><b>보낸사람</b> : Mark Abraham <mark.j.abraham@gmail.com><br><b>받는사람</b> : Discussion list for GROMACS us <gmx-users@gromacs.org><br><b>받은날짜</b> : 2014년 9월 18일(목) 21:26:48<br><b>제목</b> : Re: [gmx-users] Problem with: Shell particles are not implemented with domain de<br><!-- original content --><div style="margin-top:5px;"><pre>On Thu, Sep 18, 2014 at 2:01 PM, Kester Wong <kester2014@ibs.re.kr> wrote:
> Dear gromacs users,
>
>
> Has anyone experienced a problem with running polarisable water model
> SWM4-NDP with the following warning: Shell particles are not implemented
> with domain decomposition?
>
That means what it says - you can't have more than one domain, which means
you can't use more than one MPI rank. Hopefully OpenMP will work to let you
use more than one core, but such code paths got very little love in the
years leading up to 5.0
> The md.log also stated the following: Number of hardware threads detected
> (12) does not match the number reported by OpenMP (1). I don't think this
> is the cause, as this message was also found in my other "working"
> calculations.
>
That suggests something is broken somewhere, but you're using prehistoric
compilers, so you're clearly not after high performance in the first place
;-)
Mark</kester2014@ibs.re.kr></pre></div></div></blockquote><font face="monospace"><span style="font-size: 12px; line-height: 18px; white-space: pre;"><br>Suppose I replace the compiler with a more recent one, e.g. 4.9.1 or 4.7.4, will it enable the calculation on one MPI rank at least?</span><br>Could you please advise if there is any other aspect of tweaks that I can do to improve my clusters' performance?<br>Since the Drude/shell system would only work on one domain, would it be wise to set the OpenMP to lesser than 16?<br>I thought the higher number of OpenMP thread used (say for example -ntomp >16), the more time we will lose for it becomes less efficient?<br><br><br></font><span style="font-family: monospace; font-size: 12px; line-height: 18px; white-space: pre;">Cheers,</span><br style="font-family: monospace;"><span style="font-family: monospace; font-size: 12px; line-height: 18px; white-space: pre;">Kester</span><font face="monospace"><span style="font-size: 12px; line-height: 18px; white-space: pre;"><br></span></font><blockquote style="font-family: Gulim; font-size: 12px; border-left-style: solid; border-left-width: 2px; margin: 0pt 0pt 0pt 0.8em; padding-left: 1em;"><div style="font-family:arial,돋움;line-height:1.5"><div style="margin-top:5px;"><pre><kester2014@ibs.re.kr></kester2014@ibs.re.kr></pre><pre><kester2014@ibs.re.kr></kester2014@ibs.re.kr></pre><pre><br></pre><pre><kester2014@ibs.re.kr>
I have tried using OpenMP and also tried a variety of -ntmpi and -ntomp
> settings.
>
> The same calculation did not work in GROMACS versions 5.0 and 5.0.1, in
> the following cluster:
>
>
> Gromacs version: VERSION 5.0.1
>
> Precision: double
>
> Memory model: 64 bit
>
> MPI library: MPI
>
> OpenMP support: enabled
>
> GPU support: disabled
>
> invsqrt routine: gmx_software_invsqrt(x)
>
> SIMD instructions: NONE
>
> FFT library: fftw-3.3.3
>
> RDTSCP usage: disabled
>
> C++11 compilation: disabled
>
> TNG support: enabled
>
> Tracing support: disabled
>
> Built on: Thu Sep 18 20:14:41 KST 2014
>
> Built by: root@master.hpc [CMAKE]
>
> Build OS/arch: Linux 2.6.18-274.7.1.el5 x86_64
>
> Build CPU vendor: GenuineIntel
>
> Build CPU brand: Intel(R) Xeon(R) CPU X3220 @ 2.40GHz
>
> Build CPU family: 6 Model: 15 Stepping: 11
>
> Build CPU features: apic clfsh cmov cx8 cx16 lahf_lm mmx msr pdcm pse sse2
> sse3 ssse3
>
> C compiler: /usr/bin/cc GNU 4.1.2
>
> C compiler flags: -Wextra -Wno-missing-field-initializers
> -Wno-sign-compare -Wpointer-arith -Wall -Wno-unused -Wunused-value
> -Wunused-parameter -fomit-frame-pointer -funroll-all-loops -O3 -DNDEBUG
>
> C++ compiler: /usr/bin/c++ GNU 4.1.2
>
> C++ compiler flags: -Wextra -Wno-missing-field-initializers
> -Wpointer-arith -Wall -Wno-unused-function -fomit-frame-pointer
> -funroll-all-loops -O3 -DNDEBUG
>
> Boost version: 1.55.0 (internal)
>
> Using 24 MPI processes
> Using 1 OpenMP thread per MPI process
>
>
> However, the same input files worked in another cluster (albeit very
> slowly, ~0.3ns/day).
>
>
> Gromacs version: VERSION 5.0
>
> Precision: single
>
> Memory model: 64 bit
>
> MPI library: MPI
>
> OpenMP support: enabled
>
> GPU support: disabled
>
> invsqrt routine: gmx_software_invsqrt(x)
>
> SIMD instructions: AVX_256
>
> FFT library: fftpack (built-in)
>
> RDTSCP usage: enabled
>
> C++11 compilation: disabled
>
> TNG support: enabled
>
> Tracing support: disabled
>
> Built on: Thu Aug 28 16:44:08 KST 2014
>
> Built by: root@kant [CMAKE]
>
> Build OS/arch: Linux 2.6.32-431.23.3.el6.x86_64 x86_64
>
> Build CPU vendor: GenuineIntel
>
> Build CPU brand: Intel(R) Core(TM) i5-4670 CPU @ 3.40GHz
>
> Build CPU family: 6 Model: 60 Stepping: 3
>
> Build CPU features: aes apic avx avx2 clfsh cmov cx8 cx16 f16c fma htt
> lahf_lm mmx msr nonstop_tsc pcid pclmuldq pdcm pdpe1gb popcnt pse rdrnd
> rdtscp sse2 sse3 sse4.1 sse4.2 ssse3 tdt x2apic
>
> C compiler: /usr/bin/gcc GNU 4.4.7
>
> C compiler flags: -mavx -Wno-maybe-uninitialized -Wextra
> -Wno-missing-field-initializers -Wno-sign-compare -Wpointer-arith -Wall
> -Wno-unused -Wunused-value -Wunused-parameter -fomit-frame-pointer
> -funroll-all-loops -Wno-array-bounds -O3 -DNDEBUG
>
> C++ compiler: /usr/bin/g++ GNU 4.4.7
>
> C++ compiler flags: -mavx -Wextra -Wno-missing-field-initializers
> -Wpointer-arith -Wall -Wno-unused-function -fomit-frame-pointer
> -funroll-all-loops -Wno-array-bounds -O3 -DNDEBUG
>
> Boost version: 1.55.0 (internal)
>
>
>
> Using 1 MPI process
>
> Using 16 OpenMP threads
>
> Regards,
> Kester
>
>
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request@gromacs.org.
>
>
--
Gromacs Users mailing list
* Please search the archive at http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before posting!
* Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
* For (un)subscribe requests visit
https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or send a mail to gmx-users-request@gromacs.org.
</kester2014@ibs.re.kr></pre></div><!-- original content --><br></div></blockquote></p>
<img src='http://mail.ibs.re.kr/checkread/NzYxNjg3/Z214LXVzZXJzQGdyb21hY3Mub3Jn/' width='1px' height='1px' />