From: Chitrak Gupta (chgupta_at_mix.wvu.edu)
Date: Thu Jul 20 2017 - 11:29:37 CDT
Hi Giacomo,
The reason I am not using the pre-compiled versions is because I need to
run Gaussian accelerated MD which as far as I know if only available in the
latest nightly build. Are you aware whether any of the pre-compiled
versions on Stampede 2 can run GaMD?
On Thu, Jul 20, 2017 at 12:20 PM, Giacomo Fiorin <giacomo.fiorin_at_gmail.com>
wrote:
> Intel vs GNU may improve the single-core performance by a few percent (and
> parallel performance by the same prefactor), but not so much the parallel
> scaling.
>
> If you're not sure about the differences between these options, why don't
> you try a precompiled version?
>
> http://www.ks.uiuc.edu/Research/namd/wiki/index.cgi?NamdAtTexas
>
> In the paths shown in there the original version of NAMD is very old, but
> in the same home folder (which is Jim Phillips') you will find up to date
> versions as well.
>
> Giacomo
>
>
>
> On Thu, Jul 20, 2017 at 12:09 PM, Chitrak Gupta <chgupta_at_mix.wvu.edu>
> wrote:
>
>> Hi Giacomo,
>>
>> Thanks for the response.. No, I don't need to use the SMP version, just
>> MPI is fine for me. I had originally tried it like this
>>
>> env MPICXX=mpicxx ./build charm++ mpi-linux-x86_64 --with-production
>>
>> That worked but the scaling was extremely poor. That's when I felt maybe
>> using iccstatic would be better. Do you have any inputs for this?
>>
>>
>> Chitrak.
>>
>> On Wed, Jul 19, 2017 at 9:14 PM, Giacomo Fiorin <giacomo.fiorin_at_gmail.com
>> > wrote:
>>
>>> Hi Chitrak, I would specify manually only one of the compilers, or
>>> neither one if possible. The architecture mpi-linux-x86_64 should already
>>> take care of calling mpicxx, and you should load the relevant environment
>>> module set the MPI library and wrapper to the Intel version.
>>>
>>> Also, check this from the NAMD release notes:
>>>
>>> *MPI-based SMP builds have worse performance than verbs or ibverbs and
>>> are not recommended, particularly for GPU-accelerated builds.*
>>>
>>> So, stick to plain MPI if you must use SMP (for example because you want
>>> to communicate more efficiently with the GPU), or otherwise skip MPI use
>>> the IBverbs option for Charm++.
>>>
>>> Giacomo
>>>
>>> On Wed, Jul 19, 2017 at 6:47 PM, Chitrak Gupta <chgupta_at_mix.wvu.edu>
>>> wrote:
>>>
>>>> Hi NAMD users
>>>>
>>>> I am trying to compile the latest nightly build version of NAMD on
>>>> Stampede 2. For building charm 6.8 beta2 I used the following command that
>>>> was recommended by the interactive session
>>>>
>>>> ./build charm++ mpi-linux-x86_64 iccstatic mpicxx smp -j16
>>>> --with-production
>>>>
>>>>
>>>> However, I am getting the following error:
>>>>
>>>> Error> Tried to specify two compilers: iccstatic and mpicxx
>>>>
>>>>
>>>>
>>>> How should I fix this?
>>>>
>>>>
>>>> Best regards,
>>>> Chitrak.
>>>>
>>>
>>>
>>>
>>> --
>>> Giacomo Fiorin
>>> Associate Professor of Research, Temple University, Philadelphia, PA
>>> Contractor, National Institutes of Health, Bethesda, MD
>>> http://goo.gl/Q3TBQU
>>> https://github.com/giacomofiorin
>>>
>>
>>
>
>
> --
> Giacomo Fiorin
> Associate Professor of Research, Temple University, Philadelphia, PA
> Contractor, National Institutes of Health, Bethesda, MD
> http://goo.gl/Q3TBQU
> https://github.com/giacomofiorin
>
This archive was generated by hypermail 2.1.6 : Sun Dec 31 2017 - 23:21:29 CST