[Pw_forum] Looking for testers (again) for QE-5.1 binaries for windows

Axel Kohlmeyer akohlmey at gmail.com
Fri Sep 5 00:15:15 CEST 2014


On Thu, Sep 4, 2014 at 10:34 PM, ehsan targholi <targholi at gmail.com> wrote:
> I'm confused. Your order is that espresso must be implemented in virtualbox?

no. i don't say this anywhere. what i say is that executables compiled
with the intel compiler and linked to the intel MKL are currently
faster on x86 machines than executables compiled with the GNU
compiler. how much faster depends on a few factors. but the matter of
the fact is that the windows executables are compiled with the GNU
compilers. i do all my work on Linux and since there are the mingw
cross-compilers i can not only compile native linux programs but also
executables for windows. i am doing this for other projects as well.
this is a *much* easier way to compile executables for windows for a
package like QE. building native executables with a native compiler
(cygwin is a bit of a hybrid) would have been much more effort.

> I have worked with this software. and the loss of efficiency is very high.

then you did something wrong.

with the proper settings and reasonable hardware, the cost of the
virtual machine is only a few percent performance.
i am regularly running linux on linux virtual machines (to build RPM
packages for different linux distributions) and they work just fine
and efficient.

> Efficiency is very important in my work. In any case, I am willing to

then running on windows is probably not the best choice. like i said,
the main purpose for providing windows binaries is to give people the
opportunity to run QE calculations where it is not easy or possible to
use linux (e.g. when teaching a class in a computer lab) or to help
people to make tests before they run calculations on a more capable
machine. windows is convenient for a lot of things, but it is not as
well suited for high-performance as linux.

axel.

> cooperate with you to test its performance.
>
>
>
> best regards
>
>
> On Thu, Sep 4, 2014 at 11:19 PM, Axel Kohlmeyer <akohlmey at gmail.com> wrote:
>>
>> On Thu, Sep 4, 2014 at 8:19 PM, siddheshwar chopra <sidhusai at gmail.com>
>> wrote:
>> > Dear Sir,
>> > I  willing to run QE on my Windows xp (with 2 GB RAM and dual core 2.7
>> > GHz
>> > cpu) and windows 7 (with 2GB  RAM and i3 2.2 GHz ) machines..
>> > Sir I would be glad if you could clarify one doubt.. Is will there be
>> > any
>> > DEGRADATION of QE performance on windows, as compared to linux?
>> > I have used cygwin and virtualbox and found severe performance
>> > degradations.
>>
>> the executables are compiled with a mingw cross-compiler on a linux
>> machine. so the performance should be equivalent to what you achieve
>> with gfortran and gcc on a linux box. compared to cygwin it is likely
>> a bit faster. at least that is what people found with a different
>> package.
>>
>>  there are a few things where windows is better than linux and some
>> other things where it is worse. it depends a lot on the specific
>> windows version. with numerical codes like QE, often the compiler and
>> optimized BLAS/LAPACK libraries is the biggest impact. with virtual
>> machines, the problem is often the limited amount of RAM. if you care
>> about performance, obtain an academic license for the intel compilers
>> with MKL and run on a linux machine. it is well worth the money (often
>> cheaper than the equivalent CPU upgrade)
>>
>> the primary motivation to provide windows binaries is to make QE more
>> accessible for classrooms and tutorials and for people that want to
>> run tests on their own machine (not everybody runs linux on their
>> laptops) before sending larger versions of the same calculation to a
>> cluster or a supercomputer. so the best performance is not the highest
>> priority.
>>
>> if you are able to make properly documented comparative benchmarks,
>> i'd be happy to get to know pass that information on to other users in
>> the documentation of the windows port.
>>
>>
>>
>>
>> > I would eagerly wait for your mail regarding the download links.
>> >
>> > Kind Regards,
>> >
>> > On 04-Sep-2014 11:27 pm, "Axel Kohlmeyer" <akohlmey at gmail.com> wrote:
>> >>
>> >> On Thu, Sep 4, 2014 at 7:52 PM, ehsan targholi <targholi at gmail.com>
>> >> wrote:
>> >> > Hi
>> >> > Dear axel
>> >> > It's a very interesting idea. I'm interested to running it in Windows
>> >> > and in
>> >> > parallel version. What should I do؟
>> >>
>> >> just wait for my e-mail telling you where to find the temporary
>> >> repository and the installer packages.
>> >>
>> >> what kind/version of windows are you running and how much RAM and how
>> >> many CPU cores do you have?
>> >>
>> >> >
>> >> >
>> >> > best regards
>> >> > E.Targholi
>> >> >
>> >> >
>> >> >
>> >> > On Thu, Sep 4, 2014 at 9:14 PM, Axel Kohlmeyer <akohlmey at gmail.com>
>> >> > wrote:
>> >> >>
>> >> >> hi everybody,
>> >> >>
>> >> >> following up on my previous request last fall, i am looking again
>> >> >> for
>> >> >> folks that are interested to run quantum espresso on windows?
>> >> >>
>> >> >> last fall some of you helped to improve the development version of
>> >> >> QE
>> >> >> so it can be compiled and run on windows machine (in serial and
>> >> >> parallel). now i finally found the time to update this to the recent
>> >> >> release 5.1 and built a new set of packages. before we unleash those
>> >> >> to the general audience, and particularly inexperienced QE noobies,
>> >> >> i
>> >> >> want to make certain that everything works reasonably well.
>> >> >>
>> >> >> i only have very limited access to windows (on a virtual machine)
>> >> >> and
>> >> >> practice with it (the compilation is done with a cross compiler on
>> >> >> linux) and thus cannot test very thoroughly (and don't have the
>> >> >> time),
>> >> >> so it seem just logical to farm this out to the QE community.
>> >> >>
>> >> >> if you are still interested, please contact me off-list with a short
>> >> >> description of what kind of setup you are willing to test on (CPU,
>> >> >> windows version, 32-bit or 64-bit, MPI or serial or both). i will
>> >> >> then
>> >> >> point you to a webpage with some instructions and a link to a
>> >> >> repository where you can download the new packages as soon as they
>> >> >> are
>> >> >> available.
>> >> >>
>> >> >> thanks,
>> >> >>     axel.
>> >> >>
>> >> >> --
>> >> >> Dr. Axel Kohlmeyer  akohlmey at gmail.com  http://goo.gl/1wk0
>> >> >> College of Science & Technology, Temple University, Philadelphia PA,
>> >> >> USA
>> >> >> International Centre for Theoretical Physics, Trieste. Italy.
>> >> >> _______________________________________________
>> >> >> Pw_forum mailing list
>> >> >> Pw_forum at pwscf.org
>> >> >> http://pwscf.org/mailman/listinfo/pw_forum
>> >> >
>> >> >
>> >> >
>> >> > _______________________________________________
>> >> > Pw_forum mailing list
>> >> > Pw_forum at pwscf.org
>> >> > http://pwscf.org/mailman/listinfo/pw_forum
>> >>
>> >>
>> >>
>> >> --
>> >> Dr. Axel Kohlmeyer  akohlmey at gmail.com  http://goo.gl/1wk0
>> >> College of Science & Technology, Temple University, Philadelphia PA,
>> >> USA
>> >> International Centre for Theoretical Physics, Trieste. Italy.
>> >>
>> >> _______________________________________________
>> >> Pw_forum mailing list
>> >> Pw_forum at pwscf.org
>> >> http://pwscf.org/mailman/listinfo/pw_forum
>> >
>> >
>> > _______________________________________________
>> > Pw_forum mailing list
>> > Pw_forum at pwscf.org
>> > http://pwscf.org/mailman/listinfo/pw_forum
>>
>>
>>
>> --
>> Dr. Axel Kohlmeyer  akohlmey at gmail.com  http://goo.gl/1wk0
>> College of Science & Technology, Temple University, Philadelphia PA, USA
>> International Centre for Theoretical Physics, Trieste. Italy.
>>
>> _______________________________________________
>> Pw_forum mailing list
>> Pw_forum at pwscf.org
>> http://pwscf.org/mailman/listinfo/pw_forum
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum



-- 
Dr. Axel Kohlmeyer  akohlmey at gmail.com  http://goo.gl/1wk0
College of Science & Technology, Temple University, Philadelphia PA, USA
International Centre for Theoretical Physics, Trieste. Italy.




More information about the users mailing list