[gmx-users] Build on OSX with 4.6beta1

Szilárd Páll szilard.pall at cbr.su.se
Mon Dec 3 20:18:40 CET 2012


Hi,

Preferably we should avoid starting an OS flame-war here. However, before
this thread turns into an ode to Apple, I have to say: Mac OS X is/can be a
pain both for development and computational use. As long as you use it to
code for iPhone or write some Mac app in Xcode, it's probably excellent.
However, believe it or not, we have spent considerably more time working
around "features" and peculiarities (i.e bugs) of the OS X development
environment toolchain just to get code to compile and run than we did for
Windows!

Therefore, I would personally very much like to suggest all Mac users to
install Linux on their machine, compile without hassle and get much better
performance (at least with default development tools). Unfortunately, most
people would probably send me to hell, so instead we've tried and will try
make things work as well as possible.

Blaming everything on the build system is quite unfair, though. Although it
is not perfect, we have to deal with an inherently complex situation which
is not a choice, but more of a necessity. HPC hardware and software is
becoming increasingly complex and heterogeneous. While a few years ago you
could get close to peak performance by compiling GROMACS with SSE2 enabled
(which the majority of CPUs supported), nowadays you yourself would have to
consider SSE2, SSSE3, SSE4.1, SSE4a, AVX 128/256 bit Intel and AMD flavor
w/wo FMA3 and FMA4 (and I haven't even mentioned GPUs and multi-threading)
*if* we were not doing everything automatically at compile-/run-time
ourselves!

You probably don't realize, but we are (stupidly enough :) still trying to
make sure that the user can just type "cmake . && make" and get close to
optimal performance, but this is not simple at all. Keep in mind that
without all the automation you'd have to figure out yourself what CPU
acceleration to use with which compiler options (multiply by two, one set
for CPUs another one for GPUs), which compilers are broken (there are a
few), etc.

To conclude, I think I can safely say that issues on Linux OS-es with
reasonably new compiler toolchain will, for obvious reasons, have the best
support. We will certainly not ignore issues on other OS-es, but in general
we should rather spend time on improving build experience and performance
for the 95% than working around buggy compilers.

Cheers,
--
Szilárd

On Mon, Dec 3, 2012 at 4:53 PM, Justin Lemkul <jalemkul at vt.edu> wrote:

>
>
> On 12/3/12 10:48 AM, Albert wrote:
>
>> On 12/03/2012 04:40 PM, Justin Lemkul wrote:
>>
>>> I think every OS has something to complain about, though Linux seems to
>>> be the
>>> most friendly.  Yes, Mac has its quirks and annoyances (as does Windows,
>>> which
>>> is becoming more and more common in the scientific realm), but I would
>>> disagree that it's a complete nightmare.  I've used a Mac for everything
>>> since
>>> 2005 and have always found a workaround for any issue.  I think the main
>>> problem here is an increasingly complex build system that has to take
>>> into
>>> account breathtaking complexity in terms of hardware, OS, compilers, etc
>>> and
>>> it just needs a bit of tweaking.
>>>
>>> -Justin
>>>
>>
>> well, probably each person has different opinions. I've just have a talk
>> with
>> friends around me who are also work on computational work. They all hold
>> the
>> same opions with me. If we would like to use the macOS for computational
>> work,
>> in most cases, in fact we are working on LINUX environment instead of
>> macOS.
>> They introduced tools to use Linux under macOS such as macPORT. In
>> essence, they
>> are not native macOS at all and they are always this or that problem
>> there, and
>> you have to waste lot of time working on them.
>>
>> Personally, I would never consider and suggest to use macOS for
>> computational
>> work. It is no better than Windows for office work and it is no better
>> than
>> Linux for computational work, and I find no reasons to use a macOS.
>>
>
>
> To each his own.  I was hoping my post would re-focus on the fact that
> we're dealing with a build system issue rather than suggesting someone
> trash their computer (which Gromacs should support and will for the
> foreseeable future). Let's allow the original discussion to continue,
> please, rather than hijack the thread.  I'm sure there are plenty of places
> on the Internet where people trash Apple products ;)
>
> -Justin
>
>
> --
> ==============================**==========
>
> Justin A. Lemkul, Ph.D.
> Research Scientist
> Department of Biochemistry
> Virginia Tech
> Blacksburg, VA
> jalemkul[at]vt.edu | (540) 231-9080
> http://www.bevanlab.biochem.**vt.edu/Pages/Personal/justin<http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin>
>
> ==============================**==========
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/**mailman/listinfo/gmx-users<http://lists.gromacs.org/mailman/listinfo/gmx-users>
> * Please search the archive at http://www.gromacs.org/**
> Support/Mailing_Lists/Search<http://www.gromacs.org/Support/Mailing_Lists/Search>before posting!
> * Please don't post (un)subscribe requests to the list. Use the www
> interface or send it to gmx-users-request at gromacs.org.
> * Can't post? Read http://www.gromacs.org/**Support/Mailing_Lists<http://www.gromacs.org/Support/Mailing_Lists>
>



More information about the gromacs.org_gmx-users mailing list