[gmx-users] does gromacs-4.6 support intel core Quad Q9550 cpu?

Mark Abraham mark.j.abraham at gmail.com
Thu Sep 25 22:30:47 CEST 2014


On Thu, Sep 25, 2014 at 9:49 PM, Vedat Durmaz <durmaz at zib.de> wrote:

> :)  well, you convinced me finally. after having had screwed up all my
> courage, i just gave it a try:
>
> cmake .. -DCMAKE_INSTALL_PREFIX=$path
>
> and everything worked surprisingly fine with make.


Great!


> i guess i was still traumatized by the attempt to install gromacs on some
> cray xc30 a couple of month ago!
>

... where I bet someone thought it was a good idea to have a front end node
that was different from the target host ;-) Saving a thousand bucks is
worth wasting a lot of person and compute time, right? :-D

Mark


> vedat
>
>
>
> Am 25.09.2014 um 18:09 schrieb Mark Abraham:
>
>  On Thu, Sep 25, 2014 at 4:19 PM, Vedat Durmaz <durmaz at zib.de> wrote:
>>
>>  i guess that's it, mark. thanks. following mirco's hint and checking the
>>> log file indeed reveals that "rdtscp" was used upon compiling the
>>> binaries.
>>>
>>> so i will compile it myself on the affected machine. you say i don't
>>> need to set any special cmake variable?
>>>
>>
>> "Nothing is required if building on the target machine - GROMACS cmake
>> configuration does all the required detection"
>>
>>
>>  shouldn't i at least set the
>>> "GMX_USE_RDTSCP" variable the website you've linked below mentions?
>>>
>>
>> Not if GROMACS cmake is doing the required detection - which it does. (We
>> did improve the implementation of the detection after 4.6.5, but that is
>> an
>> issue only if you tried to do a source-based install of 4.6.5 or earlier
>> with a build host that has rdtscp and a target machine that does not - so
>> do the build on the target machine and stop worrying about it ;-) ).
>>
>> and
>>
>>> if so, how would i do that? like this?:
>>>
>>> cmake ... -DGMX_USE_RDTSCP=OFF
>>>
>>>  Yes, but you don't need if it you are building on the target machine,
>> because GROMACS cmake configuration does all the required detection, and
>> this gets turned off automatically if appropriate.
>>
>>
>>  and what about using SSE4.1? you've written: "Yes, and use the SSE4.1
>>> speed advantages ..." do i need to set some proper cmake variable in
>>> order to use this feature?
>>>
>>>  No, if you are building on the target machine, because GROMACS cmake
>> configuration does all the required detection, and turns on SSE4.1 support
>> if that would suit the build host.
>>
>> :-)
>>
>> Mark
>>
>>
>>  vedat
>>>
>>>
>>>
>>> Am 25.09.2014 um 15:28 schrieb Mark Abraham:
>>>
>>>> On Thu, Sep 25, 2014 at 12:16 PM, Vedat Durmaz <durmaz at zib.de> wrote:
>>>>
>>>>  hi guys,
>>>>>
>>>>> sorry for disturbing!
>>>>>
>>>>> the pc of a student here has ubuntu 14.04 installed on it along with
>>>>> the
>>>>> gromacs version 4.6.5 debian/ubuntu binaries from the ubuntu
>>>>>
>>>> repositories.
>>>
>>>> when we start mdrun, we get an german error message saying:
>>>>>
>>>>> "ungültiger maschinenbefehl" (something like "invalid machine
>>>>> command").
>>>>> when searching the internet i got the feeling that this has something
>>>>> to
>>>>> do with the cpu.
>>>>>
>>>>> cat /proc/cpuinfo yields (among others):
>>>>>
>>>>> ...
>>>>> flags        : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca
>>>>> cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall
>>>>> nx lm constant_tsc arch_perfmon pebs bts rep_good nopl aperfmperf pni
>>>>> dtes64 monitor ds_cpl vmx smx est tm2 ssse3 cx16 xtpr pdcm sse4_1 xsave
>>>>> lahf_lm dtherm tpr_shadow vnmi flexpriority
>>>>> ...
>>>>>
>>>>> can anybody confirm that this cpu with the given properties/flag is not
>>>>> supported by gromacs 4.6 debian binaries?
>>>>>
>>>> That cpu does not appear to support an instruction (rdtscp) that is used
>>>>
>>> by
>>>
>>>> GROMACS in doing timing if that instruction was supported on the build
>>>> machine (mentioned at
>>>>
>>>>  http://www.gromacs.org/Documentation/Installation_
>>> Instructions_4.6#4.3.1._Portability_aspects
>>>
>>>> ).
>>>>
>>>>
>>>>  if so, would compiling it from
>>>>> the sources remedy the problem?
>>>>>
>>>> Yes, and use the SSE4.1 speed advantages that the cpu does have, unlike
>>>>
>>> the
>>>
>>>> least-common-denominator build that is targeted by the
>>>> ubuntu-gromacs-package maintainer. If you want mdrun to run optimally,
>>>> compile it on the machine you want to run it on.
>>>>
>>>>
>>>>  if so, which flag must be set for
>>>>> builing/compiling gromacs accordingly?
>>>>>
>>>>>  Nothing is required if building on the target machine - GROMACS cmake
>>>> configuration does all the required detection, but it can't detect the
>>>> capabilities of a machine it hasn't yet seen...
>>>>
>>>> Mark
>>>>
>>>> thanks a lot & take care
>>>>
>>>>> vedat
>>>>>
>>>>> --
>>>>> 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 at 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 at 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 at gromacs.org.
>


More information about the gromacs.org_gmx-users mailing list