Hi Miroslav,
I’ve detected the bug in the code and a fix will be released in an upcoming patch so keep an eye out for this update.
I noticed that you are using ir6=1. Unfortunately this is not yet supported in the GPU implementation of COM distance restraints. An error message is going to be added to prevent it’s use for now. Ross and I would like to know why you use the r6 averaging as opposed to the mass centered averaging? We don’t have any experience with it and would like some more information so we can assess how high a priority implementing it should be. Is this integral to your simulation and what kind of systems require this type of averaging?
Thanks,
Perri Needham
On Jun 17, 2015, at 11:50 AM, Miroslav Krepl <krepl.seznam.cz> wrote:
> Dear all,
>
> I sent the topology, CDF restart file, Amber input and the NMR
> restraints of the problematic system (although no other system I tested
> so far worked on GPU either) in the attachment.
>
> Some more debug I forgot to mention:
>
> I tried including/exluding the COM restraints (iat=-1), but I could find
> no relation. With very few restraints it worked and it did not matter
> whether they were COM or normal. With more restraints, it failed.
>
> However, I did indeed observe that the Ile, Leu and Phe side-chain atoms
> (i.e. those that are usually heavy on the COM definition) were the first
> to blow up with the first integration step. But I am not sure whether
> this is consequential. It could be random.
>
> Thank you!
>
> Best wishes,
>
> Miroslav Krepl
>
> On Wed, 2015-06-17 at 08:00 -0700, Ross Walker wrote:
>> There is most likely a hard limit in the code - we can take a look but can you send me (by private mail if need be) the necessary input files to reproduce this failure.
>>
>> Thank you.
>>
>> All the best
>> Ross
>>
>>> On Jun 17, 2015, at 7:23 AM, David A Case <case.biomaps.rutgers.edu> wrote:
>>>
>>> On Tue, Jun 16, 2015, Miroslav Krepl wrote:
>>>>
>>>> since the com restraints were recently implemented into the GPU version,
>>>> I have tried to run a simple explicit solvent simulations of some
>>>> proteins from the PDB database while using the experimental NMR
>>>> restraints (nmropt=1; distance based restraints, about 2000 of them).
>>>
>>> Certainly sounds like a bug.
>>>
>>>>
>>>> 5. It seems that the number of NMR restraints might be the key. I have
>>>> tried reducing the number to only few (1, 4 etc.) and it appreared to be
>>>> working. However, I was unable to determine the exact number of
>>>> restraints beyond which it starts to crash.
>>>
>>> I'm guessing it is more that some nmr restraints make use of the COM facility
>>> and some don't. Can you see if the crashes are related to the presence or
>>> absence of restraints with iat=-1?
>>>
>>> But we shouldn't dump any more debugging on you, since you've already done a
>>> lot. I'm travelling right now, but any developer with access to a GPU, (and
>>> with knowledge about the GPU code!) should be able to test this way:
>>>
>>> cd $AMBERHOME/test/nmr
>>> export TESTsander="../../bin/pmemd.cuda"
>>> ./Run.nmr
>>>
>>> Thanks for the report....regards...dac
>>>
>>>
>>> _______________________________________________
>>> AMBER mailing list
>>> AMBER.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>
> <sys.zip>_______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Jul 16 2015 - 13:00:03 PDT
