Hi,
I'm not sure what's going on here, but it's probably
architecture-related; as far as I know none of the Amber devs has easy
access to a Power9 system (I certainly don't), so it may be hard for
us to reproduce. The errors only happen with mdgx? Not
sander/pmemd/cpptraj/etc?
-Dan
On Mon, Nov 22, 2021 at 7:54 PM Derek M Shore <des2037.med.cornell.edu> wrote:
>
> Thanks very much for your suggestions, Prof. Case!
>
> I think your suggestions that there is a problem with DetectBinaryFile() makes a lot of sense; the reason is because the majority of my errors are related to mdgx, and most have the same initial error:
> >ReadRst >> Attempted to read inpcrd as a NetCDF
> >ReadRst >> restart file, but failed.
>
> Here is a list of all the test that fail with this error:
> cd mdgx/1p7e && ./Test.1p7e
> cd mdgx/MLE && ./Test.mle
> cd mdgx/Settle && ./Test.settle
> cd mdgx/CMAP && ./Test.cmap
> cd mdgx/MCBarostat && ./Test.mcb
> cd mdgx/TIP4P && ./Test.tip4p
> cd mdgx/ForceCalc && ./Test.frcc
> cd mdgx/AddEP && ./Test.addep
> cd mdgx/Triclinic && ./Test.triclinic
> cd mdgx/Peptides && ./Test.peptides
> cd mdgx/Topology && ./Test.prmtop_vs
> cd mdgx/Topology && ./Test.tip4p
> cd mdgx/Topology && ./Test.tip5p
> cd mdgx/Topology && ./Test.methanol
>
> (many of the above test scripts have multiple parts/failures)
>
> I am attaching the first inpcrd (for Test.1p7e); it's not a binary file, and it looks normal to me. Just for fun, I tried loading a inpcrd & prmtop created by CHARMM-GUI for a completely different system (changing mdgx.in as needed), and I got the same error. (the point was to see if mdgx could handle an arbitrary inpcrd/prmtop, and it could not.
>
> Any advice or suggestions on how to resolve is most appreciated. Please let me know if I can provide any additional info that could be helpful.
>
> Best,
> Derek
>
> Derek M. Shore, Ph.D.
> Scientific Computing Technology Engineer
> Institute for Computational Biomedicine
> Department of Physiology and Biophysics
>
> Weill Cornell Medicine
> Department of Physiology and Biophysics
> 1300 York Avenue, Room LC501A
> New York, NY 10065
> T: 212-746-6386
> des2037.med.cornell.edu
>
>
>
> On 11/20/21, 9:03 AM, "David A Case" <david.case.rutgers.edu> wrote:
>
> On Fri, Nov 19, 2021, Derek M Shore wrote:
> >
> >I am trying to compile/test Amber20 (tools + GPU-accelerated MD), on Summit
> >at ORNL (IBM POWER9, NVIDIA V100s). The compile seems to run without a
> >problem. When I `make test.serial`, most of the tests pass, but I see
> >several errors similar to this one:
> >*/amber20/AmberTools/test/mdgx/1p7e$ ./Test.1p7e
> >ReadRst >> Attempted to read inpcrd as a NetCDF
> >ReadRst >> restart file, but failed.
>
> Do the errors all happen with mdgx? If so, you may be able to ignore
> them for the time being, if you are not planning to use mdgx.
>
> The inpcrd file in the test you cite is indeed a formatted restart file, and
> not a NetCDF file. So the problem may be with mdgx, and not with NetCDF.
> (This is just speculation: it depends on whether you see the same error with
> pememd.) Maybe Dave Cerutti has some insight into how mdgx decides how to
> try to open a restart file, but it's odd that this fails only on Summit.
>
> My WAG is that the DetectBinaryFile() may not be working on the Summit file
> system. To help, you could (a) comment out the final line of the
> /amber20/AmberTools/test/mdgx/1p7e/Test.1p7e script, and (b) post (as an
> attachment!) the inpcrd file that the test creates.
>
> Thanks for the report....dac
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> https://urldefense.proofpoint.com/v2/url?u=http-3A__lists.ambermd.org_mailman_listinfo_amber&d=DwICAg&c=lb62iw4YL4RFalcE2hQUQealT9-RXrryqt9KZX2qu2s&r=X5Sc7H5NSWV3agvbxBXvARVirdVlt5y-nKiETDcG3R8&m=falYyFfqOlpxG-cP5tpVOJ7PZYcZ5dOFs05fU6fr2QePx8L5-Y5BZPaTT2I7V9aD&s=K4IXuCOT6BVfLkDTOj0iL-jNKt-mNfqCyb5pwrrnI_8&e=
>
> _______________________________________________
> 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 Wed Dec 01 2021 - 06:00:04 PST
