On Tue, Feb 18, 2025, VERONICA MARTIN HERNANDEZ via AMBER wrote:
>They have recompiled CUDA and the issue was solved,
It's unclear what problem was solved: do you mean that timings (and the
actual results) for Amber24 now match what you expect?
> but we've got another error:
>
>Loading /soft/genoa/modulefiles/amber24_cuda12.2.2
> Loading requirement: GCCcore/11.3.0 zlib/1.2.12-GCCcore-11.3.0
> binutils/2.38-GCCcore-11.3.0 /soft/genoa/EB/modules/all/GCC/11.3.0
> numactl/2.0.14-GCCcore-11.3.0 XZ/5.2.5-GCCcore-11.3.0
> libxml2/2.9.13-GCCcore-11.3.0 libpciaccess/0.16-GCCcore-11.3.0
> hwloc/2.7.1-GCCcore-11.3.0 OpenSSL/1.1 libevent/2.1.12-GCCcore-11.3.0
> UCX/1.12.1-GCCcore-11.3.0 libfabric/1.15.1-GCCcore-11.3.0
> PMIx/4.1.2-GCCcore-11.3.0 UCC/1.0.0-GCCcore-11.3.0
> /soft/genoa/EB/modules/all/OpenMPI/4.1.4-GCC-11.3.0
> /soft/genoa/EB/modules/all/CUDA/12.2.2
>Error: invalid argument launching kernel kgBuildSpecial2RestNBPreList
>
>I don't know what causes this because the administrator of the cluster says
>that he doesn't have that problem. Maybe I'm doing something wrong??
It looks like the above error arises from some "module load" command, and
not from Amber; is that correct? Since we have no infomation on how the
module was built, I don't think folks on this list will be able to help.
If the administrator of the cluster doesn't have this problem, the clear
step is to try to find out what you are doing that is different from what
the administrator is doing.
...good luck...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sat Feb 22 2025 - 16:00:03 PST
