Re: [AMBER] Issue with running NPT ensemble simulations on AMBER 22 using RTX 4500 ada

From: Tru Huynh via AMBER <amber.ambermd.org>
Date: Wed, 14 Feb 2024 22:56:01 +0100

Hi

sbgrid is providing amber 18 and 22 (https://www.sbgrid.org/software/titles/amber)
precompiled as ready to use "bacl box" in "capsule"
(https://www.sbgrid.org/wiki/capsules.md) or module
(https://www.sbgrid.org/wiki/modulefiles.md). End user don't get details
on the build process, one can report issue to their bugs.sbgrid.org support email.

[tru.sillage bin]$ pwd
/programs/x86_64-linux/amber/22/bin
[tru.sillage bin]$ ll pmemd.cuda*
lrwxrwxrwx. 1 1004 1004 15 Nov 13 12:20 pmemd.cuda -> pmemd.cuda_SPFP
-rwxr-xr-x. 1 1004 1004 141144128 Jan 7 2023 pmemd.cuda_DPFP
-rwxr-xr-x. 1 1004 1004 143316256 Jan 7 2023 pmemd.cuda_DPFP.MPI
lrwxrwxrwx. 1 1004 1004 19 Nov 13 12:20 pmemd.cuda.MPI -> pmemd.cuda_SPFP.MPI
-rwxr-xr-x. 1 1004 1004 120138720 Jan 7 2023 pmemd.cuda_SPFP
-rwxr-xr-x. 1 1004 1004 122310072 Jan 7 2023 pmemd.cuda_SPFP.MPI

[tru.sillage bin]$ ldd pmemd.cuda_SPFP
        linux-vdso.so.1 => (0x00007ffbffffe000)
        libemil.so => /programs/x86_64-linux/amber/22//lib/libemil.so (0x00007efbf7bb3000)
        libnetcdff.so.6 => /programs/x86_64-linux/amber/22//lib/libnetcdff.so.6 (0x00007efbf7940000)
        libkmmd.so => /programs/x86_64-linux/amber/22//lib/libkmmd.so (0x00007efbf7735000)
        libcufft.so.10 => /programs/x86_64-linux/cuda/11.6.2/lib64/libcufft.so.10 (0x00007efbeeca8000)
        libdl.so.2 => /lib64/libdl.so.2 (0x00007efbeeaa4000)
        libcurand.so.10 => /programs/x86_64-linux/cuda/11.6.2/lib64/libcurand.so.10 (0x00007efbe8eb4000)
        libcublas.so.11 => /programs/x86_64-linux/cuda/11.6.2/lib64/libcublas.so.11 (0x00007efbdf736000)
        libnetcdf.so.13 => /programs/x86_64-linux/amber/22//lib/libnetcdf.so.13 (0x00007efbdf480000)
        libz.so.1 => /programs/x86_64-linux/amber/22/amber_extlib/zlib-1.2.12-grpx/lib/libz.so.1 (0x00007efbdf268000)
        libarpack.so => /programs/x86_64-linux/amber/22//lib/libarpack.so (0x00007efbdf04a000)
        liblapack.so => /programs/x86_64-linux/amber/22//lib/liblapack.so (0x00007efbded58000)
        libblas.so => /programs/x86_64-linux/amber/22//lib/libblas.so (0x00007efbdeb2d000)
        libm.so.6 => /lib64/libm.so.6 (0x00007efbde82b000)
        libpthread.so.0 => /lib64/libpthread.so.0 (0x00007efbde60f000)
        librt.so.1 => /lib64/librt.so.1 (0x00007efbde407000)
        libstdc++.so.6 => /programs/x86_64-linux/gcc/8.4.0/gcc_extlib/gcc-8.4.0-nycd/lib64/libstdc++.so.6 (0x00007efbde042000)
        libgfortran.so.5 => /programs/x86_64-linux/gcc/8.4.0/gcc_extlib/gcc-8.4.0-nycd/lib64/libgfortran.so.5 (0x00007efbddbc9000)
        libgcc_s.so.1 => /programs/x86_64-linux/gcc/8.4.0/gcc_extlib/gcc-8.4.0-nycd/lib64/libgcc_s.so.1 (0x00007efbdd9b2000)
        libquadmath.so.0 => /programs/x86_64-linux/gcc/8.4.0/gcc_extlib/gcc-8.4.0-nycd/lib64/libquadmath.so.0 (0x00007efbdd773000)
        libc.so.6 => /lib64/libc.so.6 (0x00007efbdd3a5000)
        /lib64/ld-linux-x86-64.so.2 (0x00007efbf7ddd000)
        libcublasLt.so.11 => /programs/x86_64-linux/cuda/11.6.2/lib64/libcublasLt.so.11 (0x00007efbc833f000)

So they have been compiled for cuda 11.6.2 with gcc 8.4.0.

Maybe the RTX 4500 ADA is not compatible with the provided cuda environment
and would require a rebuild from sbgrid with cuda 12.x ? or something is messing
with the expected libraries or is incompatible with them on your RTX 4500 ADA machine?

just my 2 cents

Tru
-- 
Tru Huynh (PhD) | mailto:tru.pasteur.fr | tel +33 1 45 68 87 37
https://research.pasteur.fr/en/team/structural-bioinformatics/
Institut Pasteur, 25-28 rue du Docteur Roux, 75724 Paris CEDEX 15 France  
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Feb 14 2024 - 14:00:02 PST
Custom Search