Does the test suite pass?
You can run a small QM/MM test from the test suite manually, e.g. alanine dipeptide:
cd $AMBERHOME/test/qmmm_Quick/QMMM_MD
export TESTsander=sander.quick.cuda
./Run.aladip.hf_sto-3g
All the best,
Andy
—
Dr. Andreas W. Goetz
Associate Research Scientist
San Diego Supercomputer Center
Tel: +1-858-822-4771
Email: agoetz.sdsc.edu
Web: www.awgoetz.de
On Jan 15, 2025, at 6:06 PM, Matt Goff <matt.goff.utah.edu> wrote:
Hmm interesting. I’m not running anything else on that card when I run quick. My simulation is a tetranucleotide, but it is in explicit solvent so maybe that’s driving the memory too high. I’ll build something smaller (alanine dipeptide maybe?) and see if that runs.
Get Outlook for iOS<
https://urldefense.com/v3/__https://aka.ms/o0ukef__;!!Mih3wA!GQoXUYhqvJ8syer9q1rpU41pMRXSUe1tKURfuKNMvW3yzHheFtbVl3tOAVPyPEFLAxyYrg1G4bSPe4joaBIr$>
________________________________
From: Goetz, Andreas <awgoetz.ucsd.edu>
Sent: Wednesday, January 15, 2025 7:00:19 PM
To: Matt Goff <matt.goff.utah.edu>; AMBER Mailing List <amber.ambermd.org>
Cc: Manathunga Mudiyanselage, Madushanka <manathun.msu.edu>; Kurt O'Hearn <ohearnku.msu.edu>
Subject: Re: [AMBER] Quick Cuda
Hi Matt,
We have tested QUICK (both standalon quick.cuda for QM calculations and sander.quick.cuda for QM/MM molecular dynamics in AmberTools) on RTX 6000 Ada cards. Tests were done with CUDA 12.0 and Ubuntu 20 LTS.
What you see could be a compilation issue or due to insufficient GPU memory. Are you running anything else on the GPU at the same time?
All the best,
Andy
—
Dr. Andreas W. Goetz
Associate Research Scientist
San Diego Supercomputer Center
Tel: +1-858-822-4771
Email: agoetz.sdsc.edu
Web: www.awgoetz.de<https://urldefense.com/v3/__http://www.awgoetz.de__;!!Mih3wA!GQoXUYhqvJ8syer9q1rpU41pMRXSUe1tKURfuKNMvW3yzHheFtbVl3tOAVPyPEFLAxyYrg1G4bSPe7gFmqgM$>
On Jan 15, 2025, at 4:16 PM, Matt Goff via AMBER <amber.ambermd.org> wrote:
Hey Guys,
I'm trying to run sander.quick.cuda and get the following error. Any
idea how to fix it? For reference, I'm running Ubuntu 24.04 with an
Nvidia RTX 6000 ada. As far as I can tell, I've pointed it to use Cuda
12.4 that is compatible with Amber, although 12.6 exists on my system
as well by default. Is this a bug? Or have I compiled or run something
wrong?
Best,
Matt
cudaMemcpy cuda_buffer_type :: Download failed!: an illegal memory
access was encountered in
/opt/amber/amber24_src/AmberTools/src/quick/src/cuda/gpu_type.h at line
847
cudaMemcpy cuda_buffer_type :: Download failed!: an illegal memory
access was encountered in
/opt/amber/amber24_src/AmberTools/src/quick/src/cuda/gpu_type.h at line
847
Program received signal SIGSEGV: Segmentation fault - invalid memory
reference.
Backtrace for this error:
#0 0x789723623e59 in ???
#1 0x789723622e75 in ???
#2 0x78972324531f in ???
at
./signal/../sysdeps/unix/sysv/linux/x86_64/libc_sigaction.c:0
#3 0x789726cc5a8d in ???
#4 0x7897266bb543 in ???
#5 0x7897266bc56d in ???
#6 0x7897266c006c in ???
#7 0x7897267c8ccd in ???
#8 0x7897266dbce3 in ???
#9 0x7897266a906b in ???
#10 0x7897266a984c in ???
#11 0x58620e35e092 in ???
#12 0x58620e3005f2 in ???
#13 0x58620e18d3e8 in ???
#14 0x58620e224dd3 in ???
#15 0x58620e1d5239 in ???
#16 0x58620e1cd989 in ???
#17 0x58620e1cd9ee in ???
#18 0x78972322a1c9 in __libc_start_call_main
at ../sysdeps/nptl/libc_start_call_main.h:58
#19 0x78972322a28a in __libc_start_main_impl
at ../csu/libc-start.c:360
#20 0x58620dfeef14 in ???
#21 0xffffffffffffffff in ???
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
https://urldefense.com/v3/__http://lists.ambermd.org/mailman/listinfo/amber__;!!Mih3wA!FSDcHu3BvcrpXOiHWu7dOqLybj6UqXLv-Tt3nANeezWzLhtDojOrdvEDlNQcbASNYSGXhrQiE-GT-g$
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jan 15 2025 - 18:30:03 PST
