Re: [AMBER] cudaGetDeviceCount failed unknown error AMBER16

From: Ross Walker <ross.rosswalker.co.uk>
Date: Wed, 12 Jul 2017 15:44:58 -0400

Hi Hector,

Are you running AMBER 16 here with the latest patches?

What does nvidia-smi report?

What does nvcc -v report?

You shouldn't need to do any of these hacks, especially the soft link to the libcuda.so library - that's just going to mess things up and complicate debugging more.

All the best
Ross

> On Jul 12, 2017, at 3:16 PM, Hector Velazquez <HVelazquez.linkp.com> wrote:
>
> Hello AMBER development team,
>
>
> I am trying to compile the GPU version of AMBER with an NVIDIA GTX-1080. Everything appears to be fine until I run "make test" where are all the tests fail with the following error:
>
>
> cudaGetDeviceCount failed unknown error
>
>
> I found the link where this error is mentioned:
>
>
> https://mailman.ucsd.edu/pipermail/mccammon-support-l/2015/000673.html
>
>
> However, re-running as sudo per their "quick and dirty" solution doesn't work, and the link to the longer term solution is dead.
>
>
> My serial and parallel builds seem to be fine.
>
>
> I also noticed the following during the "make install" step:
>
>
> nvcc warning : The 'compute_20', 'sm_20', and 'sm_21' architectures are deprecated, and may be removed in a future release (Use -Wno-deprecated-gpu-targets to suppress warning).
>
>
> Since it's just a warning, I don't think it matters, but I figured I would mention it just in case.
>
>
> I did some more digging around on google and found the following link:
>
>
> https://devtalk.nvidia.com/default/topic/617414/-solved-cuda-driver-version-is-insufficient-for-cuda-runtime-version-fedora-18-rpmfusion-driver/
>
>
> This thread says to do the following:
>
> 1. $ cd /usr/lib[64]/nvidia
> 2. $ sudo ln -s libcuda.so.1 libcuda.so
> 3. $ sudo ldconfig
>
>
> I did this, but I'm still experiencing the same error.
>
>
> Here is so information from the log from "make test"
>
>
>
> cd gb_ala3/ && ./Run.igb1_ntc1_min DPFP /storage/Programs/Amber/amber16/include/netcdf.mod
> cudaGetDeviceCount failed unknown error
> ./Run.igb1_ntc1_min: Program error
> make[3]: [test.pmemd.cuda.gb.serial] Error 1 (ignored)
> cd gb_ala3/ && ./Run.irest1_ntt0_igb1_ntc1 DPFP /storage/Programs/Amber/amber16/include/netcdf.mod
> cudaGetDeviceCount failed unknown error
> ./Run.irest1_ntt0_igb1_ntc1: Program error
> make[3]: [test.pmemd.cuda.gb.serial] Error 1 (ignored)
> cd gb_ala3/ && ./Run.irest0_ntt0_igb1_ntc1_hotstart DPFP /storage/Programs/Amber/amber16/include/netcdf.mod
> cudaGetDeviceCount failed unknown error
> ./Run.irest0_ntt0_igb1_ntc1_hotstart: Program error
> make[3]: [test.pmemd.cuda.gb.serial] Error 1 (ignored)
>
>
>
>
>
>
>
> Thanks for your time,
>
>
> Hector A. Velazquez
>
> Computational Chemistry and Cheminformatics Specialist
>
> NewLink Genetics
>
> 515-296-5555
>
> Ext. 2909
>
> Confidentiality Notice: This e-mail transmission contains information from NewLink Genetics that may be confidential and/or proprietary. If you are not the intended recipient, be aware that any disclosure, copying, distribution or use of the contents of this information is prohibited. Unintended distribution shall not constitute waiver of this or any other privilege and confidentiality. If you have received this e-mail transmission in error, please reply to the sender and delete the message and its contents from your system. Thank you.
> _______________________________________________
> 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 Jul 12 2017 - 13:00:03 PDT
Custom Search