Hi Brian,
Can you try one thing for me quickly. Delete your amber14 directory. Then untar things from the original .tar.bz2 files and cd $AMBERHOME; ./configure cuda gnu
But in this case when it asks you about applying updates - say NO. Then compile everything and see what happens. I am wondering if there might be a bug introduced with one of the latest updates.
All the best
Ross
> On Mar 26, 2015, at 7:41 AM, Brian Radak <brian.radak.accts.gmail.com> wrote:
>
> The nvidia install seems to differentiate those as 32/64 bit libraries.
> Anyway, I've been including both now.
>
> Noting how obnoxious things can be with stray library files, etc. I just
> re-installed /everything/ from a clean install. There are now no signs
> of different driver or cuda toolkit versions on my computer.
>
> I installed both the driver and cuda 6.5 with the run script from the
> nvidia website (this was a slightly nuanced affair on Ubuntu 14.04 given
> my hardware). However, I still get the same error:
>
> cudaMemcpyToSymbol: SetSim copy to cSim failed invalid device symbol
>
> I also tried manually running the cuda/4096wat test; unfortunately there
> wasn't much more helpful output than that above:
>
> |------------------- GPU DEVICE INFO --------------------
> |
> | CUDA_VISIBLE_DEVICES: not set
> | CUDA Capable Devices Detected: 1
> | CUDA Device ID in use: 0
> | CUDA Device Name: Quadro K2200
> | CUDA Device Global Mem Size: 4095 MB
> | CUDA Device Num Multiprocessors: 5
> | CUDA Device Core Freq: 1.12 GHz
> |
> |--------------------------------------------------------
>
>
> | Conditional Compilation Defines Used:
> | PUBFFT
> | BINTRAJ
> | CUDA
> | EMIL
>
> | Largest sphere to fit in unit cell has radius = 24.800
>
>
>
>
> On 03/25/2015 01:49 PM, Jason Swails wrote:
>>> On Mar 25, 2015, at 2:46 PM, Brian Radak <brian.radak.accts.gmail.com> wrote:
>>>
>>> Ah yes, the compiling equivalent of "turning it off and on again."
>>> Unfortunately it looks like the result is the same.
>>>
>>> Double checking my CUDA_HOME, etc. settings:
>>>
>>>>> nvcc -V
>>> nvcc: NVIDIA (R) Cuda compiler driver
>>> Copyright (c) 2005-2014 NVIDIA Corporation
>>> Built on Thu_Jul_17_21:41:27_CDT_2014
>>> Cuda compilation tools, release 6.5, V6.5.12
>>>>> echo $CUDA_HOME
>>> /usr/local/cuda-6.5
>>>>> echo $LD_LIBRARY_PATH
>>> /home/radak/software/amber14/lib:/usr/local/cuda-6.5/lib64:/usr/local/cuda-6.5/lib
>>>
>>> The CUDA install instructions I found suggested the "lib64" addition above,
>>> but the AMBER install msg only suggests "lib". Is that all cool?
>> Both is probably better (my LD_LIBRARY_PATH is ...:/opt/cuda/6.5/lib:/opt/cuda/6.5/lib64:...). $CUDA_HOME/lib on my machine only contains the runtime libraries -- all of the other libraries (like libcufft and libcublas) are in lib64, and at least cufft is needed for pmemd.cuda...
>>
>> HTH,
>> Jason
>>
>> --
>> Jason M. Swails
>> BioMaPS,
>> Rutgers University
>> Postdoctoral Researcher
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>
> --
> Brian Radak
> Postdoctoral Scholar
> Gordon Center for Integrative Science, W323A
> Department of Biochemistry & Molecular Biology
> University of Chicago
> 929 E. 57th St.
> Chicago, IL 60637-1454
> Tel: 773/834-2812
> email: radak.uchicago.edu
> _______________________________________________
> 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 Thu Mar 26 2015 - 09:00:03 PDT
