From: David Hardy (dhardy_at_ks.uiuc.edu)
Date: Wed Jan 16 2013 - 11:07:48 CST
Dear Evi,
Sorry, this is one is actually a NAMD software issue. Dating back to the earliest NAMD CUDA kernel implementations, the code uses static variables (with global file scope). Now for SMP, these variables are made thread safe using the "static __thread" declaration, which works on Linux, but not for Windows or MacOSX builds. I'll look into implementing a fix for this.
Best regards,
Dave
On Jan 16, 2013, at 10:29 AM, Paraskevi Gkeka wrote:
> Dear NAMD users,
>
> we are trying to compile NAMD with GPUs on an Apple Cluster. We have
> successfully installed charm++ and run config script to install the
> program:
>
> ./config MacOSX-x86_64-g++ --charm-arch mpi-darwin-ppc --with-cuda
> --cuda-prefix /usr/local/cuda
> cd MacOSX-x86_64-g++
> make
>
> and got the following error message for several instances:
> src/ComputeNonbondedCUDA.C:1022: error: thread-local storage not
> supported for this target
> make: *** [obj/ComputeNonbondedCUDA.o] Error 1
>
> Do you have a work-around for this?
> Thank you in advance for any suggestions.
> Evi
>
> --
> ------------------------------
> Dr. Paraskevi Gkeka
> Postdoctoral Fellow
> Biomedical Research Foundation
> Academy of Athens
> 4, Soranou Ephessiou
> 115 27
> Athens
> Greece
> Office: +302106597568
> Cell: +306974350829
> -------------------------------
-- David J. Hardy, Ph.D. Theoretical and Computational Biophysics Beckman Institute, University of Illinois dhardy_at_ks.uiuc.edu http://www.ks.uiuc.edu/~dhardy/
This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:20:51 CST