[mvapich-discuss] segmentation fault with GPUDirect

[Jonathan Perkins]

Hello Enrico.  In our case when building OMB we do not need to specify the
--with-cuda when building.  I believe that when specifying this option the
order of library inclusion is different between your build and our build.
With our build the interception of the CUDA allocation routines works
without using LD_PRELOAD.  However with your build it looks like you need
to include it.

For more information please take a look at the following link:
http://mvapich.cse.ohio-state.edu/userguide/gdr/2.2b/#_example_use_of_ld_preload

Please let us know if this helps.

On Mon, Apr 11, 2016 at 1:21 PM Enrico Calore <enrico.calore at fe.infn.it>
wrote:

> On 04/11/2016 07:18 PM, Jiri Kraus wrote:
> > Hi Enrico,
> >
> > Are you enabling CUDA-aware Features by setting the env. var.
> MV2_USE_CUDA=1?
>
> Yes :)
>
>
> Enrico
>
>
> _______________________________________________
> mvapich-discuss mailing list
> mvapich-discuss at cse.ohio-state.edu
> http://mailman.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20160411/4664d513/attachment.html>