[mvapich-discuss] --enable-cuda=basic when building with PGI compiler set

Maxim Milakov mmilakov at nvidia.com
Fri Jun 7 07:02:53 EDT 2013


Hi Sreeram,

It is great, thanks a lot!

Max.


On 06/07/2013 03:01 PM, sreeram potluri wrote:
> Maxim,
>
> Yes, you will still be able to use device_ptr in MPI_Send and other 
> MPI communication calls. It is only some datatype related 
> optimizations that are disabled with this flag and they should not 
> affect the way you can use MPI calls with device pointers.
>
> Sreeram Potluri
>
> On Fri, Jun 7, 2013 at 6:25 AM, Maxim Milakov <mmilakov at nvidia.com 
> <mailto:mmilakov at nvidia.com>> wrote:
>
>     Hi,
>
>     I am building MVAPICH2 1.9 using PGI 13.5 accelerator:
>
>     CPPFLAGS="-D x86_64
>     -D__align__\(n\)=__attribute__\(\(aligned\(n\)\)\)
>     -D__location__\(a\)=__annotate__\(a\) -DCUDARTAPI=" F77=pgf77
>     FC=pgf90 CC=pgcc CXX=pgcpp ./configure
>     --prefix=/home/max/software/pgi_specific/mvapich2 --enable-f77
>     --enable-fc --enable-cxx --enable-cuda=basic
>     --with-cuda=$CUDA_PATH --disable-mcast --enable-shared
>     --disable-static
>
>     I am specifying --enable-cuda=basic as configure script fails
>     otherwise (and "basic" was suggested here, in the user group some
>     time ago).
>
>     I am wondering in what degree I am limiting myself with this
>     "--enable-cuda=basic"... Will I be able to pass device pointer to
>     mpi_send?
>
>     Thanks!
>
>     -- 
>
>     Best regards
>     Maxim Milakov
>
>     HPC Developer Technology Engineer
>     NVIDIA Russia
>     12 Dvintsev St., bld. C
>     127018 Moscow, Russia
>     +7 495 9810300, ext.10806  <tel:%2B7%20495%209810300%2C%20ext.10806>
>
>     ------------------------------------------------------------------------
>     This email message is for the sole use of the intended
>     recipient(s) and may contain confidential information.  Any
>     unauthorized review, use, disclosure or distribution is
>     prohibited.  If you are not the intended recipient, please contact
>     the sender by reply email and destroy all copies of the original
>     message.
>     ------------------------------------------------------------------------
>
>     _______________________________________________
>     mvapich-discuss mailing list
>     mvapich-discuss at cse.ohio-state.edu
>     <mailto:mvapich-discuss at cse.ohio-state.edu>
>     http://mail.cse.ohio-state.edu/mailman/listinfo/mvapich-discuss
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.cse.ohio-state.edu/pipermail/mvapich-discuss/attachments/20130607/07b3c3cb/attachment.html


More information about the mvapich-discuss mailing list