I can go in and fix the __shfl in a number of places, but not all. I think
I'll do that after some beers. The code will go faster, but also be a bit
snarky.
On Fri, Jun 22, 2018 at 5:08 PM Jamie Smith <
jsmith.crackofdawn.onmicrosoft.com> wrote:
> The CMake build system gives a somewhat different error message, many many
> times (almost 50MB of logs per build):
>
> /iscratch/jenkins/home/workspace/Amber/AmberLinux/src/pmemd/src/cuda/kNLCINE.h(528):
> warning: function "__shfl(unsigned int, int, int)"
> /opt/cuda/9.0/include/sm_30_intrinsics.hpp(167): here was declared
> deprecated ("__shfl() is deprecated in favor of __shfl_sync() and may be
> removed in a future release (Use -Wno-deprecated-declarations to suppress
> this warning).")
>
> It looks like kNLCLINE.h uses __shfl() all over the place. Is this
> something we're able to fix? Because it would certainly be nice. I had to
> buy one of those momentum-scrolling mice just so I could scroll through
> Jenkins logs more quickly.
>
> -----Original Message-----
> From: David Cerutti [mailto:dscerutti.gmail.com]
> Sent: Friday, June 22, 2018 12:20 PM
> To: AMBER Developers Mailing List <amber-developers.ambermd.org>
> Subject: [AMBER-Developers] Deprecated messages in CUDA compilations....
> STOP!
>
> Is anyone else seeing a plethora of messages like this one?
>
> ptxas
> /tmp/tmpxft_000023ca_00000000-5_kCalculatePMENonbondEnergy.compute_61.ptx,
> line 204535; w\
> arning : Instruction 'shfl' without '.sync' is deprecated since PTX ISA
> version 6.0 and will be\ discontinued in a future PTX ISA version
>
> This is happening even though there compilation flags include
> -Wno-deprecated-declarations. It is literally finding a needle in the
> haystack to figure out what line of code is wrong having to sift through
> these mountains of deprecation warnings. Yes, shfl is deprecated.
> But how do I shut that warning off, for real?
>
> Dave
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Fri Jun 22 2018 - 14:30:04 PDT
