Build fails with CUDA v8.0

[xing-liu]

I tried to build RAJA on a POWER8 system, but it failed with the following error messages.

[ 15%] Built target lulesh-OMP.exe [ 16%] Building NVCC (Device) object test/LULESH-v1.0/LULESH-v1.0_RAJA-variants/CMakeFiles/lulesh-RAJA-parallel.exe.dir/lulesh-RAJA-parallel.exe_generated_luleshRAJA-parallel.cxx.o /home/mstars/RAJA-develop/include/RAJA/exec-cuda/reduce_cuda.hxx(269): error: cannot overload functions distinguished by return type alone

1 error detected in the compilation of "/tmp/tmpxft_0000fe77_00000000-5_luleshRAJA-parallel.cpp4.ii". CMake Error at lulesh-RAJA-parallel.exe_generated_luleshRAJA-parallel.cxx.o.cmake:260 (message): Error generating file /home/mstars/RAJA-develop/build/test/LULESH-v1.0/LULESH-v1.0_RAJA-variants/CMakeFiles/lulesh-RAJA-parallel.exe.dir//./lulesh-RAJA-parallel.exe_generated_luleshRAJA-parallel.cxx.o

make[2]: * [test/LULESH-v1.0/LULESH-v1.0_RAJA-variants/CMakeFiles/lulesh-RAJA-parallel.exe.dir/lulesh-RAJA-parallel.exe_generated_luleshRAJA-parallel.cxx.o] Error 1 make[1]: * [test/LULESH-v1.0/LULESH-v1.0_RAJA-variants/CMakeFiles/lulesh-RAJA-parallel.exe.dir/all] Error 2 make: *\ [all] Error 2

[DavidPoliakoff]

Hi Xing,

Thanks for the bug report. I wasn't able to reproduce this using just nvcc and CUDA8.0. Could you point me at the commit of RAJA you were using and the version of LULESH? Currently building RAJA doesn't build LULESH, so I'm having a little difficulty reproducing the bug.

[xing-liu]

Hi David,

Thanks for the reply. I used the RAJA develop branch and checked out it around one month ago. Which version did you try? Does that version support RAJA::Tile?

Xing

[DavidPoliakoff]

Hi Xing,

If you're using the forallN constructs (are you part of the VBL effort?), I'm actually not entirely familiar with that segment of the code, so I'll see whether @ajkunen or @trws are able to chime in. Sorry I can't be of more help on that front.

Best,

David

[ajkunen]

Hey Xing,

The Tile policy only works with forallN, not forall…. and I don’t believe that LULESH uses any forallN’s.

-Adam

From: David Poliakoff notifications@github.com<mailto:notifications@github.com> Reply-To: LLNL/RAJA reply@reply.github.com<mailto:reply@reply.github.com> Date: Thursday, September 1, 2016 at 12:14 PM To: LLNL/RAJA RAJA@noreply.github.com<mailto:RAJA@noreply.github.com> Cc: Adam Kunen kunen1@llnl.gov<mailto:kunen1@llnl.gov>, Mention mention@noreply.github.com<mailto:mention@noreply.github.com> Subject: Re: [LLNL/RAJA] Build fails with CUDA v8.0 (#122)

Hi Xing,

If you're using the forallN constructs (are you part of the VBL effort?), I'm actually not entirely familiar with that segment of the code, so I'll see whether @ajkunenhttps://github.com/ajkunen or @trwshttps://github.com/trws are able to chime in. Sorry I can't be of more help on that front.

Best,

David

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/LLNL/RAJA/issues/122#issuecomment-244182142, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ARGiBDY-Bt_73PYthkVFHsybejh69JwLks5qlyQCgaJpZM4JzAgP.

[rhornung67]

Correct. LULESH uses no forallN traversals. They are used in Kripke. —Rich

From: Adam Kunen notifications@github.com<mailto:notifications@github.com> Reply-To: LLNL/RAJA reply@reply.github.com<mailto:reply@reply.github.com> Date: Thursday, September 1, 2016 at 12:16 PM To: LLNL/RAJA RAJA@noreply.github.com<mailto:RAJA@noreply.github.com> Subject: Re: [LLNL/RAJA] Build fails with CUDA v8.0 (#122)

Hey Xing,

The Tile policy only works with forallN, not forall…. and I don’t believe that LULESH uses any forallN’s.

-Adam

From: David Poliakoff notifications@github.com<mailto:notifications@github.commailto:notifications@github.com> Reply-To: LLNL/RAJA reply@reply.github.com<mailto:reply@reply.github.commailto:reply@reply.github.com> Date: Thursday, September 1, 2016 at 12:14 PM To: LLNL/RAJA RAJA@noreply.github.com<mailto:RAJA@noreply.github.commailto:RAJA@noreply.github.com> Cc: Adam Kunen kunen1@llnl.gov<mailto:kunen1@llnl.govmailto:kunen1@llnl.gov>, Mention mention@noreply.github.com<mailto:mention@noreply.github.commailto:mention@noreply.github.com> Subject: Re: [LLNL/RAJA] Build fails with CUDA v8.0 (#122)

Hi Xing,

If you're using the forallN constructs (are you part of the VBL effort?), I'm actually not entirely familiar with that segment of the code, so I'll see whether @ajkunenhttps://github.com/ajkunen or @trwshttps://github.com/trws are able to chime in. Sorry I can't be of more help on that front.

Best,

David

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHubhttps://github.com/LLNL/RAJA/issues/122#issuecomment-244182142, or mute the threadhttps://github.com/notifications/unsubscribe-auth/ARGiBDY-Bt_73PYthkVFHsybejh69JwLks5qlyQCgaJpZM4JzAgP.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/LLNL/RAJA/issues/122#issuecomment-244182841, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AMNRnMfeH7ZLquEBBWwD7Bz8taVQ30xyks5qlySggaJpZM4JzAgP.

[DavidPoliakoff]

Xing,

It might be valuable for us to hear what you're trying to achieve so we can steer you towards making that happen. There's a known shortcoming of documentation for RAJA, currently an unfortunate amount of it is "raise an issue on Github and talk to the devs." If you tell us what you're trying to achieve (compiling code that was given to you and getting errors? Trying to optimize code and seeing a construct you thought would work not work?) we could try to get you going.

[xing-liu]

David,

I'm helping LLNL to prepare an application for Sierra, which is using RAJA for GPU parallization. I talked to Jeff on the phone, and he asked me to create an issue on git if I have any questions about RAJA.

Xing

Xing

[trws]

Right, and we're happy to help, but we need to know what the question is @xing-liu. I'm pretty sure the compile error you're seeing will be fixed by PR #91, it has to do with the addition of double-precision atomics to the CUDA API in 8.0. For the time being, you should either build with 7.5 or try a branch incorporating the fixes in #91.

All current branches (everything but example or reproducer branches I think) should support Tile for the ForallN constructs.

[rhornung67]

@xing-liu or @trws or @DavidPoliakoff is this issue resolved?

[davidbeckingsale]

I'm closing this since we don't have any information about the actual bug. Please re-open with more details if necessary.