[Trilinos-Users] CUDA compilation

[Bart Janssens]

Yes, the errors are mostly fixable, but it's basically working around
compiler bugs while I'm not even writing any Cuda code myself, so I was
just wondering if it's possible to use the regular compiler when building
against a Trilinos with Cuda. I do want to use the Cuda backend with Tpetra.

Op zo 13 aug. 2017 om 12:18 schreef Dmitry N. Mikushin <maemarcus at gmail.com>

> Do you have -std=c++11 among used NVIDIA compiler options?
>
> 2017-08-13 1:48 GMT+03:00 Bart Janssens <bart at bartjanssens.org>:
>
>> Hi,
>>
>> I'm following
>> https://trilinos.org/about/documentation/building-with-cuda-support/ to
>> build Trilinos with CUDA. Compilation of Trilinos itself seems to succeed,
>> but I'm wondering: do I have to compile my own code also using the
>> nvcc_wrapper? I get lots of errors like "__device__ does not name a type"
>> if I don't, but if I do compile with the wrapper it seems to have
>> difficulties with some of my C++11 template code.
>>
>> Kind regards,
>>
>> Bart
>>
>> _______________________________________________
>> Trilinos-Users mailing list
>> Trilinos-Users at trilinos.org
>> https://trilinos.org/mailman/listinfo/trilinos-users
>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://trilinos.org/pipermail/trilinos-users/attachments/20170813/1a466872/attachment.html>