[Trilinos-Users] Trilinos with GPU

[Hoemmen, Mark]

Tpetra uses Kokkos::Cuda for shared-memory parallelism on NVIDIA GPUs.  This, in turn, uses raw CUDA.  It does not use Thrust.

See tpetra/core/example for examples, including one for simple finite elements.

mfh






On 6/6/17, 3:11 PM, "Trilinos-Users on behalf of trilinos-users-request at trilinos.org" <trilinos-users-bounces at trilinos.org on behalf of trilinos-users-request at trilinos.org> wrote:

>Date: Mon, 29 May 2017 18:34:04 +0200
>From: Deepak Garg <deepak.garg at ingv.it>
>To: trilinos-users at trilinos.org
>Subject: [Trilinos-Users] Trilinos with GPU
>Message-ID:
>	<CAJFVxiAPbehvofWqcjDdvXp5zb_K7LTcw7naF2Q+S5dWqfaE3Q at mail.gmail.com>
>Content-Type: text/plain; charset="utf-8"
>
>Hi there
>
>I have a parallel finite element code based on openmpi, epetra and belos
>packages of trilinos. I want to convert it into a hybrid parallel code
>using tpetra(cpu) and cuda (gpu).
>
>Does tpetra supports Thrust library for shared memory parallelism?
>Are there any examples and documentation within trilinos which can explain
>how to do hybrid programming for finite elements?
>
>Any suggestion will be a great help.
>Thanks in advance
>Deepak Garg
>