[Trilinos-Users] Trilinos-Users Digest, Vol 116, Issue 9

[Hoemmen, Mark]

On 4/8/15, 12:00 PM, "trilinos-users-request at software.sandia.gov"
<trilinos-users-request at software.sandia.gov> wrote:
>Message: 2
>Date: Wed, 8 Apr 2015 11:33:29 -0400
>From: Zheng Da <zhengda1936 at gmail.com>
>To: "trilinos-users at software.sandia.gov"
>	<trilinos-users at software.sandia.gov>
>Subject: [Trilinos-Users] load a large sparse matrix to a Tpetra CRS
>	matrix	in parallel
>Message-ID:
>	<CAFLer81x1LjUnQqFOsE0M7U_xRENj9D0MX-5WLxPBPzqPso+JA at mail.gmail.com>
>Content-Type: text/plain; charset=UTF-8
>
>Hello,
>
>I want to load a very large sparse matrix (60M rows and columns and
>3.6B non-zero entries) to Tpetra Crs_matrix. Can I load the sparse
>matrix to Tpetra Crs_matrix with multiple threads?
>
>Right now, I'm using insertGlobalValues() to insert non-zero entries
>in each row. It's very slow. I tried to use OpenMP to parallelize it
>and the program crashed. Is there a faster way to load the sparse
>matrix? Currently, the sparse matrix is stored in my own format.

See Tpetra Lesson 07 (in tpetra/core/example) in the public git repository
for an example of thread-parallel sparse matrix construction.

mfh