systemml-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Matthias Boehm (JIRA)" <j...@apache.org>
Subject [jira] [Resolved] (SYSTEMML-1955) Ultra-sparse matrix multiply for large ultra-sparse matrices
Date Sat, 14 Oct 2017 06:21:00 GMT

     [ https://issues.apache.org/jira/browse/SYSTEMML-1955?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Matthias Boehm resolved SYSTEMML-1955.
--------------------------------------
       Resolution: Done
         Assignee: Matthias Boehm
    Fix Version/s: SystemML 1.0

> Ultra-sparse matrix multiply for large ultra-sparse matrices
> ------------------------------------------------------------
>
>                 Key: SYSTEMML-1955
>                 URL: https://issues.apache.org/jira/browse/SYSTEMML-1955
>             Project: SystemML
>          Issue Type: Sub-task
>            Reporter: Matthias Boehm
>            Assignee: Matthias Boehm
>             Fix For: SystemML 1.0
>
>
> Our code path for ultra-sparse matrix multiplies has a condition on the sparsity but
also the absolute number of non-zeros in order to avoid unnecessary overhead in CP where the
entire matrix is a single block. However, there are use cases, where this approach leads to
large allocated dense outputs although the output is sparse as well.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

Mime
View raw message