You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@mahout.apache.org by "Dmitriy Lyubimov (JIRA)" <ji...@apache.org> on 2015/06/08 22:58:01 UTC
[jira] [Updated] (MAHOUT-1724) Optimizations of matrix-matrix
in-core multiplication based on structural flavors
[ https://issues.apache.org/jira/browse/MAHOUT-1724?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Dmitriy Lyubimov updated MAHOUT-1724:
-------------------------------------
Description:
Current times() operation suffers a lot. Especially when e.g. column-major iterations are done over a matrix backed by row-major sequential sparse vectors, things are dropping orders of magnitude in performance.
This issue finds optimal traversal gemm kernel order for current matrix; or rewrites matrix prior to operation to enable optimal traversals.
PR https://github.com/apache/mahout/pull/135
was:
Current times() operation suffers a lot. Especially when e.g. column-major iterations are done over a matrix backed by row-major sequential sparse vectors, things are dropping orders of magnitude in performance.
This issue finds optimal traversal gemm kernel order for current matrix; or rewrites matrix prior to operation to enable optimal traversals.
> Optimizations of matrix-matrix in-core multiplication based on structural flavors
> ---------------------------------------------------------------------------------
>
> Key: MAHOUT-1724
> URL: https://issues.apache.org/jira/browse/MAHOUT-1724
> Project: Mahout
> Issue Type: Improvement
> Reporter: Dmitriy Lyubimov
> Assignee: Dmitriy Lyubimov
> Fix For: 0.10.2
>
>
> Current times() operation suffers a lot. Especially when e.g. column-major iterations are done over a matrix backed by row-major sequential sparse vectors, things are dropping orders of magnitude in performance.
> This issue finds optimal traversal gemm kernel order for current matrix; or rewrites matrix prior to operation to enable optimal traversals.
> PR https://github.com/apache/mahout/pull/135
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)