Rajesh Bordawekar created SPARK-13144:
-----------------------------------------
Summary: Enabling efficient and transparent use of GPUs for
accelerating MLLib functions
Key: SPARK-13144
URL: https://issues.apache.org/jira/browse/SPARK-13144
Project: Spark
Issue Type: Improvement
Reporter: Rajesh Bordawekar
This JIRA is associated with transperently exploiting accelerators such as GPUs
for improving MLLib performance. The key constraint of this approach is to
enable acceleration of MLLib kernels without any change in the code. The
proposed work initially focusses only on the GPUs and covers the following
aspects: (1) Developing and integrating GPU kernels to accelerate key MLLib
functions for multi-node multi-cpu environment- for both in-core and
out-of-core scenario, (2) Designing and implementing runtime support for data
layout and efficient data transfer between Spark nodes, executors, individual
threads and multiple GPUs, and (3) Designing multi-GPU resource management and
information system.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
