-
Notifications
You must be signed in to change notification settings - Fork 73
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
How to make BIDMat as the GPU Backend for Spark? #22
Comments
The challenge is to include all the jars with native libs that BIDMat uses. These are currently: All of the native libs except HDF5 can be bundled into a Jar, and will automatically be unpacked when the calling class is loaded. The JVM will need access to a tmp directory to do the unpacking. HD5 doesnt support this (or didnt as of the version we use). I dont remember if it can run at all without the native libs. I dont think so. It may not be too relevant if running on Spark where I/O is managed by Spark. Does Spark have LZ4 support already? It might be better to make a single assembly jar with all of the dependent jars and native libs are included. Otherwise someone would have to commit to maintaining repo copies of the dependencies. |
@jcanny 【?】 means I am not sure, could you please kindly confirm it? Tks. |
We should talk on the phone so I"m clear on what you're trying to do -John
|
I have a team working on this now, so it should be resolved over the next few months. |
@jcanny Hi. Any resolution on that? do you have open source code for coupling BidMat with Spark? thanks. |
@jcanny I wanna leverage the GPU Resources in Spark.
4.0.0 org.apache apache 14 org.apache.spark spark-parent 1.1.0 pom Spark Project Parent POM http://spark.apache.org/ Apache 2.0 License http://www.apache.org/licenses/LICENSE-2.0.html repo scm:git:[email protected]:apache/spark.git scm:git:https://git-wip-us.apache.org/repos/asf/spark.git scm:git:[email protected]:apache/spark.git v1.1.0-rc4 matei Matei Zaharia [email protected] http://www.cs.berkeley.edu/~matei Apache Software Foundation http://spark.apache.org JIRA https://issues.apache.org/jira/browse/SPARKFor example use GPU to do some Matrix Computation.
I am thinking about how to configure to ake BIDMat as the GPU Backend for Spark?
Likely, I use Maven, How to add sth in POM.XML (Attached)
The text was updated successfully, but these errors were encountered: