marin-ma opened a new pull request, #12090:
URL: https://github.com/apache/gluten/pull/12090

   Currently, GPULock allows only one Spark task to acquire GPU resources at a 
time, which effectively makes GPU offloading run in single-threaded mode within 
a Spark executor. However, when testing the TPCDS queries, we observed that GPU 
utilization and memory usage can be very low (< 20%), meaning GPU resources are 
underutilized.
   
   To improve GPU utilization, this PR adds support for running multiple GPU 
tasks concurrently through the configuration 
`spark.gluten.sql.columnar.backend.velox.cudf.concurrentGpuTasks`. The default 
value is 1.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to