Copilot commented on code in PR #13115:
URL: https://github.com/apache/gluten/pull/13115#discussion_r4120530738
##########
cpp/velox/compute/VeloxBackend.cc:
##########
@@ -254,6 +255,10 @@ void VeloxBackend::init(
std::make_unique<folly::CPUThreadPoolExecutor>(ioThreads,
folly::CPUThreadPoolExecutor::makeLifoSemQueue());
}
+ const auto hashTableBuildThreads =
+ std::max<int32_t>(1, backendConf_->get<int32_t>(kHashTableBuildThreads,
numTaskSlotsPerExecutor));
+ hashTableBuildExecutor_ =
std::make_unique<folly::CPUThreadPoolExecutor>(hashTableBuildThreads);
Review Comment:
The pool is constructed unconditionally during backend initialization, so
every Velox backend creates this extra CPU pool—even applications that never
execute a parallel broadcast build. This contradicts the stated lazy creation
and adds avoidable thread/resource overhead at executor startup; retain the
configured size but create the pool in a call-once accessor, following
`getReaderThreadPool()`, and reset it during teardown.
--
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]
