That memory allocation is happening at the bottom of the call stack is not 
ideal -- we should not write any more such kernels as we're just creating 
technical debt. 

Let's add a TODO to return to these and enable them to write into preallocated 
memory, otherwise pipelining (invoking multiple kernels in a row without 
intermediate allocations) is not possible

[ Full content available at: https://github.com/apache/arrow/pull/2461 ]
This message was relayed via gitbox.apache.org for [email protected]

Reply via email to