Details
-
Improvement
-
Status: Resolved
-
Major
-
Resolution: Fixed
-
None
Description
Per the review comments of https://github.com/apache/arrow/pull/7607
"Instead, we should add all the kernel variants to the same Function instance and then the Dispatch* methods should select the kernel with the maximum SIMD level as set on the Kernel object. That was the idea of the simd_level parameter"
We should delegate the SIMD sum kernels to function instance
Attachments
Issue Links
- links to