-
Type:
Enhancement
-
Resolution: Unresolved
-
Priority:
P4
-
Affects Version/s: None
-
Component/s: hotspot
Lucene folks ran into the absence of high-performance dot product support:
https://github.com/apache/lucene/pull/13572
They apparently end up calling into native implementations, where GCC is able to auto-vectorize to hardware instructions like AArch64 SDOT:
https://developer.arm.com/documentation/ddi0602/2025-12/SME-Instructions/SDOT--4-way--multiple-and-indexed-vector---Multi-vector-signed-integer-dot-product-by-indexed-element-
This might be a good addition to Vector APIs.
https://github.com/apache/lucene/pull/13572
They apparently end up calling into native implementations, where GCC is able to auto-vectorize to hardware instructions like AArch64 SDOT:
https://developer.arm.com/documentation/ddi0602/2025-12/SME-Instructions/SDOT--4-way--multiple-and-indexed-vector---Multi-vector-signed-integer-dot-product-by-indexed-element-
This might be a good addition to Vector APIs.