shyjsarah commented on code in PR #82:
URL:
https://github.com/apache/paimon-vector-index/pull/82#discussion_r3911366324
##########
core/src/pq.rs:
##########
@@ -370,6 +376,85 @@ impl ProductQuantizer {
);
}
+ /// Blocked batch encode for the IVF-PQ add path.
+ pub(crate) fn encode_batch_blocked(&self, data: &[f32], n: usize, codes:
&mut [u8]) {
Review Comment:
**Suggestion (non-blocking):** The fixed `n >= 32` SGEMM threshold may be
too low for wide Rayon pools, where small batches are split into tiny SGEMM
calls. In a local 32-thread benchmark, 32/64-row batches were slower than the
per-vector path. Could the threshold account for the pool width, or could we
add benchmarks covering small batches with different thread counts?
--
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]
