Add release notes for PR #681#683
Conversation
r-devulap
requested review from
MarkWolters,
ashkrisk,
jshook and
tlwillke
as code owners
|
Before you submit for review:
If you did not complete any of these, then please explain below. |
| per-element scalar writes with bulk operations, significantly reducing | ||
| serialization overhead. |
There was a problem hiding this comment.
Have we observed index build time improvements on systems with a single NUMA node? If not we might want to edit the release notes to reflect that.
| @@ -0,0 +1,28 @@ | |||
| ## Performance Improvements: Vector Serialization | |||
There was a problem hiding this comment.
Release notes should not contain any ##-level tags, refer the Release Notes Readme.
| per-element scalar writes with bulk operations, significantly reducing | ||
| serialization overhead. | ||
|
|
||
| ### Key Changes |
There was a problem hiding this comment.
I think this section can be omitted given that it discusses implementation details. The summary already mentions that per-element scalar writes are replaced by bulk operations, and anyone interested in exactly how it's done can refer the PR.
|
|
||
| ### Index Build Improvements | ||
| - **openai-1536-1m:** ~57% faster (105.59s → 45.22s) | ||
| - **openai-3072-1m:** ~38% faster (164.63s → 102.75s) |
There was a problem hiding this comment.
If we mention hard numbers we should probably mention the hardware we're using.
| - **openai-1536-1m:** ~57% faster (105.59s → 45.22s) | ||
| - **openai-3072-1m:** ~38% faster (164.63s → 102.75s) | ||
|
|
||
| ### Microbenchmark Summary |
There was a problem hiding this comment.
IMO microbenchmarks are unnecessary for the release notes, but I don't have strong feelings about it.
2 participants
Adding release notes for #681