You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Following #262, I executed the version regression benchmark suite. The current results show significant write performance regression after release 0.8.3.
I subsequently developed an automated bisection capability in the LmdbJava Benchmarks project. This identified the regression:
Commit Date Score Change Message
========================================================================================================================
be2a15b 2023-02-04 00:28:16 89.07 [maven-release-plugin] prepare release lmdbjava-0.8.3
1cc7e6c 2023-12-04 07:42:08 94.49 +6.09% Test on Java 21
f17b63f 2023-12-05 01:34:17 91.57 -3.09% [maven-release-plugin] prepare release lmdbjava-0.9.0
b5dfb25 2025-02-16 23:37:16 91.38 -0.21% Uplift lmdb from 0.9.29 to 0.9.31 (#245)
* 68f0a44 2025-02-16 23:46:30 128.63 +40.76% Add DbiFlags#MDB_UNSIGNEDKEY to allow to compare byte array, Byt...
928cf7b 2025-02-20 01:26:13 129.12 +0.38% Add url element to pom.xml to satisfy OSSRH deployment rules
dc24f4b 2025-10-31 00:20:59 122.03 -5.49% Generalise benchmark link
The regression emerged in commit 68f0a44, which modifies comparator logic. This represents a 40.76% write performance degradation from the previous commit.
Given @at055612 is undertaking significant comparator changes in #249 and #250, this regression is likely to be addressed in that work. The bisection tooling is now available to verify performance impact of changes.
Following #262, I executed the version regression benchmark suite. The current results show significant write performance regression after release 0.8.3.
I subsequently developed an automated bisection capability in the LmdbJava Benchmarks project. This identified the regression:
The regression emerged in commit 68f0a44, which modifies comparator logic. This represents a 40.76% write performance degradation from the previous commit.
Given @at055612 is undertaking significant comparator changes in #249 and #250, this regression is likely to be addressed in that work. The bisection tooling is now available to verify performance impact of changes.