On Thu, 11 Jun 2026 02:26:32 GMT, Quan Anh Mai <[email protected]> wrote:
>> Noticed that Shenandoah tests are failing in current lworld. This should fix >> most of the issues ahead of mainline integration. There are still plenty of >> TODO items we would need to resolve, but this at very least gives us a >> baseline to improve on. >> >> Additional testing: >> - [x] Linux x86_64 server fastdebug, `hotspot_gc_shenandoah` >> - [x] Linux x86_64 server fastdebug, `hotspot_valhalla jdk_valhalla` >> - [x] Linux x86_64 server fastdebug, `hotspot_valhalla jdk_valhalla` with >> `-XX:+UseShenandoahGC` >> - [x] Linux x86_64 server fastdebug, `hotspot_valhalla jdk_valhalla` with >> `-XX:+UseShenandoahGC -XX:ShenandoahGCHeuristics=aggressive` >> >> --------- >> - [x] I confirm that I make this contribution in accordance with the >> [OpenJDK Interim AI Policy](https://openjdk.org/legal/ai). > > src/hotspot/share/opto/inlinetypenode.cpp line 1381: > >> 1379: do_atomic = false; >> 1380: } >> 1381: if (do_atomic && UseShenandoahGC && vk->contains_oops()) { > > Can you instead reject flattening if the member type contains an oop? Maybe, I would really like to mark the paths where we need more work, like this. It follows the similar precedents in current Valhalla codebase, I think. ------------- PR Review Comment: https://git.openjdk.org/valhalla/pull/1971#discussion_r3393711398
