RFR: 8290892: C2: Intrinsify Reference.reachabilityFence [v3]

Emanuel Peter epeter at openjdk.org
Mon Sep 8 13:34:30 UTC 2025


On Wed, 3 Sep 2025 20:28:40 GMT, Vladimir Ivanov <vlivanov at openjdk.org> wrote:

>> Can you quickly comment why you changed this?
>
> Some call nodes inspected during `expand_reachability_fences` demonstrate this IR shape where some exception table projections are directly attached to the call node.
> 
> Looks like a missed case in `CallNode::extract_projections` we simply never hit before.

Alright, sounds good! Do you think this could have happened somehow, i.e. was this a bug that we could somehow reproduce?

>> The arguments are less important for me.
>
> There are 2 types of methods here: internal ones (used solely in `reachability.cpp`) and those which are called from loop optimization code (`optimize_reachability_fences` and `eliminate_reachability_fences`). 
> 
> IMO it's counter-productive to repeatedly spell out what "RF" means inside `reachability.cpp`, so I kept the names intact. I split the declarations into public and private ones to stress the distinction.

Great, the private/public split works for me :)

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/25315#discussion_r2330151757
PR Review Comment: https://git.openjdk.org/jdk/pull/25315#discussion_r2330213859


More information about the hotspot-compiler-dev mailing list