RFR: 8324751: C2 SuperWord: Aliasing Analysis runtime check [v11]

Vladimir Kozlov kvn at openjdk.org
Mon Aug 18 14:41:23 UTC 2025


On Mon, 18 Aug 2025 07:49:22 GMT, Emanuel Peter <epeter at openjdk.org> wrote:

>> It is a DIAGNOSTIC flag, which allowed me to demonstrate the performance in a JMH benchmark. You asked for that benchmark back when I first introduced multiversioning with https://github.com/openjdk/jdk/pull/22016 . I'm also fine removing the flag completely now. Or just making it develop. What do you think is best?
>
> It is now used in the JMH benchmark, I'd have to remove it there too:
> https://github.com/openjdk/jdk/pull/24278/files#diff-93288fabe20d76b9df3fb5601e4d8600a46f438fe4b9c4ef92d702fdffa1c8c9R225-R230

It is not about diagnostic vs develop. It is about to keep it or not.

It is fine to have this during development of these changes to prove that we need to optimize slow loops too. 
But now we know that we need to do the optimization. The only reason is to keep it is for debugging some future  failures.  You can keep it if you think it is very useful for debugging.

My main concern is we have too many flags we never use and we are adding more.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/24278#discussion_r2282607286


More information about the hotspot-compiler-dev mailing list