On Wed, 2 Apr 2025 11:43:40 GMT, Maurizio Cimadamore <mcimadam...@openjdk.org> wrote:
>> I didn't mean to say "this is not true, so use a `static final` instead". >> What I meant was more "let's not dive into VM-specific details now" (see my >> other comment on the topic), and just focus on semantics instead. Then let's >> put all the constant-folding related stuff somewhere else. I liked using >> non-static final fields! > > Basically: drop all references to "performance" and "constant folding" > everywhere. And then introduce a section dedicated to that topic which talks > a bit about how the content of a SV is trusted not to change by the JVM, > which enables constant-folding optimizations. The reason I'm suggesting this is that what's constant-folded is kind of a moving target, so we don't want to say/promise too much. For now, we can say that if the SV, or the stable function/collection is stored in a static/final field then its content will be constant-folded. But we know that there's other stuff in the pipeline (e.g. [this JEP](https://openjdk.org/jeps/8349536)) which will affect this picture, and enable for more constant-folding when accessing SV and stable functions/collections. ------------- PR Review Comment: https://git.openjdk.org/jdk/pull/23972#discussion_r2024659484
