On Mon, 3 Nov 2025 12:57:46 GMT, Chen Liang <[email protected]> wrote:
> > I imagine this problem could be solved in C2? > > The root cause here is C2 is generating subpar code due to it not able to > obtain specialized/inlined profiling data. We must fix it by updating the > profiling data structure. I don't think this is a problem for static methods that don't call non-static methods (that's the problem of Objects::equals etc). The only case that could be relevant is when a bytecode can result in an exception (e.g., div by zero). But without seeing any conclusive benchmarks, that really shouldn't be a blocker. ------------- PR Comment: https://git.openjdk.org/jdk/pull/28081#issuecomment-3480480052
