RFR: 8367005: ImageReader refactor caused performance regressions for startup and footprint [v2]

[Claes Redestad]

On Thu, 11 Sep 2025 11:56:30 GMT, Claes Redestad <redestad at openjdk.org> wrote:

>> I had thought of these micro optimizations, but I thought we preferred clarity unless performance was proven to be an issue. I see contains("<single-char>") in other code - should we be changing that?
>
> I think replacing `startsWith(String)` with `indexOf(char)` is dubious, since the latter is O(n) and the former O(1) 
> 
> If you _really_ need to squeeze out everything on startup then `(x.length() > 0 && x.charAt(0) == '/')` is slightly faster in the interpreter (the difference more or less disappears after JIT compilation). I'd keep using `startsWith("/")`.

(Yes, `x.lastIndexOf('/', 0) == 0` is even faster for the interpreter, but let's keep it civil in these trying times.)

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

PR Review Comment: https://git.openjdk.org/jdk/pull/27203#discussion_r2340467282