RFR: 8255274: [PPC64, s390] wrong StringLatin1.indexOf version matched

[Martin Doerr]

On Fri, 23 Oct 2020 09:07:41 GMT, Richard Reingruber <rrich at openjdk.org> wrote:

>> Thanks for fixing this!
>
> The fix looks correct.
> But you are adding another 20 lines to an already humonguous file by duplicating a complex instruction form and the diff between the 2 variants is very small. This makes the file hardly readable... at least for humans.
> Wouldn't it be possible to have just one instruction form that matches both variants and feed the predicate expression to the `is_byte` parameter of `string_indexof_char`?

Thanks for the reviews.
@reinrich That would be nice, but I'm not aware of an easy way to do this. I can't use "((StrIndexOfCharNode*)n)->encoding()" in the "ins_encode" because the Ideal node is only visible during "match".

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

PR: https://git.openjdk.java.net/jdk/pull/820