RFR: 8347405: MergeStores with reverse bytes order value

[kuaiwei]

On Fri, 10 Jan 2025 11:14:09 GMT, Emanuel Peter <epeter at openjdk.org> wrote:

>> Ah, maybe that is because you have not `big-endian` machine to test on. We could leave this to someone who cares about big-endian - and they could also adjust the tests accordingly.
>
> Suggestion:
> 
> #ifdef VM_LITTLE_ENDIAN
>     // For now, we only implement Reverse order for little-endian, and only for bytes.
>     } else if (memory_size == 1 &&
>                Matcher::match_rule_supported(Op_ReverseBytesI) &&
>                Matcher::match_rule_supported(Op_ReverseBytesL)) {
>       _value_order = DataOrder::Reverse;  // First pair has Reverse order.
> #endif

> Ah, maybe that is because you have not `big-endian` machine to test on. We could leave this to someone who cares about big-endian - and they could also adjust the tests accordingly.

Yes, I'm not sure work correctly on big endian machine. So I only enable it for little endian mode.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/23030#discussion_r1910332094