On Tue, 19 May 2026 07:40:52 GMT, Tobias Hartmann <[email protected]> wrote:

>> src/hotspot/share/opto/inlinetypenode.cpp line 2331:
>> 
>>> 2329:   }
>>> 2330: 
>>> 2331:   Node* shift_val = igvn.intcon(offset << LogBitsPerByte);
>> 
>> The way values are packed and then read or written in the x86 and aarch64 
>> back-end seems to imply little-endian.  For big-endian, it seems like we 
>> need to pack in different order, or force the back-end to byte-swap.
>
> Yes, current code around flattening is all little-endian. Once flattening is 
> implemented on a big-endian platform, we need to adjust this.

(resolved)

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

PR Review Comment: https://git.openjdk.org/jdk/pull/31122#discussion_r3320969048

Reply via email to