[master] RFR: JDK-8325104: Lilliput: Shrink Classpointers [v2]

[Thomas Stuefe]

On Sat, 10 Feb 2024 14:49:00 GMT, Thomas Stuefe <stuefe at openjdk.org> wrote:

>> src/hotspot/cpu/x86/sharedRuntime_x86.cpp line 80:
>> 
>>> 78:   // because it could be larger than 32 bits in a 64-bit vm. See markWord.hpp.
>>> 79:   if (UseCompactObjectHeaders) {
>>> 80:     STATIC_ASSERT(markWord::hash_mask_compact < nth_bit(32));
>> 
>> This makes me think if the i-hash should go/stay in its original position, and we could get rid of hash_code_compact and related changes in Lilliput? Or would that not work for some reason?
>
> It would, but we would have one unused bit in the front, three in the back. I am not sure that is an improvement.

(offlist agreed with Roman to shift hash and nKlass thus that all unused bits are clustered in the MSB.)

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

PR Review Comment: https://git.openjdk.org/lilliput/pull/128#discussion_r1486171845