RFR: 8321137: Reconsider ICStub alignment [v2]
Aleksey Shipilev
shade at openjdk.org
Tue Jan 16 09:56:49 UTC 2024
On Tue, 16 Jan 2024 09:44:09 GMT, Andrew Haley <aph at openjdk.org> wrote:
>> Aleksey Shipilev has updated the pull request incrementally with one additional commit since the last revision:
>>
>> Inline new_ic_stub
>
> src/hotspot/share/code/icBuffer.hpp line 69:
>
>> 67: // cache coherency on some architectures to order the updates to ICStub and setting
>> 68: // the destination to the ICStub. Note that cache lines size might be larger than
>> 69: // CodeEntryAlignment that is a normal alignment for CodeBlobs.
>
> This is a paragraph that raises more questions than it answers. The relationship between ordering and instruction cache line coherency is pretty tenuous even on x86 these days, isn't it? I'd not say "For extra correctness/safety," but "To be cautious,".
Sure. How does the new comment look?
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/17277#discussion_r1453184007
More information about the hotspot-dev
mailing list