RFR: Save/load nmethod without going through CodeBuffer [v2]

Vladimir Kozlov kvn at openjdk.org
Wed Mar 19 16:34:25 UTC 2025


On Wed, 19 Mar 2025 07:04:09 GMT, Ashutosh Mehra <asmehra at openjdk.org> wrote:

>>> do you know why "New CDS + AOT" results for SpringBoot is worse than before: geo mean 532.28 vs 512.17?
>> 
>> I noticed this but at the moment I don't know the reason for this increase in startup time.
>> The `SC Load Time` has gone down on expected lines. So may be there are other factors playing a role here.
>> I am trying to get my earlier prototype working with latest premain to see if I still get the same numbers as before.
>
>> I am trying to get my earlier prototype working with latest premain to see if I still get the same numbers as before.
> 
> @vnkozlov I used the first prototype but I am not getting same numbers as mentioned in https://github.com/openjdk/leyden/pull/27#issue-2731492066
> The startup time and `SC Load time` numbers are comparable to what I am getting with the current patch. So I don't think the current patch is any worse than the earlier prototype (and it shouldn't be!).

@ashu-mehra  how complex to do the same for adapters?

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

PR Comment: https://git.openjdk.org/leyden/pull/27#issuecomment-2737312131


More information about the leyden-dev mailing list