[foreign-memaccess+abi] RFR: MemorySegmentPool + Allocator [v13]

Maurizio Cimadamore mcimadamore at openjdk.org
Tue Dec 19 11:07:09 UTC 2023


On Mon, 4 Sep 2023 17:59:58 GMT, Maurizio Cimadamore <mcimadamore at openjdk.org> wrote:

>> Radoslaw Smogura has updated the pull request incrementally with one additional commit since the last revision:
>> 
>>   Make acquire of parent (pool scope) optional - 5ns per ops / down.
>>   
>>   ```StrLenTest.panama_strlen_arena                               5  avgt   10   35.290 ?   1.756  ns/op
>>     StrLenTest.panama_strlen_arena                              20  avgt   10   54.778 ?   3.094  ns/op
>>     StrLenTest.panama_strlen_arena                             100  avgt    3   97.301 ? 137.266  ns/op
>>     StrLenTest.panama_strlen_memsegmentpool_allocator            5  avgt   10   46.160 ?   0.654  ns/op
>>     StrLenTest.panama_strlen_memsegmentpool_allocator           20  avgt   10   52.966 ?   1.095  ns/op
>>     StrLenTest.panama_strlen_memsegmentpool_allocator          100  avgt   10   66.377 ?   0.878  ns/op
>>     StrLenTest.panama_strlen_memsegmentpool_allocator_copy       5  avgt   10   47.412 ?   1.073  ns/op
>>     StrLenTest.panama_strlen_memsegmentpool_allocator_copy      20  avgt   10   52.349 ?   1.296  ns/op
>>     StrLenTest.panama_strlen_memsegmentpool_allocator_copy     100  avgt   10   79.623 ?   1.616  ns/op
>>     StrLenTest.panama_strlen_prefix                              5  avgt   10   27.352 ?   0.659  ns/op
>>     StrLenTest.panama_strlen_prefix                             20  avgt   10   33.634 ?   0.081  ns/op
>>     StrLenTest.panama_strlen_prefix                            100  avgt   10   49.974 ?   0.097  ns/op
>>     StrLenTest.panama_strlen_unsafe                              5  avgt   10   51.266 ?   0.558  ns/op
>>     StrLenTest.panama_strlen_unsafe                             20  avgt   10   63.194 ?   0.360  ns/op
>>     StrLenTest.panama_strlen_unsafe                            100  avgt   10   60.800 ?   0.821  ns/op
>>     StrLenTest.panama_strlen_unsafe_trivial                      5  avgt   10   27.343 ?   0.094  ns/op
>>     StrLenTest.panama_strlen_unsafe_trivial                     20  avgt   10   32.997 ?   0.383  ns/op
>>     StrLenTest.panama_strlen_unsafe_trivial                    100  avgt   10   49.065 ?   0.756  ns/op
>>    ```
>
> Ping: does this PR needs to be open? (Perhaps it might be useful to update and make it a draft JEP, as an example of how the allocator API could be used?)

> Hi @mcimadamore - I hope you have good day. I wonder if this PR should be kept or closed. I remember that plan was to introduce it after stabilisation of API, but not sure if this plan have not changed?

Hi @rsmogura - while we do have plans to include more allocators/arenas out of the box, I'm not too sure a PR would be the best way to keep this initiative alive. I'd say let's close the PR (otherwise folks might thing this is going to get integrated soon, which might be misleading), and maybe drop the code in some branch, where it can be kept in sync with mainline?

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

PR Comment: https://git.openjdk.org/panama-foreign/pull/509#issuecomment-1862554171


More information about the panama-dev mailing list